kernel-aufs.patch

   1 SPDX-License-Identifier: GPL-2.0
   2 aufs6.x-rcN kbuild patch
   3
   4 diff --git a/fs/Kconfig b/fs/Kconfig
   5 index 2685a4d0d353..a2df3d4bb3aa 100644
   6 --- a/fs/Kconfig
   7 +++ b/fs/Kconfig
   8 @@ -319,6 +319,7 @@ source "fs/sysv/Kconfig"
   9  source "fs/ufs/Kconfig"
  10  source "fs/erofs/Kconfig"
  11  source "fs/vboxsf/Kconfig"
  12 +source "fs/aufs/Kconfig"
  13
  14  endif # MISC_FILESYSTEMS
  15
  16 diff --git a/fs/Makefile b/fs/Makefile
  17 index 4dea17840761..9795cbfdd0bf 100644
  18 --- a/fs/Makefile
  19 +++ b/fs/Makefile
  20 @@ -137,3 +137,4 @@ obj-$(CONFIG_EFIVAR_FS)             += efivarfs/
  21  obj-$(CONFIG_EROFS_FS)         += erofs/
  22  obj-$(CONFIG_VBOXSF_FS)                += vboxsf/
  23  obj-$(CONFIG_ZONEFS_FS)                += zonefs/
  24 +obj-$(CONFIG_AUFS_FS)           += aufs/
  25 SPDX-License-Identifier: GPL-2.0
  26 aufs6.x-rcN base patch
  27
  28 diff --git a/MAINTAINERS b/MAINTAINERS
  29 index f61eb221415b..b192cff97244 100644
  30 --- a/MAINTAINERS
  31 +++ b/MAINTAINERS
  32 @@ -3522,6 +3522,19 @@ F:       include/uapi/linux/audit.h
  33  F:     kernel/audit*
  34  F:     lib/*audit.c
  35
  36 +AUFS (advanced multi layered unification filesystem) FILESYSTEM
  37 +M:     "J. R. Okajima" <hooanon05g@gmail.com>
  38 +L:     aufs-users@lists.sourceforge.net (members only)
  39 +L:     linux-unionfs@vger.kernel.org
  40 +S:     Supported
  41 +W:     http://aufs.sourceforge.net
  42 +T:     git://github.com/sfjro/aufs4-linux.git
  43 +F:     Documentation/ABI/testing/debugfs-aufs
  44 +F:     Documentation/ABI/testing/sysfs-aufs
  45 +F:     Documentation/filesystems/aufs/
  46 +F:     fs/aufs/
  47 +F:     include/uapi/linux/aufs_type.h
  48 +
  49  AUXILIARY DISPLAY DRIVERS
  50  M:     Miguel Ojeda <ojeda@kernel.org>
  51  S:     Maintained
  52 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
  53 index 1518a6423279..7a9928c6db9d 100644
  54 --- a/drivers/block/loop.c
  55 +++ b/drivers/block/loop.c
  56 @@ -645,6 +645,24 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
  57         goto done;
  58  }
  59
  60 +/*
  61 + * for AUFS
  62 + * no get/put for file.
  63 + */
  64 +struct file *loop_backing_file(struct super_block *sb)
  65 +{
  66 +       struct file *ret;
  67 +       struct loop_device *l;
  68 +
  69 +       ret = NULL;
  70 +       if (MAJOR(sb->s_dev) == LOOP_MAJOR) {
  71 +               l = sb->s_bdev->bd_disk->private_data;
  72 +               ret = l->lo_backing_file;
  73 +       }
  74 +       return ret;
  75 +}
  76 +EXPORT_SYMBOL_GPL(loop_backing_file);
  77 +
  78  /* loop sysfs attributes */
  79
  80  static ssize_t loop_attr_show(struct device *dev, char *page,
  81 diff --git a/fs/dcache.c b/fs/dcache.c
  82 index 52e6d5fdab6b..519321f32f95 100644
  83 --- a/fs/dcache.c
  84 +++ b/fs/dcache.c
  85 @@ -1345,7 +1345,7 @@ enum d_walk_ret {
  86   *
  87   * The @enter() callbacks are called with d_lock held.
  88   */
  89 -static void d_walk(struct dentry *parent, void *data,
  90 +void d_walk(struct dentry *parent, void *data,
  91                    enum d_walk_ret (*enter)(void *, struct dentry *))
  92  {
  93         struct dentry *this_parent;
  94 diff --git a/fs/fcntl.c b/fs/fcntl.c
  95 index 146c9ab0cd4b..c980a0524118 100644
  96 --- a/fs/fcntl.c
  97 +++ b/fs/fcntl.c
  98 @@ -33,7 +33,7 @@
  99
 100  #define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
 101
 102 -static int setfl(int fd, struct file * filp, unsigned long arg)
 103 +int setfl(int fd, struct file *filp, unsigned long arg)
 104  {
 105         struct inode * inode = file_inode(filp);
 106         int error = 0;
 107 @@ -63,6 +63,8 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
 108
 109         if (filp->f_op->check_flags)
 110                 error = filp->f_op->check_flags(arg);
 111 +       if (!error && filp->f_op->setfl)
 112 +               error = filp->f_op->setfl(filp, arg);
 113         if (error)
 114                 return error;
 115
 116 diff --git a/fs/namespace.c b/fs/namespace.c
 117 index ab467ee58341..8cd34f760f63 100644
 118 --- a/fs/namespace.c
 119 +++ b/fs/namespace.c
 120 @@ -963,6 +963,12 @@ static inline int check_mnt(struct mount *mnt)
 121         return mnt->mnt_ns == current->nsproxy->mnt_ns;
 122  }
 123
 124 +/* for aufs, CONFIG_AUFS_BR_FUSE */
 125 +int is_current_mnt_ns(struct vfsmount *mnt)
 126 +{
 127 +       return check_mnt(real_mount(mnt));
 128 +}
 129 +
 130  /*
 131   * vfsmount lock must be held for write
 132   */
 133 diff --git a/fs/splice.c b/fs/splice.c
 134 index 5969b7a1d353..21db9b057d60 100644
 135 --- a/fs/splice.c
 136 +++ b/fs/splice.c
 137 @@ -756,8 +756,8 @@ static int warn_unsupported(struct file *file, const char *op)
 138  /*
 139   * Attempt to initiate a splice from pipe to file.
 140   */
 141 -static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
 142 -                          loff_t *ppos, size_t len, unsigned int flags)
 143 +long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
 144 +                   loff_t *ppos, size_t len, unsigned int flags)
 145  {
 146         if (unlikely(!out->f_op->splice_write))
 147                 return warn_unsupported(out, "write");
 148 @@ -767,9 +767,9 @@ static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
 149  /*
 150   * Attempt to initiate a splice from a file to a pipe.
 151   */
 152 -static long do_splice_to(struct file *in, loff_t *ppos,
 153 -                        struct pipe_inode_info *pipe, size_t len,
 154 -                        unsigned int flags)
 155 +long do_splice_to(struct file *in, loff_t *ppos,
 156 +                 struct pipe_inode_info *pipe, size_t len,
 157 +                 unsigned int flags)
 158  {
 159         unsigned int p_space;
 160         int ret;
 161 diff --git a/include/linux/fs.h b/include/linux/fs.h
 162 index 066555ad1bf8..2d30def9a580 100644
 163 --- a/include/linux/fs.h
 164 +++ b/include/linux/fs.h
 165 @@ -1383,6 +1383,7 @@ extern void fasync_free(struct fasync_struct *);
 166  /* can be called from interrupts */
 167  extern void kill_fasync(struct fasync_struct **, int, int);
 168
 169 +extern int setfl(int fd, struct file *filp, unsigned long arg);
 170  extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
 171  extern int f_setown(struct file *filp, unsigned long arg, int force);
 172  extern void f_delown(struct file *filp);
 173 @@ -2110,6 +2111,7 @@ struct file_operations {
 174         ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
 175         unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
 176         int (*check_flags)(int);
 177 +       int (*setfl)(struct file *, unsigned long);
 178         int (*flock) (struct file *, int, struct file_lock *);
 179         ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
 180         ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
 181 diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
 182 index 1f1099dac3f0..81d10b8e9dc0 100644
 183 --- a/include/linux/lockdep.h
 184 +++ b/include/linux/lockdep.h
 185 @@ -248,6 +248,8 @@ static inline int lockdep_match_key(struct lockdep_map *lock,
 186         return lock->key == key;
 187  }
 188
 189 +struct lock_class *lockdep_hlock_class(struct held_lock *hlock);
 190 +
 191  /*
 192   * Acquire a lock.
 193   *
 194 diff --git a/include/linux/mnt_namespace.h b/include/linux/mnt_namespace.h
 195 index 8f882f5881e8..6b9808f09843 100644
 196 --- a/include/linux/mnt_namespace.h
 197 +++ b/include/linux/mnt_namespace.h
 198 @@ -7,12 +7,15 @@ struct mnt_namespace;
 199  struct fs_struct;
 200  struct user_namespace;
 201  struct ns_common;
 202 +struct vfsmount;
 203
 204  extern struct mnt_namespace *copy_mnt_ns(unsigned long, struct mnt_namespace *,
 205                 struct user_namespace *, struct fs_struct *);
 206  extern void put_mnt_ns(struct mnt_namespace *ns);
 207  extern struct ns_common *from_mnt_ns(struct mnt_namespace *);
 208
 209 +extern int is_current_mnt_ns(struct vfsmount *mnt);
 210 +
 211  extern const struct file_operations proc_mounts_operations;
 212  extern const struct file_operations proc_mountinfo_operations;
 213  extern const struct file_operations proc_mountstats_operations;
 214 diff --git a/include/linux/splice.h b/include/linux/splice.h
 215 index a55179fd60fc..8e21c53cf883 100644
 216 --- a/include/linux/splice.h
 217 +++ b/include/linux/splice.h
 218 @@ -93,4 +93,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
 219
 220  extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
 221  extern const struct pipe_buf_operations default_pipe_buf_ops;
 222 +
 223 +extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
 224 +                          loff_t *ppos, size_t len, unsigned int flags);
 225 +extern long do_splice_to(struct file *in, loff_t *ppos,
 226 +                        struct pipe_inode_info *pipe, size_t len,
 227 +                        unsigned int flags);
 228  #endif
 229 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
 230 index e3375bc40dad..c1f0aadde539 100644
 231 --- a/kernel/locking/lockdep.c
 232 +++ b/kernel/locking/lockdep.c
 233 @@ -217,7 +217,7 @@ unsigned long max_lock_class_idx;
 234  struct lock_class lock_classes[MAX_LOCKDEP_KEYS];
 235  DECLARE_BITMAP(lock_classes_in_use, MAX_LOCKDEP_KEYS);
 236
 237 -static inline struct lock_class *hlock_class(struct held_lock *hlock)
 238 +inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
 239  {
 240         unsigned int class_idx = hlock->class_idx;
 241
 242 @@ -238,6 +238,7 @@ static inline struct lock_class *hlock_class(struct held_lock *hlock)
 243          */
 244         return lock_classes + class_idx;
 245  }
 246 +#define hlock_class(hlock) lockdep_hlock_class(hlock)
 247
 248  #ifdef CONFIG_LOCK_STAT
 249  static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
 250 SPDX-License-Identifier: GPL-2.0
 251 aufs6.x-rcN mmap patch
 252
 253 diff --git a/fs/proc/base.c b/fs/proc/base.c
 254 index 9e479d7d202b..986c2ae12f8b 100644
 255 --- a/fs/proc/base.c
 256 +++ b/fs/proc/base.c
 257 @@ -2218,7 +2218,7 @@ static int map_files_get_link(struct dentry *dentry, struct path *path)
 258         rc = -ENOENT;
 259         vma = find_exact_vma(mm, vm_start, vm_end);
 260         if (vma && vma->vm_file) {
 261 -               *path = vma->vm_file->f_path;
 262 +               *path = vma_pr_or_file(vma)->f_path;
 263                 path_get(path);
 264                 rc = 0;
 265         }
 266 diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
 267 index 4d3493579458..42edd9a42c78 100644
 268 --- a/fs/proc/nommu.c
 269 +++ b/fs/proc/nommu.c
 270 @@ -39,7 +39,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
 271         file = region->vm_file;
 272
 273         if (file) {
 274 -               struct inode *inode = file_inode(region->vm_file);
 275 +               struct inode *inode;
 276 +
 277 +               file = vmr_pr_or_file(region);
 278 +               inode = file_inode(file);
 279                 dev = inode->i_sb->s_dev;
 280                 ino = inode->i_ino;
 281         }
 282 diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
 283 index e35a0398db63..d9080329fd7a 100644
 284 --- a/fs/proc/task_mmu.c
 285 +++ b/fs/proc/task_mmu.c
 286 @@ -285,7 +285,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma)
 287         const char *name = NULL;
 288
 289         if (file) {
 290 -               struct inode *inode = file_inode(vma->vm_file);
 291 +               struct inode *inode;
 292 +
 293 +               file = vma_pr_or_file(vma);
 294 +               inode = file_inode(file);
 295                 dev = inode->i_sb->s_dev;
 296                 ino = inode->i_ino;
 297                 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
 298 @@ -1942,7 +1945,7 @@ static int show_numa_map(struct seq_file *m, void *v)
 299         struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
 300         struct vm_area_struct *vma = v;
 301         struct numa_maps *md = &numa_priv->md;
 302 -       struct file *file = vma->vm_file;
 303 +       struct file *file = vma_pr_or_file(vma);
 304         struct mm_struct *mm = vma->vm_mm;
 305         struct mempolicy *pol;
 306         char buffer[64];
 307 diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
 308 index 2fd06f52b6a4..b01030da424f 100644
 309 --- a/fs/proc/task_nommu.c
 310 +++ b/fs/proc/task_nommu.c
 311 @@ -150,7 +150,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
 312         file = vma->vm_file;
 313
 314         if (file) {
 315 -               struct inode *inode = file_inode(vma->vm_file);
 316 +               struct inode *inode;
 317 +
 318 +               file = vma_pr_or_file(vma);
 319 +               inode = file_inode(file);
 320                 dev = inode->i_sb->s_dev;
 321                 ino = inode->i_ino;
 322                 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
 323 diff --git a/include/linux/mm.h b/include/linux/mm.h
 324 index f3f196e4d66d..831812fd4d07 100644
 325 --- a/include/linux/mm.h
 326 +++ b/include/linux/mm.h
 327 @@ -2059,6 +2059,43 @@ static inline void unmap_shared_mapping_range(struct address_space *mapping,
 328         unmap_mapping_range(mapping, holebegin, holelen, 0);
 329  }
 330
 331 +#if IS_ENABLED(CONFIG_AUFS_FS)
 332 +extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
 333 +extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
 334 +                                     int);
 335 +extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
 336 +extern void vma_do_fput(struct vm_area_struct *, const char[], int);
 337 +
 338 +#define vma_file_update_time(vma)      vma_do_file_update_time(vma, __func__, \
 339 +                                                               __LINE__)
 340 +#define vma_pr_or_file(vma)            vma_do_pr_or_file(vma, __func__, \
 341 +                                                         __LINE__)
 342 +#define vma_get_file(vma)              vma_do_get_file(vma, __func__, __LINE__)
 343 +#define vma_fput(vma)                  vma_do_fput(vma, __func__, __LINE__)
 344 +
 345 +#ifndef CONFIG_MMU
 346 +extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
 347 +extern void vmr_do_fput(struct vm_region *, const char[], int);
 348 +
 349 +#define vmr_pr_or_file(region)         vmr_do_pr_or_file(region, __func__, \
 350 +                                                         __LINE__)
 351 +#define vmr_fput(region)               vmr_do_fput(region, __func__, __LINE__)
 352 +#endif /* !CONFIG_MMU */
 353 +
 354 +#else
 355 +
 356 +#define vma_file_update_time(vma)      file_update_time((vma)->vm_file)
 357 +#define vma_pr_or_file(vma)            (vma)->vm_file
 358 +#define vma_get_file(vma)              get_file((vma)->vm_file)
 359 +#define vma_fput(vma)                  fput((vma)->vm_file)
 360 +
 361 +#ifndef CONFIG_MMU
 362 +#define vmr_pr_or_file(region)         (region)->vm_file
 363 +#define vmr_fput(region)               fput((region)->vm_file)
 364 +#endif /* !CONFIG_MMU */
 365 +
 366 +#endif /* CONFIG_AUFS_FS */
 367 +
 368  extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
 369                 void *buf, int len, unsigned int gup_flags);
 370  extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
 371 diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
 372 index 3b8475007734..7a2de3f818bc 100644
 373 --- a/include/linux/mm_types.h
 374 +++ b/include/linux/mm_types.h
 375 @@ -504,6 +504,9 @@ struct vm_region {
 376         unsigned long   vm_top;         /* region allocated to here */
 377         unsigned long   vm_pgoff;       /* the offset in vm_file corresponding to vm_start */
 378         struct file     *vm_file;       /* the backing file or NULL */
 379 +#if IS_ENABLED(CONFIG_AUFS_FS)
 380 +       struct file     *vm_prfile;     /* the virtual backing file or NULL */
 381 +#endif
 382
 383         int             vm_usage;       /* region usage count (access under nommu_region_sem) */
 384         bool            vm_icache_flushed : 1; /* true if the icache has been flushed for
 385 @@ -575,6 +578,9 @@ struct vm_area_struct {
 386         unsigned long vm_pgoff;         /* Offset (within vm_file) in PAGE_SIZE
 387                                            units */
 388         struct file * vm_file;          /* File we map to (can be NULL). */
 389 +#if IS_ENABLED(CONFIG_AUFS_FS)
 390 +       struct file *vm_prfile;         /* shadow of vm_file */
 391 +#endif
 392         void * vm_private_data;         /* was vm_pte (shared mem) */
 393
 394  #ifdef CONFIG_ANON_VMA_NAME
 395 diff --git a/kernel/fork.c b/kernel/fork.c
 396 index 9f7fe3541897..f9e44ca88a7d 100644
 397 --- a/kernel/fork.c
 398 +++ b/kernel/fork.c
 399 @@ -664,7 +664,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
 400                 if (file) {
 401                         struct address_space *mapping = file->f_mapping;
 402
 403 -                       get_file(file);
 404 +                       vma_get_file(tmp);
 405                         i_mmap_lock_write(mapping);
 406                         if (tmp->vm_flags & VM_SHARED)
 407                                 mapping_allow_writable(mapping);
 408 diff --git a/mm/Makefile b/mm/Makefile
 409 index 8e105e5b3e29..c218d042dbce 100644
 410 --- a/mm/Makefile
 411 +++ b/mm/Makefile
 412 @@ -138,3 +138,4 @@ obj-$(CONFIG_IO_MAPPING) += io-mapping.o
 413  obj-$(CONFIG_HAVE_BOOTMEM_INFO_NODE) += bootmem_info.o
 414  obj-$(CONFIG_GENERIC_IOREMAP) += ioremap.o
 415  obj-$(CONFIG_SHRINKER_DEBUG) += shrinker_debug.o
 416 +obj-$(CONFIG_AUFS_FS:m=y) += prfile.o
 417 diff --git a/mm/filemap.c b/mm/filemap.c
 418 index c4d4ace9cc70..257da81286d1 100644
 419 --- a/mm/filemap.c
 420 +++ b/mm/filemap.c
 421 @@ -3437,7 +3437,7 @@ vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf)
 422         vm_fault_t ret = VM_FAULT_LOCKED;
 423
 424         sb_start_pagefault(mapping->host->i_sb);
 425 -       file_update_time(vmf->vma->vm_file);
 426 +       vma_file_update_time(vmf->vma);
 427         folio_lock(folio);
 428         if (folio->mapping != mapping) {
 429                 folio_unlock(folio);
 430 diff --git a/mm/mmap.c b/mm/mmap.c
 431 index 87d929316d57..fac904cd41cd 100644
 432 --- a/mm/mmap.c
 433 +++ b/mm/mmap.c
 434 @@ -139,7 +139,7 @@ static void remove_vma(struct vm_area_struct *vma)
 435         if (vma->vm_ops && vma->vm_ops->close)
 436                 vma->vm_ops->close(vma);
 437         if (vma->vm_file)
 438 -               fput(vma->vm_file);
 439 +               vma_fput(vma);
 440         mpol_put(vma_policy(vma));
 441         vm_area_free(vma);
 442  }
 443 @@ -589,7 +589,7 @@ inline int vma_expand(struct ma_state *mas, struct vm_area_struct *vma,
 444         if (remove_next) {
 445                 if (file) {
 446                         uprobe_munmap(next, next->vm_start, next->vm_end);
 447 -                       fput(file);
 448 +                       vma_fput(vma);
 449                 }
 450                 if (next->anon_vma)
 451                         anon_vma_merge(vma, next);
 452 @@ -830,7 +830,7 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 453  again:
 454                 if (file) {
 455                         uprobe_munmap(next, next->vm_start, next->vm_end);
 456 -                       fput(file);
 457 +                       vma_fput(vma);
 458                 }
 459                 if (next->anon_vma)
 460                         anon_vma_merge(vma, next);
 461 @@ -2224,7 +2224,7 @@ int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
 462                 goto out_free_mpol;
 463
 464         if (new->vm_file)
 465 -               get_file(new->vm_file);
 466 +               vma_get_file(new);
 467
 468         if (new->vm_ops && new->vm_ops->open)
 469                 new->vm_ops->open(new);
 470 @@ -2246,7 +2246,7 @@ int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
 471         if (new->vm_ops && new->vm_ops->close)
 472                 new->vm_ops->close(new);
 473         if (new->vm_file)
 474 -               fput(new->vm_file);
 475 +               vma_fput(new);
 476         unlink_anon_vmas(new);
 477   out_free_mpol:
 478         mpol_put(vma_policy(new));
 479 @@ -2746,7 +2746,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
 480         if (vma->vm_ops && vma->vm_ops->close)
 481                 vma->vm_ops->close(vma);
 482  unmap_and_free_vma:
 483 -       fput(vma->vm_file);
 484 +       vma_fput(vma);
 485         vma->vm_file = NULL;
 486
 487         /* Undo any partial mapping done by a device driver. */
 488 @@ -2813,6 +2813,9 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
 489         unsigned long populate = 0;
 490         unsigned long ret = -EINVAL;
 491         struct file *file;
 492 +#if IS_ENABLED(CONFIG_AUFS_FS)
 493 +       struct file *prfile;
 494 +#endif
 495
 496         pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/mm/remap_file_pages.rst.\n",
 497                      current->comm, current->pid);
 498 @@ -2871,10 +2874,34 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
 499         if (vma->vm_flags & VM_LOCKED)
 500                 flags |= MAP_LOCKED;
 501
 502 +#if IS_ENABLED(CONFIG_AUFS_FS)
 503 +       vma_get_file(vma);
 504 +       file = vma->vm_file;
 505 +       prfile = vma->vm_prfile;
 506 +       ret = do_mmap(vma->vm_file, start, size,
 507 +                       prot, flags, pgoff, &populate, NULL);
 508 +       if (!IS_ERR_VALUE(ret) && file && prfile) {
 509 +               struct vm_area_struct *new_vma;
 510 +
 511 +               new_vma = find_vma(mm, ret);
 512 +               if (!new_vma->vm_prfile)
 513 +                       new_vma->vm_prfile = prfile;
 514 +               if (new_vma != vma)
 515 +                       get_file(prfile);
 516 +       }
 517 +       /*
 518 +        * two fput()s instead of vma_fput(vma),
 519 +        * coz vma may not be available anymore.
 520 +        */
 521 +       fput(file);
 522 +       if (prfile)
 523 +               fput(prfile);
 524 +#else
 525         file = get_file(vma->vm_file);
 526         ret = do_mmap(vma->vm_file, start, size,
 527                         prot, flags, pgoff, &populate, NULL);
 528         fput(file);
 529 +#endif /* CONFIG_AUFS_FS */
 530  out:
 531         mmap_write_unlock(mm);
 532         if (populate)
 533 @@ -3222,7 +3249,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
 534                 if (anon_vma_clone(new_vma, vma))
 535                         goto out_free_mempol;
 536                 if (new_vma->vm_file)
 537 -                       get_file(new_vma->vm_file);
 538 +                       vma_get_file(new_vma);
 539                 if (new_vma->vm_ops && new_vma->vm_ops->open)
 540                         new_vma->vm_ops->open(new_vma);
 541                 if (vma_link(mm, new_vma))
 542 diff --git a/mm/nommu.c b/mm/nommu.c
 543 index 214c70e1d059..dccbb939d2a9 100644
 544 --- a/mm/nommu.c
 545 +++ b/mm/nommu.c
 546 @@ -523,7 +523,7 @@ static void __put_nommu_region(struct vm_region *region)
 547                 up_write(&nommu_region_sem);
 548
 549                 if (region->vm_file)
 550 -                       fput(region->vm_file);
 551 +                       vmr_fput(region);
 552
 553                 /* IO memory and memory shared directly out of the pagecache
 554                  * from ramfs/tmpfs mustn't be released here */
 555 @@ -653,7 +653,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
 556         if (vma->vm_ops && vma->vm_ops->close)
 557                 vma->vm_ops->close(vma);
 558         if (vma->vm_file)
 559 -               fput(vma->vm_file);
 560 +               vma_fput(vma);
 561         put_nommu_region(vma->vm_region);
 562         vm_area_free(vma);
 563  }
 564 @@ -1164,7 +1164,7 @@ unsigned long do_mmap(struct file *file,
 565                                         goto error_just_free;
 566                                 }
 567                         }
 568 -                       fput(region->vm_file);
 569 +                       vmr_fput(region);
 570                         kmem_cache_free(vm_region_jar, region);
 571                         region = pregion;
 572                         result = start;
 573 @@ -1242,10 +1242,11 @@ error_just_free:
 574  error:
 575         mas_destroy(&mas);
 576         if (region->vm_file)
 577 -               fput(region->vm_file);
 578 +                vmr_fput(region);
 579         kmem_cache_free(vm_region_jar, region);
 580         if (vma->vm_file)
 581                 fput(vma->vm_file);
 582 +                vmr_fput(vma);
 583         vm_area_free(vma);
 584         return ret;
 585
 586 diff --git a/mm/prfile.c b/mm/prfile.c
 587 new file mode 100644
 588 index 000000000000..8f820a235364
 589 --- /dev/null
 590 +++ b/mm/prfile.c
 591 @@ -0,0 +1,86 @@
 592 +// SPDX-License-Identifier: GPL-2.0
 593 +/*
 594 + * Mainly for aufs which mmap(2) different file and wants to print different
 595 + * path in /proc/PID/maps.
 596 + * Call these functions via macros defined in linux/mm.h.
 597 + *
 598 + * See Documentation/filesystems/aufs/design/06mmap.txt
 599 + *
 600 + * Copyright (c) 2014-2022 Junjro R. Okajima
 601 + * Copyright (c) 2014 Ian Campbell
 602 + */
 603 +
 604 +#include <linux/mm.h>
 605 +#include <linux/file.h>
 606 +#include <linux/fs.h>
 607 +
 608 +/* #define PRFILE_TRACE */
 609 +static inline void prfile_trace(struct file *f, struct file *pr,
 610 +                             const char func[], int line, const char func2[])
 611 +{
 612 +#ifdef PRFILE_TRACE
 613 +       if (pr)
 614 +               pr_info("%s:%d: %s, %pD2\n", func, line, func2, f);
 615 +#endif
 616 +}
 617 +
 618 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
 619 +                            int line)
 620 +{
 621 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 622 +
 623 +       prfile_trace(f, pr, func, line, __func__);
 624 +       file_update_time(f);
 625 +       if (f && pr)
 626 +               file_update_time(pr);
 627 +}
 628 +
 629 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
 630 +                              int line)
 631 +{
 632 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 633 +
 634 +       prfile_trace(f, pr, func, line, __func__);
 635 +       return (f && pr) ? pr : f;
 636 +}
 637 +
 638 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
 639 +{
 640 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 641 +
 642 +       prfile_trace(f, pr, func, line, __func__);
 643 +       get_file(f);
 644 +       if (f && pr)
 645 +               get_file(pr);
 646 +}
 647 +
 648 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
 649 +{
 650 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 651 +
 652 +       prfile_trace(f, pr, func, line, __func__);
 653 +       fput(f);
 654 +       if (f && pr)
 655 +               fput(pr);
 656 +}
 657 +
 658 +#ifndef CONFIG_MMU
 659 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
 660 +                              int line)
 661 +{
 662 +       struct file *f = region->vm_file, *pr = region->vm_prfile;
 663 +
 664 +       prfile_trace(f, pr, func, line, __func__);
 665 +       return (f && pr) ? pr : f;
 666 +}
 667 +
 668 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
 669 +{
 670 +       struct file *f = region->vm_file, *pr = region->vm_prfile;
 671 +
 672 +       prfile_trace(f, pr, func, line, __func__);
 673 +       fput(f);
 674 +       if (f && pr)
 675 +               fput(pr);
 676 +}
 677 +#endif /* !CONFIG_MMU */
 678 SPDX-License-Identifier: GPL-2.0
 679 aufs6.x-rcN standalone patch
 680
 681 diff --git a/fs/dcache.c b/fs/dcache.c
 682 index 519321f32f95..267e0c65914a 100644
 683 --- a/fs/dcache.c
 684 +++ b/fs/dcache.c
 685 @@ -1450,6 +1450,7 @@ void d_walk(struct dentry *parent, void *data,
 686         seq = 1;
 687         goto again;
 688  }
 689 +EXPORT_SYMBOL_GPL(d_walk);
 690
 691  struct check_mount {
 692         struct vfsmount *mnt;
 693 @@ -3052,6 +3053,7 @@ void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
 694
 695         write_sequnlock(&rename_lock);
 696  }
 697 +EXPORT_SYMBOL_GPL(d_exchange);
 698
 699  /**
 700   * d_ancestor - search for an ancestor
 701 diff --git a/fs/exec.c b/fs/exec.c
 702 index ab913243a367..d84461fddf7d 100644
 703 --- a/fs/exec.c
 704 +++ b/fs/exec.c
 705 @@ -111,6 +111,7 @@ bool path_noexec(const struct path *path)
 706         return (path->mnt->mnt_flags & MNT_NOEXEC) ||
 707                (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
 708  }
 709 +EXPORT_SYMBOL_GPL(path_noexec);
 710
 711  #ifdef CONFIG_USELIB
 712  /*
 713 diff --git a/fs/fcntl.c b/fs/fcntl.c
 714 index c980a0524118..40087d11a974 100644
 715 --- a/fs/fcntl.c
 716 +++ b/fs/fcntl.c
 717 @@ -86,6 +86,7 @@ int setfl(int fd, struct file *filp, unsigned long arg)
 718   out:
 719         return error;
 720  }
 721 +EXPORT_SYMBOL_GPL(setfl);
 722
 723  static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
 724                       int force)
 725 diff --git a/fs/file_table.c b/fs/file_table.c
 726 index dd88701e54a9..ecdb7817b074 100644
 727 --- a/fs/file_table.c
 728 +++ b/fs/file_table.c
 729 @@ -198,6 +198,7 @@ struct file *alloc_empty_file(int flags, const struct cred *cred)
 730         }
 731         return ERR_PTR(-ENFILE);
 732  }
 733 +EXPORT_SYMBOL_GPL(alloc_empty_file);
 734
 735  /*
 736   * Variant of alloc_empty_file() that doesn't check and modify nr_files.
 737 diff --git a/fs/namespace.c b/fs/namespace.c
 738 index 8cd34f760f63..799a94205906 100644
 739 --- a/fs/namespace.c
 740 +++ b/fs/namespace.c
 741 @@ -571,6 +571,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
 742         mnt_dec_writers(real_mount(mnt));
 743         preempt_enable();
 744  }
 745 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
 746
 747  /**
 748   * mnt_drop_write - give up write access to a mount
 749 @@ -968,6 +969,7 @@ int is_current_mnt_ns(struct vfsmount *mnt)
 750  {
 751         return check_mnt(real_mount(mnt));
 752  }
 753 +EXPORT_SYMBOL_GPL(is_current_mnt_ns);
 754
 755  /*
 756   * vfsmount lock must be held for write
 757 @@ -2165,6 +2167,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
 758         }
 759         return 0;
 760  }
 761 +EXPORT_SYMBOL_GPL(iterate_mounts);
 762
 763  static void lock_mnt_tree(struct mount *mnt)
 764  {
 765 diff --git a/fs/notify/group.c b/fs/notify/group.c
 766 index 1de6631a3925..3008eb37a18d 100644
 767 --- a/fs/notify/group.c
 768 +++ b/fs/notify/group.c
 769 @@ -100,6 +100,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
 770  {
 771         refcount_inc(&group->refcnt);
 772  }
 773 +EXPORT_SYMBOL_GPL(fsnotify_get_group);
 774
 775  /*
 776   * Drop a reference to a group.  Free it if it's through.
 777 diff --git a/fs/open.c b/fs/open.c
 778 index 82c1a28b3308..e7bd6f1c9474 100644
 779 --- a/fs/open.c
 780 +++ b/fs/open.c
 781 @@ -66,6 +66,7 @@ int do_truncate(struct user_namespace *mnt_userns, struct dentry *dentry,
 782         inode_unlock(dentry->d_inode);
 783         return ret;
 784  }
 785 +EXPORT_SYMBOL_GPL(do_truncate);
 786
 787  long vfs_truncate(const struct path *path, loff_t length)
 788  {
 789 diff --git a/fs/read_write.c b/fs/read_write.c
 790 index 7a2ff6157eda..76f5d4d08687 100644
 791 --- a/fs/read_write.c
 792 +++ b/fs/read_write.c
 793 @@ -477,6 +477,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
 794         inc_syscr(current);
 795         return ret;
 796  }
 797 +EXPORT_SYMBOL_GPL(vfs_read);
 798
 799  static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
 800  {
 801 @@ -592,6 +593,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
 802         file_end_write(file);
 803         return ret;
 804  }
 805 +EXPORT_SYMBOL_GPL(vfs_write);
 806
 807  /* file_ppos returns &file->f_pos or NULL if file is stream */
 808  static inline loff_t *file_ppos(struct file *file)
 809 diff --git a/fs/splice.c b/fs/splice.c
 810 index 21db9b057d60..abcd14f52c1e 100644
 811 --- a/fs/splice.c
 812 +++ b/fs/splice.c
 813 @@ -763,6 +763,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
 814                 return warn_unsupported(out, "write");
 815         return out->f_op->splice_write(pipe, out, ppos, len, flags);
 816  }
 817 +EXPORT_SYMBOL_GPL(do_splice_from);
 818
 819  /*
 820   * Attempt to initiate a splice from a file to a pipe.
 821 @@ -792,6 +793,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
 822                 return warn_unsupported(in, "read");
 823         return in->f_op->splice_read(in, ppos, pipe, len, flags);
 824  }
 825 +EXPORT_SYMBOL_GPL(do_splice_to);
 826
 827  /**
 828   * splice_direct_to_actor - splices data directly between two non-pipes
 829 diff --git a/fs/xattr.c b/fs/xattr.c
 830 index adab9a70b536..b2dc4fc0b25b 100644
 831 --- a/fs/xattr.c
 832 +++ b/fs/xattr.c
 833 @@ -407,6 +407,7 @@ vfs_getxattr_alloc(struct user_namespace *mnt_userns, struct dentry *dentry,
 834         *xattr_value = value;
 835         return error;
 836  }
 837 +EXPORT_SYMBOL_GPL(vfs_getxattr_alloc);
 838
 839  ssize_t
 840  __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
 841 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
 842 index c1f0aadde539..278e0a322d20 100644
 843 --- a/kernel/locking/lockdep.c
 844 +++ b/kernel/locking/lockdep.c
 845 @@ -238,6 +238,7 @@ inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
 846          */
 847         return lock_classes + class_idx;
 848  }
 849 +EXPORT_SYMBOL_GPL(lockdep_hlock_class);
 850  #define hlock_class(hlock) lockdep_hlock_class(hlock)
 851
 852  #ifdef CONFIG_LOCK_STAT
 853 diff --git a/kernel/task_work.c b/kernel/task_work.c
 854 index 065e1ef8fc8d..c623c6f0c645 100644
 855 --- a/kernel/task_work.c
 856 +++ b/kernel/task_work.c
 857 @@ -182,3 +182,4 @@ void task_work_run(void)
 858                 } while (work);
 859         }
 860  }
 861 +EXPORT_SYMBOL_GPL(task_work_run);
 862 diff --git a/security/security.c b/security/security.c
 863 index d1571900a8c7..c8333d54e590 100644
 864 --- a/security/security.c
 865 +++ b/security/security.c
 866 @@ -1195,6 +1195,7 @@ int security_path_rmdir(const struct path *dir, struct dentry *dentry)
 867                 return 0;
 868         return call_int_hook(path_rmdir, 0, dir, dentry);
 869  }
 870 +EXPORT_SYMBOL_GPL(security_path_rmdir);
 871
 872  int security_path_unlink(const struct path *dir, struct dentry *dentry)
 873  {
 874 @@ -1211,6 +1212,7 @@ int security_path_symlink(const struct path *dir, struct dentry *dentry,
 875                 return 0;
 876         return call_int_hook(path_symlink, 0, dir, dentry, old_name);
 877  }
 878 +EXPORT_SYMBOL_GPL(security_path_symlink);
 879
 880  int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
 881                        struct dentry *new_dentry)
 882 @@ -1219,6 +1221,7 @@ int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
 883                 return 0;
 884         return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
 885  }
 886 +EXPORT_SYMBOL_GPL(security_path_link);
 887
 888  int security_path_rename(const struct path *old_dir, struct dentry *old_dentry,
 889                          const struct path *new_dir, struct dentry *new_dentry,
 890 @@ -1239,6 +1242,7 @@ int security_path_truncate(const struct path *path)
 891                 return 0;
 892         return call_int_hook(path_truncate, 0, path);
 893  }
 894 +EXPORT_SYMBOL_GPL(security_path_truncate);
 895
 896  int security_path_chmod(const struct path *path, umode_t mode)
 897  {
 898 @@ -1246,6 +1250,7 @@ int security_path_chmod(const struct path *path, umode_t mode)
 899                 return 0;
 900         return call_int_hook(path_chmod, 0, path, mode);
 901  }
 902 +EXPORT_SYMBOL_GPL(security_path_chmod);
 903
 904  int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
 905  {
 906 @@ -1253,6 +1258,7 @@ int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
 907                 return 0;
 908         return call_int_hook(path_chown, 0, path, uid, gid);
 909  }
 910 +EXPORT_SYMBOL_GPL(security_path_chown);
 911
 912  int security_path_chroot(const struct path *path)
 913  {
 914 @@ -1353,6 +1359,7 @@ int security_inode_permission(struct inode *inode, int mask)
 915                 return 0;
 916         return call_int_hook(inode_permission, 0, inode, mask);
 917  }
 918 +EXPORT_SYMBOL_GPL(security_inode_permission);
 919
 920  int security_inode_setattr(struct user_namespace *mnt_userns,
 921                            struct dentry *dentry, struct iattr *attr)
 922 @@ -1593,6 +1600,7 @@ int security_file_permission(struct file *file, int mask)
 923
 924         return fsnotify_perm(file, mask);
 925  }
 926 +EXPORT_SYMBOL_GPL(security_file_permission);
 927
 928  int security_file_alloc(struct file *file)
 929  {
 930 @@ -1726,6 +1734,7 @@ int security_file_truncate(struct file *file)
 931  {
 932         return call_int_hook(file_truncate, 0, file);
 933  }
 934 +EXPORT_SYMBOL_GPL(security_file_truncate);
 935
 936  int security_task_alloc(struct task_struct *task, unsigned long clone_flags)
 937  {
 938 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
 939 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs     1970-01-01 01:00:00.000000000 +0100
 940 +++ linux/Documentation/ABI/testing/debugfs-aufs        2022-11-05 23:02:18.955889283 +0100
 941 @@ -0,0 +1,55 @@
 942 +What:          /debug/aufs/si_<id>/
 943 +Date:          March 2009
 944 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 945 +Description:
 946 +               Under /debug/aufs, a directory named si_<id> is created
 947 +               per aufs mount, where <id> is a unique id generated
 948 +               internally.
 949 +
 950 +What:          /debug/aufs/si_<id>/plink
 951 +Date:          Apr 2013
 952 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 953 +Description:
 954 +               It has three lines and shows the information about the
 955 +               pseudo-link. The first line is a single number
 956 +               representing a number of buckets. The second line is a
 957 +               number of pseudo-links per buckets (separated by a
 958 +               blank). The last line is a single number representing a
 959 +               total number of psedo-links.
 960 +               When the aufs mount option 'noplink' is specified, it
 961 +               will show "1\n0\n0\n".
 962 +
 963 +What:          /debug/aufs/si_<id>/xib
 964 +Date:          March 2009
 965 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 966 +Description:
 967 +               It shows the consumed blocks by xib (External Inode Number
 968 +               Bitmap), its block size and file size.
 969 +               When the aufs mount option 'noxino' is specified, it
 970 +               will be empty. About XINO files, see the aufs manual.
 971 +
 972 +What:          /debug/aufs/si_<id>/xi<branch-index>
 973 +Date:          March 2009
 974 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 975 +Description:
 976 +               It shows the consumed blocks by xino (External Inode Number
 977 +               Translation Table), its link count, block size and file
 978 +               size.
 979 +               Due to the file size limit, there may exist multiple
 980 +               xino files per branch.  In this case, "-N" is added to
 981 +               the filename and it corresponds to the index of the
 982 +               internal xino array.  "-0" is omitted.
 983 +               When the aufs mount option 'noxino' is specified, Those
 984 +               entries won't exist.  About XINO files, see the aufs
 985 +               manual.
 986 +
 987 +What:          /debug/aufs/si_<id>/xigen
 988 +Date:          March 2009
 989 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 990 +Description:
 991 +               It shows the consumed blocks by xigen (External Inode
 992 +               Generation Table), its block size and file size.
 993 +               If CONFIG_AUFS_EXPORT is disabled, this entry will not
 994 +               be created.
 995 +               When the aufs mount option 'noxino' is specified, it
 996 +               will be empty. About XINO files, see the aufs manual.
 997 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
 998 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs       1970-01-01 01:00:00.000000000 +0100
 999 +++ linux/Documentation/ABI/testing/sysfs-aufs  2022-11-05 23:02:18.955889283 +0100
1000 @@ -0,0 +1,31 @@
1001 +What:          /sys/fs/aufs/si_<id>/
1002 +Date:          March 2009
1003 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
1004 +Description:
1005 +               Under /sys/fs/aufs, a directory named si_<id> is created
1006 +               per aufs mount, where <id> is a unique id generated
1007 +               internally.
1008 +
1009 +What:          /sys/fs/aufs/si_<id>/br<idx>
1010 +Date:          March 2009
1011 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
1012 +Description:
1013 +               It shows the abolute path of a member directory (which
1014 +               is called branch) in aufs, and its permission.
1015 +
1016 +What:          /sys/fs/aufs/si_<id>/brid<idx>
1017 +Date:          July 2013
1018 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
1019 +Description:
1020 +               It shows the id of a member directory (which is called
1021 +               branch) in aufs.
1022 +
1023 +What:          /sys/fs/aufs/si_<id>/xi_path
1024 +Date:          March 2009
1025 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
1026 +Description:
1027 +               It shows the abolute path of XINO (External Inode Number
1028 +               Bitmap, Translation Table and Generation Table) file
1029 +               even if it is the default path.
1030 +               When the aufs mount option 'noxino' is specified, it
1031 +               will be empty. About XINO files, see the aufs manual.
1032 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1033 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt  1970-01-01 01:00:00.000000000 +0100
1034 +++ linux/Documentation/filesystems/aufs/design/01intro.txt     2022-11-05 23:02:18.955889283 +0100
1035 @@ -0,0 +1,171 @@
1036 +
1037 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1038 +#
1039 +# This program is free software; you can redistribute it and/or modify
1040 +# it under the terms of the GNU General Public License as published by
1041 +# the Free Software Foundation; either version 2 of the License, or
1042 +# (at your option) any later version.
1043 +#
1044 +# This program is distributed in the hope that it will be useful,
1045 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1046 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1047 +# GNU General Public License for more details.
1048 +#
1049 +# You should have received a copy of the GNU General Public License
1050 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1051 +
1052 +Introduction
1053 +----------------------------------------
1054 +
1055 +aufs [ei ju: ef es] | /ey-yoo-ef-es/ | [a u f s]
1056 +1. abbrev. for "advanced multi-layered unification filesystem".
1057 +2. abbrev. for "another unionfs".
1058 +3. abbrev. for "auf das" in German which means "on the" in English.
1059 +   Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1060 +   But "Filesystem aufs Filesystem" is hard to understand.
1061 +4. abbrev. for "African Urban Fashion Show".
1062 +
1063 +AUFS is a filesystem with features:
1064 +- multi layered stackable unification filesystem, the member directory
1065 +  is called as a branch.
1066 +- branch permission and attribute, 'readonly', 'real-readonly',
1067 +  'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1068 +  combination.
1069 +- internal "file copy-on-write".
1070 +- logical deletion, whiteout.
1071 +- dynamic branch manipulation, adding, deleting and changing permission.
1072 +- allow bypassing aufs, user's direct branch access.
1073 +- external inode number translation table and bitmap which maintains the
1074 +  persistent aufs inode number.
1075 +- seekable directory, including NFS readdir.
1076 +- file mapping, mmap and sharing pages.
1077 +- pseudo-link, hardlink over branches.
1078 +- loopback mounted filesystem as a branch.
1079 +- several policies to select one among multiple writable branches.
1080 +- revert a single systemcall when an error occurs in aufs.
1081 +- and more...
1082 +
1083 +
1084 +Multi Layered Stackable Unification Filesystem
1085 +----------------------------------------------------------------------
1086 +Most people already knows what it is.
1087 +It is a filesystem which unifies several directories and provides a
1088 +merged single directory. When users access a file, the access will be
1089 +passed/re-directed/converted (sorry, I am not sure which English word is
1090 +correct) to the real file on the member filesystem. The member
1091 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1092 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1093 +readonly branch is handled by creating 'whiteout' on the upper writable
1094 +branch.
1095 +
1096 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1097 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1098 +different approaches to implement the merged-view.
1099 +The former tries putting it into VFS, and the latter implements as a
1100 +separate filesystem.
1101 +(If I misunderstand about these implementations, please let me know and
1102 +I shall correct it. Because it is a long time ago when I read their
1103 +source files last time).
1104 +
1105 +UnionMount's approach will be able to small, but may be hard to share
1106 +branches between several UnionMount since the whiteout in it is
1107 +implemented in the inode on branch filesystem and always
1108 +shared. According to Bharata's post, readdir does not seems to be
1109 +finished yet.
1110 +There are several missing features known in this implementations such as
1111 +- for users, the inode number may change silently. eg. copy-up.
1112 +- link(2) may break by copy-up.
1113 +- read(2) may get an obsoleted filedata (fstat(2) too).
1114 +- fcntl(F_SETLK) may be broken by copy-up.
1115 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1116 +  open(O_RDWR).
1117 +
1118 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1119 +merged into mainline. This is another implementation of UnionMount as a
1120 +separated filesystem. All the limitations and known problems which
1121 +UnionMount are equally inherited to "overlay" filesystem.
1122 +
1123 +Unionfs has a longer history. When I started implementing a stackable
1124 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1125 +inode, dentry and file objects and they have an array pointing lower
1126 +same kind objects. After contributing many patches for Unionfs, I
1127 +re-started my project AUFS (Jun 2006).
1128 +
1129 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1130 +implemented my own ideas, approaches and enhancements and it became
1131 +totally different one.
1132 +
1133 +Comparing DM snapshot and fs based implementation
1134 +- the number of bytes to be copied between devices is much smaller.
1135 +- the type of filesystem must be one and only.
1136 +- the fs must be writable, no readonly fs, even for the lower original
1137 +  device. so the compression fs will not be usable. but if we use
1138 +  loopback mount, we may address this issue.
1139 +  for instance,
1140 +       mount /cdrom/squashfs.img /sq
1141 +       losetup /sq/ext2.img
1142 +       losetup /somewhere/cow
1143 +       dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1144 +- it will be difficult (or needs more operations) to extract the
1145 +  difference between the original device and COW.
1146 +- DM snapshot-merge may help a lot when users try merging. in the
1147 +  fs-layer union, users will use rsync(1).
1148 +
1149 +You may want to read my old paper "Filesystems in LiveCD"
1150 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1151 +
1152 +
1153 +Several characters/aspects/persona of aufs
1154 +----------------------------------------------------------------------
1155 +
1156 +Aufs has several characters, aspects or persona.
1157 +1. a filesystem, callee of VFS helper
1158 +2. sub-VFS, caller of VFS helper for branches
1159 +3. a virtual filesystem which maintains persistent inode number
1160 +4. reader/writer of files on branches such like an application
1161 +
1162 +1. Callee of VFS Helper
1163 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1164 +unlink(2) from an application reaches sys_unlink() kernel function and
1165 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1166 +calls filesystem specific unlink operation. Actually aufs implements the
1167 +unlink operation but it behaves like a redirector.
1168 +
1169 +2. Caller of VFS Helper for Branches
1170 +aufs_unlink() passes the unlink request to the branch filesystem as if
1171 +it were called from VFS. So the called unlink operation of the branch
1172 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1173 +every necessary pre/post operation for the branch filesystem.
1174 +- acquire the lock for the parent dir on a branch
1175 +- lookup in a branch
1176 +- revalidate dentry on a branch
1177 +- mnt_want_write() for a branch
1178 +- vfs_unlink() for a branch
1179 +- mnt_drop_write() for a branch
1180 +- release the lock on a branch
1181 +
1182 +3. Persistent Inode Number
1183 +One of the most important issue for a filesystem is to maintain inode
1184 +numbers. This is particularly important to support exporting a
1185 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1186 +backend block device for its own. But some storage is necessary to
1187 +keep and maintain the inode numbers. It may be a large space and may not
1188 +suit to keep in memory. Aufs rents some space from its first writable
1189 +branch filesystem (by default) and creates file(s) on it. These files
1190 +are created by aufs internally and removed soon (currently) keeping
1191 +opened.
1192 +Note: Because these files are removed, they are totally gone after
1193 +      unmounting aufs. It means the inode numbers are not persistent
1194 +      across unmount or reboot. I have a plan to make them really
1195 +      persistent which will be important for aufs on NFS server.
1196 +
1197 +4. Read/Write Files Internally (copy-on-write)
1198 +Because a branch can be readonly, when you write a file on it, aufs will
1199 +"copy-up" it to the upper writable branch internally. And then write the
1200 +originally requested thing to the file. Generally kernel doesn't
1201 +open/read/write file actively. In aufs, even a single write may cause a
1202 +internal "file copy". This behaviour is very similar to cp(1) command.
1203 +
1204 +Some people may think it is better to pass such work to user space
1205 +helper, instead of doing in kernel space. Actually I am still thinking
1206 +about it. But currently I have implemented it in kernel space.
1207 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1208 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1209 +++ linux/Documentation/filesystems/aufs/design/02struct.txt    2022-11-05 23:02:18.955889283 +0100
1210 @@ -0,0 +1,258 @@
1211 +
1212 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1213 +#
1214 +# This program is free software; you can redistribute it and/or modify
1215 +# it under the terms of the GNU General Public License as published by
1216 +# the Free Software Foundation; either version 2 of the License, or
1217 +# (at your option) any later version.
1218 +#
1219 +# This program is distributed in the hope that it will be useful,
1220 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1221 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1222 +# GNU General Public License for more details.
1223 +#
1224 +# You should have received a copy of the GNU General Public License
1225 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1226 +
1227 +Basic Aufs Internal Structure
1228 +
1229 +Superblock/Inode/Dentry/File Objects
1230 +----------------------------------------------------------------------
1231 +As like an ordinary filesystem, aufs has its own
1232 +superblock/inode/dentry/file objects. All these objects have a
1233 +dynamically allocated array and store the same kind of pointers to the
1234 +lower filesystem, branch.
1235 +For example, when you build a union with one readwrite branch and one
1236 +readonly, mounted /au, /rw and /ro respectively.
1237 +- /au = /rw + /ro
1238 +- /ro/fileA exists but /rw/fileA
1239 +
1240 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1241 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1242 +- [0] = NULL (because /rw/fileA doesn't exist)
1243 +- [1] = /ro/fileA
1244 +
1245 +This style of an array is essentially same to the aufs
1246 +superblock/inode/dentry/file objects.
1247 +
1248 +Because aufs supports manipulating branches, ie. add/delete/change
1249 +branches dynamically, these objects has its own generation. When
1250 +branches are changed, the generation in aufs superblock is
1251 +incremented. And a generation in other object are compared when it is
1252 +accessed. When a generation in other objects are obsoleted, aufs
1253 +refreshes the internal array.
1254 +
1255 +
1256 +Superblock
1257 +----------------------------------------------------------------------
1258 +Additionally aufs superblock has some data for policies to select one
1259 +among multiple writable branches, XIB files, pseudo-links and kobject.
1260 +See below in detail.
1261 +About the policies which supports copy-down a directory, see
1262 +wbr_policy.txt too.
1263 +
1264 +
1265 +Branch and XINO(External Inode Number Translation Table)
1266 +----------------------------------------------------------------------
1267 +Every branch has its own xino (external inode number translation table)
1268 +file. The xino file is created and unlinked by aufs internally. When two
1269 +members of a union exist on the same filesystem, they share the single
1270 +xino file.
1271 +The struct of a xino file is simple, just a sequence of aufs inode
1272 +numbers which is indexed by the lower inode number.
1273 +In the above sample, assume the inode number of /ro/fileA is i111 and
1274 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1275 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1276 +
1277 +When the inode numbers are not contiguous, the xino file will be sparse
1278 +which has a hole in it and doesn't consume as much disk space as it
1279 +might appear. If your branch filesystem consumes disk space for such
1280 +holes, then you should specify 'xino=' option at mounting aufs.
1281 +
1282 +Aufs has a mount option to free the disk blocks for such holes in XINO
1283 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1284 +meet a problem of disk shortage due to XINO files, then you should try
1285 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1286 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1287 +the holes in XINO files.
1288 +
1289 +Also a writable branch has three kinds of "whiteout bases". All these
1290 +are existed when the branch is joined to aufs, and their names are
1291 +whiteout-ed doubly, so that users will never see their names in aufs
1292 +hierarchy.
1293 +1. a regular file which will be hardlinked to all whiteouts.
1294 +2. a directory to store a pseudo-link.
1295 +3. a directory to store an "orphan"-ed file temporary.
1296 +
1297 +1. Whiteout Base
1298 +   When you remove a file on a readonly branch, aufs handles it as a
1299 +   logical deletion and creates a whiteout on the upper writable branch
1300 +   as a hardlink of this file in order not to consume inode on the
1301 +   writable branch.
1302 +2. Pseudo-link Dir
1303 +   See below, Pseudo-link.
1304 +3. Step-Parent Dir
1305 +   When "fileC" exists on the lower readonly branch only and it is
1306 +   opened and removed with its parent dir, and then user writes
1307 +   something into it, then aufs copies-up fileC to this
1308 +   directory. Because there is no other dir to store fileC. After
1309 +   creating a file under this dir, the file is unlinked.
1310 +
1311 +Because aufs supports manipulating branches, ie. add/delete/change
1312 +dynamically, a branch has its own id. When the branch order changes,
1313 +aufs finds the new index by searching the branch id.
1314 +
1315 +
1316 +Pseudo-link
1317 +----------------------------------------------------------------------
1318 +Assume "fileA" exists on the lower readonly branch only and it is
1319 +hardlinked to "fileB" on the branch. When you write something to fileA,
1320 +aufs copies-up it to the upper writable branch. Additionally aufs
1321 +creates a hardlink under the Pseudo-link Directory of the writable
1322 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1323 +simple list. If fileB is read after unlinking fileA, aufs returns
1324 +filedata from the pseudo-link instead of the lower readonly
1325 +branch. Because the pseudo-link is based upon the inode, to keep the
1326 +inode number by xino (see above) is essentially necessary.
1327 +
1328 +All the hardlinks under the Pseudo-link Directory of the writable branch
1329 +should be restored in a proper location later. Aufs provides a utility
1330 +to do this. The userspace helpers executed at remounting and unmounting
1331 +aufs by default.
1332 +During this utility is running, it puts aufs into the pseudo-link
1333 +maintenance mode. In this mode, only the process which began the
1334 +maintenance mode (and its child processes) is allowed to operate in
1335 +aufs. Some other processes which are not related to the pseudo-link will
1336 +be allowed to run too, but the rest have to return an error or wait
1337 +until the maintenance mode ends. If a process already acquires an inode
1338 +mutex (in VFS), it has to return an error.
1339 +
1340 +
1341 +XIB(external inode number bitmap)
1342 +----------------------------------------------------------------------
1343 +Addition to the xino file per a branch, aufs has an external inode number
1344 +bitmap in a superblock object. It is also an internal file such like a
1345 +xino file.
1346 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1347 +not.
1348 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1349 +
1350 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1351 +reduce the number of consumed disk blocks for these files.
1352 +
1353 +
1354 +Virtual or Vertical Dir, and Readdir in Userspace
1355 +----------------------------------------------------------------------
1356 +In order to support multiple layers (branches), aufs readdir operation
1357 +constructs a virtual dir block on memory. For readdir, aufs calls
1358 +vfs_readdir() internally for each dir on branches, merges their entries
1359 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1360 +object. So the file object has its entry list until it is closed. The
1361 +entry list will be updated when the file position is zero and becomes
1362 +obsoleted. This decision is made in aufs automatically.
1363 +
1364 +The dynamically allocated memory block for the name of entries has a
1365 +unit of 512 bytes (by default) and stores the names contiguously (no
1366 +padding). Another block for each entry is handled by kmem_cache too.
1367 +During building dir blocks, aufs creates hash list and judging whether
1368 +the entry is whiteouted by its upper branch or already listed.
1369 +The merged result is cached in the corresponding inode object and
1370 +maintained by a customizable life-time option.
1371 +
1372 +Some people may call it can be a security hole or invite DoS attack
1373 +since the opened and once readdir-ed dir (file object) holds its entry
1374 +list and becomes a pressure for system memory. But I'd say it is similar
1375 +to files under /proc or /sys. The virtual files in them also holds a
1376 +memory page (generally) while they are opened. When an idea to reduce
1377 +memory for them is introduced, it will be applied to aufs too.
1378 +For those who really hate this situation, I've developed readdir(3)
1379 +library which operates this merging in userspace. You just need to set
1380 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1381 +kernel space for readdir(3).
1382 +
1383 +
1384 +Workqueue
1385 +----------------------------------------------------------------------
1386 +Aufs sometimes requires privilege access to a branch. For instance,
1387 +in copy-up/down operation. When a user process is going to make changes
1388 +to a file which exists in the lower readonly branch only, and the mode
1389 +of one of ancestor directories may not be writable by a user
1390 +process. Here aufs copy-up the file with its ancestors and they may
1391 +require privilege to set its owner/group/mode/etc.
1392 +This is a typical case of a application character of aufs (see
1393 +Introduction).
1394 +
1395 +Aufs uses workqueue synchronously for this case. It creates its own
1396 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1397 +passes the request to call mkdir or write (for example), and wait for
1398 +its completion. This approach solves a problem of a signal handler
1399 +simply.
1400 +If aufs didn't adopt the workqueue and changed the privilege of the
1401 +process, then the process may receive the unexpected SIGXFSZ or other
1402 +signals.
1403 +
1404 +Also aufs uses the system global workqueue ("events" kernel thread) too
1405 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1406 +whiteout base and etc. This is unrelated to a privilege.
1407 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1408 +superblock at the beginning, at the same time waits for the completion
1409 +of all queued asynchronous tasks.
1410 +
1411 +
1412 +Whiteout
1413 +----------------------------------------------------------------------
1414 +The whiteout in aufs is very similar to Unionfs's. That is represented
1415 +by its filename. UnionMount takes an approach of a file mode, but I am
1416 +afraid several utilities (find(1) or something) will have to support it.
1417 +
1418 +Basically the whiteout represents "logical deletion" which stops aufs to
1419 +lookup further, but also it represents "dir is opaque" which also stop
1420 +further lookup.
1421 +
1422 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1423 +In order to make several functions in a single systemcall to be
1424 +revertible, aufs adopts an approach to rename a directory to a temporary
1425 +unique whiteouted name.
1426 +For example, in rename(2) dir where the target dir already existed, aufs
1427 +renames the target dir to a temporary unique whiteouted name before the
1428 +actual rename on a branch, and then handles other actions (make it opaque,
1429 +update the attributes, etc). If an error happens in these actions, aufs
1430 +simply renames the whiteouted name back and returns an error. If all are
1431 +succeeded, aufs registers a function to remove the whiteouted unique
1432 +temporary name completely and asynchronously to the system global
1433 +workqueue.
1434 +
1435 +
1436 +Copy-up
1437 +----------------------------------------------------------------------
1438 +It is a well-known feature or concept.
1439 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1440 +internally and makes change to the new file on the upper writable branch.
1441 +When the trigger systemcall does not update the timestamps of the parent
1442 +dir, aufs reverts it after copy-up.
1443 +
1444 +
1445 +Move-down (aufs3.9 and later)
1446 +----------------------------------------------------------------------
1447 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1448 +the lower readonly branch to the upper writable branch when a user
1449 +changes something about the file.
1450 +"Move-down" is an opposite action of copy-up. Basically this action is
1451 +ran manually instead of automatically and internally.
1452 +For desgin and implementation, aufs has to consider these issues.
1453 +- whiteout for the file may exist on the lower branch.
1454 +- ancestor directories may not exist on the lower branch.
1455 +- diropq for the ancestor directories may exist on the upper branch.
1456 +- free space on the lower branch will reduce.
1457 +- another access to the file may happen during moving-down, including
1458 +  UDBA (see "Revalidate Dentry and UDBA").
1459 +- the file should not be hard-linked nor pseudo-linked. they should be
1460 +  handled by auplink utility later.
1461 +
1462 +Sometimes users want to move-down a file from the upper writable branch
1463 +to the lower readonly or writable branch. For instance,
1464 +- the free space of the upper writable branch is going to run out.
1465 +- create a new intermediate branch between the upper and lower branch.
1466 +- etc.
1467 +
1468 +For this purpose, use "aumvdown" command in aufs-util.git.
1469 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1470 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt    1970-01-01 01:00:00.000000000 +0100
1471 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt       2022-11-05 23:02:18.955889283 +0100
1472 @@ -0,0 +1,85 @@
1473 +
1474 +# Copyright (C) 2015-2022 Junjiro R. Okajima
1475 +#
1476 +# This program is free software; you can redistribute it and/or modify
1477 +# it under the terms of the GNU General Public License as published by
1478 +# the Free Software Foundation; either version 2 of the License, or
1479 +# (at your option) any later version.
1480 +#
1481 +# This program is distributed in the hope that it will be useful,
1482 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1483 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1484 +# GNU General Public License for more details.
1485 +#
1486 +# You should have received a copy of the GNU General Public License
1487 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1488 +
1489 +Support for a branch who has its ->atomic_open()
1490 +----------------------------------------------------------------------
1491 +The filesystems who implement its ->atomic_open() are not majority. For
1492 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1493 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1494 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1495 +sure whether all filesystems who have ->atomic_open() behave like this,
1496 +but NFSv4 surely returns the error.
1497 +
1498 +In order to support ->atomic_open() for aufs, there are a few
1499 +approaches.
1500 +
1501 +A. Introduce aufs_atomic_open()
1502 +   - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1503 +     branch fs.
1504 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1505 +   an aufs user Pip Cet's approach
1506 +   - calls aufs_create(), VFS finish_open() and notify_change().
1507 +   - pass fake-mode to finish_open(), and then correct the mode by
1508 +     notify_change().
1509 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1510 +   - no aufs_atomic_open().
1511 +   - aufs_lookup() registers the TID to an aufs internal object.
1512 +   - aufs_create() does nothing when the matching TID is registered, but
1513 +     registers the mode.
1514 +   - aufs_open() calls branch fs's ->atomic_open() when the matching
1515 +     TID is registered.
1516 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1517 +   credential
1518 +   - no aufs_atomic_open().
1519 +   - aufs_create() registers the TID to an internal object. this info
1520 +     represents "this process created this file just now."
1521 +   - when aufs gets EACCES from branch fs's ->open(), then confirm the
1522 +     registered TID and re-try open() with superuser's credential.
1523 +
1524 +Pros and cons for each approach.
1525 +
1526 +A.
1527 +   - straightforward but highly depends upon VFS internal.
1528 +   - the atomic behavaiour is kept.
1529 +   - some of parameters such as nameidata are hard to reproduce for
1530 +     branch fs.
1531 +   - large overhead.
1532 +B.
1533 +   - easy to implement.
1534 +   - the atomic behavaiour is lost.
1535 +C.
1536 +   - the atomic behavaiour is kept.
1537 +   - dirty and tricky.
1538 +   - VFS checks whether the file is created correctly after calling
1539 +     ->create(), which means this approach doesn't work.
1540 +D.
1541 +   - easy to implement.
1542 +   - the atomic behavaiour is lost.
1543 +   - to open a file with superuser's credential and give it to a user
1544 +     process is a bad idea, since the file object keeps the credential
1545 +     in it. It may affect LSM or something. This approach doesn't work
1546 +     either.
1547 +
1548 +The approach A is ideal, but it hard to implement. So here is a
1549 +variation of A, which is to be implemented.
1550 +
1551 +A-1. Introduce aufs_atomic_open()
1552 +     - calls branch fs ->atomic_open() if exists. otherwise calls
1553 +       vfs_create() and finish_open().
1554 +     - the demerit is that the several checks after branch fs
1555 +       ->atomic_open() are lost. in the ordinary case, the checks are
1556 +       done by VFS:do_last(), lookup_open() and atomic_open(). some can
1557 +       be implemented in aufs, but not all I am afraid.
1558 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1559 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1560 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt    2022-11-05 23:02:18.959222617 +0100
1561 @@ -0,0 +1,113 @@
1562 +
1563 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1564 +#
1565 +# This program is free software; you can redistribute it and/or modify
1566 +# it under the terms of the GNU General Public License as published by
1567 +# the Free Software Foundation; either version 2 of the License, or
1568 +# (at your option) any later version.
1569 +#
1570 +# This program is distributed in the hope that it will be useful,
1571 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1572 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1573 +# GNU General Public License for more details.
1574 +#
1575 +# You should have received a copy of the GNU General Public License
1576 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1577 +
1578 +Lookup in a Branch
1579 +----------------------------------------------------------------------
1580 +Since aufs has a character of sub-VFS (see Introduction), it operates
1581 +lookup for branches as VFS does. It may be a heavy work. But almost all
1582 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1583 +directly connected to its parent. Digging down the directory hierarchy
1584 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1585 +aufs calls it.
1586 +
1587 +When a branch is a remote filesystem, aufs basically relies upon its
1588 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1589 +them.
1590 +For d_revalidate, aufs implements three levels of revalidate tests. See
1591 +"Revalidate Dentry and UDBA" in detail.
1592 +
1593 +
1594 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1595 +----------------------------------------------------------------------
1596 +Let's try case study.
1597 +- aufs has two branches, upper readwrite and lower readonly.
1598 +  /au = /rw + /ro
1599 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1600 +- user invoked "chmod a+rx /au/dirA"
1601 +- the internal copy-up is activated and "/rw/dirA" is created and its
1602 +  permission bits are set to world readable.
1603 +- then "/au/dirA" becomes world readable?
1604 +
1605 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1606 +or it may be a natively readonly filesystem. If aufs respects the lower
1607 +branch, it should not respond readdir request from other users. But user
1608 +allowed it by chmod. Should really aufs rejects showing the entries
1609 +under /ro/dirA?
1610 +
1611 +To be honest, I don't have a good solution for this case. So aufs
1612 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1613 +users.
1614 +When dirperm1 is specified, aufs checks only the highest one for the
1615 +directory permission, and shows the entries. Otherwise, as usual, checks
1616 +every dir existing on all branches and rejects the request.
1617 +
1618 +As a side effect, dirperm1 option improves the performance of aufs
1619 +because the number of permission check is reduced when the number of
1620 +branch is many.
1621 +
1622 +
1623 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1624 +----------------------------------------------------------------------
1625 +Generally VFS helpers re-validate a dentry as a part of lookup.
1626 +0. digging down the directory hierarchy.
1627 +1. lock the parent dir by its i_mutex.
1628 +2. lookup the final (child) entry.
1629 +3. revalidate it.
1630 +4. call the actual operation (create, unlink, etc.)
1631 +5. unlock the parent dir
1632 +
1633 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1634 +called. Actually aufs implements it and checks the dentry on a branch is
1635 +still valid.
1636 +But it is not enough. Because aufs has to release the lock for the
1637 +parent dir on a branch at the end of ->lookup() (step 2) and
1638 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1639 +held by VFS.
1640 +If the file on a branch is changed directly, eg. bypassing aufs, after
1641 +aufs released the lock, then the subsequent operation may cause
1642 +something unpleasant result.
1643 +
1644 +This situation is a result of VFS architecture, ->lookup() and
1645 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1646 +design from VFS's point of view. It is just not suitable for sub-VFS
1647 +character in aufs.
1648 +
1649 +Aufs supports such case by three level of revalidation which is
1650 +selectable by user.
1651 +1. Simple Revalidate
1652 +   Addition to the native flow in VFS's, confirm the child-parent
1653 +   relationship on the branch just after locking the parent dir on the
1654 +   branch in the "actual operation" (step 4). When this validation
1655 +   fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1656 +   checks the validation of the dentry on branches.
1657 +2. Monitor Changes Internally by Inotify/Fsnotify
1658 +   Addition to above, in the "actual operation" (step 4) aufs re-lookup
1659 +   the dentry on the branch, and returns EBUSY if it finds different
1660 +   dentry.
1661 +   Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1662 +   during it is in cache. When the event is notified, aufs registers a
1663 +   function to kernel 'events' thread by schedule_work(). And the
1664 +   function sets some special status to the cached aufs dentry and inode
1665 +   private data. If they are not cached, then aufs has nothing to
1666 +   do. When the same file is accessed through aufs (step 0-3) later,
1667 +   aufs will detect the status and refresh all necessary data.
1668 +   In this mode, aufs has to ignore the event which is fired by aufs
1669 +   itself.
1670 +3. No Extra Validation
1671 +   This is the simplest test and doesn't add any additional revalidation
1672 +   test, and skip the revalidation in step 4. It is useful and improves
1673 +   aufs performance when system surely hide the aufs branches from user,
1674 +   by over-mounting something (or another method).
1675 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1676 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1677 +++ linux/Documentation/filesystems/aufs/design/04branch.txt    2022-11-05 23:02:18.959222617 +0100
1678 @@ -0,0 +1,74 @@
1679 +
1680 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1681 +#
1682 +# This program is free software; you can redistribute it and/or modify
1683 +# it under the terms of the GNU General Public License as published by
1684 +# the Free Software Foundation; either version 2 of the License, or
1685 +# (at your option) any later version.
1686 +#
1687 +# This program is distributed in the hope that it will be useful,
1688 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1689 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1690 +# GNU General Public License for more details.
1691 +#
1692 +# You should have received a copy of the GNU General Public License
1693 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1694 +
1695 +Branch Manipulation
1696 +
1697 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1698 +and changing its permission/attribute, there are a lot of works to do.
1699 +
1700 +
1701 +Add a Branch
1702 +----------------------------------------------------------------------
1703 +o Confirm the adding dir exists outside of aufs, including loopback
1704 +  mount, and its various attributes.
1705 +o Initialize the xino file and whiteout bases if necessary.
1706 +  See struct.txt.
1707 +
1708 +o Check the owner/group/mode of the directory
1709 +  When the owner/group/mode of the adding directory differs from the
1710 +  existing branch, aufs issues a warning because it may impose a
1711 +  security risk.
1712 +  For example, when a upper writable branch has a world writable empty
1713 +  top directory, a malicious user can create any files on the writable
1714 +  branch directly, like copy-up and modify manually. If something like
1715 +  /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1716 +  writable branch, and the writable branch is world-writable, then a
1717 +  malicious guy may create /etc/passwd on the writable branch directly
1718 +  and the infected file will be valid in aufs.
1719 +  I am afraid it can be a security issue, but aufs can do nothing except
1720 +  producing a warning.
1721 +
1722 +
1723 +Delete a Branch
1724 +----------------------------------------------------------------------
1725 +o Confirm the deleting branch is not busy
1726 +  To be general, there is one merit to adopt "remount" interface to
1727 +  manipulate branches. It is to discard caches. At deleting a branch,
1728 +  aufs checks the still cached (and connected) dentries and inodes. If
1729 +  there are any, then they are all in-use. An inode without its
1730 +  corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1731 +
1732 +  For the cached one, aufs checks whether the same named entry exists on
1733 +  other branches.
1734 +  If the cached one is a directory, because aufs provides a merged view
1735 +  to users, as long as one dir is left on any branch aufs can show the
1736 +  dir to users. In this case, the branch can be removed from aufs.
1737 +  Otherwise aufs rejects deleting the branch.
1738 +
1739 +  If any file on the deleting branch is opened by aufs, then aufs
1740 +  rejects deleting.
1741 +
1742 +
1743 +Modify the Permission of a Branch
1744 +----------------------------------------------------------------------
1745 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1746 +  See struct.txt.
1747 +
1748 +o rw --> ro: Confirm the modifying branch is not busy
1749 +  Aufs rejects the request if any of these conditions are true.
1750 +  - a file on the branch is mmap-ed.
1751 +  - a regular file on the branch is opened for write and there is no
1752 +    same named entry on the upper branch.
1753 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1754 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt     1970-01-01 01:00:00.000000000 +0100
1755 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt        2022-11-05 23:02:18.959222617 +0100
1756 @@ -0,0 +1,64 @@
1757 +
1758 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1759 +#
1760 +# This program is free software; you can redistribute it and/or modify
1761 +# it under the terms of the GNU General Public License as published by
1762 +# the Free Software Foundation; either version 2 of the License, or
1763 +# (at your option) any later version.
1764 +#
1765 +# This program is distributed in the hope that it will be useful,
1766 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1767 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1768 +# GNU General Public License for more details.
1769 +#
1770 +# You should have received a copy of the GNU General Public License
1771 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1772 +
1773 +Policies to Select One among Multiple Writable Branches
1774 +----------------------------------------------------------------------
1775 +When the number of writable branch is more than one, aufs has to decide
1776 +the target branch for file creation or copy-up. By default, the highest
1777 +writable branch which has the parent (or ancestor) dir of the target
1778 +file is chosen (top-down-parent policy).
1779 +By user's request, aufs implements some other policies to select the
1780 +writable branch, for file creation several policies, round-robin,
1781 +most-free-space, and other policies. For copy-up, top-down-parent,
1782 +bottom-up-parent, bottom-up and others.
1783 +
1784 +As expected, the round-robin policy selects the branch in circular. When
1785 +you have two writable branches and creates 10 new files, 5 files will be
1786 +created for each branch. mkdir(2) systemcall is an exception. When you
1787 +create 10 new directories, all will be created on the same branch.
1788 +And the most-free-space policy selects the one which has most free
1789 +space among the writable branches. The amount of free space will be
1790 +checked by aufs internally, and users can specify its time interval.
1791 +
1792 +The policies for copy-up is more simple,
1793 +top-down-parent is equivalent to the same named on in create policy,
1794 +bottom-up-parent selects the writable branch where the parent dir
1795 +exists and the nearest upper one from the copyup-source,
1796 +bottom-up selects the nearest upper writable branch from the
1797 +copyup-source, regardless the existence of the parent dir.
1798 +
1799 +There are some rules or exceptions to apply these policies.
1800 +- If there is a readonly branch above the policy-selected branch and
1801 +  the parent dir is marked as opaque (a variation of whiteout), or the
1802 +  target (creating) file is whiteout-ed on the upper readonly branch,
1803 +  then the result of the policy is ignored and the target file will be
1804 +  created on the nearest upper writable branch than the readonly branch.
1805 +- If there is a writable branch above the policy-selected branch and
1806 +  the parent dir is marked as opaque or the target file is whiteouted
1807 +  on the branch, then the result of the policy is ignored and the target
1808 +  file will be created on the highest one among the upper writable
1809 +  branches who has diropq or whiteout. In case of whiteout, aufs removes
1810 +  it as usual.
1811 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1812 +  They try selecting the branch where the source exists as possible
1813 +  since copyup a large file will take long time. If it can't be,
1814 +  ie. the branch where the source exists is readonly, then they will
1815 +  follow the copyup policy.
1816 +- There is an exception for rename(2) when the target exists.
1817 +  If the rename target exists, aufs compares the index of the branches
1818 +  where the source and the target exists and selects the higher
1819 +  one. If the selected branch is readonly, then aufs follows the
1820 +  copyup policy.
1821 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot linux/Documentation/filesystems/aufs/design/06dirren.dot
1822 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot 1970-01-01 01:00:00.000000000 +0100
1823 +++ linux/Documentation/filesystems/aufs/design/06dirren.dot    2022-11-05 23:02:18.959222617 +0100
1824 @@ -0,0 +1,44 @@
1825 +
1826 +// to view this graph, run dot(1) command in GRAPHVIZ.
1827 +//
1828 +// This program is free software; you can redistribute it and/or modify
1829 +// it under the terms of the GNU General Public License as published by
1830 +// the Free Software Foundation; either version 2 of the License, or
1831 +// (at your option) any later version.
1832 +//
1833 +// This program is distributed in the hope that it will be useful,
1834 +// but WITHOUT ANY WARRANTY; without even the implied warranty of
1835 +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1836 +// GNU General Public License for more details.
1837 +//
1838 +// You should have received a copy of the GNU General Public License
1839 +// along with this program.  If not, see <http://www.gnu.org/licenses/>.
1840 +
1841 +digraph G {
1842 +node [shape=box];
1843 +whinfo [label="detailed info file\n(lower_brid_root-hinum, h_inum, namelen, old name)"];
1844 +
1845 +node [shape=oval];
1846 +
1847 +aufs_rename -> whinfo [label="store/remove"];
1848 +
1849 +node [shape=oval];
1850 +inode_list [label="h_inum list in branch\ncache"];
1851 +
1852 +node [shape=box];
1853 +whinode [label="h_inum list file"];
1854 +
1855 +node [shape=oval];
1856 +brmgmt [label="br_add/del/mod/umount"];
1857 +
1858 +brmgmt -> inode_list [label="create/remove"];
1859 +brmgmt -> whinode [label="load/store"];
1860 +
1861 +inode_list -> whinode [style=dashed,dir=both];
1862 +
1863 +aufs_rename -> inode_list [label="add/del"];
1864 +
1865 +aufs_lookup -> inode_list [label="search"];
1866 +
1867 +aufs_lookup -> whinfo [label="load/remove"];
1868 +}
1869 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt linux/Documentation/filesystems/aufs/design/06dirren.txt
1870 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt 1970-01-01 01:00:00.000000000 +0100
1871 +++ linux/Documentation/filesystems/aufs/design/06dirren.txt    2022-11-05 23:02:18.959222617 +0100
1872 @@ -0,0 +1,102 @@
1873 +
1874 +# Copyright (C) 2017-2022 Junjiro R. Okajima
1875 +#
1876 +# This program is free software; you can redistribute it and/or modify
1877 +# it under the terms of the GNU General Public License as published by
1878 +# the Free Software Foundation; either version 2 of the License, or
1879 +# (at your option) any later version.
1880 +#
1881 +# This program is distributed in the hope that it will be useful,
1882 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1883 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1884 +# GNU General Public License for more details.
1885 +#
1886 +# You should have received a copy of the GNU General Public License
1887 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1888 +
1889 +Special handling for renaming a directory (DIRREN)
1890 +----------------------------------------------------------------------
1891 +First, let's assume we have a simple usecase.
1892 +
1893 +- /u = /rw + /ro
1894 +- /rw/dirA exists
1895 +- /ro/dirA and /ro/dirA/file exist too
1896 +- there is no dirB on both branches
1897 +- a user issues rename("dirA", "dirB")
1898 +
1899 +Now, what should aufs behave against this rename(2)?
1900 +There are a few possible cases.
1901 +
1902 +A. returns EROFS.
1903 +   since dirA exists on a readonly branch which cannot be renamed.
1904 +B. returns EXDEV.
1905 +   it is possible to copy-up dirA (only the dir itself), but the child
1906 +   entries ("file" in this case) should not be. it must be a bad
1907 +   approach to copy-up recursively.
1908 +C. returns a success.
1909 +   even the branch /ro is readonly, aufs tries renaming it. Obviously it
1910 +   is a violation of aufs' policy.
1911 +D. construct an extra information which indicates that /ro/dirA should
1912 +   be handled as the name of dirB.
1913 +   overlayfs has a similar feature called REDIRECT.
1914 +
1915 +Until now, aufs implements the case B only which returns EXDEV, and
1916 +expects the userspace application behaves like mv(1) which tries
1917 +issueing rename(2) recursively.
1918 +
1919 +A new aufs feature called DIRREN is introduced which implements the case
1920 +D. There are several "extra information" added.
1921 +
1922 +1. detailed info per renamed directory
1923 +   path: /rw/dirB/$AUFS_WH_DR_INFO_PFX.<lower branch-id>
1924 +2. the inode-number list of directories on a branch
1925 +   path: /rw/dirB/$AUFS_WH_DR_BRHINO
1926 +
1927 +The filename of "detailed info per directory" represents the lower
1928 +branch, and its format is
1929 +- a type of the branch id
1930 +  one of these.
1931 +  + uuid (not implemented yet)
1932 +  + fsid
1933 +  + dev
1934 +- the inode-number of the branch root dir
1935 +
1936 +And it contains these info in a single regular file.
1937 +- magic number
1938 +- branch's inode-number of the logically renamed dir
1939 +- the name of the before-renamed dir
1940 +
1941 +The "detailed info per directory" file is created in aufs rename(2), and
1942 +loaded in any lookup.
1943 +The info is considered in lookup for the matching case only. Here
1944 +"matching" means that the root of branch (in the info filename) is same
1945 +to the current looking-up branch. After looking-up the before-renamed
1946 +name, the inode-number is compared. And the matched dentry is used.
1947 +
1948 +The "inode-number list of directories" is a regular file which contains
1949 +simply the inode-numbers on the branch. The file is created or updated
1950 +in removing the branch, and loaded in adding the branch. Its lifetime is
1951 +equal to the branch.
1952 +The list is referred in lookup, and when the current target inode is
1953 +found in the list, the aufs tries loading the "detailed info per
1954 +directory" and get the changed and valid name of the dir.
1955 +
1956 +Theoretically these "extra informaiton" may be able to be put into XATTR
1957 +in the dir inode. But aufs doesn't choose this way because
1958 +1. XATTR may not be supported by the branch (or its configuration)
1959 +2. XATTR may have its size limit.
1960 +3. XATTR may be less easy to convert than a regular file, when the
1961 +   format of the info is changed in the future.
1962 +At the same time, I agree that the regular file approach is much slower
1963 +than XATTR approach. So, in the future, aufs may take the XATTR or other
1964 +better approach.
1965 +
1966 +This DIRREN feature is enabled by aufs configuration, and is activated
1967 +by a new mount option.
1968 +
1969 +For the more complicated case, there is a work with UDBA option, which
1970 +is to dected the direct access to the branches (by-passing aufs) and to
1971 +maintain the cashes in aufs. Since a single cached aufs dentry may
1972 +contains two names, before- and after-rename, the name comparision in
1973 +UDBA handler may not work correctly. In this case, the behaviour will be
1974 +equivalen to udba=reval case.
1975 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1976 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt   1970-01-01 01:00:00.000000000 +0100
1977 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt      2022-11-05 23:02:18.959222617 +0100
1978 @@ -0,0 +1,118 @@
1979 +
1980 +# Copyright (C) 2011-2022 Junjiro R. Okajima
1981 +#
1982 +# This program is free software; you can redistribute it and/or modify
1983 +# it under the terms of the GNU General Public License as published by
1984 +# the Free Software Foundation; either version 2 of the License, or
1985 +# (at your option) any later version.
1986 +#
1987 +# This program is distributed in the hope that it will be useful,
1988 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1989 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1990 +# GNU General Public License for more details.
1991 +#
1992 +# You should have received a copy of the GNU General Public License
1993 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1994 +
1995 +File-based Hierarchical Storage Management (FHSM)
1996 +----------------------------------------------------------------------
1997 +Hierarchical Storage Management (or HSM) is a well-known feature in the
1998 +storage world. Aufs provides this feature as file-based with multiple
1999 +writable branches, based upon the principle of "Colder, the Lower".
2000 +Here the word "colder" means that the less used files, and "lower" means
2001 +that the position in the order of the stacked branches vertically.
2002 +These multiple writable branches are prioritized, ie. the topmost one
2003 +should be the fastest drive and be used heavily.
2004 +
2005 +o Characters in aufs FHSM story
2006 +- aufs itself and a new branch attribute.
2007 +- a new ioctl interface to move-down and to establish a connection with
2008 +  the daemon ("move-down" is a converse of "copy-up").
2009 +- userspace tool and daemon.
2010 +
2011 +The userspace daemon establishes a connection with aufs and waits for
2012 +the notification. The notified information is very similar to struct
2013 +statfs containing the number of consumed blocks and inodes.
2014 +When the consumed blocks/inodes of a branch exceeds the user-specified
2015 +upper watermark, the daemon activates its move-down process until the
2016 +consumed blocks/inodes reaches the user-specified lower watermark.
2017 +
2018 +The actual move-down is done by aufs based upon the request from
2019 +user-space since we need to maintain the inode number and the internal
2020 +pointer arrays in aufs.
2021 +
2022 +Currently aufs FHSM handles the regular files only. Additionally they
2023 +must not be hard-linked nor pseudo-linked.
2024 +
2025 +
2026 +o Cowork of aufs and the user-space daemon
2027 +  During the userspace daemon established the connection, aufs sends a
2028 +  small notification to it whenever aufs writes something into the
2029 +  writable branch. But it may cost high since aufs issues statfs(2)
2030 +  internally. So user can specify a new option to cache the
2031 +  info. Actually the notification is controlled by these factors.
2032 +  + the specified cache time.
2033 +  + classified as "force" by aufs internally.
2034 +  Until the specified time expires, aufs doesn't send the info
2035 +  except the forced cases. When aufs decide forcing, the info is always
2036 +  notified to userspace.
2037 +  For example, the number of free inodes is generally large enough and
2038 +  the shortage of it happens rarely. So aufs doesn't force the
2039 +  notification when creating a new file, directory and others. This is
2040 +  the typical case which aufs doesn't force.
2041 +  When aufs writes the actual filedata and the files consumes any of new
2042 +  blocks, the aufs forces notifying.
2043 +
2044 +
2045 +o Interfaces in aufs
2046 +- New branch attribute.
2047 +  + fhsm
2048 +    Specifies that the branch is managed by FHSM feature. In other word,
2049 +    participant in the FHSM.
2050 +    When nofhsm is set to the branch, it will not be the source/target
2051 +    branch of the move-down operation. This attribute is set
2052 +    independently from coo and moo attributes, and if you want full
2053 +    FHSM, you should specify them as well.
2054 +- New mount option.
2055 +  + fhsm_sec
2056 +    Specifies a second to suppress many less important info to be
2057 +    notified.
2058 +- New ioctl.
2059 +  + AUFS_CTL_FHSM_FD
2060 +    create a new file descriptor which userspace can read the notification
2061 +    (a subset of struct statfs) from aufs.
2062 +- Module parameter 'brs'
2063 +  It has to be set to 1. Otherwise the new mount option 'fhsm' will not
2064 +  be set.
2065 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
2066 +  When there are two or more branches with fhsm attributes,
2067 +  /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
2068 +  terminates it. As a result of remounting and branch-manipulation, the
2069 +  number of branches with fhsm attribute can be one. In this case,
2070 +  /sbin/mount.aufs will terminate the user-space daemon.
2071 +
2072 +
2073 +Finally the operation is done as these steps in kernel-space.
2074 +- make sure that,
2075 +  + no one else is using the file.
2076 +  + the file is not hard-linked.
2077 +  + the file is not pseudo-linked.
2078 +  + the file is a regular file.
2079 +  + the parent dir is not opaqued.
2080 +- find the target writable branch.
2081 +- make sure the file is not whiteout-ed by the upper (than the target)
2082 +  branch.
2083 +- make the parent dir on the target branch.
2084 +- mutex lock the inode on the branch.
2085 +- unlink the whiteout on the target branch (if exists).
2086 +- lookup and create the whiteout-ed temporary name on the target branch.
2087 +- copy the file as the whiteout-ed temporary name on the target branch.
2088 +- rename the whiteout-ed temporary name to the original name.
2089 +- unlink the file on the source branch.
2090 +- maintain the internal pointer array and the external inode number
2091 +  table (XINO).
2092 +- maintain the timestamps and other attributes of the parent dir and the
2093 +  file.
2094 +
2095 +And of course, in every step, an error may happen. So the operation
2096 +should restore the original file state after an error happens.
2097 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
2098 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt   1970-01-01 01:00:00.000000000 +0100
2099 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt      2022-11-05 23:02:18.959222617 +0100
2100 @@ -0,0 +1,72 @@
2101 +
2102 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2103 +#
2104 +# This program is free software; you can redistribute it and/or modify
2105 +# it under the terms of the GNU General Public License as published by
2106 +# the Free Software Foundation; either version 2 of the License, or
2107 +# (at your option) any later version.
2108 +#
2109 +# This program is distributed in the hope that it will be useful,
2110 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2111 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2112 +# GNU General Public License for more details.
2113 +#
2114 +# You should have received a copy of the GNU General Public License
2115 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2116 +
2117 +mmap(2) -- File Memory Mapping
2118 +----------------------------------------------------------------------
2119 +In aufs, the file-mapped pages are handled by a branch fs directly, no
2120 +interaction with aufs. It means aufs_mmap() calls the branch fs's
2121 +->mmap().
2122 +This approach is simple and good, but there is one problem.
2123 +Under /proc, several entries show the mmapped files by its path (with
2124 +device and inode number), and the printed path will be the path on the
2125 +branch fs's instead of virtual aufs's.
2126 +This is not a problem in most cases, but some utilities lsof(1) (and its
2127 +user) may expect the path on aufs.
2128 +
2129 +To address this issue, aufs adds a new member called vm_prfile in struct
2130 +vm_area_struct (and struct vm_region). The original vm_file points to
2131 +the file on the branch fs in order to handle everything correctly as
2132 +usual. The new vm_prfile points to a virtual file in aufs, and the
2133 +show-functions in procfs refers to vm_prfile if it is set.
2134 +Also we need to maintain several other places where touching vm_file
2135 +such like
2136 +- fork()/clone() copies vma and the reference count of vm_file is
2137 +  incremented.
2138 +- merging vma maintains the ref count too.
2139 +
2140 +This is not a good approach. It just fakes the printed path. But it
2141 +leaves all behaviour around f_mapping unchanged. This is surely an
2142 +advantage.
2143 +Actually aufs had adopted another complicated approach which calls
2144 +generic_file_mmap() and handles struct vm_operations_struct. In this
2145 +approach, aufs met a hard problem and I could not solve it without
2146 +switching the approach.
2147 +
2148 +There may be one more another approach which is
2149 +- bind-mount the branch-root onto the aufs-root internally
2150 +- grab the new vfsmount (ie. struct mount)
2151 +- lazy-umount the branch-root internally
2152 +- in open(2) the aufs-file, open the branch-file with the hidden
2153 +  vfsmount (instead of the original branch's vfsmount)
2154 +- ideally this "bind-mount and lazy-umount" should be done atomically,
2155 +  but it may be possible from userspace by the mount helper.
2156 +
2157 +Adding the internal hidden vfsmount and using it in opening a file, the
2158 +file path under /proc will be printed correctly. This approach looks
2159 +smarter, but is not possible I am afraid.
2160 +- aufs-root may be bind-mount later. when it happens, another hidden
2161 +  vfsmount will be required.
2162 +- it is hard to get the chance to bind-mount and lazy-umount
2163 +  + in kernel-space, FS can have vfsmount in open(2) via
2164 +    file->f_path, and aufs can know its vfsmount. But several locks are
2165 +    already acquired, and if aufs tries to bind-mount and lazy-umount
2166 +    here, then it may cause a deadlock.
2167 +  + in user-space, bind-mount doesn't invoke the mount helper.
2168 +- since /proc shows dev and ino, aufs has to give vma these info. it
2169 +  means a new member vm_prinode will be necessary. this is essentially
2170 +  equivalent to vm_prfile described above.
2171 +
2172 +I have to give up this "looks-smater" approach.
2173 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2174 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt  1970-01-01 01:00:00.000000000 +0100
2175 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt     2022-11-05 23:02:18.959222617 +0100
2176 @@ -0,0 +1,94 @@
2177 +
2178 +# Copyright (C) 2014-2022 Junjiro R. Okajima
2179 +#
2180 +# This program is free software; you can redistribute it and/or modify
2181 +# it under the terms of the GNU General Public License as published by
2182 +# the Free Software Foundation; either version 2 of the License, or
2183 +# (at your option) any later version.
2184 +#
2185 +# This program is distributed in the hope that it will be useful,
2186 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2187 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2188 +# GNU General Public License for more details.
2189 +#
2190 +# You should have received a copy of the GNU General Public License
2191 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2192 +
2193 +Listing XATTR/EA and getting the value
2194 +----------------------------------------------------------------------
2195 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2196 +shows the values from the topmost existing file. This behaviour is good
2197 +for the non-dir entries since the bahaviour exactly matches the shown
2198 +information. But for the directories, aufs considers all the same named
2199 +entries on the lower branches. Which means, if one of the lower entry
2200 +rejects readdir call, then aufs returns an error even if the topmost
2201 +entry allows it. This behaviour is necessary to respect the branch fs's
2202 +security, but can make users confused since the user-visible standard
2203 +attributes don't match the behaviour.
2204 +To address this issue, aufs has a mount option called dirperm1 which
2205 +checks the permission for the topmost entry only, and ignores the lower
2206 +entry's permission.
2207 +
2208 +A similar issue can happen around XATTR.
2209 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2210 +always set. Otherwise these very unpleasant situation would happen.
2211 +- listxattr(2) may return the duplicated entries.
2212 +- users may not be able to remove or reset the XATTR forever,
2213 +
2214 +
2215 +XATTR/EA support in the internal (copy,move)-(up,down)
2216 +----------------------------------------------------------------------
2217 +Generally the extended attributes of inode are categorized as these.
2218 +- "security" for LSM and capability.
2219 +- "system" for posix ACL, 'acl' mount option is required for the branch
2220 +  fs generally.
2221 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2222 +- "user" for userspace, 'user_xattr' mount option is required for the
2223 +  branch fs generally.
2224 +
2225 +Moreover there are some other categories. Aufs handles these rather
2226 +unpopular categories as the ordinary ones, ie. there is no special
2227 +condition nor exception.
2228 +
2229 +In copy-up, the support for XATTR on the dst branch may differ from the
2230 +src branch. In this case, the copy-up operation will get an error and
2231 +the original user operation which triggered the copy-up will fail. It
2232 +can happen that even all copy-up will fail.
2233 +When both of src and dst branches support XATTR and if an error occurs
2234 +during copying XATTR, then the copy-up should fail obviously. That is a
2235 +good reason and aufs should return an error to userspace. But when only
2236 +the src branch support that XATTR, aufs should not return an error.
2237 +For example, the src branch supports ACL but the dst branch doesn't
2238 +because the dst branch may natively un-support it or temporary
2239 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2240 +may NOT return an error even if the XATTR is not supported. It is
2241 +totally up to the branch fs.
2242 +
2243 +Anyway when the aufs internal copy-up gets an error from the dst branch
2244 +fs, then aufs tries removing the just copied entry and returns the error
2245 +to the userspace. The worst case of this situation will be all copy-up
2246 +will fail.
2247 +
2248 +For the copy-up operation, there two basic approaches.
2249 +- copy the specified XATTR only (by category above), and return the
2250 +  error unconditionally if it happens.
2251 +- copy all XATTR, and ignore the error on the specified category only.
2252 +
2253 +In order to support XATTR and to implement the correct behaviour, aufs
2254 +chooses the latter approach and introduces some new branch attributes,
2255 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2256 +They correspond to the XATTR namespaces (see above). Additionally, to be
2257 +convenient, "icex" is also provided which means all "icex*" attributes
2258 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2259 +
2260 +The meaning of these attributes is to ignore the error from setting
2261 +XATTR on that branch.
2262 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2263 +error from the dst branch according to the specified attributes.
2264 +
2265 +Some XATTR may have its default value. The default value may come from
2266 +the parent dir or the environment. If the default value is set at the
2267 +file creating-time, it will be overwritten by copy-up.
2268 +Some contradiction may happen I am afraid.
2269 +Do we need another attribute to stop copying XATTR? I am unsure. For
2270 +now, aufs implements the branch attributes to ignore the error.
2271 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2272 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2273 +++ linux/Documentation/filesystems/aufs/design/07export.txt    2022-11-05 23:02:18.959222617 +0100
2274 @@ -0,0 +1,58 @@
2275 +
2276 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2277 +#
2278 +# This program is free software; you can redistribute it and/or modify
2279 +# it under the terms of the GNU General Public License as published by
2280 +# the Free Software Foundation; either version 2 of the License, or
2281 +# (at your option) any later version.
2282 +#
2283 +# This program is distributed in the hope that it will be useful,
2284 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2285 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2286 +# GNU General Public License for more details.
2287 +#
2288 +# You should have received a copy of the GNU General Public License
2289 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2290 +
2291 +Export Aufs via NFS
2292 +----------------------------------------------------------------------
2293 +Here is an approach.
2294 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2295 +  generation.
2296 +- iget_locked(): initialize aufs inode generation for a new inode, and
2297 +  store it in xigen file.
2298 +- destroy_inode(): increment aufs inode generation and store it in xigen
2299 +  file. it is necessary even if it is not unlinked, because any data of
2300 +  inode may be changed by UDBA.
2301 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2302 +  build file handle by
2303 +  + branch id (4 bytes)
2304 +  + superblock generation (4 bytes)
2305 +  + inode number (4 or 8 bytes)
2306 +  + parent dir inode number (4 or 8 bytes)
2307 +  + inode generation (4 bytes))
2308 +  + return value of exportfs_encode_fh() for the parent on a branch (4
2309 +    bytes)
2310 +  + file handle for a branch (by exportfs_encode_fh())
2311 +- fh_to_dentry():
2312 +  + find the index of a branch from its id in handle, and check it is
2313 +    still exist in aufs.
2314 +  + 1st level: get the inode number from handle and search it in cache.
2315 +  + 2nd level: if not found in cache, get the parent inode number from
2316 +    the handle and search it in cache. and then open the found parent
2317 +    dir, find the matching inode number by vfs_readdir() and get its
2318 +    name, and call lookup_one_len() for the target dentry.
2319 +  + 3rd level: if the parent dir is not cached, call
2320 +    exportfs_decode_fh() for a branch and get the parent on a branch,
2321 +    build a pathname of it, convert it a pathname in aufs, call
2322 +    path_lookup(). now aufs gets a parent dir dentry, then handle it as
2323 +    the 2nd level.
2324 +  + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2325 +    for every branch, but not itself. to get this, (currently) aufs
2326 +    searches in current->nsproxy->mnt_ns list. it may not be a good
2327 +    idea, but I didn't get other approach.
2328 +  + test the generation of the gotten inode.
2329 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2330 +  convert it into ESTALE for NFSD.
2331 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2332 +  lookup_one_len(), vfs_getattr(), encode_fh() and others.
2333 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2334 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt   1970-01-01 01:00:00.000000000 +0100
2335 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt      2022-11-05 23:02:18.959222617 +0100
2336 @@ -0,0 +1,52 @@
2337 +
2338 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2339 +#
2340 +# This program is free software; you can redistribute it and/or modify
2341 +# it under the terms of the GNU General Public License as published by
2342 +# the Free Software Foundation; either version 2 of the License, or
2343 +# (at your option) any later version.
2344 +#
2345 +# This program is distributed in the hope that it will be useful,
2346 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2347 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2348 +# GNU General Public License for more details.
2349 +#
2350 +# You should have received a copy of the GNU General Public License
2351 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2352 +
2353 +Show Whiteout Mode (shwh)
2354 +----------------------------------------------------------------------
2355 +Generally aufs hides the name of whiteouts. But in some cases, to show
2356 +them is very useful for users. For instance, creating a new middle layer
2357 +(branch) by merging existing layers.
2358 +
2359 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2360 +When you have three branches,
2361 +- Bottom: 'system', squashfs (underlying base system), read-only
2362 +- Middle: 'mods', squashfs, read-only
2363 +- Top: 'overlay', ram (tmpfs), read-write
2364 +
2365 +The top layer is loaded at boot time and saved at shutdown, to preserve
2366 +the changes made to the system during the session.
2367 +When larger changes have been made, or smaller changes have accumulated,
2368 +the size of the saved top layer data grows. At this point, it would be
2369 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2370 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2371 +restoring save and load speed.
2372 +
2373 +This merging is simplified by the use of another aufs mount, of just the
2374 +two overlay branches using the 'shwh' option.
2375 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2376 +       aufs /livesys/merge_union
2377 +
2378 +A merged view of these two branches is then available at
2379 +/livesys/merge_union, and the new feature is that the whiteouts are
2380 +visible!
2381 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2382 +writing to all branches. Also the default mode for all branches is 'ro'.
2383 +It is now possible to save the combined contents of the two overlay
2384 +branches to a new squashfs, e.g.:
2385 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2386 +
2387 +This new squashfs archive can be stored on the boot device and the
2388 +initramfs will use it to replace the old one at the next boot.
2389 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2390 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt  1970-01-01 01:00:00.000000000 +0100
2391 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt     2022-11-05 23:02:18.959222617 +0100
2392 @@ -0,0 +1,47 @@
2393 +
2394 +# Copyright (C) 2010-2022 Junjiro R. Okajima
2395 +#
2396 +# This program is free software; you can redistribute it and/or modify
2397 +# it under the terms of the GNU General Public License as published by
2398 +# the Free Software Foundation; either version 2 of the License, or
2399 +# (at your option) any later version.
2400 +#
2401 +# This program is distributed in the hope that it will be useful,
2402 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2403 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2404 +# GNU General Public License for more details.
2405 +#
2406 +# You should have received a copy of the GNU General Public License
2407 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2408 +
2409 +Dynamically customizable FS operations
2410 +----------------------------------------------------------------------
2411 +Generally FS operations (struct inode_operations, struct
2412 +address_space_operations, struct file_operations, etc.) are defined as
2413 +"static const", but it never means that FS have only one set of
2414 +operation. Some FS have multiple sets of them. For instance, ext2 has
2415 +three sets, one for XIP, for NOBH, and for normal.
2416 +Since aufs overrides and redirects these operations, sometimes aufs has
2417 +to change its behaviour according to the branch FS type. More importantly
2418 +VFS acts differently if a function (member in the struct) is set or
2419 +not. It means aufs should have several sets of operations and select one
2420 +among them according to the branch FS definition.
2421 +
2422 +In order to solve this problem and not to affect the behaviour of VFS,
2423 +aufs defines these operations dynamically. For instance, aufs defines
2424 +dummy direct_IO function for struct address_space_operations, but it may
2425 +not be set to the address_space_operations actually. When the branch FS
2426 +doesn't have it, aufs doesn't set it to its address_space_operations
2427 +while the function definition itself is still alive. So the behaviour
2428 +itself will not change, and it will return an error when direct_IO is
2429 +not set.
2430 +
2431 +The lifetime of these dynamically generated operation object is
2432 +maintained by aufs branch object. When the branch is removed from aufs,
2433 +the reference counter of the object is decremented. When it reaches
2434 +zero, the dynamically generated operation object will be freed.
2435 +
2436 +This approach is designed to support AIO (io_submit), Direct I/O and
2437 +XIP (DAX) mainly.
2438 +Currently this approach is applied to address_space_operations for
2439 +regular files only.
2440 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2441 --- /usr/share/empty/Documentation/filesystems/aufs/README      1970-01-01 01:00:00.000000000 +0100
2442 +++ linux/Documentation/filesystems/aufs/README 2023-02-20 21:05:51.959693785 +0100
2443 @@ -0,0 +1,408 @@
2444 +
2445 +Aufs6 -- advanced multi layered unification filesystem version 6.x
2446 +http://aufs.sf.net
2447 +Junjiro R. Okajima
2448 +
2449 +
2450 +0. Introduction
2451 +----------------------------------------
2452 +In the early days, aufs was entirely re-designed and re-implemented
2453 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2454 +improvements and implementations, it became totally different from
2455 +Unionfs while keeping the basic features.
2456 +Later, Unionfs Version 2.x series began taking some of the same
2457 +approaches to aufs1's.
2458 +Unionfs was being developed by Professor Erez Zadok at Stony Brook
2459 +University and his team.
2460 +
2461 +Aufs6 supports linux-v6.0 and later, try aufs6.0 branch in
2462 +aufs-linux.git or aufs-standalone.git.
2463 +If you want older kernel version support,
2464 +- for linux-v5.x series, try aufs-linux.git or aufs-standalone.git
2465 +- for linux-v4.x series, try aufs4-linux.git or aufs4-standalone.git
2466 +- for linux-v3.x series, try aufs3-linux.git or aufs3-standalone.git
2467 +- for linux-v2.6.16 and later, try aufs2-2.6.git, aufs2-standalone.git
2468 +  or aufs1 from CVS on SourceForge.
2469 +
2470 +Note: the name of aufs5-linux.git and aufs5-standalone.git on github
2471 +      were changed. Now they are aufs-linux.git and
2472 +      aufs-standalone.git and they contain aufs5 and later branches.
2473 +
2474 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2475 +      According to Christoph Hellwig, linux rejects all union-type
2476 +      filesystems but UnionMount.
2477 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2478 +
2479 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2480 +    UnionMount, and he pointed out an issue around a directory mutex
2481 +    lock and aufs addressed it. But it is still unsure whether aufs will
2482 +    be merged (or any other union solution).
2483 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2484 +
2485 +
2486 +1. Features
2487 +----------------------------------------
2488 +- unite several directories into a single virtual filesystem. The member
2489 +  directory is called as a branch.
2490 +- you can specify the permission flags to the branch, which are 'readonly',
2491 +  'readwrite' and 'whiteout-able.'
2492 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2493 +  readonly branch are modifiable logically.
2494 +- dynamic branch manipulation, add, del.
2495 +- etc...
2496 +
2497 +Also there are many enhancements in aufs, such as:
2498 +- test only the highest one for the directory permission (dirperm1)
2499 +- copyup on open (coo=)
2500 +- 'move' policy for copy-up between two writable branches, after
2501 +  checking free space.
2502 +- xattr, acl
2503 +- readdir(3) in userspace.
2504 +- keep inode number by external inode number table
2505 +- keep the timestamps of file/dir in internal copyup operation
2506 +- seekable directory, supporting NFS readdir.
2507 +- whiteout is hardlinked in order to reduce the consumption of inodes
2508 +  on branch
2509 +- do not copyup, nor create a whiteout when it is unnecessary
2510 +- revert a single systemcall when an error occurs in aufs
2511 +- remount interface instead of ioctl
2512 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2513 +- loopback mounted filesystem as a branch
2514 +- kernel thread for removing the dir who has a plenty of whiteouts
2515 +- support copyup sparse file (a file which has a 'hole' in it)
2516 +- default permission flags for branches
2517 +- selectable permission flags for ro branch, whether whiteout can
2518 +  exist or not
2519 +- export via NFS.
2520 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2521 +- support multiple writable branches, some policies to select one
2522 +  among multiple writable branches.
2523 +- a new semantics for link(2) and rename(2) to support multiple
2524 +  writable branches.
2525 +- no glibc changes are required.
2526 +- pseudo hardlink (hardlink over branches)
2527 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2528 +  including NFS or remote filesystem branch.
2529 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2530 +- and more...
2531 +
2532 +Currently these features are dropped temporary from aufs6.
2533 +See design/08plan.txt in detail.
2534 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2535 +  (robr)
2536 +- statistics of aufs thread (/sys/fs/aufs/stat)
2537 +
2538 +Features or just an idea in the future (see also design/*.txt),
2539 +- reorder the branch index without del/re-add.
2540 +- permanent xino files for NFSD
2541 +- an option for refreshing the opened files after add/del branches
2542 +- light version, without branch manipulation. (unnecessary?)
2543 +- copyup in userspace
2544 +- inotify in userspace
2545 +- readv/writev
2546 +
2547 +
2548 +2. Download
2549 +----------------------------------------
2550 +There are three GIT trees for aufs6, aufs-linux.git,
2551 +aufs-standalone.git, and aufs-util.git.
2552 +While the aufs-util is always necessary, you need either of aufs-linux
2553 +or aufs-standalone.
2554 +
2555 +The aufs-linux tree includes the whole linux mainline GIT tree,
2556 +git://git.kernel.org/.../torvalds/linux.git.
2557 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2558 +build aufs6 as an external kernel module.
2559 +Several extra patches are not included in this tree. Only
2560 +aufs-standalone tree contains them. They are described in the later
2561 +section "Configuration and Compilation."
2562 +
2563 +On the other hand, the aufs-standalone tree has only aufs source files
2564 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2565 +But you need to apply all aufs patches manually.
2566 +
2567 +You will find GIT branches whose name is in form of "aufs6.x" where "x"
2568 +represents the linux kernel version, "linux-6.x". For instance,
2569 +"aufs6.0" is for linux-6.0. For latest "linux-6.x-rcN", use
2570 +"aufs6.x-rcN" branch.
2571 +
2572 +o aufs-linux tree
2573 +$ git clone --reference /your/linux/git/tree \
2574 +       git://github.com/sfjro/aufs-linux.git aufs-linux.git
2575 +- if you don't have linux GIT tree, then remove "--reference ..."
2576 +$ cd aufs-linux.git
2577 +$ git checkout origin/aufs6.0
2578 +
2579 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2580 +leave the patch-work to GIT.
2581 +$ cd /your/linux/git/tree
2582 +$ git remote add aufs git://github.com/sfjro/aufs-linux.git
2583 +$ git fetch aufs
2584 +$ git checkout -b my6.0 v6.0
2585 +$ (add your local change...)
2586 +$ git pull aufs aufs6.0
2587 +- now you have v6.0 + your_changes + aufs6.0 in you my6.0 branch.
2588 +- you may need to solve some conflicts between your_changes and
2589 +  aufs6.0. in this case, git-rerere is recommended so that you can
2590 +  solve the similar conflicts automatically when you upgrade to 6.1 or
2591 +  later in the future.
2592 +
2593 +o aufs-standalone tree
2594 +$ git clone git://github.com/sfjro/aufs-standalone.git aufs-standalone.git
2595 +$ cd aufs-standalone.git
2596 +$ git checkout origin/aufs6.0
2597 +
2598 +o aufs-util tree
2599 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2600 +- note that the public aufs-util.git is on SourceForge instead of
2601 +  GitHUB.
2602 +$ cd aufs-util.git
2603 +$ git checkout origin/aufs6.0
2604 +
2605 +Note: The 6.x-rcN branch is to be used with `rc' kernel versions ONLY.
2606 +The minor version number, 'x' in '6.x', of aufs may not always
2607 +follow the minor version number of the kernel.
2608 +Because changes in the kernel that cause the use of a new
2609 +minor version number do not always require changes to aufs-util.
2610 +
2611 +Since aufs-util has its own minor version number, you may not be
2612 +able to find a GIT branch in aufs-util for your kernel's
2613 +exact minor version number.
2614 +In this case, you should git-checkout the branch for the
2615 +nearest lower number.
2616 +
2617 +For (an unreleased) example:
2618 +If you are using "linux-6.10" and the "aufs6.10" branch
2619 +does not exist in aufs-util repository, then "aufs6.9", "aufs6.8"
2620 +or something numerically smaller is the branch for your kernel.
2621 +
2622 +Also you can view all branches by
2623 +       $ git branch -a
2624 +
2625 +
2626 +3. Configuration and Compilation
2627 +----------------------------------------
2628 +Make sure you have git-checkout'ed the correct branch.
2629 +
2630 +For aufs-linux tree,
2631 +- enable CONFIG_AUFS_FS.
2632 +- set other aufs configurations if necessary.
2633 +- for aufs5.13 and later
2634 +  Because aufs is not only an ordinary filesystem (callee of VFS), but
2635 +  also a caller of VFS functions for branch filesystems, subclassing of
2636 +  the internal locks for LOCKDEP is necessary. LOCKDEP is a debugging
2637 +  feature of linux kernel. If you enable CONFIG_LOCKDEP, then you will
2638 +  need to customize some LOCKDEP numbers. Here are what I use on my
2639 +  test environment.
2640 +       CONFIG_LOCKDEP_BITS=21
2641 +       CONFIG_LOCKDEP_CHAINS_BITS=21
2642 +       CONFIG_LOCKDEP_STACK_TRACE_BITS=24
2643 +  Also you will need to expand some constant values in LOCKDEP. Refer
2644 +  to lockdep-debug.patch in aufs-standalone.git.
2645 +
2646 +For aufs-standalone tree,
2647 +There are several ways to build.
2648 +
2649 +1.
2650 +- apply ./aufs6-kbuild.patch to your kernel source files.
2651 +- apply ./aufs6-base.patch too.
2652 +- apply ./aufs6-mmap.patch too.
2653 +- apply ./aufs6-standalone.patch too, if you have a plan to set
2654 +  CONFIG_AUFS_FS=m. otherwise you don't need ./aufs-standalone.patch.
2655 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2656 +  kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2657 +- enable CONFIG_AUFS_FS, you can select either
2658 +  =m or =y.
2659 +- and build your kernel as usual.
2660 +- install the built kernel.
2661 +- install the header files too by "make headers_install" to the
2662 +  directory where you specify. By default, it is $PWD/usr.
2663 +  "make help" shows a brief note for headers_install.
2664 +- and reboot your system.
2665 +
2666 +2.
2667 +- module only (CONFIG_AUFS_FS=m).
2668 +- apply ./aufs6-base.patch to your kernel source files.
2669 +- apply ./aufs6-mmap.patch too.
2670 +- apply ./aufs6-standalone.patch too.
2671 +- build your kernel, don't forget "make headers_install", and reboot.
2672 +- edit ./config.mk and set other aufs configurations if necessary.
2673 +  Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2674 +  every aufs configurations.
2675 +- build the module by simple "make".
2676 +- you can specify ${KDIR} make variable which points to your kernel
2677 +  source tree.
2678 +- install the files
2679 +  + run "make install" to install the aufs module, or copy the built
2680 +    $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2681 +  + run "make install_headers" (instead of headers_install) to install
2682 +    the modified aufs header file (you can specify DESTDIR which is
2683 +    available in aufs standalone version's Makefile only), or copy
2684 +    $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2685 +    you like manually. By default, the target directory is $PWD/usr.
2686 +- no need to apply aufs6-kbuild.patch, nor copying source files to your
2687 +  kernel source tree.
2688 +
2689 +Note: The header file aufs_type.h is necessary to build aufs-util
2690 +      as well as "make headers_install" in the kernel source tree.
2691 +      headers_install is subject to be forgotten, but it is essentially
2692 +      necessary, not only for building aufs-util.
2693 +      You may not meet problems without headers_install in some older
2694 +      version though.
2695 +
2696 +And then,
2697 +- read README in aufs-util, build and install it
2698 +- note that your distribution may contain an obsoleted version of
2699 +  aufs_type.h in /usr/include/linux or something. When you build aufs
2700 +  utilities, make sure that your compiler refers the correct aufs header
2701 +  file which is built by "make headers_install."
2702 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2703 +  then run "make install_ulib" too. And refer to the aufs manual in
2704 +  detail.
2705 +
2706 +There several other patches in aufs-standalone.git. They are all
2707 +optional. When you meet some problems, they will help you.
2708 +- aufs6-loopback.patch
2709 +  Supports a nested loopback mount in a branch-fs. This patch is
2710 +  unnecessary until aufs produces a message like "you may want to try
2711 +  another patch for loopback file".
2712 +- vfs-ino.patch
2713 +  Modifies a system global kernel internal function get_next_ino() in
2714 +  order to stop assigning 0 for an inode-number. Not directly related to
2715 +  aufs, but recommended generally.
2716 +- tmpfs-idr.patch
2717 +  Keeps the tmpfs inode number as the lowest value. Effective to reduce
2718 +  the size of aufs XINO files for tmpfs branch. Also it prevents the
2719 +  duplication of inode number, which is important for backup tools and
2720 +  other utilities. When you find aufs XINO files for tmpfs branch
2721 +  growing too much, try this patch.
2722 +- lockdep-debug.patch
2723 +  Similar to some kernel configurations for LOCKDEP (see the top of
2724 +  this section), you will need expand some constants in LOCKDEP for
2725 +  aufs if you enable CONFIG_LOCKDEP.
2726 +
2727 +
2728 +4. Usage
2729 +----------------------------------------
2730 +At first, make sure aufs-util are installed, and please read the aufs
2731 +manual, aufs.5 in aufs-util.git tree.
2732 +$ man -l aufs.5
2733 +
2734 +And then,
2735 +$ mkdir /tmp/rw /tmp/aufs
2736 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2737 +
2738 +Here is another example. The result is equivalent.
2739 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2740 +  Or
2741 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2742 +# mount -o remount,append:${HOME} /tmp/aufs
2743 +
2744 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2745 +you modify a file under /tmp/aufs, the one on your home directory is
2746 +not affected, instead the same named file will be newly created under
2747 +/tmp/rw. And all of your modification to a file will be applied to
2748 +the one under /tmp/rw. This is called the file based Copy on Write
2749 +(COW) method.
2750 +Aufs mount options are described in aufs.5.
2751 +If you run chroot or something and make your aufs as a root directory,
2752 +then you need to customize the shutdown script. See the aufs manual in
2753 +detail.
2754 +
2755 +Additionally, there are some sample usages of aufs which are a
2756 +diskless system with network booting, and LiveCD over NFS.
2757 +See sample dir in CVS tree on SourceForge.
2758 +
2759 +
2760 +5. Contact
2761 +----------------------------------------
2762 +When you have any problems or strange behaviour in aufs, please let me
2763 +know with:
2764 +- /proc/mounts (instead of the output of mount(8))
2765 +- /sys/module/aufs/*
2766 +- /sys/fs/aufs/* (if you have them)
2767 +- /debug/aufs/* (if you have them)
2768 +- linux kernel version
2769 +  if your kernel is not plain, for example modified by distributor,
2770 +  the url where i can download its source is necessary too.
2771 +- aufs version which was printed at loading the module or booting the
2772 +  system, instead of the date you downloaded.
2773 +- configuration (define/undefine CONFIG_AUFS_xxx)
2774 +- kernel configuration or /proc/config.gz (if you have it)
2775 +- LSM (linux security module, if you are using)
2776 +- behaviour which you think to be incorrect
2777 +- actual operation, reproducible one is better
2778 +- mailto: aufs-users at lists.sourceforge.net
2779 +
2780 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2781 +and Feature Requests) on SourceForge. Please join and write to
2782 +aufs-users ML.
2783 +
2784 +
2785 +6. Acknowledgements
2786 +----------------------------------------
2787 +Thanks to everyone who have tried and are using aufs, whoever
2788 +have reported a bug or any feedback.
2789 +
2790 +Especially donators:
2791 +Tomas Matejicek(slax.org) made a donation (much more than once).
2792 +       Since Apr 2010, Tomas M (the author of Slax and Linux Live
2793 +       scripts) is making "doubling" donations.
2794 +       Unfortunately I cannot list all of the donators, but I really
2795 +       appreciate.
2796 +       It ends Aug 2010, but the ordinary donation URL is still available.
2797 +       <http://sourceforge.net/donate/index.php?group_id=167503>
2798 +Dai Itasaka made a donation (2007/8).
2799 +Chuck Smith made a donation (2008/4, 10 and 12).
2800 +Henk Schoneveld made a donation (2008/9).
2801 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2802 +Francois Dupoux made a donation (2008/11).
2803 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2804 +       aufs2 GIT tree (2009/2).
2805 +William Grant made a donation (2009/3).
2806 +Patrick Lane made a donation (2009/4).
2807 +The Mail Archive (mail-archive.com) made donations (2009/5).
2808 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2809 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2810 +Pavel Pronskiy made a donation (2011/2).
2811 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2812 +       Networks (Ed Wildgoose) made a donation for hardware (2011/3).
2813 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2814 +11).
2815 +Sam Liddicott made a donation (2011/9).
2816 +Era Scarecrow made a donation (2013/4).
2817 +Bor Ratajc made a donation (2013/4).
2818 +Alessandro Gorreta made a donation (2013/4).
2819 +POIRETTE Marc made a donation (2013/4).
2820 +Alessandro Gorreta made a donation (2013/4).
2821 +lauri kasvandik made a donation (2013/5).
2822 +"pemasu from Finland" made a donation (2013/7).
2823 +The Parted Magic Project made a donation (2013/9 and 11).
2824 +Pavel Barta made a donation (2013/10).
2825 +Nikolay Pertsev made a donation (2014/5).
2826 +James B made a donation (2014/7, 2015/7, and 2021/12).
2827 +Stefano Di Biase made a donation (2014/8).
2828 +Daniel Epellei made a donation (2015/1).
2829 +OmegaPhil made a donation (2016/1, 2018/4).
2830 +Tomasz Szewczyk made a donation (2016/4).
2831 +James Burry made a donation (2016/12).
2832 +Carsten Rose made a donation (2018/9).
2833 +Porteus Kiosk made a donation (2018/10).
2834 +Enya Quetzalli Gomez Rodriguez made a donation (2022/5).
2835 +
2836 +Thank you very much.
2837 +Donations are always, including future donations, very important and
2838 +helpful for me to keep on developing aufs.
2839 +
2840 +
2841 +7.
2842 +----------------------------------------
2843 +If you are an experienced user, no explanation is needed. Aufs is
2844 +just a linux filesystem.
2845 +
2846 +
2847 +Enjoy!
2848 +
2849 +# Local variables: ;
2850 +# mode: text;
2851 +# End: ;
2852 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2853 --- /usr/share/empty/fs/aufs/aufs.h     1970-01-01 01:00:00.000000000 +0100
2854 +++ linux/fs/aufs/aufs.h        2022-11-05 23:02:18.959222617 +0100
2855 @@ -0,0 +1,62 @@
2856 +/* SPDX-License-Identifier: GPL-2.0 */
2857 +/*
2858 + * Copyright (C) 2005-2022 Junjiro R. Okajima
2859 + *
2860 + * This program is free software; you can redistribute it and/or modify
2861 + * it under the terms of the GNU General Public License as published by
2862 + * the Free Software Foundation; either version 2 of the License, or
2863 + * (at your option) any later version.
2864 + *
2865 + * This program is distributed in the hope that it will be useful,
2866 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2867 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2868 + * GNU General Public License for more details.
2869 + *
2870 + * You should have received a copy of the GNU General Public License
2871 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
2872 + */
2873 +
2874 +/*
2875 + * all header files
2876 + */
2877 +
2878 +#ifndef __AUFS_H__
2879 +#define __AUFS_H__
2880 +
2881 +#ifdef __KERNEL__
2882 +
2883 +#define AuStub(type, name, body, ...) \
2884 +       static inline type name(__VA_ARGS__) { body; }
2885 +
2886 +#define AuStubVoid(name, ...) \
2887 +       AuStub(void, name, , __VA_ARGS__)
2888 +#define AuStubInt0(name, ...) \
2889 +       AuStub(int, name, return 0, __VA_ARGS__)
2890 +
2891 +#include "debug.h"
2892 +
2893 +#include "branch.h"
2894 +#include "cpup.h"
2895 +#include "dcsub.h"
2896 +#include "dbgaufs.h"
2897 +#include "dentry.h"
2898 +#include "dir.h"
2899 +#include "dirren.h"
2900 +#include "dynop.h"
2901 +#include "file.h"
2902 +#include "fstype.h"
2903 +#include "hbl.h"
2904 +#include "inode.h"
2905 +#include "lcnt.h"
2906 +#include "loop.h"
2907 +#include "module.h"
2908 +#include "opts.h"
2909 +#include "rwsem.h"
2910 +#include "super.h"
2911 +#include "sysaufs.h"
2912 +#include "vfsub.h"
2913 +#include "whout.h"
2914 +#include "wkq.h"
2915 +
2916 +#endif /* __KERNEL__ */
2917 +#endif /* __AUFS_H__ */
2918 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2919 --- /usr/share/empty/fs/aufs/branch.c   1970-01-01 01:00:00.000000000 +0100
2920 +++ linux/fs/aufs/branch.c      2022-11-05 23:02:18.959222617 +0100
2921 @@ -0,0 +1,1427 @@
2922 +// SPDX-License-Identifier: GPL-2.0
2923 +/*
2924 + * Copyright (C) 2005-2022 Junjiro R. Okajima
2925 + *
2926 + * This program is free software; you can redistribute it and/or modify
2927 + * it under the terms of the GNU General Public License as published by
2928 + * the Free Software Foundation; either version 2 of the License, or
2929 + * (at your option) any later version.
2930 + *
2931 + * This program is distributed in the hope that it will be useful,
2932 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2933 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2934 + * GNU General Public License for more details.
2935 + *
2936 + * You should have received a copy of the GNU General Public License
2937 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
2938 + */
2939 +
2940 +/*
2941 + * branch management
2942 + */
2943 +
2944 +#include <linux/compat.h>
2945 +#include <linux/statfs.h>
2946 +#include "aufs.h"
2947 +
2948 +/*
2949 + * free a single branch
2950 + */
2951 +static void au_br_do_free(struct au_branch *br)
2952 +{
2953 +       int i;
2954 +       struct au_wbr *wbr;
2955 +       struct au_dykey **key;
2956 +
2957 +       au_hnotify_fin_br(br);
2958 +       /* always, regardless the mount option */
2959 +       au_dr_hino_free(&br->br_dirren);
2960 +       au_xino_put(br);
2961 +
2962 +       AuLCntZero(au_lcnt_read(&br->br_nfiles, /*do_rev*/0));
2963 +       au_lcnt_fin(&br->br_nfiles, /*do_sync*/0);
2964 +       AuLCntZero(au_lcnt_read(&br->br_count, /*do_rev*/0));
2965 +       au_lcnt_fin(&br->br_count, /*do_sync*/0);
2966 +
2967 +       wbr = br->br_wbr;
2968 +       if (wbr) {
2969 +               for (i = 0; i < AuBrWh_Last; i++)
2970 +                       dput(wbr->wbr_wh[i]);
2971 +               AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2972 +               AuRwDestroy(&wbr->wbr_wh_rwsem);
2973 +       }
2974 +
2975 +       if (br->br_fhsm) {
2976 +               au_br_fhsm_fin(br->br_fhsm);
2977 +               au_kfree_try_rcu(br->br_fhsm);
2978 +       }
2979 +
2980 +       key = br->br_dykey;
2981 +       for (i = 0; i < AuBrDynOp; i++, key++)
2982 +               if (*key)
2983 +                       au_dy_put(*key);
2984 +               else
2985 +                       break;
2986 +
2987 +       /* recursive lock, s_umount of branch's */
2988 +       /* synchronize_rcu(); */ /* why? */
2989 +       lockdep_off();
2990 +       path_put(&br->br_path);
2991 +       lockdep_on();
2992 +       au_kfree_rcu(wbr);
2993 +       au_lcnt_wait_for_fin(&br->br_nfiles);
2994 +       au_lcnt_wait_for_fin(&br->br_count);
2995 +       /* I don't know why, but percpu_refcount requires this */
2996 +       /* synchronize_rcu(); */
2997 +       au_kfree_rcu(br);
2998 +}
2999 +
3000 +/*
3001 + * frees all branches
3002 + */
3003 +void au_br_free(struct au_sbinfo *sbinfo)
3004 +{
3005 +       aufs_bindex_t bmax;
3006 +       struct au_branch **br;
3007 +
3008 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
3009 +
3010 +       bmax = sbinfo->si_bbot + 1;
3011 +       br = sbinfo->si_branch;
3012 +       while (bmax--)
3013 +               au_br_do_free(*br++);
3014 +}
3015 +
3016 +/*
3017 + * find the index of a branch which is specified by @br_id.
3018 + */
3019 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
3020 +{
3021 +       aufs_bindex_t bindex, bbot;
3022 +
3023 +       bbot = au_sbbot(sb);
3024 +       for (bindex = 0; bindex <= bbot; bindex++)
3025 +               if (au_sbr_id(sb, bindex) == br_id)
3026 +                       return bindex;
3027 +       return -1;
3028 +}
3029 +
3030 +/* ---------------------------------------------------------------------- */
3031 +
3032 +/*
3033 + * add a branch
3034 + */
3035 +
3036 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
3037 +                       struct dentry *h_root)
3038 +{
3039 +       if (unlikely(h_adding == h_root
3040 +                    || au_test_loopback_overlap(sb, h_adding)))
3041 +               return 1;
3042 +       if (h_adding->d_sb != h_root->d_sb)
3043 +               return 0;
3044 +       return au_test_subdir(h_adding, h_root)
3045 +               || au_test_subdir(h_root, h_adding);
3046 +}
3047 +
3048 +/*
3049 + * returns a newly allocated branch. @new_nbranch is a number of branches
3050 + * after adding a branch.
3051 + */
3052 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
3053 +                                    int perm)
3054 +{
3055 +       struct au_branch *add_branch;
3056 +       struct dentry *root;
3057 +       struct inode *inode;
3058 +       int err;
3059 +
3060 +       err = -ENOMEM;
3061 +       add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
3062 +       if (unlikely(!add_branch))
3063 +               goto out;
3064 +       add_branch->br_xino = au_xino_alloc(/*nfile*/1);
3065 +       if (unlikely(!add_branch->br_xino))
3066 +               goto out_br;
3067 +       err = au_hnotify_init_br(add_branch, perm);
3068 +       if (unlikely(err))
3069 +               goto out_xino;
3070 +
3071 +       if (au_br_writable(perm)) {
3072 +               /* may be freed separately at changing the branch permission */
3073 +               add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
3074 +                                            GFP_NOFS);
3075 +               if (unlikely(!add_branch->br_wbr))
3076 +                       goto out_hnotify;
3077 +       }
3078 +
3079 +       if (au_br_fhsm(perm)) {
3080 +               err = au_fhsm_br_alloc(add_branch);
3081 +               if (unlikely(err))
3082 +                       goto out_wbr;
3083 +       }
3084 +
3085 +       root = sb->s_root;
3086 +       err = au_sbr_realloc(au_sbi(sb), new_nbranch, /*may_shrink*/0);
3087 +       if (!err)
3088 +               err = au_di_realloc(au_di(root), new_nbranch, /*may_shrink*/0);
3089 +       if (!err) {
3090 +               inode = d_inode(root);
3091 +               err = au_hinode_realloc(au_ii(inode), new_nbranch,
3092 +                                       /*may_shrink*/0);
3093 +       }
3094 +       if (!err)
3095 +               return add_branch; /* success */
3096 +
3097 +out_wbr:
3098 +       au_kfree_rcu(add_branch->br_wbr);
3099 +out_hnotify:
3100 +       au_hnotify_fin_br(add_branch);
3101 +out_xino:
3102 +       au_xino_put(add_branch);
3103 +out_br:
3104 +       au_kfree_rcu(add_branch);
3105 +out:
3106 +       return ERR_PTR(err);
3107 +}
3108 +
3109 +/*
3110 + * test if the branch permission is legal or not.
3111 + */
3112 +static int test_br(struct inode *inode, int brperm, char *path)
3113 +{
3114 +       int err;
3115 +
3116 +       err = (au_br_writable(brperm) && IS_RDONLY(inode));
3117 +       if (!err)
3118 +               goto out;
3119 +
3120 +       err = -EINVAL;
3121 +       pr_err("write permission for readonly mount or inode, %s\n", path);
3122 +
3123 +out:
3124 +       return err;
3125 +}
3126 +
3127 +/*
3128 + * returns:
3129 + * 0: success, the caller will add it
3130 + * plus: success, it is already unified, the caller should ignore it
3131 + * minus: error
3132 + */
3133 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
3134 +{
3135 +       int err;
3136 +       aufs_bindex_t bbot, bindex;
3137 +       struct dentry *root, *h_dentry;
3138 +       struct inode *inode, *h_inode;
3139 +
3140 +       root = sb->s_root;
3141 +       bbot = au_sbbot(sb);
3142 +       if (unlikely(bbot >= 0
3143 +                    && au_find_dbindex(root, add->path.dentry) >= 0)) {
3144 +               err = 1;
3145 +               if (!remount) {
3146 +                       err = -EINVAL;
3147 +                       pr_err("%s duplicated\n", add->pathname);
3148 +               }
3149 +               goto out;
3150 +       }
3151 +
3152 +       err = -ENOSPC; /* -E2BIG; */
3153 +       if (unlikely(AUFS_BRANCH_MAX <= add->bindex
3154 +                    || AUFS_BRANCH_MAX - 1 <= bbot)) {
3155 +               pr_err("number of branches exceeded %s\n", add->pathname);
3156 +               goto out;
3157 +       }
3158 +
3159 +       err = -EDOM;
3160 +       if (unlikely(add->bindex < 0 || bbot + 1 < add->bindex)) {
3161 +               pr_err("bad index %d\n", add->bindex);
3162 +               goto out;
3163 +       }
3164 +
3165 +       inode = d_inode(add->path.dentry);
3166 +       err = -ENOENT;
3167 +       if (unlikely(!inode->i_nlink)) {
3168 +               pr_err("no existence %s\n", add->pathname);
3169 +               goto out;
3170 +       }
3171 +
3172 +       err = -EINVAL;
3173 +       if (unlikely(inode->i_sb == sb)) {
3174 +               pr_err("%s must be outside\n", add->pathname);
3175 +               goto out;
3176 +       }
3177 +
3178 +       if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
3179 +               pr_err("unsupported filesystem, %s (%s)\n",
3180 +                      add->pathname, au_sbtype(inode->i_sb));
3181 +               goto out;
3182 +       }
3183 +
3184 +       if (unlikely(inode->i_sb->s_stack_depth)) {
3185 +               pr_err("already stacked, %s (%s)\n",
3186 +                      add->pathname, au_sbtype(inode->i_sb));
3187 +               goto out;
3188 +       }
3189 +
3190 +       err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3191 +       if (unlikely(err))
3192 +               goto out;
3193 +
3194 +       if (bbot < 0)
3195 +               return 0; /* success */
3196 +
3197 +       err = -EINVAL;
3198 +       for (bindex = 0; bindex <= bbot; bindex++)
3199 +               if (unlikely(test_overlap(sb, add->path.dentry,
3200 +                                         au_h_dptr(root, bindex)))) {
3201 +                       pr_err("%s is overlapped\n", add->pathname);
3202 +                       goto out;
3203 +               }
3204 +
3205 +       err = 0;
3206 +       if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3207 +               h_dentry = au_h_dptr(root, 0);
3208 +               h_inode = d_inode(h_dentry);
3209 +               if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3210 +                   || !uid_eq(h_inode->i_uid, inode->i_uid)
3211 +                   || !gid_eq(h_inode->i_gid, inode->i_gid))
3212 +                       pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3213 +                               add->pathname,
3214 +                               i_uid_read(inode), i_gid_read(inode),
3215 +                               (inode->i_mode & S_IALLUGO),
3216 +                               i_uid_read(h_inode), i_gid_read(h_inode),
3217 +                               (h_inode->i_mode & S_IALLUGO));
3218 +       }
3219 +
3220 +out:
3221 +       return err;
3222 +}
3223 +
3224 +/*
3225 + * initialize or clean the whiteouts for an adding branch
3226 + */
3227 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3228 +                        int new_perm)
3229 +{
3230 +       int err, old_perm;
3231 +       aufs_bindex_t bindex;
3232 +       struct inode *h_inode;
3233 +       struct au_wbr *wbr;
3234 +       struct au_hinode *hdir;
3235 +       struct dentry *h_dentry;
3236 +
3237 +       err = vfsub_mnt_want_write(au_br_mnt(br));
3238 +       if (unlikely(err))
3239 +               goto out;
3240 +
3241 +       wbr = br->br_wbr;
3242 +       old_perm = br->br_perm;
3243 +       br->br_perm = new_perm;
3244 +       hdir = NULL;
3245 +       h_inode = NULL;
3246 +       bindex = au_br_index(sb, br->br_id);
3247 +       if (0 <= bindex) {
3248 +               hdir = au_hi(d_inode(sb->s_root), bindex);
3249 +               au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
3250 +       } else {
3251 +               h_dentry = au_br_dentry(br);
3252 +               h_inode = d_inode(h_dentry);
3253 +               inode_lock_nested(h_inode, AuLsc_I_PARENT);
3254 +       }
3255 +       if (!wbr)
3256 +               err = au_wh_init(br, sb);
3257 +       else {
3258 +               wbr_wh_write_lock(wbr);
3259 +               err = au_wh_init(br, sb);
3260 +               wbr_wh_write_unlock(wbr);
3261 +       }
3262 +       if (hdir)
3263 +               au_hn_inode_unlock(hdir);
3264 +       else
3265 +               inode_unlock(h_inode);
3266 +       vfsub_mnt_drop_write(au_br_mnt(br));
3267 +       br->br_perm = old_perm;
3268 +
3269 +       if (!err && wbr && !au_br_writable(new_perm)) {
3270 +               au_kfree_rcu(wbr);
3271 +               br->br_wbr = NULL;
3272 +       }
3273 +
3274 +out:
3275 +       return err;
3276 +}
3277 +
3278 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3279 +                      int perm)
3280 +{
3281 +       int err;
3282 +       struct kstatfs kst;
3283 +       struct au_wbr *wbr;
3284 +
3285 +       wbr = br->br_wbr;
3286 +       au_rw_init(&wbr->wbr_wh_rwsem);
3287 +       atomic_set(&wbr->wbr_wh_running, 0);
3288 +
3289 +       /*
3290 +        * a limit for rmdir/rename a dir
3291 +        * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3292 +        */
3293 +       err = vfs_statfs(&br->br_path, &kst);
3294 +       if (unlikely(err))
3295 +               goto out;
3296 +       err = -EINVAL;
3297 +       if (kst.f_namelen >= NAME_MAX)
3298 +               err = au_br_init_wh(sb, br, perm);
3299 +       else
3300 +               pr_err("%pd(%s), unsupported namelen %ld\n",
3301 +                      au_br_dentry(br),
3302 +                      au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3303 +
3304 +out:
3305 +       return err;
3306 +}
3307 +
3308 +/* initialize a new branch */
3309 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3310 +                     struct au_opt_add *add)
3311 +{
3312 +       int err;
3313 +       struct au_branch *brbase;
3314 +       struct file *xf;
3315 +       struct inode *h_inode;
3316 +
3317 +       err = 0;
3318 +       br->br_perm = add->perm;
3319 +       br->br_path = add->path; /* set first, path_get() later */
3320 +       spin_lock_init(&br->br_dykey_lock);
3321 +       au_lcnt_init(&br->br_nfiles, /*release*/NULL);
3322 +       au_lcnt_init(&br->br_count, /*release*/NULL);
3323 +       br->br_id = au_new_br_id(sb);
3324 +       AuDebugOn(br->br_id < 0);
3325 +
3326 +       /* always, regardless the given option */
3327 +       err = au_dr_br_init(sb, br, &add->path);
3328 +       if (unlikely(err))
3329 +               goto out_err;
3330 +
3331 +       if (au_br_writable(add->perm)) {
3332 +               err = au_wbr_init(br, sb, add->perm);
3333 +               if (unlikely(err))
3334 +                       goto out_err;
3335 +       }
3336 +
3337 +       if (au_opt_test(au_mntflags(sb), XINO)) {
3338 +               brbase = au_sbr(sb, 0);
3339 +               xf = au_xino_file(brbase->br_xino, /*idx*/-1);
3340 +               AuDebugOn(!xf);
3341 +               h_inode = d_inode(add->path.dentry);
3342 +               err = au_xino_init_br(sb, br, h_inode->i_ino, &xf->f_path);
3343 +               if (unlikely(err)) {
3344 +                       AuDebugOn(au_xino_file(br->br_xino, /*idx*/-1));
3345 +                       goto out_err;
3346 +               }
3347 +       }
3348 +
3349 +       sysaufs_br_init(br);
3350 +       path_get(&br->br_path);
3351 +       goto out; /* success */
3352 +
3353 +out_err:
3354 +       memset(&br->br_path, 0, sizeof(br->br_path));
3355 +out:
3356 +       return err;
3357 +}
3358 +
3359 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3360 +                            struct au_branch *br, aufs_bindex_t bbot,
3361 +                            aufs_bindex_t amount)
3362 +{
3363 +       struct au_branch **brp;
3364 +
3365 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
3366 +
3367 +       brp = sbinfo->si_branch + bindex;
3368 +       memmove(brp + 1, brp, sizeof(*brp) * amount);
3369 +       *brp = br;
3370 +       sbinfo->si_bbot++;
3371 +       if (unlikely(bbot < 0))
3372 +               sbinfo->si_bbot = 0;
3373 +}
3374 +
3375 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3376 +                            aufs_bindex_t bbot, aufs_bindex_t amount)
3377 +{
3378 +       struct au_hdentry *hdp;
3379 +
3380 +       AuRwMustWriteLock(&dinfo->di_rwsem);
3381 +
3382 +       hdp = au_hdentry(dinfo, bindex);
3383 +       memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3384 +       au_h_dentry_init(hdp);
3385 +       dinfo->di_bbot++;
3386 +       if (unlikely(bbot < 0))
3387 +               dinfo->di_btop = 0;
3388 +}
3389 +
3390 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3391 +                            aufs_bindex_t bbot, aufs_bindex_t amount)
3392 +{
3393 +       struct au_hinode *hip;
3394 +
3395 +       AuRwMustWriteLock(&iinfo->ii_rwsem);
3396 +
3397 +       hip = au_hinode(iinfo, bindex);
3398 +       memmove(hip + 1, hip, sizeof(*hip) * amount);
3399 +       au_hinode_init(hip);
3400 +       iinfo->ii_bbot++;
3401 +       if (unlikely(bbot < 0))
3402 +               iinfo->ii_btop = 0;
3403 +}
3404 +
3405 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3406 +                        aufs_bindex_t bindex)
3407 +{
3408 +       struct dentry *root, *h_dentry;
3409 +       struct inode *root_inode, *h_inode;
3410 +       aufs_bindex_t bbot, amount;
3411 +
3412 +       root = sb->s_root;
3413 +       root_inode = d_inode(root);
3414 +       bbot = au_sbbot(sb);
3415 +       amount = bbot + 1 - bindex;
3416 +       h_dentry = au_br_dentry(br);
3417 +       au_sbilist_lock();
3418 +       au_br_do_add_brp(au_sbi(sb), bindex, br, bbot, amount);
3419 +       au_br_do_add_hdp(au_di(root), bindex, bbot, amount);
3420 +       au_br_do_add_hip(au_ii(root_inode), bindex, bbot, amount);
3421 +       au_set_h_dptr(root, bindex, dget(h_dentry));
3422 +       h_inode = d_inode(h_dentry);
3423 +       au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3424 +       au_sbilist_unlock();
3425 +}
3426 +
3427 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3428 +{
3429 +       int err;
3430 +       aufs_bindex_t bbot, add_bindex;
3431 +       struct dentry *root, *h_dentry;
3432 +       struct inode *root_inode;
3433 +       struct au_branch *add_branch;
3434 +
3435 +       root = sb->s_root;
3436 +       root_inode = d_inode(root);
3437 +       IMustLock(root_inode);
3438 +       IiMustWriteLock(root_inode);
3439 +       err = test_add(sb, add, remount);
3440 +       if (unlikely(err < 0))
3441 +               goto out;
3442 +       if (err) {
3443 +               err = 0;
3444 +               goto out; /* success */
3445 +       }
3446 +
3447 +       bbot = au_sbbot(sb);
3448 +       add_branch = au_br_alloc(sb, bbot + 2, add->perm);
3449 +       err = PTR_ERR(add_branch);
3450 +       if (IS_ERR(add_branch))
3451 +               goto out;
3452 +
3453 +       err = au_br_init(add_branch, sb, add);
3454 +       if (unlikely(err)) {
3455 +               au_br_do_free(add_branch);
3456 +               goto out;
3457 +       }
3458 +
3459 +       add_bindex = add->bindex;
3460 +       sysaufs_brs_del(sb, add_bindex);        /* remove successors */
3461 +       au_br_do_add(sb, add_branch, add_bindex);
3462 +       sysaufs_brs_add(sb, add_bindex);        /* append successors */
3463 +       dbgaufs_brs_add(sb, add_bindex, /*topdown*/0);  /* rename successors */
3464 +
3465 +       h_dentry = add->path.dentry;
3466 +       if (!add_bindex) {
3467 +               au_cpup_attr_all(root_inode, /*force*/1);
3468 +               sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3469 +       } else
3470 +               au_add_nlink(root_inode, d_inode(h_dentry));
3471 +
3472 +out:
3473 +       return err;
3474 +}
3475 +
3476 +/* ---------------------------------------------------------------------- */
3477 +
3478 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3479 +                                      unsigned long long max __maybe_unused,
3480 +                                      void *arg)
3481 +{
3482 +       unsigned long long n;
3483 +       struct file **p, *f;
3484 +       struct hlist_bl_head *files;
3485 +       struct hlist_bl_node *pos;
3486 +       struct au_finfo *finfo;
3487 +
3488 +       n = 0;
3489 +       p = a;
3490 +       files = &au_sbi(sb)->si_files;
3491 +       hlist_bl_lock(files);
3492 +       hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
3493 +               f = finfo->fi_file;
3494 +               if (file_count(f)
3495 +                   && !special_file(file_inode(f)->i_mode)) {
3496 +                       get_file(f);
3497 +                       *p++ = f;
3498 +                       n++;
3499 +                       AuDebugOn(n > max);
3500 +               }
3501 +       }
3502 +       hlist_bl_unlock(files);
3503 +
3504 +       return n;
3505 +}
3506 +
3507 +static struct file **au_farray_alloc(struct super_block *sb,
3508 +                                    unsigned long long *max)
3509 +{
3510 +       struct au_sbinfo *sbi;
3511 +
3512 +       sbi = au_sbi(sb);
3513 +       *max = au_lcnt_read(&sbi->si_nfiles, /*do_rev*/1);
3514 +       return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3515 +}
3516 +
3517 +static void au_farray_free(struct file **a, unsigned long long max)
3518 +{
3519 +       unsigned long long ull;
3520 +
3521 +       for (ull = 0; ull < max; ull++)
3522 +               if (a[ull])
3523 +                       fput(a[ull]);
3524 +       kvfree(a);
3525 +}
3526 +
3527 +/* ---------------------------------------------------------------------- */
3528 +
3529 +/*
3530 + * delete a branch
3531 + */
3532 +
3533 +/* to show the line number, do not make it inlined function */
3534 +#define AuVerbose(do_info, fmt, ...) do { \
3535 +       if (do_info) \
3536 +               pr_info(fmt, ##__VA_ARGS__); \
3537 +} while (0)
3538 +
3539 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t btop,
3540 +                        aufs_bindex_t bbot)
3541 +{
3542 +       return (inode && !S_ISDIR(inode->i_mode)) || btop == bbot;
3543 +}
3544 +
3545 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t btop,
3546 +                        aufs_bindex_t bbot)
3547 +{
3548 +       return au_test_ibusy(d_inode(dentry), btop, bbot);
3549 +}
3550 +
3551 +/*
3552 + * test if the branch is deletable or not.
3553 + */
3554 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3555 +                           unsigned int sigen, const unsigned int verbose)
3556 +{
3557 +       int err, i, j, ndentry;
3558 +       aufs_bindex_t btop, bbot;
3559 +       struct au_dcsub_pages dpages;
3560 +       struct au_dpage *dpage;
3561 +       struct dentry *d;
3562 +
3563 +       err = au_dpages_init(&dpages, GFP_NOFS);
3564 +       if (unlikely(err))
3565 +               goto out;
3566 +       err = au_dcsub_pages(&dpages, root, NULL, NULL);
3567 +       if (unlikely(err))
3568 +               goto out_dpages;
3569 +
3570 +       for (i = 0; !err && i < dpages.ndpage; i++) {
3571 +               dpage = dpages.dpages + i;
3572 +               ndentry = dpage->ndentry;
3573 +               for (j = 0; !err && j < ndentry; j++) {
3574 +                       d = dpage->dentries[j];
3575 +                       AuDebugOn(au_dcount(d) <= 0);
3576 +                       if (!au_digen_test(d, sigen)) {
3577 +                               di_read_lock_child(d, AuLock_IR);
3578 +                               if (unlikely(au_dbrange_test(d))) {
3579 +                                       di_read_unlock(d, AuLock_IR);
3580 +                                       continue;
3581 +                               }
3582 +                       } else {
3583 +                               di_write_lock_child(d);
3584 +                               if (unlikely(au_dbrange_test(d))) {
3585 +                                       di_write_unlock(d);
3586 +                                       continue;
3587 +                               }
3588 +                               err = au_reval_dpath(d, sigen);
3589 +                               if (!err)
3590 +                                       di_downgrade_lock(d, AuLock_IR);
3591 +                               else {
3592 +                                       di_write_unlock(d);
3593 +                                       break;
3594 +                               }
3595 +                       }
3596 +
3597 +                       /* AuDbgDentry(d); */
3598 +                       btop = au_dbtop(d);
3599 +                       bbot = au_dbbot(d);
3600 +                       if (btop <= bindex
3601 +                           && bindex <= bbot
3602 +                           && au_h_dptr(d, bindex)
3603 +                           && au_test_dbusy(d, btop, bbot)) {
3604 +                               err = -EBUSY;
3605 +                               AuVerbose(verbose, "busy %pd\n", d);
3606 +                               AuDbgDentry(d);
3607 +                       }
3608 +                       di_read_unlock(d, AuLock_IR);
3609 +               }
3610 +       }
3611 +
3612 +out_dpages:
3613 +       au_dpages_free(&dpages);
3614 +out:
3615 +       return err;
3616 +}
3617 +
3618 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3619 +                          unsigned int sigen, const unsigned int verbose)
3620 +{
3621 +       int err;
3622 +       unsigned long long max, ull;
3623 +       struct inode *i, **array;
3624 +       aufs_bindex_t btop, bbot;
3625 +
3626 +       array = au_iarray_alloc(sb, &max);
3627 +       err = PTR_ERR(array);
3628 +       if (IS_ERR(array))
3629 +               goto out;
3630 +
3631 +       err = 0;
3632 +       AuDbg("b%d\n", bindex);
3633 +       for (ull = 0; !err && ull < max; ull++) {
3634 +               i = array[ull];
3635 +               if (unlikely(!i))
3636 +                       break;
3637 +               if (i->i_ino == AUFS_ROOT_INO)
3638 +                       continue;
3639 +
3640 +               /* AuDbgInode(i); */
3641 +               if (au_iigen(i, NULL) == sigen)
3642 +                       ii_read_lock_child(i);
3643 +               else {
3644 +                       ii_write_lock_child(i);
3645 +                       err = au_refresh_hinode_self(i);
3646 +                       au_iigen_dec(i);
3647 +                       if (!err)
3648 +                               ii_downgrade_lock(i);
3649 +                       else {
3650 +                               ii_write_unlock(i);
3651 +                               break;
3652 +                       }
3653 +               }
3654 +
3655 +               btop = au_ibtop(i);
3656 +               bbot = au_ibbot(i);
3657 +               if (btop <= bindex
3658 +                   && bindex <= bbot
3659 +                   && au_h_iptr(i, bindex)
3660 +                   && au_test_ibusy(i, btop, bbot)) {
3661 +                       err = -EBUSY;
3662 +                       AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3663 +                       AuDbgInode(i);
3664 +               }
3665 +               ii_read_unlock(i);
3666 +       }
3667 +       au_iarray_free(array, max);
3668 +
3669 +out:
3670 +       return err;
3671 +}
3672 +
3673 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3674 +                             const unsigned int verbose)
3675 +{
3676 +       int err;
3677 +       unsigned int sigen;
3678 +
3679 +       sigen = au_sigen(root->d_sb);
3680 +       DiMustNoWaiters(root);
3681 +       IiMustNoWaiters(d_inode(root));
3682 +       di_write_unlock(root);
3683 +       err = test_dentry_busy(root, bindex, sigen, verbose);
3684 +       if (!err)
3685 +               err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3686 +       di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3687 +
3688 +       return err;
3689 +}
3690 +
3691 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3692 +                        struct file **to_free, int *idx)
3693 +{
3694 +       int err;
3695 +       unsigned char matched, root;
3696 +       aufs_bindex_t bindex, bbot;
3697 +       struct au_fidir *fidir;
3698 +       struct au_hfile *hfile;
3699 +
3700 +       err = 0;
3701 +       root = IS_ROOT(file->f_path.dentry);
3702 +       if (root) {
3703 +               get_file(file);
3704 +               to_free[*idx] = file;
3705 +               (*idx)++;
3706 +               goto out;
3707 +       }
3708 +
3709 +       matched = 0;
3710 +       fidir = au_fi(file)->fi_hdir;
3711 +       AuDebugOn(!fidir);
3712 +       bbot = au_fbbot_dir(file);
3713 +       for (bindex = au_fbtop(file); bindex <= bbot; bindex++) {
3714 +               hfile = fidir->fd_hfile + bindex;
3715 +               if (!hfile->hf_file)
3716 +                       continue;
3717 +
3718 +               if (hfile->hf_br->br_id == br_id) {
3719 +                       matched = 1;
3720 +                       break;
3721 +               }
3722 +       }
3723 +       if (matched)
3724 +               err = -EBUSY;
3725 +
3726 +out:
3727 +       return err;
3728 +}
3729 +
3730 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3731 +                         struct file **to_free, int opened)
3732 +{
3733 +       int err, idx;
3734 +       unsigned long long ull, max;
3735 +       aufs_bindex_t btop;
3736 +       struct file *file, **array;
3737 +       struct dentry *root;
3738 +       struct au_hfile *hfile;
3739 +
3740 +       array = au_farray_alloc(sb, &max);
3741 +       err = PTR_ERR(array);
3742 +       if (IS_ERR(array))
3743 +               goto out;
3744 +
3745 +       err = 0;
3746 +       idx = 0;
3747 +       root = sb->s_root;
3748 +       di_write_unlock(root);
3749 +       for (ull = 0; ull < max; ull++) {
3750 +               file = array[ull];
3751 +               if (unlikely(!file))
3752 +                       break;
3753 +
3754 +               /* AuDbg("%pD\n", file); */
3755 +               fi_read_lock(file);
3756 +               btop = au_fbtop(file);
3757 +               if (!d_is_dir(file->f_path.dentry)) {
3758 +                       hfile = &au_fi(file)->fi_htop;
3759 +                       if (hfile->hf_br->br_id == br_id)
3760 +                               err = -EBUSY;
3761 +               } else
3762 +                       err = test_dir_busy(file, br_id, to_free, &idx);
3763 +               fi_read_unlock(file);
3764 +               if (unlikely(err))
3765 +                       break;
3766 +       }
3767 +       di_write_lock_child(root);
3768 +       au_farray_free(array, max);
3769 +       AuDebugOn(idx > opened);
3770 +
3771 +out:
3772 +       return err;
3773 +}
3774 +
3775 +static void br_del_file(struct file **to_free, unsigned long long opened,
3776 +                       aufs_bindex_t br_id)
3777 +{
3778 +       unsigned long long ull;
3779 +       aufs_bindex_t bindex, btop, bbot, bfound;
3780 +       struct file *file;
3781 +       struct au_fidir *fidir;
3782 +       struct au_hfile *hfile;
3783 +
3784 +       for (ull = 0; ull < opened; ull++) {
3785 +               file = to_free[ull];
3786 +               if (unlikely(!file))
3787 +                       break;
3788 +
3789 +               /* AuDbg("%pD\n", file); */
3790 +               AuDebugOn(!d_is_dir(file->f_path.dentry));
3791 +               bfound = -1;
3792 +               fidir = au_fi(file)->fi_hdir;
3793 +               AuDebugOn(!fidir);
3794 +               fi_write_lock(file);
3795 +               btop = au_fbtop(file);
3796 +               bbot = au_fbbot_dir(file);
3797 +               for (bindex = btop; bindex <= bbot; bindex++) {
3798 +                       hfile = fidir->fd_hfile + bindex;
3799 +                       if (!hfile->hf_file)
3800 +                               continue;
3801 +
3802 +                       if (hfile->hf_br->br_id == br_id) {
3803 +                               bfound = bindex;
3804 +                               break;
3805 +                       }
3806 +               }
3807 +               AuDebugOn(bfound < 0);
3808 +               au_set_h_fptr(file, bfound, NULL);
3809 +               if (bfound == btop) {
3810 +                       for (btop++; btop <= bbot; btop++)
3811 +                               if (au_hf_dir(file, btop)) {
3812 +                                       au_set_fbtop(file, btop);
3813 +                                       break;
3814 +                               }
3815 +               }
3816 +               fi_write_unlock(file);
3817 +       }
3818 +}
3819 +
3820 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3821 +                            const aufs_bindex_t bindex,
3822 +                            const aufs_bindex_t bbot)
3823 +{
3824 +       struct au_branch **brp, **p;
3825 +
3826 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
3827 +
3828 +       brp = sbinfo->si_branch + bindex;
3829 +       if (bindex < bbot)
3830 +               memmove(brp, brp + 1, sizeof(*brp) * (bbot - bindex));
3831 +       sbinfo->si_branch[0 + bbot] = NULL;
3832 +       sbinfo->si_bbot--;
3833 +
3834 +       p = au_krealloc(sbinfo->si_branch, sizeof(*p) * bbot, AuGFP_SBILIST,
3835 +                       /*may_shrink*/1);
3836 +       if (p)
3837 +               sbinfo->si_branch = p;
3838 +       /* harmless error */
3839 +}
3840 +
3841 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3842 +                            const aufs_bindex_t bbot)
3843 +{
3844 +       struct au_hdentry *hdp, *p;
3845 +
3846 +       AuRwMustWriteLock(&dinfo->di_rwsem);
3847 +
3848 +       hdp = au_hdentry(dinfo, bindex);
3849 +       if (bindex < bbot)
3850 +               memmove(hdp, hdp + 1, sizeof(*hdp) * (bbot - bindex));
3851 +       /* au_h_dentry_init(au_hdentry(dinfo, bbot); */
3852 +       dinfo->di_bbot--;
3853 +
3854 +       p = au_krealloc(dinfo->di_hdentry, sizeof(*p) * bbot, AuGFP_SBILIST,
3855 +                       /*may_shrink*/1);
3856 +       if (p)
3857 +               dinfo->di_hdentry = p;
3858 +       /* harmless error */
3859 +}
3860 +
3861 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3862 +                            const aufs_bindex_t bbot)
3863 +{
3864 +       struct au_hinode *hip, *p;
3865 +
3866 +       AuRwMustWriteLock(&iinfo->ii_rwsem);
3867 +
3868 +       hip = au_hinode(iinfo, bindex);
3869 +       if (bindex < bbot)
3870 +               memmove(hip, hip + 1, sizeof(*hip) * (bbot - bindex));
3871 +       /* au_hinode_init(au_hinode(iinfo, bbot)); */
3872 +       iinfo->ii_bbot--;
3873 +
3874 +       p = au_krealloc(iinfo->ii_hinode, sizeof(*p) * bbot, AuGFP_SBILIST,
3875 +                       /*may_shrink*/1);
3876 +       if (p)
3877 +               iinfo->ii_hinode = p;
3878 +       /* harmless error */
3879 +}
3880 +
3881 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3882 +                        struct au_branch *br)
3883 +{
3884 +       aufs_bindex_t bbot;
3885 +       struct au_sbinfo *sbinfo;
3886 +       struct dentry *root, *h_root;
3887 +       struct inode *inode, *h_inode;
3888 +       struct au_hinode *hinode;
3889 +
3890 +       SiMustWriteLock(sb);
3891 +
3892 +       root = sb->s_root;
3893 +       inode = d_inode(root);
3894 +       sbinfo = au_sbi(sb);
3895 +       bbot = sbinfo->si_bbot;
3896 +
3897 +       h_root = au_h_dptr(root, bindex);
3898 +       hinode = au_hi(inode, bindex);
3899 +       h_inode = au_igrab(hinode->hi_inode);
3900 +       au_hiput(hinode);
3901 +
3902 +       au_sbilist_lock();
3903 +       au_br_do_del_brp(sbinfo, bindex, bbot);
3904 +       au_br_do_del_hdp(au_di(root), bindex, bbot);
3905 +       au_br_do_del_hip(au_ii(inode), bindex, bbot);
3906 +       au_sbilist_unlock();
3907 +
3908 +       /* ignore an error */
3909 +       au_dr_br_fin(sb, br); /* always, regardless the mount option */
3910 +
3911 +       dput(h_root);
3912 +       iput(h_inode);
3913 +       au_br_do_free(br);
3914 +}
3915 +
3916 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3917 +                                  unsigned long long max, void *arg)
3918 +{
3919 +       return max;
3920 +}
3921 +
3922 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3923 +{
3924 +       int err, rerr, i;
3925 +       unsigned long long opened;
3926 +       unsigned int mnt_flags;
3927 +       aufs_bindex_t bindex, bbot, br_id;
3928 +       unsigned char do_wh, verbose;
3929 +       struct au_branch *br;
3930 +       struct au_wbr *wbr;
3931 +       struct dentry *root;
3932 +       struct file **to_free;
3933 +
3934 +       err = 0;
3935 +       opened = 0;
3936 +       to_free = NULL;
3937 +       root = sb->s_root;
3938 +       bindex = au_find_dbindex(root, del->h_path.dentry);
3939 +       if (bindex < 0) {
3940 +               if (remount)
3941 +                       goto out; /* success */
3942 +               err = -ENOENT;
3943 +               pr_err("%s no such branch\n", del->pathname);
3944 +               goto out;
3945 +       }
3946 +       AuDbg("bindex b%d\n", bindex);
3947 +
3948 +       err = -EBUSY;
3949 +       mnt_flags = au_mntflags(sb);
3950 +       verbose = !!au_opt_test(mnt_flags, VERBOSE);
3951 +       bbot = au_sbbot(sb);
3952 +       if (unlikely(!bbot)) {
3953 +               AuVerbose(verbose, "no more branches left\n");
3954 +               goto out;
3955 +       }
3956 +
3957 +       br = au_sbr(sb, bindex);
3958 +       AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3959 +       if (unlikely(au_lcnt_read(&br->br_count, /*do_rev*/1))) {
3960 +               AuVerbose(verbose, "br %pd2 is busy now\n", del->h_path.dentry);
3961 +               goto out;
3962 +       }
3963 +
3964 +       br_id = br->br_id;
3965 +       opened = au_lcnt_read(&br->br_nfiles, /*do_rev*/1);
3966 +       if (unlikely(opened)) {
3967 +               to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3968 +               err = PTR_ERR(to_free);
3969 +               if (IS_ERR(to_free))
3970 +                       goto out;
3971 +
3972 +               err = test_file_busy(sb, br_id, to_free, opened);
3973 +               if (unlikely(err)) {
3974 +                       AuVerbose(verbose, "%llu file(s) opened\n", opened);
3975 +                       goto out;
3976 +               }
3977 +       }
3978 +
3979 +       wbr = br->br_wbr;
3980 +       do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3981 +       if (do_wh) {
3982 +               /* instead of WbrWhMustWriteLock(wbr) */
3983 +               SiMustWriteLock(sb);
3984 +               for (i = 0; i < AuBrWh_Last; i++) {
3985 +                       dput(wbr->wbr_wh[i]);
3986 +                       wbr->wbr_wh[i] = NULL;
3987 +               }
3988 +       }
3989 +
3990 +       err = test_children_busy(root, bindex, verbose);
3991 +       if (unlikely(err)) {
3992 +               if (do_wh)
3993 +                       goto out_wh;
3994 +               goto out;
3995 +       }
3996 +
3997 +       err = 0;
3998 +       if (to_free) {
3999 +               /*
4000 +                * now we confirmed the branch is deletable.
4001 +                * let's free the remaining opened dirs on the branch.
4002 +                */
4003 +               di_write_unlock(root);
4004 +               br_del_file(to_free, opened, br_id);
4005 +               di_write_lock_child(root);
4006 +       }
4007 +
4008 +       sysaufs_brs_del(sb, bindex);    /* remove successors */
4009 +       dbgaufs_xino_del(br);           /* remove one */
4010 +       au_br_do_del(sb, bindex, br);
4011 +       sysaufs_brs_add(sb, bindex);    /* append successors */
4012 +       dbgaufs_brs_add(sb, bindex, /*topdown*/1);      /* rename successors */
4013 +
4014 +       if (!bindex) {
4015 +               au_cpup_attr_all(d_inode(root), /*force*/1);
4016 +               sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
4017 +       } else
4018 +               au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
4019 +       if (au_opt_test(mnt_flags, PLINK))
4020 +               au_plink_half_refresh(sb, br_id);
4021 +
4022 +       goto out; /* success */
4023 +
4024 +out_wh:
4025 +       /* revert */
4026 +       rerr = au_br_init_wh(sb, br, br->br_perm);
4027 +       if (rerr)
4028 +               pr_warn("failed re-creating base whiteout, %s. (%d)\n",
4029 +                       del->pathname, rerr);
4030 +out:
4031 +       if (to_free)
4032 +               au_farray_free(to_free, opened);
4033 +       return err;
4034 +}
4035 +
4036 +/* ---------------------------------------------------------------------- */
4037 +
4038 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
4039 +{
4040 +       int err;
4041 +       aufs_bindex_t btop, bbot;
4042 +       struct aufs_ibusy ibusy;
4043 +       struct inode *inode, *h_inode;
4044 +
4045 +       err = -EPERM;
4046 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
4047 +               goto out;
4048 +
4049 +       err = copy_from_user(&ibusy, arg, sizeof(ibusy));
4050 +       if (!err)
4051 +               /* VERIFY_WRITE */
4052 +               err = !access_ok(&arg->h_ino, sizeof(arg->h_ino));
4053 +       if (unlikely(err)) {
4054 +               err = -EFAULT;
4055 +               AuTraceErr(err);
4056 +               goto out;
4057 +       }
4058 +
4059 +       err = -EINVAL;
4060 +       si_read_lock(sb, AuLock_FLUSH);
4061 +       if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbbot(sb)))
4062 +               goto out_unlock;
4063 +
4064 +       err = 0;
4065 +       ibusy.h_ino = 0; /* invalid */
4066 +       inode = ilookup(sb, ibusy.ino);
4067 +       if (!inode
4068 +           || inode->i_ino == AUFS_ROOT_INO
4069 +           || au_is_bad_inode(inode))
4070 +               goto out_unlock;
4071 +
4072 +       ii_read_lock_child(inode);
4073 +       btop = au_ibtop(inode);
4074 +       bbot = au_ibbot(inode);
4075 +       if (btop <= ibusy.bindex && ibusy.bindex <= bbot) {
4076 +               h_inode = au_h_iptr(inode, ibusy.bindex);
4077 +               if (h_inode && au_test_ibusy(inode, btop, bbot))
4078 +                       ibusy.h_ino = h_inode->i_ino;
4079 +       }
4080 +       ii_read_unlock(inode);
4081 +       iput(inode);
4082 +
4083 +out_unlock:
4084 +       si_read_unlock(sb);
4085 +       if (!err) {
4086 +               err = __put_user(ibusy.h_ino, &arg->h_ino);
4087 +               if (unlikely(err)) {
4088 +                       err = -EFAULT;
4089 +                       AuTraceErr(err);
4090 +               }
4091 +       }
4092 +out:
4093 +       return err;
4094 +}
4095 +
4096 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
4097 +{
4098 +       return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
4099 +}
4100 +
4101 +#ifdef CONFIG_COMPAT
4102 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
4103 +{
4104 +       return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
4105 +}
4106 +#endif
4107 +
4108 +/* ---------------------------------------------------------------------- */
4109 +
4110 +/*
4111 + * change a branch permission
4112 + */
4113 +
4114 +static void au_warn_ima(void)
4115 +{
4116 +#ifdef CONFIG_IMA
4117 +       /* since it doesn't support mark_files_ro() */
4118 +       AuWarn1("RW -> RO makes IMA to produce wrong message\n");
4119 +#endif
4120 +}
4121 +
4122 +static int do_need_sigen_inc(int a, int b)
4123 +{
4124 +       return au_br_whable(a) && !au_br_whable(b);
4125 +}
4126 +
4127 +static int need_sigen_inc(int old, int new)
4128 +{
4129 +       return do_need_sigen_inc(old, new)
4130 +               || do_need_sigen_inc(new, old);
4131 +}
4132 +
4133 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
4134 +{
4135 +       int err, do_warn;
4136 +       unsigned int mnt_flags;
4137 +       unsigned long long ull, max;
4138 +       aufs_bindex_t br_id;
4139 +       unsigned char verbose, writer;
4140 +       struct file *file, *hf, **array;
4141 +       struct au_hfile *hfile;
4142 +       struct inode *h_inode;
4143 +
4144 +       mnt_flags = au_mntflags(sb);
4145 +       verbose = !!au_opt_test(mnt_flags, VERBOSE);
4146 +
4147 +       array = au_farray_alloc(sb, &max);
4148 +       err = PTR_ERR(array);
4149 +       if (IS_ERR(array))
4150 +               goto out;
4151 +
4152 +       do_warn = 0;
4153 +       br_id = au_sbr_id(sb, bindex);
4154 +       for (ull = 0; ull < max; ull++) {
4155 +               file = array[ull];
4156 +               if (unlikely(!file))
4157 +                       break;
4158 +
4159 +               /* AuDbg("%pD\n", file); */
4160 +               fi_read_lock(file);
4161 +               if (unlikely(au_test_mmapped(file))) {
4162 +                       err = -EBUSY;
4163 +                       AuVerbose(verbose, "mmapped %pD\n", file);
4164 +                       AuDbgFile(file);
4165 +                       FiMustNoWaiters(file);
4166 +                       fi_read_unlock(file);
4167 +                       goto out_array;
4168 +               }
4169 +
4170 +               hfile = &au_fi(file)->fi_htop;
4171 +               hf = hfile->hf_file;
4172 +               if (!d_is_reg(file->f_path.dentry)
4173 +                   || !(file->f_mode & FMODE_WRITE)
4174 +                   || hfile->hf_br->br_id != br_id
4175 +                   || !(hf->f_mode & FMODE_WRITE))
4176 +                       array[ull] = NULL;
4177 +               else {
4178 +                       do_warn = 1;
4179 +                       get_file(file);
4180 +               }
4181 +
4182 +               FiMustNoWaiters(file);
4183 +               fi_read_unlock(file);
4184 +               fput(file);
4185 +       }
4186 +
4187 +       err = 0;
4188 +       if (do_warn)
4189 +               au_warn_ima();
4190 +
4191 +       for (ull = 0; ull < max; ull++) {
4192 +               file = array[ull];
4193 +               if (!file)
4194 +                       continue;
4195 +
4196 +               /* todo: already flushed? */
4197 +               /*
4198 +                * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4199 +                * approach which resets f_mode and calls mnt_drop_write() and
4200 +                * file_release_write() for each file, because the branch
4201 +                * attribute in aufs world is totally different from the native
4202 +                * fs rw/ro mode.
4203 +                */
4204 +               /* fi_read_lock(file); */
4205 +               hfile = &au_fi(file)->fi_htop;
4206 +               hf = hfile->hf_file;
4207 +               /* fi_read_unlock(file); */
4208 +               spin_lock(&hf->f_lock);
4209 +               writer = !!(hf->f_mode & FMODE_WRITER);
4210 +               hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4211 +               spin_unlock(&hf->f_lock);
4212 +               if (writer) {
4213 +                       h_inode = file_inode(hf);
4214 +                       if (hf->f_mode & FMODE_READ)
4215 +                               i_readcount_inc(h_inode);
4216 +                       put_write_access(h_inode);
4217 +                       __mnt_drop_write(hf->f_path.mnt);
4218 +               }
4219 +       }
4220 +
4221 +out_array:
4222 +       au_farray_free(array, max);
4223 +out:
4224 +       AuTraceErr(err);
4225 +       return err;
4226 +}
4227 +
4228 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4229 +             int *do_refresh)
4230 +{
4231 +       int err, rerr;
4232 +       aufs_bindex_t bindex;
4233 +       struct dentry *root;
4234 +       struct au_branch *br;
4235 +       struct au_br_fhsm *bf;
4236 +
4237 +       root = sb->s_root;
4238 +       bindex = au_find_dbindex(root, mod->h_root);
4239 +       if (bindex < 0) {
4240 +               if (remount)
4241 +                       return 0; /* success */
4242 +               err = -ENOENT;
4243 +               pr_err("%s no such branch\n", mod->path);
4244 +               goto out;
4245 +       }
4246 +       AuDbg("bindex b%d\n", bindex);
4247 +
4248 +       err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4249 +       if (unlikely(err))
4250 +               goto out;
4251 +
4252 +       br = au_sbr(sb, bindex);
4253 +       AuDebugOn(mod->h_root != au_br_dentry(br));
4254 +       if (br->br_perm == mod->perm)
4255 +               return 0; /* success */
4256 +
4257 +       /* pre-allocate for non-fhsm --> fhsm */
4258 +       bf = NULL;
4259 +       if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4260 +               err = au_fhsm_br_alloc(br);
4261 +               if (unlikely(err))
4262 +                       goto out;
4263 +               bf = br->br_fhsm;
4264 +               br->br_fhsm = NULL;
4265 +       }
4266 +
4267 +       if (au_br_writable(br->br_perm)) {
4268 +               /* remove whiteout base */
4269 +               err = au_br_init_wh(sb, br, mod->perm);
4270 +               if (unlikely(err))
4271 +                       goto out_bf;
4272 +
4273 +               if (!au_br_writable(mod->perm)) {
4274 +                       /* rw --> ro, file might be mmapped */
4275 +                       DiMustNoWaiters(root);
4276 +                       IiMustNoWaiters(d_inode(root));
4277 +                       di_write_unlock(root);
4278 +                       err = au_br_mod_files_ro(sb, bindex);
4279 +                       /* aufs_write_lock() calls ..._child() */
4280 +                       di_write_lock_child(root);
4281 +
4282 +                       if (unlikely(err)) {
4283 +                               rerr = -ENOMEM;
4284 +                               br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4285 +                                                    GFP_NOFS);
4286 +                               if (br->br_wbr)
4287 +                                       rerr = au_wbr_init(br, sb, br->br_perm);
4288 +                               if (unlikely(rerr)) {
4289 +                                       AuIOErr("nested error %d (%d)\n",
4290 +                                               rerr, err);
4291 +                                       br->br_perm = mod->perm;
4292 +                               }
4293 +                       }
4294 +               }
4295 +       } else if (au_br_writable(mod->perm)) {
4296 +               /* ro --> rw */
4297 +               err = -ENOMEM;
4298 +               br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4299 +               if (br->br_wbr) {
4300 +                       err = au_wbr_init(br, sb, mod->perm);
4301 +                       if (unlikely(err)) {
4302 +                               au_kfree_rcu(br->br_wbr);
4303 +                               br->br_wbr = NULL;
4304 +                       }
4305 +               }
4306 +       }
4307 +       if (unlikely(err))
4308 +               goto out_bf;
4309 +
4310 +       if (au_br_fhsm(br->br_perm)) {
4311 +               if (!au_br_fhsm(mod->perm)) {
4312 +                       /* fhsm --> non-fhsm */
4313 +                       au_br_fhsm_fin(br->br_fhsm);
4314 +                       au_kfree_rcu(br->br_fhsm);
4315 +                       br->br_fhsm = NULL;
4316 +               }
4317 +       } else if (au_br_fhsm(mod->perm))
4318 +               /* non-fhsm --> fhsm */
4319 +               br->br_fhsm = bf;
4320 +
4321 +       *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4322 +       br->br_perm = mod->perm;
4323 +       goto out; /* success */
4324 +
4325 +out_bf:
4326 +       au_kfree_try_rcu(bf);
4327 +out:
4328 +       AuTraceErr(err);
4329 +       return err;
4330 +}
4331 +
4332 +/* ---------------------------------------------------------------------- */
4333 +
4334 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4335 +{
4336 +       int err;
4337 +       struct kstatfs kstfs;
4338 +
4339 +       err = vfs_statfs(&br->br_path, &kstfs);
4340 +       if (!err) {
4341 +               stfs->f_blocks = kstfs.f_blocks;
4342 +               stfs->f_bavail = kstfs.f_bavail;
4343 +               stfs->f_files = kstfs.f_files;
4344 +               stfs->f_ffree = kstfs.f_ffree;
4345 +       }
4346 +
4347 +       return err;
4348 +}
4349 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4350 --- /usr/share/empty/fs/aufs/branch.h   1970-01-01 01:00:00.000000000 +0100
4351 +++ linux/fs/aufs/branch.h      2022-11-05 23:02:18.959222617 +0100
4352 @@ -0,0 +1,375 @@
4353 +/* SPDX-License-Identifier: GPL-2.0 */
4354 +/*
4355 + * Copyright (C) 2005-2022 Junjiro R. Okajima
4356 + *
4357 + * This program is free software; you can redistribute it and/or modify
4358 + * it under the terms of the GNU General Public License as published by
4359 + * the Free Software Foundation; either version 2 of the License, or
4360 + * (at your option) any later version.
4361 + *
4362 + * This program is distributed in the hope that it will be useful,
4363 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4364 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
4365 + * GNU General Public License for more details.
4366 + *
4367 + * You should have received a copy of the GNU General Public License
4368 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
4369 + */
4370 +
4371 +/*
4372 + * branch filesystems and xino for them
4373 + */
4374 +
4375 +#ifndef __AUFS_BRANCH_H__
4376 +#define __AUFS_BRANCH_H__
4377 +
4378 +#ifdef __KERNEL__
4379 +
4380 +#include <linux/mount.h>
4381 +#include "dirren.h"
4382 +#include "dynop.h"
4383 +#include "lcnt.h"
4384 +#include "rwsem.h"
4385 +#include "super.h"
4386 +
4387 +/* ---------------------------------------------------------------------- */
4388 +
4389 +/* a xino file */
4390 +struct au_xino {
4391 +       struct file             **xi_file;
4392 +       unsigned int            xi_nfile;
4393 +
4394 +       struct {
4395 +               spinlock_t              spin;
4396 +               ino_t                   *array;
4397 +               int                     total;
4398 +               /* reserved for future use */
4399 +               /* unsigned long        *bitmap; */
4400 +               wait_queue_head_t       wqh;
4401 +       } xi_nondir;
4402 +
4403 +       struct mutex            xi_mtx; /* protects xi_file array */
4404 +       struct hlist_bl_head    xi_writing;
4405 +
4406 +       atomic_t                xi_truncating;
4407 +
4408 +       struct kref             xi_kref;
4409 +};
4410 +
4411 +/* File-based Hierarchical Storage Management */
4412 +struct au_br_fhsm {
4413 +#ifdef CONFIG_AUFS_FHSM
4414 +       struct mutex            bf_lock;
4415 +       unsigned long           bf_jiffy;
4416 +       struct aufs_stfs        bf_stfs;
4417 +       int                     bf_readable;
4418 +#endif
4419 +};
4420 +
4421 +/* members for writable branch only */
4422 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4423 +struct au_wbr {
4424 +       struct au_rwsem         wbr_wh_rwsem;
4425 +       struct dentry           *wbr_wh[AuBrWh_Last];
4426 +       atomic_t                wbr_wh_running;
4427 +#define wbr_whbase             wbr_wh[AuBrWh_BASE]     /* whiteout base */
4428 +#define wbr_plink              wbr_wh[AuBrWh_PLINK]    /* pseudo-link dir */
4429 +#define wbr_orph               wbr_wh[AuBrWh_ORPH]     /* dir for orphans */
4430 +
4431 +       /* mfs mode */
4432 +       unsigned long long      wbr_bytes;
4433 +};
4434 +
4435 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4436 +#define AuBrDynOp (AuDyLast * 4)
4437 +
4438 +#ifdef CONFIG_AUFS_HFSNOTIFY
4439 +/* support for asynchronous destruction */
4440 +struct au_br_hfsnotify {
4441 +       struct fsnotify_group   *hfsn_group;
4442 +};
4443 +#endif
4444 +
4445 +/* sysfs entries */
4446 +struct au_brsysfs {
4447 +       char                    name[16];
4448 +       struct attribute        attr;
4449 +};
4450 +
4451 +enum {
4452 +       AuBrSysfs_BR,
4453 +       AuBrSysfs_BRID,
4454 +       AuBrSysfs_Last
4455 +};
4456 +
4457 +/* protected by superblock rwsem */
4458 +struct au_branch {
4459 +       struct au_xino          *br_xino;
4460 +
4461 +       aufs_bindex_t           br_id;
4462 +
4463 +       int                     br_perm;
4464 +       struct path             br_path;
4465 +       spinlock_t              br_dykey_lock;
4466 +       struct au_dykey         *br_dykey[AuBrDynOp];
4467 +       au_lcnt_t               br_nfiles;      /* opened files */
4468 +       au_lcnt_t               br_count;       /* in-use for other */
4469 +
4470 +       struct au_wbr           *br_wbr;
4471 +       struct au_br_fhsm       *br_fhsm;
4472 +
4473 +#ifdef CONFIG_AUFS_HFSNOTIFY
4474 +       struct au_br_hfsnotify  *br_hfsn;
4475 +#endif
4476 +
4477 +#ifdef CONFIG_SYSFS
4478 +       /* entries under sysfs per mount-point */
4479 +       struct au_brsysfs       br_sysfs[AuBrSysfs_Last];
4480 +#endif
4481 +
4482 +#ifdef CONFIG_DEBUG_FS
4483 +       struct dentry            *br_dbgaufs; /* xino */
4484 +#endif
4485 +
4486 +       struct au_dr_br         br_dirren;
4487 +};
4488 +
4489 +/* ---------------------------------------------------------------------- */
4490 +
4491 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4492 +{
4493 +       return br->br_path.mnt;
4494 +}
4495 +
4496 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4497 +{
4498 +       return br->br_path.dentry;
4499 +}
4500 +
4501 +static inline struct user_namespace *au_br_userns(struct au_branch *br)
4502 +{
4503 +       return mnt_user_ns(br->br_path.mnt);
4504 +}
4505 +
4506 +static inline struct super_block *au_br_sb(struct au_branch *br)
4507 +{
4508 +       return au_br_mnt(br)->mnt_sb;
4509 +}
4510 +
4511 +static inline int au_br_rdonly(struct au_branch *br)
4512 +{
4513 +       return (sb_rdonly(au_br_sb(br))
4514 +               || !au_br_writable(br->br_perm))
4515 +               ? -EROFS : 0;
4516 +}
4517 +
4518 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4519 +{
4520 +#ifdef CONFIG_AUFS_HNOTIFY
4521 +       return !(brperm & AuBrPerm_RR);
4522 +#else
4523 +       return 0;
4524 +#endif
4525 +}
4526 +
4527 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4528 +{
4529 +       int err, exec_flag;
4530 +
4531 +       err = 0;
4532 +       exec_flag = oflag & __FMODE_EXEC;
4533 +       if (unlikely(exec_flag && path_noexec(&br->br_path)))
4534 +               err = -EACCES;
4535 +
4536 +       return err;
4537 +}
4538 +
4539 +static inline void au_xino_get(struct au_branch *br)
4540 +{
4541 +       struct au_xino *xi;
4542 +
4543 +       xi = br->br_xino;
4544 +       if (xi)
4545 +               kref_get(&xi->xi_kref);
4546 +}
4547 +
4548 +static inline int au_xino_count(struct au_branch *br)
4549 +{
4550 +       int v;
4551 +       struct au_xino *xi;
4552 +
4553 +       v = 0;
4554 +       xi = br->br_xino;
4555 +       if (xi)
4556 +               v = kref_read(&xi->xi_kref);
4557 +
4558 +       return v;
4559 +}
4560 +
4561 +/* ---------------------------------------------------------------------- */
4562 +
4563 +/* branch.c */
4564 +struct au_sbinfo;
4565 +void au_br_free(struct au_sbinfo *sinfo);
4566 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4567 +struct au_opt_add;
4568 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4569 +struct au_opt_del;
4570 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4571 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4572 +#ifdef CONFIG_COMPAT
4573 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4574 +#endif
4575 +struct au_opt_mod;
4576 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4577 +             int *do_refresh);
4578 +struct aufs_stfs;
4579 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4580 +
4581 +/* xino.c */
4582 +static const loff_t au_loff_max = LLONG_MAX;
4583 +
4584 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry);
4585 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
4586 +                           int wbrtop);
4587 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
4588 +                            struct file *copy_src);
4589 +struct au_xi_new {
4590 +       struct au_xino *xi;     /* switch between xino and xigen */
4591 +       int idx;
4592 +       struct path *base;
4593 +       struct file *copy_src;
4594 +};
4595 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew);
4596 +
4597 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4598 +                ino_t *ino);
4599 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4600 +                 ino_t ino);
4601 +ssize_t xino_fread(struct file *file, void *buf, size_t size, loff_t *pos);
4602 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos);
4603 +
4604 +int au_xib_trunc(struct super_block *sb);
4605 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin);
4606 +
4607 +struct au_xino *au_xino_alloc(unsigned int nfile);
4608 +int au_xino_put(struct au_branch *br);
4609 +struct file *au_xino_file1(struct au_xino *xi);
4610 +
4611 +struct au_opt_xino;
4612 +void au_xino_clr(struct super_block *sb);
4613 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount);
4614 +struct file *au_xino_def(struct super_block *sb);
4615 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4616 +                   struct path *base);
4617 +
4618 +ino_t au_xino_new_ino(struct super_block *sb);
4619 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4620 +
4621 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
4622 +                      ino_t h_ino, int idx);
4623 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4624 +                     int *idx);
4625 +
4626 +int au_xino_path(struct seq_file *seq, struct file *file);
4627 +
4628 +/* ---------------------------------------------------------------------- */
4629 +
4630 +/* @idx is signed to accept -1 meaning the first file */
4631 +static inline struct file *au_xino_file(struct au_xino *xi, int idx)
4632 +{
4633 +       struct file *file;
4634 +
4635 +       file = NULL;
4636 +       if (!xi)
4637 +               goto out;
4638 +
4639 +       if (idx >= 0) {
4640 +               if (idx < xi->xi_nfile)
4641 +                       file = xi->xi_file[idx];
4642 +       } else
4643 +               file = au_xino_file1(xi);
4644 +
4645 +out:
4646 +       return file;
4647 +}
4648 +
4649 +/* ---------------------------------------------------------------------- */
4650 +
4651 +/* Superblock to branch */
4652 +static inline
4653 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4654 +{
4655 +       return au_sbr(sb, bindex)->br_id;
4656 +}
4657 +
4658 +static inline
4659 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4660 +{
4661 +       return au_br_mnt(au_sbr(sb, bindex));
4662 +}
4663 +
4664 +static inline
4665 +struct user_namespace *au_sbr_userns(struct super_block *sb, aufs_bindex_t bindex)
4666 +{
4667 +       return au_br_userns(au_sbr(sb, bindex));
4668 +}
4669 +
4670 +static inline
4671 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4672 +{
4673 +       return au_br_sb(au_sbr(sb, bindex));
4674 +}
4675 +
4676 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4677 +{
4678 +       return au_sbr(sb, bindex)->br_perm;
4679 +}
4680 +
4681 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4682 +{
4683 +       return au_br_whable(au_sbr_perm(sb, bindex));
4684 +}
4685 +
4686 +/* ---------------------------------------------------------------------- */
4687 +
4688 +#define wbr_wh_read_lock(wbr)  au_rw_read_lock(&(wbr)->wbr_wh_rwsem)
4689 +#define wbr_wh_write_lock(wbr) au_rw_write_lock(&(wbr)->wbr_wh_rwsem)
4690 +#define wbr_wh_read_trylock(wbr)       au_rw_read_trylock(&(wbr)->wbr_wh_rwsem)
4691 +#define wbr_wh_write_trylock(wbr) au_rw_write_trylock(&(wbr)->wbr_wh_rwsem)
4692 +/*
4693 +#define wbr_wh_read_trylock_nested(wbr) \
4694 +       au_rw_read_trylock_nested(&(wbr)->wbr_wh_rwsem)
4695 +#define wbr_wh_write_trylock_nested(wbr) \
4696 +       au_rw_write_trylock_nested(&(wbr)->wbr_wh_rwsem)
4697 +*/
4698 +
4699 +#define wbr_wh_read_unlock(wbr)        au_rw_read_unlock(&(wbr)->wbr_wh_rwsem)
4700 +#define wbr_wh_write_unlock(wbr)       au_rw_write_unlock(&(wbr)->wbr_wh_rwsem)
4701 +#define wbr_wh_downgrade_lock(wbr)     au_rw_dgrade_lock(&(wbr)->wbr_wh_rwsem)
4702 +
4703 +#define WbrWhMustNoWaiters(wbr)        AuRwMustNoWaiters(&(wbr)->wbr_wh_rwsem)
4704 +#define WbrWhMustAnyLock(wbr)  AuRwMustAnyLock(&(wbr)->wbr_wh_rwsem)
4705 +#define WbrWhMustWriteLock(wbr)        AuRwMustWriteLock(&(wbr)->wbr_wh_rwsem)
4706 +
4707 +/* ---------------------------------------------------------------------- */
4708 +
4709 +#ifdef CONFIG_AUFS_FHSM
4710 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4711 +{
4712 +       mutex_init(&brfhsm->bf_lock);
4713 +       brfhsm->bf_jiffy = 0;
4714 +       brfhsm->bf_readable = 0;
4715 +}
4716 +
4717 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4718 +{
4719 +       mutex_destroy(&brfhsm->bf_lock);
4720 +}
4721 +#else
4722 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4723 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4724 +#endif
4725 +
4726 +#endif /* __KERNEL__ */
4727 +#endif /* __AUFS_BRANCH_H__ */
4728 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4729 --- /usr/share/empty/fs/aufs/conf.mk    1970-01-01 01:00:00.000000000 +0100
4730 +++ linux/fs/aufs/conf.mk       2022-11-05 23:02:18.959222617 +0100
4731 @@ -0,0 +1,40 @@
4732 +# SPDX-License-Identifier: GPL-2.0
4733 +
4734 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4735 +
4736 +define AuConf
4737 +ifdef ${1}
4738 +AuConfStr += ${1}=${${1}}
4739 +endif
4740 +endef
4741 +
4742 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4743 +       SBILIST \
4744 +       HNOTIFY HFSNOTIFY \
4745 +       EXPORT INO_T_64 \
4746 +       XATTR \
4747 +       FHSM \
4748 +       RDU \
4749 +       DIRREN \
4750 +       SHWH \
4751 +       BR_RAMFS \
4752 +       BR_FUSE POLL \
4753 +       BR_HFSPLUS \
4754 +       BDEV_LOOP \
4755 +       DEBUG MAGIC_SYSRQ
4756 +$(foreach i, ${AuConfAll}, \
4757 +       $(eval $(call AuConf,CONFIG_AUFS_${i})))
4758 +
4759 +AuConfName = ${obj}/conf.str
4760 +${AuConfName}.tmp: FORCE
4761 +       @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4762 +${AuConfName}: ${AuConfName}.tmp
4763 +       @diff -q $< $@ > /dev/null 2>&1 || { \
4764 +       echo '  GEN    ' $@; \
4765 +       cp -p $< $@; \
4766 +       }
4767 +FORCE:
4768 +clean-files += ${AuConfName} ${AuConfName}.tmp
4769 +${obj}/sysfs.o: ${AuConfName}
4770 +
4771 +-include ${srctree}/${src}/conf_priv.mk
4772 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4773 --- /usr/share/empty/fs/aufs/cpup.c     1970-01-01 01:00:00.000000000 +0100
4774 +++ linux/fs/aufs/cpup.c        2023-02-20 21:05:51.959693785 +0100
4775 @@ -0,0 +1,1459 @@
4776 +// SPDX-License-Identifier: GPL-2.0
4777 +/*
4778 + * Copyright (C) 2005-2022 Junjiro R. Okajima
4779 + *
4780 + * This program is free software; you can redistribute it and/or modify
4781 + * it under the terms of the GNU General Public License as published by
4782 + * the Free Software Foundation; either version 2 of the License, or
4783 + * (at your option) any later version.
4784 + *
4785 + * This program is distributed in the hope that it will be useful,
4786 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4787 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
4788 + * GNU General Public License for more details.
4789 + *
4790 + * You should have received a copy of the GNU General Public License
4791 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
4792 + */
4793 +
4794 +/*
4795 + * copy-up functions, see wbr_policy.c for copy-down
4796 + */
4797 +
4798 +#include <linux/fs_stack.h>
4799 +#include <linux/mm.h>
4800 +#include <linux/task_work.h>
4801 +#include "aufs.h"
4802 +
4803 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4804 +{
4805 +       const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4806 +               | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4807 +
4808 +       BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4809 +
4810 +       dst->i_flags |= iflags & ~mask;
4811 +       if (au_test_fs_notime(dst->i_sb))
4812 +               dst->i_flags |= S_NOATIME | S_NOCMTIME;
4813 +}
4814 +
4815 +void au_cpup_attr_timesizes(struct inode *inode)
4816 +{
4817 +       struct inode *h_inode;
4818 +
4819 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
4820 +       fsstack_copy_attr_times(inode, h_inode);
4821 +       fsstack_copy_inode_size(inode, h_inode);
4822 +}
4823 +
4824 +void au_cpup_attr_nlink(struct inode *inode, int force)
4825 +{
4826 +       struct inode *h_inode;
4827 +       struct super_block *sb;
4828 +       aufs_bindex_t bindex, bbot;
4829 +
4830 +       sb = inode->i_sb;
4831 +       bindex = au_ibtop(inode);
4832 +       h_inode = au_h_iptr(inode, bindex);
4833 +       if (!force
4834 +           && !S_ISDIR(h_inode->i_mode)
4835 +           && au_opt_test(au_mntflags(sb), PLINK)
4836 +           && au_plink_test(inode))
4837 +               return;
4838 +
4839 +       /*
4840 +        * 0 can happen in revalidating.
4841 +        * h_inode->i_mutex may not be held here, but it is harmless since once
4842 +        * i_nlink reaches 0, it will never become positive except O_TMPFILE
4843 +        * case.
4844 +        * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4845 +        *       the incorrect link count.
4846 +        */
4847 +       set_nlink(inode, h_inode->i_nlink);
4848 +
4849 +       /*
4850 +        * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4851 +        * it may includes whplink directory.
4852 +        */
4853 +       if (S_ISDIR(h_inode->i_mode)) {
4854 +               bbot = au_ibbot(inode);
4855 +               for (bindex++; bindex <= bbot; bindex++) {
4856 +                       h_inode = au_h_iptr(inode, bindex);
4857 +                       if (h_inode)
4858 +                               au_add_nlink(inode, h_inode);
4859 +               }
4860 +       }
4861 +}
4862 +
4863 +void au_cpup_attr_changeable(struct inode *inode)
4864 +{
4865 +       struct inode *h_inode;
4866 +
4867 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
4868 +       inode->i_mode = h_inode->i_mode;
4869 +       inode->i_uid = h_inode->i_uid;
4870 +       inode->i_gid = h_inode->i_gid;
4871 +       au_cpup_attr_timesizes(inode);
4872 +       au_cpup_attr_flags(inode, h_inode->i_flags);
4873 +}
4874 +
4875 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4876 +{
4877 +       struct au_iinfo *iinfo = au_ii(inode);
4878 +
4879 +       IiMustWriteLock(inode);
4880 +
4881 +       iinfo->ii_higen = h_inode->i_generation;
4882 +       iinfo->ii_hsb1 = h_inode->i_sb;
4883 +}
4884 +
4885 +void au_cpup_attr_all(struct inode *inode, int force)
4886 +{
4887 +       struct inode *h_inode;
4888 +
4889 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
4890 +       au_cpup_attr_changeable(inode);
4891 +       if (inode->i_nlink > 0)
4892 +               au_cpup_attr_nlink(inode, force);
4893 +       inode->i_rdev = h_inode->i_rdev;
4894 +       inode->i_blkbits = h_inode->i_blkbits;
4895 +       au_cpup_igen(inode, h_inode);
4896 +}
4897 +
4898 +/* ---------------------------------------------------------------------- */
4899 +
4900 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4901 +
4902 +/* keep the timestamps of the parent dir when cpup */
4903 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4904 +                   struct path *h_path)
4905 +{
4906 +       struct inode *h_inode;
4907 +
4908 +       dt->dt_dentry = dentry;
4909 +       dt->dt_h_path = *h_path;
4910 +       h_inode = d_inode(h_path->dentry);
4911 +       dt->dt_atime = h_inode->i_atime;
4912 +       dt->dt_mtime = h_inode->i_mtime;
4913 +       /* smp_mb(); */
4914 +}
4915 +
4916 +void au_dtime_revert(struct au_dtime *dt)
4917 +{
4918 +       struct iattr attr;
4919 +       int err;
4920 +
4921 +       attr.ia_atime = dt->dt_atime;
4922 +       attr.ia_mtime = dt->dt_mtime;
4923 +       attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4924 +               | ATTR_ATIME | ATTR_ATIME_SET;
4925 +
4926 +       /* no delegation since this is a directory */
4927 +       err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4928 +       if (unlikely(err))
4929 +               pr_warn("restoring timestamps failed(%d). ignored\n", err);
4930 +}
4931 +
4932 +/* ---------------------------------------------------------------------- */
4933 +
4934 +/* internal use only */
4935 +struct au_cpup_reg_attr {
4936 +       int             valid;
4937 +       struct kstat    st;
4938 +       unsigned int    iflags; /* inode->i_flags */
4939 +};
4940 +
4941 +static noinline_for_stack
4942 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct path *h_src,
4943 +              struct au_cpup_reg_attr *h_src_attr)
4944 +{
4945 +       int err, sbits, icex;
4946 +       unsigned int mnt_flags;
4947 +       unsigned char verbose;
4948 +       struct iattr ia;
4949 +       struct path h_path;
4950 +       struct inode *h_isrc, *h_idst;
4951 +       struct kstat *h_st;
4952 +       struct au_branch *br;
4953 +
4954 +       br = au_sbr(dst->d_sb, bindex);
4955 +       h_path.mnt = au_br_mnt(br);
4956 +       h_path.dentry = au_h_dptr(dst, bindex);
4957 +       h_idst = d_inode(h_path.dentry);
4958 +       h_isrc = d_inode(h_src->dentry);
4959 +       ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4960 +               | ATTR_ATIME | ATTR_MTIME
4961 +               | ATTR_ATIME_SET | ATTR_MTIME_SET;
4962 +       if (h_src_attr && h_src_attr->valid) {
4963 +               h_st = &h_src_attr->st;
4964 +               ia.ia_uid = h_st->uid;
4965 +               ia.ia_gid = h_st->gid;
4966 +               ia.ia_atime = h_st->atime;
4967 +               ia.ia_mtime = h_st->mtime;
4968 +               if (h_idst->i_mode != h_st->mode
4969 +                   && !S_ISLNK(h_idst->i_mode)) {
4970 +                       ia.ia_valid |= ATTR_MODE;
4971 +                       ia.ia_mode = h_st->mode;
4972 +               }
4973 +               sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4974 +               au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4975 +       } else {
4976 +               ia.ia_uid = h_isrc->i_uid;
4977 +               ia.ia_gid = h_isrc->i_gid;
4978 +               ia.ia_atime = h_isrc->i_atime;
4979 +               ia.ia_mtime = h_isrc->i_mtime;
4980 +               if (h_idst->i_mode != h_isrc->i_mode
4981 +                   && !S_ISLNK(h_idst->i_mode)) {
4982 +                       ia.ia_valid |= ATTR_MODE;
4983 +                       ia.ia_mode = h_isrc->i_mode;
4984 +               }
4985 +               sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4986 +               au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4987 +       }
4988 +       /* no delegation since it is just created */
4989 +       err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4990 +
4991 +       /* is this nfs only? */
4992 +       if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4993 +               ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4994 +               ia.ia_mode = h_isrc->i_mode;
4995 +               err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4996 +       }
4997 +
4998 +       icex = br->br_perm & AuBrAttr_ICEX;
4999 +       if (!err) {
5000 +               mnt_flags = au_mntflags(dst->d_sb);
5001 +               verbose = !!au_opt_test(mnt_flags, VERBOSE);
5002 +               err = au_cpup_xattr(&h_path, h_src, icex, verbose);
5003 +       }
5004 +
5005 +       return err;
5006 +}
5007 +
5008 +/* ---------------------------------------------------------------------- */
5009 +
5010 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
5011 +                          char *buf, unsigned long blksize)
5012 +{
5013 +       int err;
5014 +       size_t sz, rbytes, wbytes;
5015 +       unsigned char all_zero;
5016 +       char *p, *zp;
5017 +       struct inode *h_inode;
5018 +       /* reduce stack usage */
5019 +       struct iattr *ia;
5020 +
5021 +       zp = page_address(ZERO_PAGE(0));
5022 +       if (unlikely(!zp))
5023 +               return -ENOMEM; /* possible? */
5024 +
5025 +       err = 0;
5026 +       all_zero = 0;
5027 +       while (len) {
5028 +               AuDbg("len %lld\n", len);
5029 +               sz = blksize;
5030 +               if (len < blksize)
5031 +                       sz = len;
5032 +
5033 +               rbytes = 0;
5034 +               /* todo: signal_pending? */
5035 +               while (!rbytes || err == -EAGAIN || err == -EINTR) {
5036 +                       rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
5037 +                       err = rbytes;
5038 +               }
5039 +               if (unlikely(err < 0))
5040 +                       break;
5041 +
5042 +               all_zero = 0;
5043 +               if (len >= rbytes && rbytes == blksize)
5044 +                       all_zero = !memcmp(buf, zp, rbytes);
5045 +               if (!all_zero) {
5046 +                       wbytes = rbytes;
5047 +                       p = buf;
5048 +                       while (wbytes) {
5049 +                               size_t b;
5050 +
5051 +                               b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
5052 +                               err = b;
5053 +                               /* todo: signal_pending? */
5054 +                               if (unlikely(err == -EAGAIN || err == -EINTR))
5055 +                                       continue;
5056 +                               if (unlikely(err < 0))
5057 +                                       break;
5058 +                               wbytes -= b;
5059 +                               p += b;
5060 +                       }
5061 +                       if (unlikely(err < 0))
5062 +                               break;
5063 +               } else {
5064 +                       loff_t res;
5065 +
5066 +                       AuLabel(hole);
5067 +                       res = vfsub_llseek(dst, rbytes, SEEK_CUR);
5068 +                       err = res;
5069 +                       if (unlikely(res < 0))
5070 +                               break;
5071 +               }
5072 +               len -= rbytes;
5073 +               err = 0;
5074 +       }
5075 +
5076 +       /* the last block may be a hole */
5077 +       if (!err && all_zero) {
5078 +               AuLabel(last hole);
5079 +
5080 +               err = 1;
5081 +               if (au_test_nfs(dst->f_path.dentry->d_sb)) {
5082 +                       /* nfs requires this step to make last hole */
5083 +                       /* is this only nfs? */
5084 +                       do {
5085 +                               /* todo: signal_pending? */
5086 +                               err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
5087 +                       } while (err == -EAGAIN || err == -EINTR);
5088 +                       if (err == 1)
5089 +                               dst->f_pos--;
5090 +               }
5091 +
5092 +               if (err == 1) {
5093 +                       ia = (void *)buf;
5094 +                       ia->ia_size = dst->f_pos;
5095 +                       ia->ia_valid = ATTR_SIZE | ATTR_FILE;
5096 +                       ia->ia_file = dst;
5097 +                       h_inode = file_inode(dst);
5098 +                       inode_lock_nested(h_inode, AuLsc_I_CHILD2);
5099 +                       /* no delegation since it is just created */
5100 +                       err = vfsub_notify_change(&dst->f_path, ia,
5101 +                                                 /*delegated*/NULL);
5102 +                       inode_unlock(h_inode);
5103 +               }
5104 +       }
5105 +
5106 +       return err;
5107 +}
5108 +
5109 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
5110 +{
5111 +       int err;
5112 +       unsigned long blksize;
5113 +       unsigned char do_kfree;
5114 +       char *buf;
5115 +       struct super_block *h_sb;
5116 +
5117 +       err = -ENOMEM;
5118 +       h_sb = file_inode(dst)->i_sb;
5119 +       blksize = h_sb->s_blocksize;
5120 +       if (!blksize || PAGE_SIZE < blksize)
5121 +               blksize = PAGE_SIZE;
5122 +       AuDbg("blksize %lu\n", blksize);
5123 +       do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
5124 +       if (do_kfree)
5125 +               buf = kmalloc(blksize, GFP_NOFS);
5126 +       else
5127 +               buf = (void *)__get_free_page(GFP_NOFS);
5128 +       if (unlikely(!buf))
5129 +               goto out;
5130 +
5131 +       if (len > (1 << 22))
5132 +               AuDbg("copying a large file %lld\n", (long long)len);
5133 +
5134 +       src->f_pos = 0;
5135 +       dst->f_pos = 0;
5136 +       err = au_do_copy_file(dst, src, len, buf, blksize);
5137 +       if (do_kfree) {
5138 +               AuDebugOn(!au_kfree_do_sz_test(blksize));
5139 +               au_kfree_do_rcu(buf);
5140 +       } else
5141 +               free_page((unsigned long)buf);
5142 +
5143 +out:
5144 +       return err;
5145 +}
5146 +
5147 +static int au_do_copy(struct file *dst, struct file *src, loff_t len)
5148 +{
5149 +       int err;
5150 +       struct super_block *h_src_sb;
5151 +       struct inode *h_src_inode;
5152 +
5153 +       h_src_inode = file_inode(src);
5154 +       h_src_sb = h_src_inode->i_sb;
5155 +
5156 +       /* XFS acquires inode_lock */
5157 +       if (!au_test_xfs(h_src_sb))
5158 +               err = au_copy_file(dst, src, len);
5159 +       else {
5160 +               inode_unlock_shared(h_src_inode);
5161 +               err = au_copy_file(dst, src, len);
5162 +               inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5163 +       }
5164 +
5165 +       return err;
5166 +}
5167 +
5168 +static int au_clone_or_copy(struct file *dst, struct file *src, loff_t len)
5169 +{
5170 +       int err;
5171 +       loff_t lo;
5172 +       struct super_block *h_src_sb;
5173 +       struct inode *h_src_inode;
5174 +
5175 +       h_src_inode = file_inode(src);
5176 +       h_src_sb = h_src_inode->i_sb;
5177 +       if (h_src_sb != file_inode(dst)->i_sb
5178 +           || !dst->f_op->remap_file_range) {
5179 +               err = au_do_copy(dst, src, len);
5180 +               goto out;
5181 +       }
5182 +
5183 +       if (!au_test_nfs(h_src_sb)) {
5184 +               inode_unlock_shared(h_src_inode);
5185 +               lo = vfsub_clone_file_range(src, dst, len);
5186 +               inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5187 +       } else
5188 +               lo = vfsub_clone_file_range(src, dst, len);
5189 +       if (lo == len) {
5190 +               err = 0;
5191 +               goto out; /* success */
5192 +       } else if (lo >= 0)
5193 +               /* todo: possible? */
5194 +               /* paritially succeeded */
5195 +               AuDbg("lo %lld, len %lld. Retrying.\n", lo, len);
5196 +       else if (lo != -EOPNOTSUPP) {
5197 +               /* older XFS has a condition in cloning */
5198 +               err = lo;
5199 +               goto out;
5200 +       }
5201 +
5202 +       /* the backend fs on NFS may not support cloning */
5203 +       err = au_do_copy(dst, src, len);
5204 +
5205 +out:
5206 +       AuTraceErr(err);
5207 +       return err;
5208 +}
5209 +
5210 +/*
5211 + * to support a sparse file which is opened with O_APPEND,
5212 + * we need to close the file.
5213 + */
5214 +static int au_cp_regular(struct au_cp_generic *cpg)
5215 +{
5216 +       int err, i;
5217 +       enum { SRC, DST };
5218 +       struct {
5219 +               aufs_bindex_t bindex;
5220 +               unsigned int flags;
5221 +               struct dentry *dentry;
5222 +               int force_wr;
5223 +               struct file *file;
5224 +       } *f, file[] = {
5225 +               {
5226 +                       .bindex = cpg->bsrc,
5227 +                       .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
5228 +               },
5229 +               {
5230 +                       .bindex = cpg->bdst,
5231 +                       .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
5232 +                       .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
5233 +               }
5234 +       };
5235 +       struct au_branch *br;
5236 +       struct super_block *sb, *h_src_sb;
5237 +       struct inode *h_src_inode;
5238 +       struct task_struct *tsk = current;
5239 +
5240 +       /* bsrc branch can be ro/rw. */
5241 +       sb = cpg->dentry->d_sb;
5242 +       f = file;
5243 +       for (i = 0; i < 2; i++, f++) {
5244 +               f->dentry = au_h_dptr(cpg->dentry, f->bindex);
5245 +               f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
5246 +                                   /*file*/NULL, f->force_wr);
5247 +               if (IS_ERR(f->file)) {
5248 +                       err = PTR_ERR(f->file);
5249 +                       if (i == SRC)
5250 +                               goto out;
5251 +                       else
5252 +                               goto out_src;
5253 +               }
5254 +       }
5255 +
5256 +       /* try stopping to update while we copyup */
5257 +       h_src_inode = d_inode(file[SRC].dentry);
5258 +       h_src_sb = h_src_inode->i_sb;
5259 +       if (!au_test_nfs(h_src_sb))
5260 +               IMustLock(h_src_inode);
5261 +       err = au_clone_or_copy(file[DST].file, file[SRC].file, cpg->len);
5262 +
5263 +       /* i wonder if we had O_NO_DELAY_FPUT flag */
5264 +       if (tsk->flags & PF_KTHREAD)
5265 +               __fput_sync(file[DST].file);
5266 +       else {
5267 +               /* it happened actually */
5268 +               fput(file[DST].file);
5269 +               /*
5270 +                * too bad.
5271 +                * we have to call both since we don't know which place the file
5272 +                * was added to.
5273 +                */
5274 +               task_work_run();
5275 +               flush_delayed_fput();
5276 +       }
5277 +       br = au_sbr(sb, file[DST].bindex);
5278 +       au_lcnt_dec(&br->br_nfiles);
5279 +
5280 +out_src:
5281 +       fput(file[SRC].file);
5282 +       br = au_sbr(sb, file[SRC].bindex);
5283 +       au_lcnt_dec(&br->br_nfiles);
5284 +out:
5285 +       return err;
5286 +}
5287 +
5288 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
5289 +                             struct au_cpup_reg_attr *h_src_attr)
5290 +{
5291 +       int err, rerr;
5292 +       loff_t l;
5293 +       struct path h_path;
5294 +       struct inode *h_src_inode, *h_dst_inode;
5295 +
5296 +       err = 0;
5297 +       h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
5298 +       l = i_size_read(h_src_inode);
5299 +       if (cpg->len == -1 || l < cpg->len)
5300 +               cpg->len = l;
5301 +       if (cpg->len) {
5302 +               /* try stopping to update while we are referencing */
5303 +               inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5304 +               au_pin_hdir_unlock(cpg->pin);
5305 +
5306 +               h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5307 +               h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
5308 +               h_src_attr->iflags = h_src_inode->i_flags;
5309 +               if (!au_test_nfs(h_src_inode->i_sb))
5310 +                       err = vfsub_getattr(&h_path, &h_src_attr->st);
5311 +               else {
5312 +                       inode_unlock_shared(h_src_inode);
5313 +                       err = vfsub_getattr(&h_path, &h_src_attr->st);
5314 +                       inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5315 +               }
5316 +               if (unlikely(err)) {
5317 +                       inode_unlock_shared(h_src_inode);
5318 +                       goto out;
5319 +               }
5320 +               h_src_attr->valid = 1;
5321 +               if (!au_test_nfs(h_src_inode->i_sb)) {
5322 +                       err = au_cp_regular(cpg);
5323 +                       inode_unlock_shared(h_src_inode);
5324 +               } else {
5325 +                       inode_unlock_shared(h_src_inode);
5326 +                       err = au_cp_regular(cpg);
5327 +               }
5328 +               rerr = au_pin_hdir_relock(cpg->pin);
5329 +               if (!err && rerr)
5330 +                       err = rerr;
5331 +       }
5332 +       if (!err && (h_src_inode->i_state & I_LINKABLE)) {
5333 +               h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5334 +               h_dst_inode = d_inode(h_path.dentry);
5335 +               spin_lock(&h_dst_inode->i_lock);
5336 +               h_dst_inode->i_state |= I_LINKABLE;
5337 +               spin_unlock(&h_dst_inode->i_lock);
5338 +       }
5339 +
5340 +out:
5341 +       return err;
5342 +}
5343 +
5344 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
5345 +                             struct inode *h_dir)
5346 +{
5347 +       int err;
5348 +       DEFINE_DELAYED_CALL(done);
5349 +       const char *sym;
5350 +
5351 +       sym = vfs_get_link(h_src, &done);
5352 +       err = PTR_ERR(sym);
5353 +       if (IS_ERR(sym))
5354 +               goto out;
5355 +
5356 +       err = vfsub_symlink(h_dir, h_path, sym);
5357 +
5358 +out:
5359 +       do_delayed_call(&done);
5360 +       return err;
5361 +}
5362 +
5363 +/*
5364 + * regardless 'acl' option, reset all ACL.
5365 + * All ACL will be copied up later from the original entry on the lower branch.
5366 + */
5367 +static int au_reset_acl(struct path *h_path, umode_t mode)
5368 +{
5369 +       int err;
5370 +       struct dentry *h_dentry;
5371 +       /* struct inode *h_inode; */
5372 +       struct user_namespace *h_userns;
5373 +
5374 +       h_userns = mnt_user_ns(h_path->mnt);
5375 +       h_dentry = h_path->dentry;
5376 +       /* h_inode = d_inode(h_dentry); */
5377 +       /* forget_all_cached_acls(h_inode)); */
5378 +       err = vfsub_remove_acl(h_userns, h_dentry, XATTR_NAME_POSIX_ACL_ACCESS);
5379 +       AuTraceErr(err);
5380 +       if (err == -EOPNOTSUPP)
5381 +               err = 0;
5382 +       if (!err)
5383 +               err = vfsub_acl_chmod(h_userns, h_dentry, mode);
5384 +
5385 +       AuTraceErr(err);
5386 +       return err;
5387 +}
5388 +
5389 +static int au_do_cpup_dir(struct au_cp_generic *cpg, struct dentry *dst_parent,
5390 +                         struct inode *h_dir, struct path *h_path)
5391 +{
5392 +       int err;
5393 +       struct inode *dir, *inode;
5394 +       struct user_namespace *h_userns;
5395 +
5396 +       h_userns = mnt_user_ns(h_path->mnt);
5397 +       err = vfsub_remove_acl(h_userns, h_path->dentry,
5398 +                              XATTR_NAME_POSIX_ACL_DEFAULT);
5399 +       AuTraceErr(err);
5400 +       if (err == -EOPNOTSUPP)
5401 +               err = 0;
5402 +       if (unlikely(err))
5403 +               goto out;
5404 +
5405 +       /*
5406 +        * strange behaviour from the users view,
5407 +        * particularly setattr case
5408 +        */
5409 +       dir = d_inode(dst_parent);
5410 +       if (au_ibtop(dir) == cpg->bdst)
5411 +               au_cpup_attr_nlink(dir, /*force*/1);
5412 +       inode = d_inode(cpg->dentry);
5413 +       au_cpup_attr_nlink(inode, /*force*/1);
5414 +
5415 +out:
5416 +       return err;
5417 +}
5418 +
5419 +static noinline_for_stack
5420 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5421 +              struct au_cpup_reg_attr *h_src_attr)
5422 +{
5423 +       int err;
5424 +       umode_t mode;
5425 +       unsigned int mnt_flags;
5426 +       unsigned char isdir, isreg, force;
5427 +       const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5428 +       struct au_dtime dt;
5429 +       struct path h_path;
5430 +       struct dentry *h_src, *h_dst, *h_parent;
5431 +       struct inode *h_inode, *h_dir;
5432 +       struct super_block *sb;
5433 +
5434 +       /* bsrc branch can be ro/rw. */
5435 +       h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5436 +       h_inode = d_inode(h_src);
5437 +       AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5438 +
5439 +       /* try stopping to be referenced while we are creating */
5440 +       h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5441 +       if (au_ftest_cpup(cpg->flags, RENAME))
5442 +               AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5443 +                                 AUFS_WH_PFX_LEN));
5444 +       h_parent = h_dst->d_parent; /* dir inode is locked */
5445 +       h_dir = d_inode(h_parent);
5446 +       IMustLock(h_dir);
5447 +       AuDebugOn(h_parent != h_dst->d_parent);
5448 +
5449 +       sb = cpg->dentry->d_sb;
5450 +       h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5451 +       if (do_dt) {
5452 +               h_path.dentry = h_parent;
5453 +               au_dtime_store(&dt, dst_parent, &h_path);
5454 +       }
5455 +       h_path.dentry = h_dst;
5456 +
5457 +       isreg = 0;
5458 +       isdir = 0;
5459 +       mode = h_inode->i_mode;
5460 +       switch (mode & S_IFMT) {
5461 +       case S_IFREG:
5462 +               isreg = 1;
5463 +               err = vfsub_create(h_dir, &h_path, 0600, /*want_excl*/true);
5464 +               if (!err)
5465 +                       err = au_do_cpup_regular(cpg, h_src_attr);
5466 +               break;
5467 +       case S_IFDIR:
5468 +               isdir = 1;
5469 +               err = vfsub_mkdir(h_dir, &h_path, mode);
5470 +               if (!err)
5471 +                       err = au_do_cpup_dir(cpg, dst_parent, h_dir, &h_path);
5472 +               break;
5473 +       case S_IFLNK:
5474 +               err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5475 +               break;
5476 +       case S_IFCHR:
5477 +       case S_IFBLK:
5478 +               AuDebugOn(!capable(CAP_MKNOD));
5479 +               fallthrough;
5480 +       case S_IFIFO:
5481 +       case S_IFSOCK:
5482 +               err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5483 +               break;
5484 +       default:
5485 +               AuIOErr("Unknown inode type 0%o\n", mode);
5486 +               err = -EIO;
5487 +       }
5488 +       if (!err)
5489 +               err = au_reset_acl(&h_path, mode);
5490 +
5491 +       mnt_flags = au_mntflags(sb);
5492 +       if (!au_opt_test(mnt_flags, UDBA_NONE)
5493 +           && !isdir
5494 +           && au_opt_test(mnt_flags, XINO)
5495 +           && (h_inode->i_nlink == 1
5496 +               || (h_inode->i_state & I_LINKABLE))
5497 +           /* todo: unnecessary? */
5498 +           /* && d_inode(cpg->dentry)->i_nlink == 1 */
5499 +           && cpg->bdst < cpg->bsrc
5500 +           && !au_ftest_cpup(cpg->flags, KEEPLINO))
5501 +               au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5502 +               /* ignore this error */
5503 +
5504 +       if (!err) {
5505 +               force = 0;
5506 +               if (isreg) {
5507 +                       force = !!cpg->len;
5508 +                       if (cpg->len == -1)
5509 +                               force = !!i_size_read(h_inode);
5510 +               }
5511 +               au_fhsm_wrote(sb, cpg->bdst, force);
5512 +       }
5513 +
5514 +       if (do_dt)
5515 +               au_dtime_revert(&dt);
5516 +       return err;
5517 +}
5518 +
5519 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5520 +{
5521 +       int err;
5522 +       struct dentry *dentry, *h_dentry, *h_parent, *parent;
5523 +       struct path h_ppath;
5524 +       struct inode *h_dir;
5525 +       aufs_bindex_t bdst;
5526 +
5527 +       dentry = cpg->dentry;
5528 +       bdst = cpg->bdst;
5529 +       h_ppath.mnt = au_sbr_mnt(dentry->d_sb, bdst);
5530 +       h_dentry = au_h_dptr(dentry, bdst);
5531 +       if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5532 +               dget(h_dentry);
5533 +               au_set_h_dptr(dentry, bdst, NULL);
5534 +               err = au_lkup_neg(dentry, bdst, /*wh*/0);
5535 +               if (!err)
5536 +                       h_path->dentry = dget(au_h_dptr(dentry, bdst));
5537 +               au_set_h_dptr(dentry, bdst, h_dentry);
5538 +       } else {
5539 +               err = 0;
5540 +               parent = dget_parent(dentry);
5541 +               h_ppath.dentry = au_h_dptr(parent, bdst);
5542 +               dput(parent);
5543 +               h_path->dentry = vfsub_lkup_one(&dentry->d_name, &h_ppath);
5544 +               if (IS_ERR(h_path->dentry))
5545 +                       err = PTR_ERR(h_path->dentry);
5546 +       }
5547 +       if (unlikely(err))
5548 +               goto out;
5549 +
5550 +       h_parent = h_dentry->d_parent; /* dir inode is locked */
5551 +       h_dir = d_inode(h_parent);
5552 +       IMustLock(h_dir);
5553 +       AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5554 +       /* no delegation since it is just created */
5555 +       err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL,
5556 +                          /*flags*/0);
5557 +       dput(h_path->dentry);
5558 +
5559 +out:
5560 +       return err;
5561 +}
5562 +
5563 +/*
5564 + * copyup the @dentry from @bsrc to @bdst.
5565 + * the caller must set the both of lower dentries.
5566 + * @len is for truncating when it is -1 copyup the entire file.
5567 + * in link/rename cases, @dst_parent may be different from the real one.
5568 + * basic->bsrc can be larger than basic->bdst.
5569 + * aufs doesn't touch the credential so
5570 + * security_inode_copy_up{,_xattr}() are unnecessary.
5571 + */
5572 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5573 +{
5574 +       int err, rerr;
5575 +       aufs_bindex_t old_ibtop;
5576 +       unsigned char isdir, plink;
5577 +       struct dentry *h_src, *h_dst, *h_parent;
5578 +       struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5579 +       struct super_block *sb;
5580 +       struct au_branch *br;
5581 +       struct path h_src_path;
5582 +       /* to reduce stack size */
5583 +       struct {
5584 +               struct au_dtime dt;
5585 +               struct path h_path;
5586 +               struct au_cpup_reg_attr h_src_attr;
5587 +       } *a;
5588 +
5589 +       err = -ENOMEM;
5590 +       a = kmalloc(sizeof(*a), GFP_NOFS);
5591 +       if (unlikely(!a))
5592 +               goto out;
5593 +       a->h_src_attr.valid = 0;
5594 +
5595 +       sb = cpg->dentry->d_sb;
5596 +       br = au_sbr(sb, cpg->bdst);
5597 +       a->h_path.mnt = au_br_mnt(br);
5598 +       h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5599 +       h_parent = h_dst->d_parent; /* dir inode is locked */
5600 +       h_dir = d_inode(h_parent);
5601 +       IMustLock(h_dir);
5602 +
5603 +       h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5604 +       inode = d_inode(cpg->dentry);
5605 +
5606 +       if (!dst_parent)
5607 +               dst_parent = dget_parent(cpg->dentry);
5608 +       else
5609 +               dget(dst_parent);
5610 +
5611 +       plink = !!au_opt_test(au_mntflags(sb), PLINK);
5612 +       dst_inode = au_h_iptr(inode, cpg->bdst);
5613 +       if (dst_inode) {
5614 +               if (unlikely(!plink)) {
5615 +                       err = -EIO;
5616 +                       AuIOErr("hi%lu(i%lu) exists on b%d "
5617 +                               "but plink is disabled\n",
5618 +                               dst_inode->i_ino, inode->i_ino, cpg->bdst);
5619 +                       goto out_parent;
5620 +               }
5621 +
5622 +               if (dst_inode->i_nlink) {
5623 +                       const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5624 +
5625 +                       h_src = au_plink_lkup(inode, cpg->bdst);
5626 +                       err = PTR_ERR(h_src);
5627 +                       if (IS_ERR(h_src))
5628 +                               goto out_parent;
5629 +                       if (unlikely(d_is_negative(h_src))) {
5630 +                               err = -EIO;
5631 +                               AuIOErr("i%lu exists on b%d "
5632 +                                       "but not pseudo-linked\n",
5633 +                                       inode->i_ino, cpg->bdst);
5634 +                               dput(h_src);
5635 +                               goto out_parent;
5636 +                       }
5637 +
5638 +                       if (do_dt) {
5639 +                               a->h_path.dentry = h_parent;
5640 +                               au_dtime_store(&a->dt, dst_parent, &a->h_path);
5641 +                       }
5642 +
5643 +                       a->h_path.dentry = h_dst;
5644 +                       delegated = NULL;
5645 +                       err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5646 +                       if (!err && au_ftest_cpup(cpg->flags, RENAME))
5647 +                               err = au_do_ren_after_cpup(cpg, &a->h_path);
5648 +                       if (do_dt)
5649 +                               au_dtime_revert(&a->dt);
5650 +                       if (unlikely(err == -EWOULDBLOCK)) {
5651 +                               pr_warn("cannot retry for NFSv4 delegation"
5652 +                                       " for an internal link\n");
5653 +                               iput(delegated);
5654 +                       }
5655 +                       dput(h_src);
5656 +                       goto out_parent;
5657 +               } else
5658 +                       /* todo: cpup_wh_file? */
5659 +                       /* udba work */
5660 +                       au_update_ibrange(inode, /*do_put_zero*/1);
5661 +       }
5662 +
5663 +       isdir = S_ISDIR(inode->i_mode);
5664 +       old_ibtop = au_ibtop(inode);
5665 +       err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5666 +       if (unlikely(err))
5667 +               goto out_rev;
5668 +       dst_inode = d_inode(h_dst);
5669 +       inode_lock_nested(dst_inode, AuLsc_I_CHILD2);
5670 +       /* todo: necessary? */
5671 +       /* au_pin_hdir_unlock(cpg->pin); */
5672 +
5673 +       h_src_path.dentry = h_src;
5674 +       h_src_path.mnt = au_sbr_mnt(sb, cpg->bsrc);
5675 +       err = cpup_iattr(cpg->dentry, cpg->bdst, &h_src_path, &a->h_src_attr);
5676 +       if (unlikely(err)) {
5677 +               /* todo: necessary? */
5678 +               /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5679 +               inode_unlock(dst_inode);
5680 +               goto out_rev;
5681 +       }
5682 +
5683 +       if (cpg->bdst < old_ibtop) {
5684 +               if (S_ISREG(inode->i_mode)) {
5685 +                       err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5686 +                       if (unlikely(err)) {
5687 +                               /* ignore an error */
5688 +                               /* au_pin_hdir_relock(cpg->pin); */
5689 +                               inode_unlock(dst_inode);
5690 +                               goto out_rev;
5691 +                       }
5692 +               }
5693 +               au_set_ibtop(inode, cpg->bdst);
5694 +       } else
5695 +               au_set_ibbot(inode, cpg->bdst);
5696 +       au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5697 +                     au_hi_flags(inode, isdir));
5698 +
5699 +       /* todo: necessary? */
5700 +       /* err = au_pin_hdir_relock(cpg->pin); */
5701 +       inode_unlock(dst_inode);
5702 +       if (unlikely(err))
5703 +               goto out_rev;
5704 +
5705 +       src_inode = d_inode(h_src);
5706 +       if (!isdir
5707 +           && (src_inode->i_nlink > 1
5708 +               || src_inode->i_state & I_LINKABLE)
5709 +           && plink)
5710 +               au_plink_append(inode, cpg->bdst, h_dst);
5711 +
5712 +       if (au_ftest_cpup(cpg->flags, RENAME)) {
5713 +               a->h_path.dentry = h_dst;
5714 +               err = au_do_ren_after_cpup(cpg, &a->h_path);
5715 +       }
5716 +       if (!err)
5717 +               goto out_parent; /* success */
5718 +
5719 +       /* revert */
5720 +out_rev:
5721 +       a->h_path.dentry = h_parent;
5722 +       au_dtime_store(&a->dt, dst_parent, &a->h_path);
5723 +       a->h_path.dentry = h_dst;
5724 +       rerr = 0;
5725 +       if (d_is_positive(h_dst)) {
5726 +               if (!isdir) {
5727 +                       /* no delegation since it is just created */
5728 +                       rerr = vfsub_unlink(h_dir, &a->h_path,
5729 +                                           /*delegated*/NULL, /*force*/0);
5730 +               } else
5731 +                       rerr = vfsub_rmdir(h_dir, &a->h_path);
5732 +       }
5733 +       au_dtime_revert(&a->dt);
5734 +       if (rerr) {
5735 +               AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5736 +               err = -EIO;
5737 +       }
5738 +out_parent:
5739 +       dput(dst_parent);
5740 +       au_kfree_rcu(a);
5741 +out:
5742 +       return err;
5743 +}
5744 +
5745 +#if 0 /* reserved */
5746 +struct au_cpup_single_args {
5747 +       int *errp;
5748 +       struct au_cp_generic *cpg;
5749 +       struct dentry *dst_parent;
5750 +};
5751 +
5752 +static void au_call_cpup_single(void *args)
5753 +{
5754 +       struct au_cpup_single_args *a = args;
5755 +
5756 +       au_pin_hdir_acquire_nest(a->cpg->pin);
5757 +       *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5758 +       au_pin_hdir_release(a->cpg->pin);
5759 +}
5760 +#endif
5761 +
5762 +/*
5763 + * prevent SIGXFSZ in copy-up.
5764 + * testing CAP_MKNOD is for generic fs,
5765 + * but CAP_FSETID is for xfs only, currently.
5766 + */
5767 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5768 +{
5769 +       int do_sio;
5770 +       struct super_block *sb;
5771 +       struct inode *h_dir;
5772 +
5773 +       do_sio = 0;
5774 +       sb = au_pinned_parent(pin)->d_sb;
5775 +       if (!au_wkq_test()
5776 +           && (!au_sbi(sb)->si_plink_maint_pid
5777 +               || au_plink_maint(sb, AuLock_NOPLM))) {
5778 +               switch (mode & S_IFMT) {
5779 +               case S_IFREG:
5780 +                       /* no condition about RLIMIT_FSIZE and the file size */
5781 +                       do_sio = 1;
5782 +                       break;
5783 +               case S_IFCHR:
5784 +               case S_IFBLK:
5785 +                       do_sio = !capable(CAP_MKNOD);
5786 +                       break;
5787 +               }
5788 +               if (!do_sio)
5789 +                       do_sio = ((mode & (S_ISUID | S_ISGID))
5790 +                                 && !capable(CAP_FSETID));
5791 +               /* this workaround may be removed in the future */
5792 +               if (!do_sio) {
5793 +                       h_dir = au_pinned_h_dir(pin);
5794 +                       do_sio = h_dir->i_mode & S_ISVTX;
5795 +               }
5796 +       }
5797 +
5798 +       return do_sio;
5799 +}
5800 +
5801 +#if 0 /* reserved */
5802 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5803 +{
5804 +       int err, wkq_err;
5805 +       struct dentry *h_dentry;
5806 +
5807 +       h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5808 +       if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5809 +               err = au_cpup_single(cpg, dst_parent);
5810 +       else {
5811 +               struct au_cpup_single_args args = {
5812 +                       .errp           = &err,
5813 +                       .cpg            = cpg,
5814 +                       .dst_parent     = dst_parent
5815 +               };
5816 +               wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5817 +               if (unlikely(wkq_err))
5818 +                       err = wkq_err;
5819 +       }
5820 +
5821 +       return err;
5822 +}
5823 +#endif
5824 +
5825 +/*
5826 + * copyup the @dentry from the first active lower branch to @bdst,
5827 + * using au_cpup_single().
5828 + */
5829 +static int au_cpup_simple(struct au_cp_generic *cpg)
5830 +{
5831 +       int err;
5832 +       unsigned int flags_orig;
5833 +       struct dentry *dentry;
5834 +
5835 +       AuDebugOn(cpg->bsrc < 0);
5836 +
5837 +       dentry = cpg->dentry;
5838 +       DiMustWriteLock(dentry);
5839 +
5840 +       err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5841 +       if (!err) {
5842 +               flags_orig = cpg->flags;
5843 +               au_fset_cpup(cpg->flags, RENAME);
5844 +               err = au_cpup_single(cpg, NULL);
5845 +               cpg->flags = flags_orig;
5846 +               if (!err)
5847 +                       return 0; /* success */
5848 +
5849 +               /* revert */
5850 +               au_set_h_dptr(dentry, cpg->bdst, NULL);
5851 +               au_set_dbtop(dentry, cpg->bsrc);
5852 +       }
5853 +
5854 +       return err;
5855 +}
5856 +
5857 +struct au_cpup_simple_args {
5858 +       int *errp;
5859 +       struct au_cp_generic *cpg;
5860 +};
5861 +
5862 +static void au_call_cpup_simple(void *args)
5863 +{
5864 +       struct au_cpup_simple_args *a = args;
5865 +
5866 +       au_pin_hdir_acquire_nest(a->cpg->pin);
5867 +       *a->errp = au_cpup_simple(a->cpg);
5868 +       au_pin_hdir_release(a->cpg->pin);
5869 +}
5870 +
5871 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5872 +{
5873 +       int err, wkq_err;
5874 +       struct dentry *dentry, *parent;
5875 +       struct file *h_file;
5876 +       struct inode *h_dir;
5877 +       struct user_namespace *h_userns;
5878 +
5879 +       dentry = cpg->dentry;
5880 +       h_file = NULL;
5881 +       if (au_ftest_cpup(cpg->flags, HOPEN)) {
5882 +               AuDebugOn(cpg->bsrc < 0);
5883 +               h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5884 +               err = PTR_ERR(h_file);
5885 +               if (IS_ERR(h_file))
5886 +                       goto out;
5887 +       }
5888 +
5889 +       parent = dget_parent(dentry);
5890 +       h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5891 +       h_userns = au_sbr_userns(dentry->d_sb, cpg->bdst);
5892 +       if (!au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC | MAY_WRITE)
5893 +           && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5894 +               err = au_cpup_simple(cpg);
5895 +       else {
5896 +               struct au_cpup_simple_args args = {
5897 +                       .errp           = &err,
5898 +                       .cpg            = cpg
5899 +               };
5900 +               wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5901 +               if (unlikely(wkq_err))
5902 +                       err = wkq_err;
5903 +       }
5904 +
5905 +       dput(parent);
5906 +       if (h_file)
5907 +               au_h_open_post(dentry, cpg->bsrc, h_file);
5908 +
5909 +out:
5910 +       return err;
5911 +}
5912 +
5913 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5914 +{
5915 +       aufs_bindex_t bsrc, bbot;
5916 +       struct dentry *dentry, *h_dentry;
5917 +
5918 +       if (cpg->bsrc < 0) {
5919 +               dentry = cpg->dentry;
5920 +               bbot = au_dbbot(dentry);
5921 +               for (bsrc = cpg->bdst + 1; bsrc <= bbot; bsrc++) {
5922 +                       h_dentry = au_h_dptr(dentry, bsrc);
5923 +                       if (h_dentry) {
5924 +                               AuDebugOn(d_is_negative(h_dentry));
5925 +                               break;
5926 +                       }
5927 +               }
5928 +               AuDebugOn(bsrc > bbot);
5929 +               cpg->bsrc = bsrc;
5930 +       }
5931 +       AuDebugOn(cpg->bsrc <= cpg->bdst);
5932 +       return au_do_sio_cpup_simple(cpg);
5933 +}
5934 +
5935 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5936 +{
5937 +       AuDebugOn(cpg->bdst <= cpg->bsrc);
5938 +       return au_do_sio_cpup_simple(cpg);
5939 +}
5940 +
5941 +/* ---------------------------------------------------------------------- */
5942 +
5943 +/*
5944 + * copyup the deleted file for writing.
5945 + */
5946 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5947 +                        struct file *file)
5948 +{
5949 +       int err;
5950 +       unsigned int flags_orig;
5951 +       aufs_bindex_t bsrc_orig;
5952 +       struct au_dinfo *dinfo;
5953 +       struct {
5954 +               struct au_hdentry *hd;
5955 +               struct dentry *h_dentry;
5956 +       } hdst, hsrc;
5957 +
5958 +       dinfo = au_di(cpg->dentry);
5959 +       AuRwMustWriteLock(&dinfo->di_rwsem);
5960 +
5961 +       bsrc_orig = cpg->bsrc;
5962 +       cpg->bsrc = dinfo->di_btop;
5963 +       hdst.hd = au_hdentry(dinfo, cpg->bdst);
5964 +       hdst.h_dentry = hdst.hd->hd_dentry;
5965 +       hdst.hd->hd_dentry = wh_dentry;
5966 +       dinfo->di_btop = cpg->bdst;
5967 +
5968 +       hsrc.h_dentry = NULL;
5969 +       if (file) {
5970 +               hsrc.hd = au_hdentry(dinfo, cpg->bsrc);
5971 +               hsrc.h_dentry = hsrc.hd->hd_dentry;
5972 +               hsrc.hd->hd_dentry = au_hf_top(file)->f_path.dentry;
5973 +       }
5974 +       flags_orig = cpg->flags;
5975 +       cpg->flags = !AuCpup_DTIME;
5976 +       err = au_cpup_single(cpg, /*h_parent*/NULL);
5977 +       cpg->flags = flags_orig;
5978 +       if (file) {
5979 +               if (!err)
5980 +                       err = au_reopen_nondir(file);
5981 +               hsrc.hd->hd_dentry = hsrc.h_dentry;
5982 +       }
5983 +       hdst.hd->hd_dentry = hdst.h_dentry;
5984 +       dinfo->di_btop = cpg->bsrc;
5985 +       cpg->bsrc = bsrc_orig;
5986 +
5987 +       return err;
5988 +}
5989 +
5990 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5991 +{
5992 +       int err;
5993 +       aufs_bindex_t bdst;
5994 +       struct au_dtime dt;
5995 +       struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5996 +       struct au_branch *br;
5997 +       struct path h_path;
5998 +
5999 +       dentry = cpg->dentry;
6000 +       bdst = cpg->bdst;
6001 +       br = au_sbr(dentry->d_sb, bdst);
6002 +       parent = dget_parent(dentry);
6003 +       h_parent = au_h_dptr(parent, bdst);
6004 +       wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
6005 +       err = PTR_ERR(wh_dentry);
6006 +       if (IS_ERR(wh_dentry))
6007 +               goto out;
6008 +
6009 +       h_path.dentry = h_parent;
6010 +       h_path.mnt = au_br_mnt(br);
6011 +       au_dtime_store(&dt, parent, &h_path);
6012 +       err = au_do_cpup_wh(cpg, wh_dentry, file);
6013 +       if (unlikely(err))
6014 +               goto out_wh;
6015 +
6016 +       dget(wh_dentry);
6017 +       h_path.dentry = wh_dentry;
6018 +       if (!d_is_dir(wh_dentry)) {
6019 +               /* no delegation since it is just created */
6020 +               err = vfsub_unlink(d_inode(h_parent), &h_path,
6021 +                                  /*delegated*/NULL, /*force*/0);
6022 +       } else
6023 +               err = vfsub_rmdir(d_inode(h_parent), &h_path);
6024 +       if (unlikely(err)) {
6025 +               AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
6026 +                       wh_dentry, err);
6027 +               err = -EIO;
6028 +       }
6029 +       au_dtime_revert(&dt);
6030 +       au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
6031 +
6032 +out_wh:
6033 +       dput(wh_dentry);
6034 +out:
6035 +       dput(parent);
6036 +       return err;
6037 +}
6038 +
6039 +struct au_cpup_wh_args {
6040 +       int *errp;
6041 +       struct au_cp_generic *cpg;
6042 +       struct file *file;
6043 +};
6044 +
6045 +static void au_call_cpup_wh(void *args)
6046 +{
6047 +       struct au_cpup_wh_args *a = args;
6048 +
6049 +       au_pin_hdir_acquire_nest(a->cpg->pin);
6050 +       *a->errp = au_cpup_wh(a->cpg, a->file);
6051 +       au_pin_hdir_release(a->cpg->pin);
6052 +}
6053 +
6054 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6055 +{
6056 +       int err, wkq_err;
6057 +       aufs_bindex_t bdst;
6058 +       struct dentry *dentry, *parent, *h_orph, *h_parent;
6059 +       struct inode *dir, *h_dir, *h_tmpdir;
6060 +       struct au_wbr *wbr;
6061 +       struct au_pin wh_pin, *pin_orig;
6062 +       struct user_namespace *h_userns;
6063 +
6064 +       dentry = cpg->dentry;
6065 +       bdst = cpg->bdst;
6066 +       parent = dget_parent(dentry);
6067 +       dir = d_inode(parent);
6068 +       h_orph = NULL;
6069 +       h_parent = NULL;
6070 +       h_dir = au_igrab(au_h_iptr(dir, bdst));
6071 +       h_tmpdir = h_dir;
6072 +       pin_orig = NULL;
6073 +       if (!h_dir->i_nlink) {
6074 +               wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
6075 +               h_orph = wbr->wbr_orph;
6076 +
6077 +               h_parent = dget(au_h_dptr(parent, bdst));
6078 +               au_set_h_dptr(parent, bdst, dget(h_orph));
6079 +               h_tmpdir = d_inode(h_orph);
6080 +               au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
6081 +
6082 +               inode_lock_nested(h_tmpdir, AuLsc_I_PARENT3);
6083 +               /* todo: au_h_open_pre()? */
6084 +
6085 +               pin_orig = cpg->pin;
6086 +               au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
6087 +                           AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
6088 +               cpg->pin = &wh_pin;
6089 +       }
6090 +
6091 +       h_userns = au_sbr_userns(dentry->d_sb, bdst);
6092 +       if (!au_test_h_perm_sio(h_userns, h_tmpdir, MAY_EXEC | MAY_WRITE)
6093 +           && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6094 +               err = au_cpup_wh(cpg, file);
6095 +       else {
6096 +               struct au_cpup_wh_args args = {
6097 +                       .errp   = &err,
6098 +                       .cpg    = cpg,
6099 +                       .file   = file
6100 +               };
6101 +               wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
6102 +               if (unlikely(wkq_err))
6103 +                       err = wkq_err;
6104 +       }
6105 +
6106 +       if (h_orph) {
6107 +               inode_unlock(h_tmpdir);
6108 +               /* todo: au_h_open_post()? */
6109 +               au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
6110 +               au_set_h_dptr(parent, bdst, h_parent);
6111 +               AuDebugOn(!pin_orig);
6112 +               cpg->pin = pin_orig;
6113 +       }
6114 +       iput(h_dir);
6115 +       dput(parent);
6116 +
6117 +       return err;
6118 +}
6119 +
6120 +/* ---------------------------------------------------------------------- */
6121 +
6122 +/*
6123 + * generic routine for both of copy-up and copy-down.
6124 + */
6125 +/* cf. revalidate function in file.c */
6126 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6127 +              int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6128 +                        struct au_pin *pin,
6129 +                        struct dentry *h_parent, void *arg),
6130 +              void *arg)
6131 +{
6132 +       int err;
6133 +       struct au_pin pin;
6134 +       struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
6135 +
6136 +       err = 0;
6137 +       parent = dget_parent(dentry);
6138 +       if (IS_ROOT(parent))
6139 +               goto out;
6140 +
6141 +       au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
6142 +                   au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
6143 +
6144 +       /* do not use au_dpage */
6145 +       real_parent = parent;
6146 +       while (1) {
6147 +               dput(parent);
6148 +               parent = dget_parent(dentry);
6149 +               h_parent = au_h_dptr(parent, bdst);
6150 +               if (h_parent)
6151 +                       goto out; /* success */
6152 +
6153 +               /* find top dir which is necessary to cpup */
6154 +               do {
6155 +                       d = parent;
6156 +                       dput(parent);
6157 +                       parent = dget_parent(d);
6158 +                       di_read_lock_parent3(parent, !AuLock_IR);
6159 +                       h_parent = au_h_dptr(parent, bdst);
6160 +                       di_read_unlock(parent, !AuLock_IR);
6161 +               } while (!h_parent);
6162 +
6163 +               if (d != real_parent)
6164 +                       di_write_lock_child3(d);
6165 +
6166 +               /* somebody else might create while we were sleeping */
6167 +               h_dentry = au_h_dptr(d, bdst);
6168 +               if (!h_dentry || d_is_negative(h_dentry)) {
6169 +                       if (h_dentry)
6170 +                               au_update_dbtop(d);
6171 +
6172 +                       au_pin_set_dentry(&pin, d);
6173 +                       err = au_do_pin(&pin);
6174 +                       if (!err) {
6175 +                               err = cp(d, bdst, &pin, h_parent, arg);
6176 +                               au_unpin(&pin);
6177 +                       }
6178 +               }
6179 +
6180 +               if (d != real_parent)
6181 +                       di_write_unlock(d);
6182 +               if (unlikely(err))
6183 +                       break;
6184 +       }
6185 +
6186 +out:
6187 +       dput(parent);
6188 +       return err;
6189 +}
6190 +
6191 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
6192 +                      struct au_pin *pin,
6193 +                      struct dentry *h_parent __maybe_unused,
6194 +                      void *arg __maybe_unused)
6195 +{
6196 +       struct au_cp_generic cpg = {
6197 +               .dentry = dentry,
6198 +               .bdst   = bdst,
6199 +               .bsrc   = -1,
6200 +               .len    = 0,
6201 +               .pin    = pin,
6202 +               .flags  = AuCpup_DTIME
6203 +       };
6204 +       return au_sio_cpup_simple(&cpg);
6205 +}
6206 +
6207 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6208 +{
6209 +       return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
6210 +}
6211 +
6212 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6213 +{
6214 +       int err;
6215 +       struct dentry *parent;
6216 +       struct inode *dir;
6217 +
6218 +       parent = dget_parent(dentry);
6219 +       dir = d_inode(parent);
6220 +       err = 0;
6221 +       if (au_h_iptr(dir, bdst))
6222 +               goto out;
6223 +
6224 +       di_read_unlock(parent, AuLock_IR);
6225 +       di_write_lock_parent(parent);
6226 +       /* someone else might change our inode while we were sleeping */
6227 +       if (!au_h_iptr(dir, bdst))
6228 +               err = au_cpup_dirs(dentry, bdst);
6229 +       di_downgrade_lock(parent, AuLock_IR);
6230 +
6231 +out:
6232 +       dput(parent);
6233 +       return err;
6234 +}
6235 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
6236 --- /usr/share/empty/fs/aufs/cpup.h     1970-01-01 01:00:00.000000000 +0100
6237 +++ linux/fs/aufs/cpup.h        2022-11-05 23:02:18.962555950 +0100
6238 @@ -0,0 +1,100 @@
6239 +/* SPDX-License-Identifier: GPL-2.0 */
6240 +/*
6241 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6242 + *
6243 + * This program is free software; you can redistribute it and/or modify
6244 + * it under the terms of the GNU General Public License as published by
6245 + * the Free Software Foundation; either version 2 of the License, or
6246 + * (at your option) any later version.
6247 + *
6248 + * This program is distributed in the hope that it will be useful,
6249 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6250 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6251 + * GNU General Public License for more details.
6252 + *
6253 + * You should have received a copy of the GNU General Public License
6254 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6255 + */
6256 +
6257 +/*
6258 + * copy-up/down functions
6259 + */
6260 +
6261 +#ifndef __AUFS_CPUP_H__
6262 +#define __AUFS_CPUP_H__
6263 +
6264 +#ifdef __KERNEL__
6265 +
6266 +#include <linux/path.h>
6267 +
6268 +struct inode;
6269 +struct file;
6270 +struct au_pin;
6271 +
6272 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
6273 +void au_cpup_attr_timesizes(struct inode *inode);
6274 +void au_cpup_attr_nlink(struct inode *inode, int force);
6275 +void au_cpup_attr_changeable(struct inode *inode);
6276 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
6277 +void au_cpup_attr_all(struct inode *inode, int force);
6278 +
6279 +/* ---------------------------------------------------------------------- */
6280 +
6281 +struct au_cp_generic {
6282 +       struct dentry   *dentry;
6283 +       aufs_bindex_t   bdst, bsrc;
6284 +       loff_t          len;
6285 +       struct au_pin   *pin;
6286 +       unsigned int    flags;
6287 +};
6288 +
6289 +/* cpup flags */
6290 +#define AuCpup_DTIME           1               /* do dtime_store/revert */
6291 +#define AuCpup_KEEPLINO                (1 << 1)        /* do not clear the lower xino,
6292 +                                                  for link(2) */
6293 +#define AuCpup_RENAME          (1 << 2)        /* rename after cpup */
6294 +#define AuCpup_HOPEN           (1 << 3)        /* call h_open_pre/post() in
6295 +                                                  cpup */
6296 +#define AuCpup_OVERWRITE       (1 << 4)        /* allow overwriting the
6297 +                                                  existing entry */
6298 +#define AuCpup_RWDST           (1 << 5)        /* force write target even if
6299 +                                                  the branch is marked as RO */
6300 +
6301 +#ifndef CONFIG_AUFS_BR_HFSPLUS
6302 +#undef AuCpup_HOPEN
6303 +#define AuCpup_HOPEN           0
6304 +#endif
6305 +
6306 +#define au_ftest_cpup(flags, name)     ((flags) & AuCpup_##name)
6307 +#define au_fset_cpup(flags, name) \
6308 +       do { (flags) |= AuCpup_##name; } while (0)
6309 +#define au_fclr_cpup(flags, name) \
6310 +       do { (flags) &= ~AuCpup_##name; } while (0)
6311 +
6312 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
6313 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
6314 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
6315 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
6316 +
6317 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6318 +              int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6319 +                        struct au_pin *pin,
6320 +                        struct dentry *h_parent, void *arg),
6321 +              void *arg);
6322 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6323 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6324 +
6325 +/* ---------------------------------------------------------------------- */
6326 +
6327 +/* keep timestamps when copyup */
6328 +struct au_dtime {
6329 +       struct dentry *dt_dentry;
6330 +       struct path dt_h_path;
6331 +       struct timespec64 dt_atime, dt_mtime;
6332 +};
6333 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
6334 +                   struct path *h_path);
6335 +void au_dtime_revert(struct au_dtime *dt);
6336 +
6337 +#endif /* __KERNEL__ */
6338 +#endif /* __AUFS_CPUP_H__ */
6339 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
6340 --- /usr/share/empty/fs/aufs/dbgaufs.c  1970-01-01 01:00:00.000000000 +0100
6341 +++ linux/fs/aufs/dbgaufs.c     2022-11-05 23:02:18.962555950 +0100
6342 @@ -0,0 +1,526 @@
6343 +// SPDX-License-Identifier: GPL-2.0
6344 +/*
6345 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6346 + *
6347 + * This program is free software; you can redistribute it and/or modify
6348 + * it under the terms of the GNU General Public License as published by
6349 + * the Free Software Foundation; either version 2 of the License, or
6350 + * (at your option) any later version.
6351 + *
6352 + * This program is distributed in the hope that it will be useful,
6353 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6354 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6355 + * GNU General Public License for more details.
6356 + *
6357 + * You should have received a copy of the GNU General Public License
6358 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6359 + */
6360 +
6361 +/*
6362 + * debugfs interface
6363 + */
6364 +
6365 +#include <linux/debugfs.h>
6366 +#include "aufs.h"
6367 +
6368 +#ifndef CONFIG_SYSFS
6369 +#error DEBUG_FS depends upon SYSFS
6370 +#endif
6371 +
6372 +static struct dentry *dbgaufs;
6373 +static const mode_t dbgaufs_mode = 0444;
6374 +
6375 +/* 20 is max digits length of ulong 64 */
6376 +struct dbgaufs_arg {
6377 +       int n;
6378 +       char a[20 * 4];
6379 +};
6380 +
6381 +/*
6382 + * common function for all XINO files
6383 + */
6384 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
6385 +                             struct file *file)
6386 +{
6387 +       void *p;
6388 +
6389 +       p = file->private_data;
6390 +       if (p) {
6391 +               /* this is struct dbgaufs_arg */
6392 +               AuDebugOn(!au_kfree_sz_test(p));
6393 +               au_kfree_do_rcu(p);
6394 +       }
6395 +       return 0;
6396 +}
6397 +
6398 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt,
6399 +                          int cnt)
6400 +{
6401 +       int err;
6402 +       struct kstat st;
6403 +       struct dbgaufs_arg *p;
6404 +
6405 +       err = -ENOMEM;
6406 +       p = kmalloc(sizeof(*p), GFP_NOFS);
6407 +       if (unlikely(!p))
6408 +               goto out;
6409 +
6410 +       err = 0;
6411 +       p->n = 0;
6412 +       file->private_data = p;
6413 +       if (!xf)
6414 +               goto out;
6415 +
6416 +       err = vfsub_getattr(&xf->f_path, &st);
6417 +       if (!err) {
6418 +               if (do_fcnt)
6419 +                       p->n = snprintf
6420 +                               (p->a, sizeof(p->a), "%d, %llux%u %lld\n",
6421 +                                cnt, st.blocks, st.blksize,
6422 +                                (long long)st.size);
6423 +               else
6424 +                       p->n = snprintf(p->a, sizeof(p->a), "%llux%u %lld\n",
6425 +                                       st.blocks, st.blksize,
6426 +                                       (long long)st.size);
6427 +               AuDebugOn(p->n >= sizeof(p->a));
6428 +       } else {
6429 +               p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
6430 +               err = 0;
6431 +       }
6432 +
6433 +out:
6434 +       return err;
6435 +}
6436 +
6437 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
6438 +                              size_t count, loff_t *ppos)
6439 +{
6440 +       struct dbgaufs_arg *p;
6441 +
6442 +       p = file->private_data;
6443 +       return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6444 +}
6445 +
6446 +/* ---------------------------------------------------------------------- */
6447 +
6448 +struct dbgaufs_plink_arg {
6449 +       int n;
6450 +       char a[];
6451 +};
6452 +
6453 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6454 +                                struct file *file)
6455 +{
6456 +       free_page((unsigned long)file->private_data);
6457 +       return 0;
6458 +}
6459 +
6460 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6461 +{
6462 +       int err, i, limit;
6463 +       unsigned long n, sum;
6464 +       struct dbgaufs_plink_arg *p;
6465 +       struct au_sbinfo *sbinfo;
6466 +       struct super_block *sb;
6467 +       struct hlist_bl_head *hbl;
6468 +
6469 +       err = -ENOMEM;
6470 +       p = (void *)get_zeroed_page(GFP_NOFS);
6471 +       if (unlikely(!p))
6472 +               goto out;
6473 +
6474 +       err = -EFBIG;
6475 +       sbinfo = inode->i_private;
6476 +       sb = sbinfo->si_sb;
6477 +       si_noflush_read_lock(sb);
6478 +       if (au_opt_test(au_mntflags(sb), PLINK)) {
6479 +               limit = PAGE_SIZE - sizeof(p->n);
6480 +
6481 +               /* the number of buckets */
6482 +               n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6483 +               p->n += n;
6484 +               limit -= n;
6485 +
6486 +               sum = 0;
6487 +               for (i = 0, hbl = sbinfo->si_plink; i < AuPlink_NHASH;
6488 +                    i++, hbl++) {
6489 +                       n = au_hbl_count(hbl);
6490 +                       sum += n;
6491 +
6492 +                       n = snprintf(p->a + p->n, limit, "%lu ", n);
6493 +                       p->n += n;
6494 +                       limit -= n;
6495 +                       if (unlikely(limit <= 0))
6496 +                               goto out_free;
6497 +               }
6498 +               p->a[p->n - 1] = '\n';
6499 +
6500 +               /* the sum of plinks */
6501 +               n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6502 +               p->n += n;
6503 +               limit -= n;
6504 +               if (unlikely(limit <= 0))
6505 +                       goto out_free;
6506 +       } else {
6507 +#define str "1\n0\n0\n"
6508 +               p->n = sizeof(str) - 1;
6509 +               strcpy(p->a, str);
6510 +#undef str
6511 +       }
6512 +       si_read_unlock(sb);
6513 +
6514 +       err = 0;
6515 +       file->private_data = p;
6516 +       goto out; /* success */
6517 +
6518 +out_free:
6519 +       free_page((unsigned long)p);
6520 +out:
6521 +       return err;
6522 +}
6523 +
6524 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6525 +                                 size_t count, loff_t *ppos)
6526 +{
6527 +       struct dbgaufs_plink_arg *p;
6528 +
6529 +       p = file->private_data;
6530 +       return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6531 +}
6532 +
6533 +static const struct file_operations dbgaufs_plink_fop = {
6534 +       .owner          = THIS_MODULE,
6535 +       .open           = dbgaufs_plink_open,
6536 +       .release        = dbgaufs_plink_release,
6537 +       .read           = dbgaufs_plink_read
6538 +};
6539 +
6540 +/* ---------------------------------------------------------------------- */
6541 +
6542 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6543 +{
6544 +       int err;
6545 +       struct au_sbinfo *sbinfo;
6546 +       struct super_block *sb;
6547 +
6548 +       sbinfo = inode->i_private;
6549 +       sb = sbinfo->si_sb;
6550 +       si_noflush_read_lock(sb);
6551 +       err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0, /*cnt*/0);
6552 +       si_read_unlock(sb);
6553 +       return err;
6554 +}
6555 +
6556 +static const struct file_operations dbgaufs_xib_fop = {
6557 +       .owner          = THIS_MODULE,
6558 +       .open           = dbgaufs_xib_open,
6559 +       .release        = dbgaufs_xi_release,
6560 +       .read           = dbgaufs_xi_read
6561 +};
6562 +
6563 +/* ---------------------------------------------------------------------- */
6564 +
6565 +#define DbgaufsXi_PREFIX "xi"
6566 +
6567 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6568 +{
6569 +       int err, idx;
6570 +       long l;
6571 +       aufs_bindex_t bindex;
6572 +       char *p, a[sizeof(DbgaufsXi_PREFIX) + 8];
6573 +       struct au_sbinfo *sbinfo;
6574 +       struct super_block *sb;
6575 +       struct au_xino *xi;
6576 +       struct file *xf;
6577 +       struct qstr *name;
6578 +       struct au_branch *br;
6579 +
6580 +       err = -ENOENT;
6581 +       name = &file->f_path.dentry->d_name;
6582 +       if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6583 +                    || memcmp(name->name, DbgaufsXi_PREFIX,
6584 +                              sizeof(DbgaufsXi_PREFIX) - 1)))
6585 +               goto out;
6586 +
6587 +       AuDebugOn(name->len >= sizeof(a));
6588 +       memcpy(a, name->name, name->len);
6589 +       a[name->len] = '\0';
6590 +       p = strchr(a, '-');
6591 +       if (p)
6592 +               *p = '\0';
6593 +       err = kstrtol(a + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6594 +       if (unlikely(err))
6595 +               goto out;
6596 +       bindex = l;
6597 +       idx = 0;
6598 +       if (p) {
6599 +               err = kstrtol(p + 1, 10, &l);
6600 +               if (unlikely(err))
6601 +                       goto out;
6602 +               idx = l;
6603 +       }
6604 +
6605 +       err = -ENOENT;
6606 +       sbinfo = inode->i_private;
6607 +       sb = sbinfo->si_sb;
6608 +       si_noflush_read_lock(sb);
6609 +       if (unlikely(bindex < 0 || bindex > au_sbbot(sb)))
6610 +               goto out_si;
6611 +       br = au_sbr(sb, bindex);
6612 +       xi = br->br_xino;
6613 +       if (unlikely(idx >= xi->xi_nfile))
6614 +               goto out_si;
6615 +       xf = au_xino_file(xi, idx);
6616 +       if (xf)
6617 +               err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1,
6618 +                                     au_xino_count(br));
6619 +
6620 +out_si:
6621 +       si_read_unlock(sb);
6622 +out:
6623 +       AuTraceErr(err);
6624 +       return err;
6625 +}
6626 +
6627 +static const struct file_operations dbgaufs_xino_fop = {
6628 +       .owner          = THIS_MODULE,
6629 +       .open           = dbgaufs_xino_open,
6630 +       .release        = dbgaufs_xi_release,
6631 +       .read           = dbgaufs_xi_read
6632 +};
6633 +
6634 +void dbgaufs_xino_del(struct au_branch *br)
6635 +{
6636 +       struct dentry *dbgaufs;
6637 +
6638 +       dbgaufs = br->br_dbgaufs;
6639 +       if (!dbgaufs)
6640 +               return;
6641 +
6642 +       br->br_dbgaufs = NULL;
6643 +       /* debugfs acquires the parent i_mutex */
6644 +       lockdep_off();
6645 +       debugfs_remove(dbgaufs);
6646 +       lockdep_on();
6647 +}
6648 +
6649 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6650 +{
6651 +       aufs_bindex_t bbot;
6652 +       struct au_branch *br;
6653 +
6654 +       if (!au_sbi(sb)->si_dbgaufs)
6655 +               return;
6656 +
6657 +       bbot = au_sbbot(sb);
6658 +       for (; bindex <= bbot; bindex++) {
6659 +               br = au_sbr(sb, bindex);
6660 +               dbgaufs_xino_del(br);
6661 +       }
6662 +}
6663 +
6664 +static void dbgaufs_br_do_add(struct super_block *sb, aufs_bindex_t bindex,
6665 +                             unsigned int idx, struct dentry *parent,
6666 +                             struct au_sbinfo *sbinfo)
6667 +{
6668 +       struct au_branch *br;
6669 +       struct dentry *d;
6670 +       /* "xi" bindex(5) "-" idx(2) NULL */
6671 +       char name[sizeof(DbgaufsXi_PREFIX) + 8];
6672 +
6673 +       if (!idx)
6674 +               snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6675 +       else
6676 +               snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d-%u",
6677 +                        bindex, idx);
6678 +       br = au_sbr(sb, bindex);
6679 +       if (br->br_dbgaufs) {
6680 +               struct qstr qstr = QSTR_INIT(name, strlen(name));
6681 +
6682 +               if (!au_qstreq(&br->br_dbgaufs->d_name, &qstr)) {
6683 +                       /* debugfs acquires the parent i_mutex */
6684 +                       lockdep_off();
6685 +                       d = debugfs_rename(parent, br->br_dbgaufs, parent,
6686 +                                          name);
6687 +                       lockdep_on();
6688 +                       if (unlikely(!d))
6689 +                               pr_warn("failed renaming %pd/%s, ignored.\n",
6690 +                                       parent, name);
6691 +               }
6692 +       } else {
6693 +               lockdep_off();
6694 +               br->br_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6695 +                                                    sbinfo, &dbgaufs_xino_fop);
6696 +               lockdep_on();
6697 +               if (unlikely(!br->br_dbgaufs))
6698 +                       pr_warn("failed creating %pd/%s, ignored.\n",
6699 +                               parent, name);
6700 +       }
6701 +}
6702 +
6703 +static void dbgaufs_br_add(struct super_block *sb, aufs_bindex_t bindex,
6704 +                          struct dentry *parent, struct au_sbinfo *sbinfo)
6705 +{
6706 +       struct au_branch *br;
6707 +       struct au_xino *xi;
6708 +       unsigned int u;
6709 +
6710 +       br = au_sbr(sb, bindex);
6711 +       xi = br->br_xino;
6712 +       for (u = 0; u < xi->xi_nfile; u++)
6713 +               dbgaufs_br_do_add(sb, bindex, u, parent, sbinfo);
6714 +}
6715 +
6716 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown)
6717 +{
6718 +       struct au_sbinfo *sbinfo;
6719 +       struct dentry *parent;
6720 +       aufs_bindex_t bbot;
6721 +
6722 +       if (!au_opt_test(au_mntflags(sb), XINO))
6723 +               return;
6724 +
6725 +       sbinfo = au_sbi(sb);
6726 +       parent = sbinfo->si_dbgaufs;
6727 +       if (!parent)
6728 +               return;
6729 +
6730 +       bbot = au_sbbot(sb);
6731 +       if (topdown)
6732 +               for (; bindex <= bbot; bindex++)
6733 +                       dbgaufs_br_add(sb, bindex, parent, sbinfo);
6734 +       else
6735 +               for (; bbot >= bindex; bbot--)
6736 +                       dbgaufs_br_add(sb, bbot, parent, sbinfo);
6737 +}
6738 +
6739 +/* ---------------------------------------------------------------------- */
6740 +
6741 +#ifdef CONFIG_AUFS_EXPORT
6742 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6743 +{
6744 +       int err;
6745 +       struct au_sbinfo *sbinfo;
6746 +       struct super_block *sb;
6747 +
6748 +       sbinfo = inode->i_private;
6749 +       sb = sbinfo->si_sb;
6750 +       si_noflush_read_lock(sb);
6751 +       err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0, /*cnt*/0);
6752 +       si_read_unlock(sb);
6753 +       return err;
6754 +}
6755 +
6756 +static const struct file_operations dbgaufs_xigen_fop = {
6757 +       .owner          = THIS_MODULE,
6758 +       .open           = dbgaufs_xigen_open,
6759 +       .release        = dbgaufs_xi_release,
6760 +       .read           = dbgaufs_xi_read
6761 +};
6762 +
6763 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6764 +{
6765 +       int err;
6766 +
6767 +       /*
6768 +        * This function is a dynamic '__init' function actually,
6769 +        * so the tiny check for si_rwsem is unnecessary.
6770 +        */
6771 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6772 +
6773 +       err = -EIO;
6774 +       sbinfo->si_dbgaufs_xigen = debugfs_create_file
6775 +               ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6776 +                &dbgaufs_xigen_fop);
6777 +       if (sbinfo->si_dbgaufs_xigen)
6778 +               err = 0;
6779 +
6780 +       return err;
6781 +}
6782 +#else
6783 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6784 +{
6785 +       return 0;
6786 +}
6787 +#endif /* CONFIG_AUFS_EXPORT */
6788 +
6789 +/* ---------------------------------------------------------------------- */
6790 +
6791 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6792 +{
6793 +       /*
6794 +        * This function is a dynamic '__fin' function actually,
6795 +        * so the tiny check for si_rwsem is unnecessary.
6796 +        */
6797 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6798 +
6799 +       debugfs_remove_recursive(sbinfo->si_dbgaufs);
6800 +       sbinfo->si_dbgaufs = NULL;
6801 +}
6802 +
6803 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6804 +{
6805 +       int err;
6806 +       char name[SysaufsSiNameLen];
6807 +
6808 +       /*
6809 +        * This function is a dynamic '__init' function actually,
6810 +        * so the tiny check for si_rwsem is unnecessary.
6811 +        */
6812 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6813 +
6814 +       err = -ENOENT;
6815 +       if (!dbgaufs) {
6816 +               AuErr1("/debug/aufs is uninitialized\n");
6817 +               goto out;
6818 +       }
6819 +
6820 +       err = -EIO;
6821 +       sysaufs_name(sbinfo, name);
6822 +       sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6823 +       if (unlikely(!sbinfo->si_dbgaufs))
6824 +               goto out;
6825 +
6826 +       /* regardless plink/noplink option */
6827 +       sbinfo->si_dbgaufs_plink = debugfs_create_file
6828 +               ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6829 +                &dbgaufs_plink_fop);
6830 +       if (unlikely(!sbinfo->si_dbgaufs_plink))
6831 +               goto out_dir;
6832 +
6833 +       /* regardless xino/noxino option */
6834 +       sbinfo->si_dbgaufs_xib = debugfs_create_file
6835 +               ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6836 +                &dbgaufs_xib_fop);
6837 +       if (unlikely(!sbinfo->si_dbgaufs_xib))
6838 +               goto out_dir;
6839 +
6840 +       err = dbgaufs_xigen_init(sbinfo);
6841 +       if (!err)
6842 +               goto out; /* success */
6843 +
6844 +out_dir:
6845 +       dbgaufs_si_fin(sbinfo);
6846 +out:
6847 +       if (unlikely(err))
6848 +               pr_err("debugfs/aufs failed\n");
6849 +       return err;
6850 +}
6851 +
6852 +/* ---------------------------------------------------------------------- */
6853 +
6854 +void dbgaufs_fin(void)
6855 +{
6856 +       debugfs_remove(dbgaufs);
6857 +}
6858 +
6859 +int __init dbgaufs_init(void)
6860 +{
6861 +       int err;
6862 +
6863 +       err = -EIO;
6864 +       dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6865 +       if (dbgaufs)
6866 +               err = 0;
6867 +       return err;
6868 +}
6869 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6870 --- /usr/share/empty/fs/aufs/dbgaufs.h  1970-01-01 01:00:00.000000000 +0100
6871 +++ linux/fs/aufs/dbgaufs.h     2022-11-05 23:02:18.962555950 +0100
6872 @@ -0,0 +1,53 @@
6873 +/* SPDX-License-Identifier: GPL-2.0 */
6874 +/*
6875 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6876 + *
6877 + * This program is free software; you can redistribute it and/or modify
6878 + * it under the terms of the GNU General Public License as published by
6879 + * the Free Software Foundation; either version 2 of the License, or
6880 + * (at your option) any later version.
6881 + *
6882 + * This program is distributed in the hope that it will be useful,
6883 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6884 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6885 + * GNU General Public License for more details.
6886 + *
6887 + * You should have received a copy of the GNU General Public License
6888 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6889 + */
6890 +
6891 +/*
6892 + * debugfs interface
6893 + */
6894 +
6895 +#ifndef __DBGAUFS_H__
6896 +#define __DBGAUFS_H__
6897 +
6898 +#ifdef __KERNEL__
6899 +
6900 +struct super_block;
6901 +struct au_sbinfo;
6902 +struct au_branch;
6903 +
6904 +#ifdef CONFIG_DEBUG_FS
6905 +/* dbgaufs.c */
6906 +void dbgaufs_xino_del(struct au_branch *br);
6907 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6908 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown);
6909 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6910 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6911 +void dbgaufs_fin(void);
6912 +int __init dbgaufs_init(void);
6913 +#else
6914 +AuStubVoid(dbgaufs_xino_del, struct au_branch *br)
6915 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6916 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex,
6917 +          int topdown)
6918 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6919 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6920 +AuStubVoid(dbgaufs_fin, void)
6921 +AuStubInt0(__init dbgaufs_init, void)
6922 +#endif /* CONFIG_DEBUG_FS */
6923 +
6924 +#endif /* __KERNEL__ */
6925 +#endif /* __DBGAUFS_H__ */
6926 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6927 --- /usr/share/empty/fs/aufs/dcsub.c    1970-01-01 01:00:00.000000000 +0100
6928 +++ linux/fs/aufs/dcsub.c       2022-11-05 23:02:18.962555950 +0100
6929 @@ -0,0 +1,225 @@
6930 +// SPDX-License-Identifier: GPL-2.0
6931 +/*
6932 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6933 + *
6934 + * This program is free software; you can redistribute it and/or modify
6935 + * it under the terms of the GNU General Public License as published by
6936 + * the Free Software Foundation; either version 2 of the License, or
6937 + * (at your option) any later version.
6938 + *
6939 + * This program is distributed in the hope that it will be useful,
6940 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6941 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6942 + * GNU General Public License for more details.
6943 + *
6944 + * You should have received a copy of the GNU General Public License
6945 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6946 + */
6947 +
6948 +/*
6949 + * sub-routines for dentry cache
6950 + */
6951 +
6952 +#include "aufs.h"
6953 +
6954 +static void au_dpage_free(struct au_dpage *dpage)
6955 +{
6956 +       int i;
6957 +       struct dentry **p;
6958 +
6959 +       p = dpage->dentries;
6960 +       for (i = 0; i < dpage->ndentry; i++)
6961 +               dput(*p++);
6962 +       free_page((unsigned long)dpage->dentries);
6963 +}
6964 +
6965 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6966 +{
6967 +       int err;
6968 +       void *p;
6969 +
6970 +       err = -ENOMEM;
6971 +       dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6972 +       if (unlikely(!dpages->dpages))
6973 +               goto out;
6974 +
6975 +       p = (void *)__get_free_page(gfp);
6976 +       if (unlikely(!p))
6977 +               goto out_dpages;
6978 +
6979 +       dpages->dpages[0].ndentry = 0;
6980 +       dpages->dpages[0].dentries = p;
6981 +       dpages->ndpage = 1;
6982 +       return 0; /* success */
6983 +
6984 +out_dpages:
6985 +       au_kfree_try_rcu(dpages->dpages);
6986 +out:
6987 +       return err;
6988 +}
6989 +
6990 +void au_dpages_free(struct au_dcsub_pages *dpages)
6991 +{
6992 +       int i;
6993 +       struct au_dpage *p;
6994 +
6995 +       p = dpages->dpages;
6996 +       for (i = 0; i < dpages->ndpage; i++)
6997 +               au_dpage_free(p++);
6998 +       au_kfree_try_rcu(dpages->dpages);
6999 +}
7000 +
7001 +static int au_dpages_append(struct au_dcsub_pages *dpages,
7002 +                           struct dentry *dentry, gfp_t gfp)
7003 +{
7004 +       int err, sz;
7005 +       struct au_dpage *dpage;
7006 +       void *p;
7007 +
7008 +       dpage = dpages->dpages + dpages->ndpage - 1;
7009 +       sz = PAGE_SIZE / sizeof(dentry);
7010 +       if (unlikely(dpage->ndentry >= sz)) {
7011 +               AuLabel(new dpage);
7012 +               err = -ENOMEM;
7013 +               sz = dpages->ndpage * sizeof(*dpages->dpages);
7014 +               p = au_kzrealloc(dpages->dpages, sz,
7015 +                                sz + sizeof(*dpages->dpages), gfp,
7016 +                                /*may_shrink*/0);
7017 +               if (unlikely(!p))
7018 +                       goto out;
7019 +
7020 +               dpages->dpages = p;
7021 +               dpage = dpages->dpages + dpages->ndpage;
7022 +               p = (void *)__get_free_page(gfp);
7023 +               if (unlikely(!p))
7024 +                       goto out;
7025 +
7026 +               dpage->ndentry = 0;
7027 +               dpage->dentries = p;
7028 +               dpages->ndpage++;
7029 +       }
7030 +
7031 +       AuDebugOn(au_dcount(dentry) <= 0);
7032 +       dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
7033 +       return 0; /* success */
7034 +
7035 +out:
7036 +       return err;
7037 +}
7038 +
7039 +/* todo: BAD approach */
7040 +/* copied from linux/fs/dcache.c */
7041 +enum d_walk_ret {
7042 +       D_WALK_CONTINUE,
7043 +       D_WALK_QUIT,
7044 +       D_WALK_NORETRY,
7045 +       D_WALK_SKIP,
7046 +};
7047 +
7048 +extern void d_walk(struct dentry *parent, void *data,
7049 +                  enum d_walk_ret (*enter)(void *, struct dentry *));
7050 +
7051 +struct ac_dpages_arg {
7052 +       int err;
7053 +       struct au_dcsub_pages *dpages;
7054 +       struct super_block *sb;
7055 +       au_dpages_test test;
7056 +       void *arg;
7057 +};
7058 +
7059 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
7060 +{
7061 +       enum d_walk_ret ret;
7062 +       struct ac_dpages_arg *arg = _arg;
7063 +
7064 +       ret = D_WALK_CONTINUE;
7065 +       if (dentry->d_sb == arg->sb
7066 +           && !IS_ROOT(dentry)
7067 +           && au_dcount(dentry) > 0
7068 +           && au_di(dentry)
7069 +           && (!arg->test || arg->test(dentry, arg->arg))) {
7070 +               arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
7071 +               if (unlikely(arg->err))
7072 +                       ret = D_WALK_QUIT;
7073 +       }
7074 +
7075 +       return ret;
7076 +}
7077 +
7078 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7079 +                  au_dpages_test test, void *arg)
7080 +{
7081 +       struct ac_dpages_arg args = {
7082 +               .err    = 0,
7083 +               .dpages = dpages,
7084 +               .sb     = root->d_sb,
7085 +               .test   = test,
7086 +               .arg    = arg
7087 +       };
7088 +
7089 +       d_walk(root, &args, au_call_dpages_append);
7090 +
7091 +       return args.err;
7092 +}
7093 +
7094 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7095 +                      int do_include, au_dpages_test test, void *arg)
7096 +{
7097 +       int err;
7098 +
7099 +       err = 0;
7100 +       write_seqlock(&rename_lock);
7101 +       spin_lock(&dentry->d_lock);
7102 +       if (do_include
7103 +           && au_dcount(dentry) > 0
7104 +           && (!test || test(dentry, arg)))
7105 +               err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7106 +       spin_unlock(&dentry->d_lock);
7107 +       if (unlikely(err))
7108 +               goto out;
7109 +
7110 +       /*
7111 +        * RCU for vfsmount is unnecessary since this is a traverse in a single
7112 +        * mount
7113 +        */
7114 +       while (!IS_ROOT(dentry)) {
7115 +               dentry = dentry->d_parent; /* rename_lock is locked */
7116 +               spin_lock(&dentry->d_lock);
7117 +               if (au_dcount(dentry) > 0
7118 +                   && (!test || test(dentry, arg)))
7119 +                       err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7120 +               spin_unlock(&dentry->d_lock);
7121 +               if (unlikely(err))
7122 +                       break;
7123 +       }
7124 +
7125 +out:
7126 +       write_sequnlock(&rename_lock);
7127 +       return err;
7128 +}
7129 +
7130 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
7131 +{
7132 +       return au_di(dentry) && dentry->d_sb == arg;
7133 +}
7134 +
7135 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7136 +                           struct dentry *dentry, int do_include)
7137 +{
7138 +       return au_dcsub_pages_rev(dpages, dentry, do_include,
7139 +                                 au_dcsub_dpages_aufs, dentry->d_sb);
7140 +}
7141 +
7142 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
7143 +{
7144 +       struct path path[2] = {
7145 +               {
7146 +                       .dentry = d1
7147 +               },
7148 +               {
7149 +                       .dentry = d2
7150 +               }
7151 +       };
7152 +
7153 +       return path_is_under(path + 0, path + 1);
7154 +}
7155 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
7156 --- /usr/share/empty/fs/aufs/dcsub.h    1970-01-01 01:00:00.000000000 +0100
7157 +++ linux/fs/aufs/dcsub.h       2022-11-05 23:02:18.962555950 +0100
7158 @@ -0,0 +1,137 @@
7159 +/* SPDX-License-Identifier: GPL-2.0 */
7160 +/*
7161 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7162 + *
7163 + * This program is free software; you can redistribute it and/or modify
7164 + * it under the terms of the GNU General Public License as published by
7165 + * the Free Software Foundation; either version 2 of the License, or
7166 + * (at your option) any later version.
7167 + *
7168 + * This program is distributed in the hope that it will be useful,
7169 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7170 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7171 + * GNU General Public License for more details.
7172 + *
7173 + * You should have received a copy of the GNU General Public License
7174 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7175 + */
7176 +
7177 +/*
7178 + * sub-routines for dentry cache
7179 + */
7180 +
7181 +#ifndef __AUFS_DCSUB_H__
7182 +#define __AUFS_DCSUB_H__
7183 +
7184 +#ifdef __KERNEL__
7185 +
7186 +#include <linux/dcache.h>
7187 +#include <linux/fs.h>
7188 +
7189 +struct au_dpage {
7190 +       int ndentry;
7191 +       struct dentry **dentries;
7192 +};
7193 +
7194 +struct au_dcsub_pages {
7195 +       int ndpage;
7196 +       struct au_dpage *dpages;
7197 +};
7198 +
7199 +/* ---------------------------------------------------------------------- */
7200 +
7201 +/* dcsub.c */
7202 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
7203 +void au_dpages_free(struct au_dcsub_pages *dpages);
7204 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
7205 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7206 +                  au_dpages_test test, void *arg);
7207 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7208 +                      int do_include, au_dpages_test test, void *arg);
7209 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7210 +                           struct dentry *dentry, int do_include);
7211 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
7212 +
7213 +/* ---------------------------------------------------------------------- */
7214 +
7215 +/*
7216 + * todo: in linux-3.13, several similar (but faster) helpers are added to
7217 + * include/linux/dcache.h. Try them (in the future).
7218 + */
7219 +
7220 +static inline int au_d_hashed_positive(struct dentry *d)
7221 +{
7222 +       int err;
7223 +       struct inode *inode = d_inode(d);
7224 +
7225 +       err = 0;
7226 +       if (unlikely(d_unhashed(d)
7227 +                    || d_is_negative(d)
7228 +                    || !inode->i_nlink))
7229 +               err = -ENOENT;
7230 +       return err;
7231 +}
7232 +
7233 +static inline int au_d_linkable(struct dentry *d)
7234 +{
7235 +       int err;
7236 +       struct inode *inode = d_inode(d);
7237 +
7238 +       err = au_d_hashed_positive(d);
7239 +       if (err
7240 +           && d_is_positive(d)
7241 +           && (inode->i_state & I_LINKABLE))
7242 +               err = 0;
7243 +       return err;
7244 +}
7245 +
7246 +static inline int au_d_alive(struct dentry *d)
7247 +{
7248 +       int err;
7249 +       struct inode *inode;
7250 +
7251 +       err = 0;
7252 +       if (!IS_ROOT(d))
7253 +               err = au_d_hashed_positive(d);
7254 +       else {
7255 +               inode = d_inode(d);
7256 +               if (unlikely(d_unlinked(d)
7257 +                            || d_is_negative(d)
7258 +                            || !inode->i_nlink))
7259 +                       err = -ENOENT;
7260 +       }
7261 +       return err;
7262 +}
7263 +
7264 +static inline int au_alive_dir(struct dentry *d)
7265 +{
7266 +       int err;
7267 +
7268 +       err = au_d_alive(d);
7269 +       if (unlikely(err || IS_DEADDIR(d_inode(d))))
7270 +               err = -ENOENT;
7271 +       return err;
7272 +}
7273 +
7274 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
7275 +{
7276 +       return a->len == b->len
7277 +               && !memcmp(a->name, b->name, a->len);
7278 +}
7279 +
7280 +/*
7281 + * by the commit
7282 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
7283 + *                     taking d_lock
7284 + * the type of d_lockref.count became int, but the inlined function d_count()
7285 + * still returns unsigned int.
7286 + * I don't know why. Maybe it is for every d_count() users?
7287 + * Anyway au_dcount() lives on.
7288 + */
7289 +static inline int au_dcount(struct dentry *d)
7290 +{
7291 +       return (int)d_count(d);
7292 +}
7293 +
7294 +#endif /* __KERNEL__ */
7295 +#endif /* __AUFS_DCSUB_H__ */
7296 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
7297 --- /usr/share/empty/fs/aufs/debug.c    1970-01-01 01:00:00.000000000 +0100
7298 +++ linux/fs/aufs/debug.c       2023-02-20 21:05:51.959693785 +0100
7299 @@ -0,0 +1,446 @@
7300 +// SPDX-License-Identifier: GPL-2.0
7301 +/*
7302 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7303 + *
7304 + * This program is free software; you can redistribute it and/or modify
7305 + * it under the terms of the GNU General Public License as published by
7306 + * the Free Software Foundation; either version 2 of the License, or
7307 + * (at your option) any later version.
7308 + *
7309 + * This program is distributed in the hope that it will be useful,
7310 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7311 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7312 + * GNU General Public License for more details.
7313 + *
7314 + * You should have received a copy of the GNU General Public License
7315 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7316 + */
7317 +
7318 +/*
7319 + * debug print functions
7320 + */
7321 +
7322 +#include <linux/iversion.h>
7323 +#include "aufs.h"
7324 +
7325 +/* Returns 0, or -errno.  arg is in kp->arg. */
7326 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
7327 +{
7328 +       int err, n;
7329 +
7330 +       err = kstrtoint(val, 0, &n);
7331 +       if (!err) {
7332 +               if (n > 0)
7333 +                       au_debug_on();
7334 +               else
7335 +                       au_debug_off();
7336 +       }
7337 +       return err;
7338 +}
7339 +
7340 +/* Returns length written or -errno.  Buffer is 4k (ie. be short!) */
7341 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
7342 +{
7343 +       atomic_t *a;
7344 +
7345 +       a = kp->arg;
7346 +       return sprintf(buffer, "%d", atomic_read(a));
7347 +}
7348 +
7349 +static const struct kernel_param_ops param_ops_atomic_t = {
7350 +       .set = param_atomic_t_set,
7351 +       .get = param_atomic_t_get
7352 +       /* void (*free)(void *arg) */
7353 +};
7354 +
7355 +atomic_t aufs_debug = ATOMIC_INIT(0);
7356 +MODULE_PARM_DESC(debug, "debug print");
7357 +module_param_named(debug, aufs_debug, atomic_t, 0664);
7358 +
7359 +DEFINE_MUTEX(au_dbg_mtx);      /* just to serialize the dbg msgs */
7360 +char *au_plevel = KERN_DEBUG;
7361 +#define dpri(fmt, ...) do {                                    \
7362 +       if ((au_plevel                                          \
7363 +            && strcmp(au_plevel, KERN_DEBUG))                  \
7364 +           || au_debug_test())                                 \
7365 +               printk("%s" fmt, au_plevel, ##__VA_ARGS__);     \
7366 +} while (0)
7367 +
7368 +/* ---------------------------------------------------------------------- */
7369 +
7370 +void au_dpri_whlist(struct au_nhash *whlist)
7371 +{
7372 +       unsigned long ul, n;
7373 +       struct hlist_head *head;
7374 +       struct au_vdir_wh *pos;
7375 +
7376 +       n = whlist->nh_num;
7377 +       head = whlist->nh_head;
7378 +       for (ul = 0; ul < n; ul++) {
7379 +               hlist_for_each_entry(pos, head, wh_hash)
7380 +                       dpri("b%d, %.*s, %d\n",
7381 +                            pos->wh_bindex,
7382 +                            pos->wh_str.len, pos->wh_str.name,
7383 +                            pos->wh_str.len);
7384 +               head++;
7385 +       }
7386 +}
7387 +
7388 +void au_dpri_vdir(struct au_vdir *vdir)
7389 +{
7390 +       unsigned long ul;
7391 +       union au_vdir_deblk_p p;
7392 +       unsigned char *o;
7393 +
7394 +       if (!vdir || IS_ERR(vdir)) {
7395 +               dpri("err %ld\n", PTR_ERR(vdir));
7396 +               return;
7397 +       }
7398 +
7399 +       dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %llu\n",
7400 +            vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
7401 +            vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
7402 +       for (ul = 0; ul < vdir->vd_nblk; ul++) {
7403 +               p.deblk = vdir->vd_deblk[ul];
7404 +               o = p.deblk;
7405 +               dpri("[%lu]: %p\n", ul, o);
7406 +       }
7407 +}
7408 +
7409 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
7410 +                       struct dentry *wh)
7411 +{
7412 +       char *n = NULL;
7413 +       int l = 0;
7414 +
7415 +       if (!inode || IS_ERR(inode)) {
7416 +               dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
7417 +               return -1;
7418 +       }
7419 +
7420 +       /* the type of i_blocks depends upon CONFIG_LBDAF */
7421 +       BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
7422 +                    && sizeof(inode->i_blocks) != sizeof(u64));
7423 +       if (wh) {
7424 +               n = (void *)wh->d_name.name;
7425 +               l = wh->d_name.len;
7426 +       }
7427 +
7428 +       dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
7429 +            " acl %p, def_acl %p,"
7430 +            " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
7431 +            bindex, inode,
7432 +            inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
7433 +            atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
7434 +            i_size_read(inode), (unsigned long long)inode->i_blocks,
7435 +            inode->i_acl, inode->i_default_acl,
7436 +            hn, (long long)timespec64_to_ns(&inode->i_ctime) & 0x0ffff,
7437 +            inode->i_mapping ? inode->i_mapping->nrpages : 0,
7438 +            inode->i_state, inode->i_flags, inode_peek_iversion(inode),
7439 +            inode->i_generation,
7440 +            l ? ", wh " : "", l, n);
7441 +       return 0;
7442 +}
7443 +
7444 +void au_dpri_inode(struct inode *inode)
7445 +{
7446 +       struct au_iinfo *iinfo;
7447 +       struct au_hinode *hi;
7448 +       aufs_bindex_t bindex;
7449 +       int err, hn;
7450 +
7451 +       err = do_pri_inode(-1, inode, -1, NULL);
7452 +       if (err || !au_test_aufs(inode->i_sb) || au_is_bad_inode(inode))
7453 +               return;
7454 +
7455 +       iinfo = au_ii(inode);
7456 +       dpri("i-1: btop %d, bbot %d, gen %d\n",
7457 +            iinfo->ii_btop, iinfo->ii_bbot, au_iigen(inode, NULL));
7458 +       if (iinfo->ii_btop < 0)
7459 +               return;
7460 +       hn = 0;
7461 +       for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot; bindex++) {
7462 +               hi = au_hinode(iinfo, bindex);
7463 +               hn = !!au_hn(hi);
7464 +               do_pri_inode(bindex, hi->hi_inode, hn, hi->hi_whdentry);
7465 +       }
7466 +}
7467 +
7468 +void au_dpri_dalias(struct inode *inode)
7469 +{
7470 +       struct dentry *d;
7471 +
7472 +       spin_lock(&inode->i_lock);
7473 +       hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
7474 +               au_dpri_dentry(d);
7475 +       spin_unlock(&inode->i_lock);
7476 +}
7477 +
7478 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
7479 +{
7480 +       struct dentry *wh = NULL;
7481 +       int hn;
7482 +       struct inode *inode;
7483 +       struct au_iinfo *iinfo;
7484 +       struct au_hinode *hi;
7485 +
7486 +       if (!dentry || IS_ERR(dentry)) {
7487 +               dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
7488 +               return -1;
7489 +       }
7490 +       /* do not call dget_parent() here */
7491 +       /* note: access d_xxx without d_lock */
7492 +       dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
7493 +            bindex, dentry, dentry,
7494 +            dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
7495 +            au_dcount(dentry), dentry->d_flags,
7496 +            d_unhashed(dentry) ? "un" : "");
7497 +       hn = -1;
7498 +       inode = NULL;
7499 +       if (d_is_positive(dentry))
7500 +               inode = d_inode(dentry);
7501 +       if (inode
7502 +           && au_test_aufs(dentry->d_sb)
7503 +           && bindex >= 0
7504 +           && !au_is_bad_inode(inode)) {
7505 +               iinfo = au_ii(inode);
7506 +               hi = au_hinode(iinfo, bindex);
7507 +               hn = !!au_hn(hi);
7508 +               wh = hi->hi_whdentry;
7509 +       }
7510 +       do_pri_inode(bindex, inode, hn, wh);
7511 +       return 0;
7512 +}
7513 +
7514 +void au_dpri_dentry(struct dentry *dentry)
7515 +{
7516 +       struct au_dinfo *dinfo;
7517 +       aufs_bindex_t bindex;
7518 +       int err;
7519 +
7520 +       err = do_pri_dentry(-1, dentry);
7521 +       if (err || !au_test_aufs(dentry->d_sb))
7522 +               return;
7523 +
7524 +       dinfo = au_di(dentry);
7525 +       if (!dinfo)
7526 +               return;
7527 +       dpri("d-1: btop %d, bbot %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
7528 +            dinfo->di_btop, dinfo->di_bbot,
7529 +            dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
7530 +            dinfo->di_tmpfile);
7531 +       if (dinfo->di_btop < 0)
7532 +               return;
7533 +       for (bindex = dinfo->di_btop; bindex <= dinfo->di_bbot; bindex++)
7534 +               do_pri_dentry(bindex, au_hdentry(dinfo, bindex)->hd_dentry);
7535 +}
7536 +
7537 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7538 +{
7539 +       char a[32];
7540 +
7541 +       if (!file || IS_ERR(file)) {
7542 +               dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7543 +               return -1;
7544 +       }
7545 +       a[0] = 0;
7546 +       if (bindex < 0
7547 +           && !IS_ERR_OR_NULL(file->f_path.dentry)
7548 +           && au_test_aufs(file->f_path.dentry->d_sb)
7549 +           && au_fi(file))
7550 +               snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7551 +                        au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7552 +       dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7553 +            bindex, file->f_mode, file->f_flags, (long)file_count(file),
7554 +            file->f_version, file->f_pos, a);
7555 +       if (!IS_ERR_OR_NULL(file->f_path.dentry))
7556 +               do_pri_dentry(bindex, file->f_path.dentry);
7557 +       return 0;
7558 +}
7559 +
7560 +void au_dpri_file(struct file *file)
7561 +{
7562 +       struct au_finfo *finfo;
7563 +       struct au_fidir *fidir;
7564 +       struct au_hfile *hfile;
7565 +       aufs_bindex_t bindex;
7566 +       int err;
7567 +
7568 +       err = do_pri_file(-1, file);
7569 +       if (err
7570 +           || IS_ERR_OR_NULL(file->f_path.dentry)
7571 +           || !au_test_aufs(file->f_path.dentry->d_sb))
7572 +               return;
7573 +
7574 +       finfo = au_fi(file);
7575 +       if (!finfo)
7576 +               return;
7577 +       if (finfo->fi_btop < 0)
7578 +               return;
7579 +       fidir = finfo->fi_hdir;
7580 +       if (!fidir)
7581 +               do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7582 +       else
7583 +               for (bindex = finfo->fi_btop;
7584 +                    bindex >= 0 && bindex <= fidir->fd_bbot;
7585 +                    bindex++) {
7586 +                       hfile = fidir->fd_hfile + bindex;
7587 +                       do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7588 +               }
7589 +}
7590 +
7591 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7592 +{
7593 +       struct vfsmount *mnt;
7594 +       struct super_block *sb;
7595 +
7596 +       if (!br || IS_ERR(br))
7597 +               goto out;
7598 +       mnt = au_br_mnt(br);
7599 +       if (!mnt || IS_ERR(mnt))
7600 +               goto out;
7601 +       sb = mnt->mnt_sb;
7602 +       if (!sb || IS_ERR(sb))
7603 +               goto out;
7604 +
7605 +       dpri("s%d: {perm 0x%x, id %d, wbr %p}, "
7606 +            "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7607 +            "xino %d\n",
7608 +            bindex, br->br_perm, br->br_id, br->br_wbr,
7609 +            au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7610 +            sb->s_flags, sb->s_count,
7611 +            atomic_read(&sb->s_active),
7612 +            !!au_xino_file(br->br_xino, /*idx*/-1));
7613 +       return 0;
7614 +
7615 +out:
7616 +       dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7617 +       return -1;
7618 +}
7619 +
7620 +void au_dpri_sb(struct super_block *sb)
7621 +{
7622 +       struct au_sbinfo *sbinfo;
7623 +       aufs_bindex_t bindex;
7624 +       int err;
7625 +       /* to reduce stack size */
7626 +       struct {
7627 +               struct vfsmount mnt;
7628 +               struct au_branch fake;
7629 +       } *a;
7630 +
7631 +       /* this function can be called from magic sysrq */
7632 +       a = kzalloc(sizeof(*a), GFP_ATOMIC);
7633 +       if (unlikely(!a)) {
7634 +               dpri("no memory\n");
7635 +               return;
7636 +       }
7637 +
7638 +       a->mnt.mnt_sb = sb;
7639 +       a->fake.br_path.mnt = &a->mnt;
7640 +       err = do_pri_br(-1, &a->fake);
7641 +       au_kfree_rcu(a);
7642 +       dpri("dev 0x%x\n", sb->s_dev);
7643 +       if (err || !au_test_aufs(sb))
7644 +               return;
7645 +
7646 +       sbinfo = au_sbi(sb);
7647 +       if (!sbinfo)
7648 +               return;
7649 +       dpri("nw %d, gen %u, kobj %d\n",
7650 +            atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7651 +            kref_read(&sbinfo->si_kobj.kref));
7652 +       for (bindex = 0; bindex <= sbinfo->si_bbot; bindex++)
7653 +               do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7654 +}
7655 +
7656 +/* ---------------------------------------------------------------------- */
7657 +
7658 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7659 +{
7660 +       struct inode *h_inode, *inode = d_inode(dentry);
7661 +       struct dentry *h_dentry;
7662 +       aufs_bindex_t bindex, bbot, bi;
7663 +
7664 +       if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7665 +               return;
7666 +
7667 +       bbot = au_dbbot(dentry);
7668 +       bi = au_ibbot(inode);
7669 +       if (bi < bbot)
7670 +               bbot = bi;
7671 +       bindex = au_dbtop(dentry);
7672 +       bi = au_ibtop(inode);
7673 +       if (bi > bindex)
7674 +               bindex = bi;
7675 +
7676 +       for (; bindex <= bbot; bindex++) {
7677 +               h_dentry = au_h_dptr(dentry, bindex);
7678 +               if (!h_dentry)
7679 +                       continue;
7680 +               h_inode = au_h_iptr(inode, bindex);
7681 +               if (unlikely(h_inode != d_inode(h_dentry))) {
7682 +                       au_debug_on();
7683 +                       AuDbg("b%d, %s:%d\n", bindex, func, line);
7684 +                       AuDbgDentry(dentry);
7685 +                       AuDbgInode(inode);
7686 +                       au_debug_off();
7687 +                       if (au_test_fuse(h_inode->i_sb))
7688 +                               WARN_ON_ONCE(1);
7689 +                       else
7690 +                               BUG();
7691 +               }
7692 +       }
7693 +}
7694 +
7695 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7696 +{
7697 +       int err, i, j;
7698 +       struct au_dcsub_pages dpages;
7699 +       struct au_dpage *dpage;
7700 +       struct dentry **dentries;
7701 +
7702 +       err = au_dpages_init(&dpages, GFP_NOFS);
7703 +       AuDebugOn(err);
7704 +       err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7705 +       AuDebugOn(err);
7706 +       for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7707 +               dpage = dpages.dpages + i;
7708 +               dentries = dpage->dentries;
7709 +               for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7710 +                       AuDebugOn(au_digen_test(dentries[j], sigen));
7711 +       }
7712 +       au_dpages_free(&dpages);
7713 +}
7714 +
7715 +void au_dbg_verify_kthread(void)
7716 +{
7717 +       if (au_wkq_test()) {
7718 +               au_dbg_blocked();
7719 +               /*
7720 +                * It may be recursive, but udba=notify between two aufs mounts,
7721 +                * where a single ro branch is shared, is not a problem.
7722 +                */
7723 +               /* WARN_ON(1); */
7724 +       }
7725 +}
7726 +
7727 +/* ---------------------------------------------------------------------- */
7728 +
7729 +int __init au_debug_init(void)
7730 +{
7731 +       aufs_bindex_t bindex;
7732 +       struct au_vdir_destr destr;
7733 +
7734 +       bindex = -1;
7735 +       AuDebugOn(bindex >= 0);
7736 +
7737 +       destr.len = -1;
7738 +       AuDebugOn(destr.len < NAME_MAX);
7739 +
7740 +#ifdef CONFIG_4KSTACKS
7741 +       pr_warn("CONFIG_4KSTACKS is defined.\n");
7742 +#endif
7743 +
7744 +       return 0;
7745 +}
7746 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7747 --- /usr/share/empty/fs/aufs/debug.h    1970-01-01 01:00:00.000000000 +0100
7748 +++ linux/fs/aufs/debug.h       2022-11-05 23:02:18.962555950 +0100
7749 @@ -0,0 +1,226 @@
7750 +/* SPDX-License-Identifier: GPL-2.0 */
7751 +/*
7752 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7753 + *
7754 + * This program is free software; you can redistribute it and/or modify
7755 + * it under the terms of the GNU General Public License as published by
7756 + * the Free Software Foundation; either version 2 of the License, or
7757 + * (at your option) any later version.
7758 + *
7759 + * This program is distributed in the hope that it will be useful,
7760 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7761 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7762 + * GNU General Public License for more details.
7763 + *
7764 + * You should have received a copy of the GNU General Public License
7765 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7766 + */
7767 +
7768 +/*
7769 + * debug print functions
7770 + */
7771 +
7772 +#ifndef __AUFS_DEBUG_H__
7773 +#define __AUFS_DEBUG_H__
7774 +
7775 +#ifdef __KERNEL__
7776 +
7777 +#include <linux/atomic.h>
7778 +#include <linux/module.h>
7779 +#include <linux/kallsyms.h>
7780 +#include <linux/sysrq.h>
7781 +
7782 +#ifdef CONFIG_AUFS_DEBUG
7783 +#define AuDebugOn(a)           BUG_ON(a)
7784 +
7785 +/* module parameter */
7786 +extern atomic_t aufs_debug;
7787 +static inline void au_debug_on(void)
7788 +{
7789 +       atomic_inc(&aufs_debug);
7790 +}
7791 +static inline void au_debug_off(void)
7792 +{
7793 +       atomic_dec_if_positive(&aufs_debug);
7794 +}
7795 +
7796 +static inline int au_debug_test(void)
7797 +{
7798 +       return atomic_read(&aufs_debug) > 0;
7799 +}
7800 +#else
7801 +#define AuDebugOn(a)           do {} while (0)
7802 +AuStubVoid(au_debug_on, void)
7803 +AuStubVoid(au_debug_off, void)
7804 +AuStubInt0(au_debug_test, void)
7805 +#endif /* CONFIG_AUFS_DEBUG */
7806 +
7807 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7808 +
7809 +/* ---------------------------------------------------------------------- */
7810 +
7811 +/* debug print */
7812 +
7813 +#define AuDbg(fmt, ...) do { \
7814 +       if (au_debug_test()) \
7815 +               pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7816 +} while (0)
7817 +#define AuLabel(l)             AuDbg(#l "\n")
7818 +#define AuIOErr(fmt, ...)      pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7819 +#define AuWarn1(fmt, ...) do { \
7820 +       static unsigned char _c; \
7821 +       if (!_c++) \
7822 +               pr_warn(fmt, ##__VA_ARGS__); \
7823 +} while (0)
7824 +
7825 +#define AuErr1(fmt, ...) do { \
7826 +       static unsigned char _c; \
7827 +       if (!_c++) \
7828 +               pr_err(fmt, ##__VA_ARGS__); \
7829 +} while (0)
7830 +
7831 +#define AuIOErr1(fmt, ...) do { \
7832 +       static unsigned char _c; \
7833 +       if (!_c++) \
7834 +               AuIOErr(fmt, ##__VA_ARGS__); \
7835 +} while (0)
7836 +
7837 +#define AuUnsupportMsg "This operation is not supported." \
7838 +                       " Please report this application to aufs-users ML."
7839 +#define AuUnsupport(fmt, ...) do { \
7840 +       pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7841 +       dump_stack(); \
7842 +} while (0)
7843 +
7844 +#define AuTraceErr(e) do { \
7845 +       if (unlikely((e) < 0)) \
7846 +               AuDbg("err %d\n", (int)(e)); \
7847 +} while (0)
7848 +
7849 +#define AuTraceErrPtr(p) do { \
7850 +       if (IS_ERR(p)) \
7851 +               AuDbg("err %ld\n", PTR_ERR(p)); \
7852 +} while (0)
7853 +
7854 +/* dirty macros for debug print, use with "%.*s" and caution */
7855 +#define AuLNPair(qstr)         (qstr)->len, (qstr)->name
7856 +
7857 +/* ---------------------------------------------------------------------- */
7858 +
7859 +struct dentry;
7860 +#ifdef CONFIG_AUFS_DEBUG
7861 +extern struct mutex au_dbg_mtx;
7862 +extern char *au_plevel;
7863 +struct au_nhash;
7864 +void au_dpri_whlist(struct au_nhash *whlist);
7865 +struct au_vdir;
7866 +void au_dpri_vdir(struct au_vdir *vdir);
7867 +struct inode;
7868 +void au_dpri_inode(struct inode *inode);
7869 +void au_dpri_dalias(struct inode *inode);
7870 +void au_dpri_dentry(struct dentry *dentry);
7871 +struct file;
7872 +void au_dpri_file(struct file *filp);
7873 +struct super_block;
7874 +void au_dpri_sb(struct super_block *sb);
7875 +
7876 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7877 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7878 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7879 +void au_dbg_verify_kthread(void);
7880 +
7881 +int __init au_debug_init(void);
7882 +
7883 +#define AuDbgWhlist(w) do { \
7884 +       mutex_lock(&au_dbg_mtx); \
7885 +       AuDbg(#w "\n"); \
7886 +       au_dpri_whlist(w); \
7887 +       mutex_unlock(&au_dbg_mtx); \
7888 +} while (0)
7889 +
7890 +#define AuDbgVdir(v) do { \
7891 +       mutex_lock(&au_dbg_mtx); \
7892 +       AuDbg(#v "\n"); \
7893 +       au_dpri_vdir(v); \
7894 +       mutex_unlock(&au_dbg_mtx); \
7895 +} while (0)
7896 +
7897 +#define AuDbgInode(i) do { \
7898 +       mutex_lock(&au_dbg_mtx); \
7899 +       AuDbg(#i "\n"); \
7900 +       au_dpri_inode(i); \
7901 +       mutex_unlock(&au_dbg_mtx); \
7902 +} while (0)
7903 +
7904 +#define AuDbgDAlias(i) do { \
7905 +       mutex_lock(&au_dbg_mtx); \
7906 +       AuDbg(#i "\n"); \
7907 +       au_dpri_dalias(i); \
7908 +       mutex_unlock(&au_dbg_mtx); \
7909 +} while (0)
7910 +
7911 +#define AuDbgDentry(d) do { \
7912 +       mutex_lock(&au_dbg_mtx); \
7913 +       AuDbg(#d "\n"); \
7914 +       au_dpri_dentry(d); \
7915 +       mutex_unlock(&au_dbg_mtx); \
7916 +} while (0)
7917 +
7918 +#define AuDbgFile(f) do { \
7919 +       mutex_lock(&au_dbg_mtx); \
7920 +       AuDbg(#f "\n"); \
7921 +       au_dpri_file(f); \
7922 +       mutex_unlock(&au_dbg_mtx); \
7923 +} while (0)
7924 +
7925 +#define AuDbgSb(sb) do { \
7926 +       mutex_lock(&au_dbg_mtx); \
7927 +       AuDbg(#sb "\n"); \
7928 +       au_dpri_sb(sb); \
7929 +       mutex_unlock(&au_dbg_mtx); \
7930 +} while (0)
7931 +
7932 +#define AuDbgSym(addr) do {                            \
7933 +       char sym[KSYM_SYMBOL_LEN];                      \
7934 +       sprint_symbol(sym, (unsigned long)addr);        \
7935 +       AuDbg("%s\n", sym);                             \
7936 +} while (0)
7937 +#else
7938 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7939 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7940 +AuStubVoid(au_dbg_verify_kthread, void)
7941 +AuStubInt0(__init au_debug_init, void)
7942 +
7943 +#define AuDbgWhlist(w)         do {} while (0)
7944 +#define AuDbgVdir(v)           do {} while (0)
7945 +#define AuDbgInode(i)          do {} while (0)
7946 +#define AuDbgDAlias(i)         do {} while (0)
7947 +#define AuDbgDentry(d)         do {} while (0)
7948 +#define AuDbgFile(f)           do {} while (0)
7949 +#define AuDbgSb(sb)            do {} while (0)
7950 +#define AuDbgSym(addr)         do {} while (0)
7951 +#endif /* CONFIG_AUFS_DEBUG */
7952 +
7953 +/* ---------------------------------------------------------------------- */
7954 +
7955 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7956 +int __init au_sysrq_init(void);
7957 +void au_sysrq_fin(void);
7958 +
7959 +#ifdef CONFIG_HW_CONSOLE
7960 +#define au_dbg_blocked() do { \
7961 +       WARN_ON(1); \
7962 +       handle_sysrq('w'); \
7963 +} while (0)
7964 +#else
7965 +AuStubVoid(au_dbg_blocked, void)
7966 +#endif
7967 +
7968 +#else
7969 +AuStubInt0(__init au_sysrq_init, void)
7970 +AuStubVoid(au_sysrq_fin, void)
7971 +AuStubVoid(au_dbg_blocked, void)
7972 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7973 +
7974 +#endif /* __KERNEL__ */
7975 +#endif /* __AUFS_DEBUG_H__ */
7976 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7977 --- /usr/share/empty/fs/aufs/dentry.c   1970-01-01 01:00:00.000000000 +0100
7978 +++ linux/fs/aufs/dentry.c      2022-11-05 23:02:18.962555950 +0100
7979 @@ -0,0 +1,1168 @@
7980 +// SPDX-License-Identifier: GPL-2.0
7981 +/*
7982 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7983 + *
7984 + * This program is free software; you can redistribute it and/or modify
7985 + * it under the terms of the GNU General Public License as published by
7986 + * the Free Software Foundation; either version 2 of the License, or
7987 + * (at your option) any later version.
7988 + *
7989 + * This program is distributed in the hope that it will be useful,
7990 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7991 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7992 + * GNU General Public License for more details.
7993 + *
7994 + * You should have received a copy of the GNU General Public License
7995 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7996 + */
7997 +
7998 +/*
7999 + * lookup and dentry operations
8000 + */
8001 +
8002 +#include <linux/iversion.h>
8003 +#include "aufs.h"
8004 +
8005 +/*
8006 + * returns positive/negative dentry, NULL or an error.
8007 + * NULL means whiteout-ed or not-found.
8008 + */
8009 +static struct dentry*
8010 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
8011 +            aufs_bindex_t bindex, struct au_do_lookup_args *args)
8012 +{
8013 +       struct dentry *h_dentry;
8014 +       struct inode *h_inode;
8015 +       struct au_branch *br;
8016 +       struct user_namespace *h_userns;
8017 +       struct path h_path;
8018 +       int wh_found, opq;
8019 +       unsigned char wh_able;
8020 +       const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
8021 +       const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
8022 +                                                         IGNORE_PERM);
8023 +
8024 +       wh_found = 0;
8025 +       br = au_sbr(dentry->d_sb, bindex);
8026 +       h_path.dentry = h_parent;
8027 +       h_path.mnt = au_br_mnt(br);
8028 +       h_userns = au_br_userns(br);
8029 +       wh_able = !!au_br_whable(br->br_perm);
8030 +       if (wh_able)
8031 +               wh_found = au_wh_test(h_userns, &h_path, &args->whname,
8032 +                                     ignore_perm);
8033 +       h_dentry = ERR_PTR(wh_found);
8034 +       if (!wh_found)
8035 +               goto real_lookup;
8036 +       if (unlikely(wh_found < 0))
8037 +               goto out;
8038 +
8039 +       /* We found a whiteout */
8040 +       /* au_set_dbbot(dentry, bindex); */
8041 +       au_set_dbwh(dentry, bindex);
8042 +       if (!allow_neg)
8043 +               return NULL; /* success */
8044 +
8045 +real_lookup:
8046 +       if (!ignore_perm)
8047 +               h_dentry = vfsub_lkup_one(args->name, &h_path);
8048 +       else
8049 +               h_dentry = au_sio_lkup_one(h_userns, args->name, &h_path);
8050 +       if (IS_ERR(h_dentry)) {
8051 +               if (PTR_ERR(h_dentry) == -ENAMETOOLONG
8052 +                   && !allow_neg)
8053 +                       h_dentry = NULL;
8054 +               goto out;
8055 +       }
8056 +
8057 +       h_inode = d_inode(h_dentry);
8058 +       if (d_is_negative(h_dentry)) {
8059 +               if (!allow_neg)
8060 +                       goto out_neg;
8061 +       } else if (wh_found
8062 +                  || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
8063 +               goto out_neg;
8064 +       else if (au_ftest_lkup(args->flags, DIRREN)
8065 +                /* && h_inode */
8066 +                && !au_dr_lkup_h_ino(args, bindex, h_inode->i_ino)) {
8067 +               AuDbg("b%d %pd ignored hi%llu\n", bindex, h_dentry,
8068 +                     (unsigned long long)h_inode->i_ino);
8069 +               goto out_neg;
8070 +       }
8071 +
8072 +       if (au_dbbot(dentry) <= bindex)
8073 +               au_set_dbbot(dentry, bindex);
8074 +       if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
8075 +               au_set_dbtop(dentry, bindex);
8076 +       au_set_h_dptr(dentry, bindex, h_dentry);
8077 +
8078 +       if (!d_is_dir(h_dentry)
8079 +           || !wh_able
8080 +           || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
8081 +               goto out; /* success */
8082 +
8083 +       h_path.dentry = h_dentry;
8084 +       inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
8085 +       opq = au_diropq_test(h_userns, &h_path);
8086 +       inode_unlock_shared(h_inode);
8087 +       if (opq > 0)
8088 +               au_set_dbdiropq(dentry, bindex);
8089 +       else if (unlikely(opq < 0)) {
8090 +               au_set_h_dptr(dentry, bindex, NULL);
8091 +               h_dentry = ERR_PTR(opq);
8092 +       }
8093 +       goto out;
8094 +
8095 +out_neg:
8096 +       dput(h_dentry);
8097 +       h_dentry = NULL;
8098 +out:
8099 +       return h_dentry;
8100 +}
8101 +
8102 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
8103 +{
8104 +       if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
8105 +                    && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
8106 +               return -EPERM;
8107 +       return 0;
8108 +}
8109 +
8110 +/*
8111 + * returns the number of lower positive dentries,
8112 + * otherwise an error.
8113 + * can be called at unlinking with @type is zero.
8114 + */
8115 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
8116 +                  unsigned int flags)
8117 +{
8118 +       int npositive, err;
8119 +       aufs_bindex_t bindex, btail, bdiropq;
8120 +       unsigned char isdir, dirperm1, dirren;
8121 +       struct au_do_lookup_args args = {
8122 +               .flags          = flags,
8123 +               .name           = &dentry->d_name
8124 +       };
8125 +       struct dentry *parent;
8126 +       struct super_block *sb;
8127 +
8128 +       sb = dentry->d_sb;
8129 +       err = au_test_shwh(sb, args.name);
8130 +       if (unlikely(err))
8131 +               goto out;
8132 +
8133 +       err = au_wh_name_alloc(&args.whname, args.name);
8134 +       if (unlikely(err))
8135 +               goto out;
8136 +
8137 +       isdir = !!d_is_dir(dentry);
8138 +       dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
8139 +       dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
8140 +       if (dirren)
8141 +               au_fset_lkup(args.flags, DIRREN);
8142 +
8143 +       npositive = 0;
8144 +       parent = dget_parent(dentry);
8145 +       btail = au_dbtaildir(parent);
8146 +       for (bindex = btop; bindex <= btail; bindex++) {
8147 +               struct dentry *h_parent, *h_dentry;
8148 +               struct inode *h_inode, *h_dir;
8149 +               struct au_branch *br;
8150 +
8151 +               h_dentry = au_h_dptr(dentry, bindex);
8152 +               if (h_dentry) {
8153 +                       if (d_is_positive(h_dentry))
8154 +                               npositive++;
8155 +                       break;
8156 +               }
8157 +               h_parent = au_h_dptr(parent, bindex);
8158 +               if (!h_parent || !d_is_dir(h_parent))
8159 +                       continue;
8160 +
8161 +               if (dirren) {
8162 +                       /* if the inum matches, then use the prepared name */
8163 +                       err = au_dr_lkup_name(&args, bindex);
8164 +                       if (unlikely(err))
8165 +                               goto out_parent;
8166 +               }
8167 +
8168 +               h_dir = d_inode(h_parent);
8169 +               inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
8170 +               h_dentry = au_do_lookup(h_parent, dentry, bindex, &args);
8171 +               inode_unlock_shared(h_dir);
8172 +               err = PTR_ERR(h_dentry);
8173 +               if (IS_ERR(h_dentry))
8174 +                       goto out_parent;
8175 +               if (h_dentry)
8176 +                       au_fclr_lkup(args.flags, ALLOW_NEG);
8177 +               if (dirperm1)
8178 +                       au_fset_lkup(args.flags, IGNORE_PERM);
8179 +
8180 +               if (au_dbwh(dentry) == bindex)
8181 +                       break;
8182 +               if (!h_dentry)
8183 +                       continue;
8184 +               if (d_is_negative(h_dentry))
8185 +                       continue;
8186 +               h_inode = d_inode(h_dentry);
8187 +               npositive++;
8188 +               if (!args.type)
8189 +                       args.type = h_inode->i_mode & S_IFMT;
8190 +               if (args.type != S_IFDIR)
8191 +                       break;
8192 +               else if (isdir) {
8193 +                       /* the type of lower may be different */
8194 +                       bdiropq = au_dbdiropq(dentry);
8195 +                       if (bdiropq >= 0 && bdiropq <= bindex)
8196 +                               break;
8197 +               }
8198 +               br = au_sbr(sb, bindex);
8199 +               if (dirren
8200 +                   && au_dr_hino_test_add(&br->br_dirren, h_inode->i_ino,
8201 +                                          /*add_ent*/NULL)) {
8202 +                       /* prepare next name to lookup */
8203 +                       err = au_dr_lkup(&args, dentry, bindex);
8204 +                       if (unlikely(err))
8205 +                               goto out_parent;
8206 +               }
8207 +       }
8208 +
8209 +       if (npositive) {
8210 +               AuLabel(positive);
8211 +               au_update_dbtop(dentry);
8212 +       }
8213 +       err = npositive;
8214 +       if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
8215 +                    && au_dbtop(dentry) < 0)) {
8216 +               err = -EIO;
8217 +               AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
8218 +                       dentry, err);
8219 +       }
8220 +
8221 +out_parent:
8222 +       dput(parent);
8223 +       au_kfree_try_rcu(args.whname.name);
8224 +       if (dirren)
8225 +               au_dr_lkup_fin(&args);
8226 +out:
8227 +       return err;
8228 +}
8229 +
8230 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
8231 +                              struct path *ppath)
8232 +{
8233 +       struct dentry *dentry;
8234 +       int wkq_err;
8235 +
8236 +       if (!au_test_h_perm_sio(userns, d_inode(ppath->dentry), MAY_EXEC))
8237 +               dentry = vfsub_lkup_one(name, ppath);
8238 +       else {
8239 +               struct vfsub_lkup_one_args args = {
8240 +                       .errp   = &dentry,
8241 +                       .name   = name,
8242 +                       .ppath  = ppath
8243 +               };
8244 +
8245 +               wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
8246 +               if (unlikely(wkq_err))
8247 +                       dentry = ERR_PTR(wkq_err);
8248 +       }
8249 +
8250 +       return dentry;
8251 +}
8252 +
8253 +/*
8254 + * lookup @dentry on @bindex which should be negative.
8255 + */
8256 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
8257 +{
8258 +       int err;
8259 +       struct dentry *parent, *h_dentry;
8260 +       struct au_branch *br;
8261 +       struct user_namespace *h_userns;
8262 +       struct path h_ppath;
8263 +
8264 +       parent = dget_parent(dentry);
8265 +       br = au_sbr(dentry->d_sb, bindex);
8266 +       h_ppath.dentry = au_h_dptr(parent, bindex);
8267 +       h_ppath.mnt = au_br_mnt(br);
8268 +       h_userns = au_br_userns(br);
8269 +       if (wh)
8270 +               h_dentry = au_whtmp_lkup(h_ppath.dentry, br, &dentry->d_name);
8271 +       else
8272 +               h_dentry = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
8273 +       err = PTR_ERR(h_dentry);
8274 +       if (IS_ERR(h_dentry))
8275 +               goto out;
8276 +       if (unlikely(d_is_positive(h_dentry))) {
8277 +               err = -EIO;
8278 +               AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
8279 +               dput(h_dentry);
8280 +               goto out;
8281 +       }
8282 +
8283 +       err = 0;
8284 +       if (bindex < au_dbtop(dentry))
8285 +               au_set_dbtop(dentry, bindex);
8286 +       if (au_dbbot(dentry) < bindex)
8287 +               au_set_dbbot(dentry, bindex);
8288 +       au_set_h_dptr(dentry, bindex, h_dentry);
8289 +
8290 +out:
8291 +       dput(parent);
8292 +       return err;
8293 +}
8294 +
8295 +/* ---------------------------------------------------------------------- */
8296 +
8297 +/* subset of struct inode */
8298 +struct au_iattr {
8299 +       unsigned long           i_ino;
8300 +       /* unsigned int         i_nlink; */
8301 +       kuid_t                  i_uid;
8302 +       kgid_t                  i_gid;
8303 +       u64                     i_version;
8304 +/*
8305 +       loff_t                  i_size;
8306 +       blkcnt_t                i_blocks;
8307 +*/
8308 +       umode_t                 i_mode;
8309 +};
8310 +
8311 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
8312 +{
8313 +       ia->i_ino = h_inode->i_ino;
8314 +       /* ia->i_nlink = h_inode->i_nlink; */
8315 +       ia->i_uid = h_inode->i_uid;
8316 +       ia->i_gid = h_inode->i_gid;
8317 +       ia->i_version = inode_query_iversion(h_inode);
8318 +/*
8319 +       ia->i_size = h_inode->i_size;
8320 +       ia->i_blocks = h_inode->i_blocks;
8321 +*/
8322 +       ia->i_mode = (h_inode->i_mode & S_IFMT);
8323 +}
8324 +
8325 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
8326 +{
8327 +       return ia->i_ino != h_inode->i_ino
8328 +               /* || ia->i_nlink != h_inode->i_nlink */
8329 +               || !uid_eq(ia->i_uid, h_inode->i_uid)
8330 +               || !gid_eq(ia->i_gid, h_inode->i_gid)
8331 +               || !inode_eq_iversion(h_inode, ia->i_version)
8332 +/*
8333 +               || ia->i_size != h_inode->i_size
8334 +               || ia->i_blocks != h_inode->i_blocks
8335 +*/
8336 +               || ia->i_mode != (h_inode->i_mode & S_IFMT);
8337 +}
8338 +
8339 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
8340 +                             struct au_branch *br)
8341 +{
8342 +       int err;
8343 +       struct au_iattr ia;
8344 +       struct inode *h_inode;
8345 +       struct dentry *h_d;
8346 +       struct super_block *h_sb;
8347 +       struct path h_ppath;
8348 +
8349 +       err = 0;
8350 +       memset(&ia, -1, sizeof(ia));
8351 +       h_sb = h_dentry->d_sb;
8352 +       h_inode = NULL;
8353 +       if (d_is_positive(h_dentry)) {
8354 +               h_inode = d_inode(h_dentry);
8355 +               au_iattr_save(&ia, h_inode);
8356 +       } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
8357 +               /* nfs d_revalidate may return 0 for negative dentry */
8358 +               /* fuse d_revalidate always return 0 for negative dentry */
8359 +               goto out;
8360 +
8361 +       /* main purpose is namei.c:cached_lookup() and d_revalidate */
8362 +       h_ppath.dentry = h_parent;
8363 +       h_ppath.mnt = au_br_mnt(br);
8364 +       h_d = vfsub_lkup_one(&h_dentry->d_name, &h_ppath);
8365 +       err = PTR_ERR(h_d);
8366 +       if (IS_ERR(h_d))
8367 +               goto out;
8368 +
8369 +       err = 0;
8370 +       if (unlikely(h_d != h_dentry
8371 +                    || d_inode(h_d) != h_inode
8372 +                    || (h_inode && au_iattr_test(&ia, h_inode))))
8373 +               err = au_busy_or_stale();
8374 +       dput(h_d);
8375 +
8376 +out:
8377 +       AuTraceErr(err);
8378 +       return err;
8379 +}
8380 +
8381 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8382 +               struct dentry *h_parent, struct au_branch *br)
8383 +{
8384 +       int err;
8385 +
8386 +       err = 0;
8387 +       if (udba == AuOpt_UDBA_REVAL
8388 +           && !au_test_fs_remote(h_dentry->d_sb)) {
8389 +               IMustLock(h_dir);
8390 +               err = (d_inode(h_dentry->d_parent) != h_dir);
8391 +       } else if (udba != AuOpt_UDBA_NONE)
8392 +               err = au_h_verify_dentry(h_dentry, h_parent, br);
8393 +
8394 +       return err;
8395 +}
8396 +
8397 +/* ---------------------------------------------------------------------- */
8398 +
8399 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
8400 +{
8401 +       int err;
8402 +       aufs_bindex_t new_bindex, bindex, bbot, bwh, bdiropq;
8403 +       struct au_hdentry tmp, *p, *q;
8404 +       struct au_dinfo *dinfo;
8405 +       struct super_block *sb;
8406 +
8407 +       DiMustWriteLock(dentry);
8408 +
8409 +       sb = dentry->d_sb;
8410 +       dinfo = au_di(dentry);
8411 +       bbot = dinfo->di_bbot;
8412 +       bwh = dinfo->di_bwh;
8413 +       bdiropq = dinfo->di_bdiropq;
8414 +       bindex = dinfo->di_btop;
8415 +       p = au_hdentry(dinfo, bindex);
8416 +       for (; bindex <= bbot; bindex++, p++) {
8417 +               if (!p->hd_dentry)
8418 +                       continue;
8419 +
8420 +               new_bindex = au_br_index(sb, p->hd_id);
8421 +               if (new_bindex == bindex)
8422 +                       continue;
8423 +
8424 +               if (dinfo->di_bwh == bindex)
8425 +                       bwh = new_bindex;
8426 +               if (dinfo->di_bdiropq == bindex)
8427 +                       bdiropq = new_bindex;
8428 +               if (new_bindex < 0) {
8429 +                       au_hdput(p);
8430 +                       p->hd_dentry = NULL;
8431 +                       continue;
8432 +               }
8433 +
8434 +               /* swap two lower dentries, and loop again */
8435 +               q = au_hdentry(dinfo, new_bindex);
8436 +               tmp = *q;
8437 +               *q = *p;
8438 +               *p = tmp;
8439 +               if (tmp.hd_dentry) {
8440 +                       bindex--;
8441 +                       p--;
8442 +               }
8443 +       }
8444 +
8445 +       dinfo->di_bwh = -1;
8446 +       if (bwh >= 0 && bwh <= au_sbbot(sb) && au_sbr_whable(sb, bwh))
8447 +               dinfo->di_bwh = bwh;
8448 +
8449 +       dinfo->di_bdiropq = -1;
8450 +       if (bdiropq >= 0
8451 +           && bdiropq <= au_sbbot(sb)
8452 +           && au_sbr_whable(sb, bdiropq))
8453 +               dinfo->di_bdiropq = bdiropq;
8454 +
8455 +       err = -EIO;
8456 +       dinfo->di_btop = -1;
8457 +       dinfo->di_bbot = -1;
8458 +       bbot = au_dbbot(parent);
8459 +       bindex = 0;
8460 +       p = au_hdentry(dinfo, bindex);
8461 +       for (; bindex <= bbot; bindex++, p++)
8462 +               if (p->hd_dentry) {
8463 +                       dinfo->di_btop = bindex;
8464 +                       break;
8465 +               }
8466 +
8467 +       if (dinfo->di_btop >= 0) {
8468 +               bindex = bbot;
8469 +               p = au_hdentry(dinfo, bindex);
8470 +               for (; bindex >= 0; bindex--, p--)
8471 +                       if (p->hd_dentry) {
8472 +                               dinfo->di_bbot = bindex;
8473 +                               err = 0;
8474 +                               break;
8475 +                       }
8476 +       }
8477 +
8478 +       return err;
8479 +}
8480 +
8481 +static void au_do_hide(struct dentry *dentry)
8482 +{
8483 +       struct inode *inode;
8484 +
8485 +       if (d_really_is_positive(dentry)) {
8486 +               inode = d_inode(dentry);
8487 +               if (!d_is_dir(dentry)) {
8488 +                       if (inode->i_nlink && !d_unhashed(dentry))
8489 +                               drop_nlink(inode);
8490 +               } else {
8491 +                       clear_nlink(inode);
8492 +                       /* stop next lookup */
8493 +                       inode->i_flags |= S_DEAD;
8494 +               }
8495 +               smp_mb(); /* necessary? */
8496 +       }
8497 +       d_drop(dentry);
8498 +}
8499 +
8500 +static int au_hide_children(struct dentry *parent)
8501 +{
8502 +       int err, i, j, ndentry;
8503 +       struct au_dcsub_pages dpages;
8504 +       struct au_dpage *dpage;
8505 +       struct dentry *dentry;
8506 +
8507 +       err = au_dpages_init(&dpages, GFP_NOFS);
8508 +       if (unlikely(err))
8509 +               goto out;
8510 +       err = au_dcsub_pages(&dpages, parent, NULL, NULL);
8511 +       if (unlikely(err))
8512 +               goto out_dpages;
8513 +
8514 +       /* in reverse order */
8515 +       for (i = dpages.ndpage - 1; i >= 0; i--) {
8516 +               dpage = dpages.dpages + i;
8517 +               ndentry = dpage->ndentry;
8518 +               for (j = ndentry - 1; j >= 0; j--) {
8519 +                       dentry = dpage->dentries[j];
8520 +                       if (dentry != parent)
8521 +                               au_do_hide(dentry);
8522 +               }
8523 +       }
8524 +
8525 +out_dpages:
8526 +       au_dpages_free(&dpages);
8527 +out:
8528 +       return err;
8529 +}
8530 +
8531 +static void au_hide(struct dentry *dentry)
8532 +{
8533 +       int err;
8534 +
8535 +       AuDbgDentry(dentry);
8536 +       if (d_is_dir(dentry)) {
8537 +               /* shrink_dcache_parent(dentry); */
8538 +               err = au_hide_children(dentry);
8539 +               if (unlikely(err))
8540 +                       AuIOErr("%pd, failed hiding children, ignored %d\n",
8541 +                               dentry, err);
8542 +       }
8543 +       au_do_hide(dentry);
8544 +}
8545 +
8546 +/*
8547 + * By adding a dirty branch, a cached dentry may be affected in various ways.
8548 + *
8549 + * a dirty branch is added
8550 + * - on the top of layers
8551 + * - in the middle of layers
8552 + * - to the bottom of layers
8553 + *
8554 + * on the added branch there exists
8555 + * - a whiteout
8556 + * - a diropq
8557 + * - a same named entry
8558 + *   + exist
8559 + *     * negative --> positive
8560 + *     * positive --> positive
8561 + *      - type is unchanged
8562 + *      - type is changed
8563 + *   + doesn't exist
8564 + *     * negative --> negative
8565 + *     * positive --> negative (rejected by au_br_del() for non-dir case)
8566 + * - none
8567 + */
8568 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8569 +                              struct au_dinfo *tmp)
8570 +{
8571 +       int err;
8572 +       aufs_bindex_t bindex, bbot;
8573 +       struct {
8574 +               struct dentry *dentry;
8575 +               struct inode *inode;
8576 +               mode_t mode;
8577 +       } orig_h, tmp_h = {
8578 +               .dentry = NULL
8579 +       };
8580 +       struct au_hdentry *hd;
8581 +       struct inode *inode, *h_inode;
8582 +       struct dentry *h_dentry;
8583 +
8584 +       err = 0;
8585 +       AuDebugOn(dinfo->di_btop < 0);
8586 +       orig_h.mode = 0;
8587 +       orig_h.dentry = au_hdentry(dinfo, dinfo->di_btop)->hd_dentry;
8588 +       orig_h.inode = NULL;
8589 +       if (d_is_positive(orig_h.dentry)) {
8590 +               orig_h.inode = d_inode(orig_h.dentry);
8591 +               orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8592 +       }
8593 +       if (tmp->di_btop >= 0) {
8594 +               tmp_h.dentry = au_hdentry(tmp, tmp->di_btop)->hd_dentry;
8595 +               if (d_is_positive(tmp_h.dentry)) {
8596 +                       tmp_h.inode = d_inode(tmp_h.dentry);
8597 +                       tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8598 +               }
8599 +       }
8600 +
8601 +       inode = NULL;
8602 +       if (d_really_is_positive(dentry))
8603 +               inode = d_inode(dentry);
8604 +       if (!orig_h.inode) {
8605 +               AuDbg("negative originally\n");
8606 +               if (inode) {
8607 +                       au_hide(dentry);
8608 +                       goto out;
8609 +               }
8610 +               AuDebugOn(inode);
8611 +               AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8612 +               AuDebugOn(dinfo->di_bdiropq != -1);
8613 +
8614 +               if (!tmp_h.inode) {
8615 +                       AuDbg("negative --> negative\n");
8616 +                       /* should have only one negative lower */
8617 +                       if (tmp->di_btop >= 0
8618 +                           && tmp->di_btop < dinfo->di_btop) {
8619 +                               AuDebugOn(tmp->di_btop != tmp->di_bbot);
8620 +                               AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8621 +                               au_set_h_dptr(dentry, dinfo->di_btop, NULL);
8622 +                               au_di_cp(dinfo, tmp);
8623 +                               hd = au_hdentry(tmp, tmp->di_btop);
8624 +                               au_set_h_dptr(dentry, tmp->di_btop,
8625 +                                             dget(hd->hd_dentry));
8626 +                       }
8627 +                       au_dbg_verify_dinode(dentry);
8628 +               } else {
8629 +                       AuDbg("negative --> positive\n");
8630 +                       /*
8631 +                        * similar to the behaviour of creating with bypassing
8632 +                        * aufs.
8633 +                        * unhash it in order to force an error in the
8634 +                        * succeeding create operation.
8635 +                        * we should not set S_DEAD here.
8636 +                        */
8637 +                       d_drop(dentry);
8638 +                       /* au_di_swap(tmp, dinfo); */
8639 +                       au_dbg_verify_dinode(dentry);
8640 +               }
8641 +       } else {
8642 +               AuDbg("positive originally\n");
8643 +               /* inode may be NULL */
8644 +               AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8645 +               if (!tmp_h.inode) {
8646 +                       AuDbg("positive --> negative\n");
8647 +                       /* or bypassing aufs */
8648 +                       au_hide(dentry);
8649 +                       if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_btop)
8650 +                               dinfo->di_bwh = tmp->di_bwh;
8651 +                       if (inode)
8652 +                               err = au_refresh_hinode_self(inode);
8653 +                       au_dbg_verify_dinode(dentry);
8654 +               } else if (orig_h.mode == tmp_h.mode) {
8655 +                       AuDbg("positive --> positive, same type\n");
8656 +                       if (!S_ISDIR(orig_h.mode)
8657 +                           && dinfo->di_btop > tmp->di_btop) {
8658 +                               /*
8659 +                                * similar to the behaviour of removing and
8660 +                                * creating.
8661 +                                */
8662 +                               au_hide(dentry);
8663 +                               if (inode)
8664 +                                       err = au_refresh_hinode_self(inode);
8665 +                               au_dbg_verify_dinode(dentry);
8666 +                       } else {
8667 +                               /* fill empty slots */
8668 +                               if (dinfo->di_btop > tmp->di_btop)
8669 +                                       dinfo->di_btop = tmp->di_btop;
8670 +                               if (dinfo->di_bbot < tmp->di_bbot)
8671 +                                       dinfo->di_bbot = tmp->di_bbot;
8672 +                               dinfo->di_bwh = tmp->di_bwh;
8673 +                               dinfo->di_bdiropq = tmp->di_bdiropq;
8674 +                               bbot = dinfo->di_bbot;
8675 +                               bindex = tmp->di_btop;
8676 +                               hd = au_hdentry(tmp, bindex);
8677 +                               for (; bindex <= bbot; bindex++, hd++) {
8678 +                                       if (au_h_dptr(dentry, bindex))
8679 +                                               continue;
8680 +                                       h_dentry = hd->hd_dentry;
8681 +                                       if (!h_dentry)
8682 +                                               continue;
8683 +                                       AuDebugOn(d_is_negative(h_dentry));
8684 +                                       h_inode = d_inode(h_dentry);
8685 +                                       AuDebugOn(orig_h.mode
8686 +                                                 != (h_inode->i_mode
8687 +                                                     & S_IFMT));
8688 +                                       au_set_h_dptr(dentry, bindex,
8689 +                                                     dget(h_dentry));
8690 +                               }
8691 +                               if (inode)
8692 +                                       err = au_refresh_hinode(inode, dentry);
8693 +                               au_dbg_verify_dinode(dentry);
8694 +                       }
8695 +               } else {
8696 +                       AuDbg("positive --> positive, different type\n");
8697 +                       /* similar to the behaviour of removing and creating */
8698 +                       au_hide(dentry);
8699 +                       if (inode)
8700 +                               err = au_refresh_hinode_self(inode);
8701 +                       au_dbg_verify_dinode(dentry);
8702 +               }
8703 +       }
8704 +
8705 +out:
8706 +       return err;
8707 +}
8708 +
8709 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8710 +{
8711 +       const struct dentry_operations *dop
8712 +               = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8713 +       static const unsigned int mask
8714 +               = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8715 +
8716 +       BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8717 +
8718 +       if (dentry->d_op == dop)
8719 +               return;
8720 +
8721 +       AuDbg("%pd\n", dentry);
8722 +       spin_lock(&dentry->d_lock);
8723 +       if (dop == &aufs_dop)
8724 +               dentry->d_flags |= mask;
8725 +       else
8726 +               dentry->d_flags &= ~mask;
8727 +       dentry->d_op = dop;
8728 +       spin_unlock(&dentry->d_lock);
8729 +}
8730 +
8731 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8732 +{
8733 +       int err, ebrange, nbr;
8734 +       unsigned int sigen;
8735 +       struct au_dinfo *dinfo, *tmp;
8736 +       struct super_block *sb;
8737 +       struct inode *inode;
8738 +
8739 +       DiMustWriteLock(dentry);
8740 +       AuDebugOn(IS_ROOT(dentry));
8741 +       AuDebugOn(d_really_is_negative(parent));
8742 +
8743 +       sb = dentry->d_sb;
8744 +       sigen = au_sigen(sb);
8745 +       err = au_digen_test(parent, sigen);
8746 +       if (unlikely(err))
8747 +               goto out;
8748 +
8749 +       nbr = au_sbbot(sb) + 1;
8750 +       dinfo = au_di(dentry);
8751 +       err = au_di_realloc(dinfo, nbr, /*may_shrink*/0);
8752 +       if (unlikely(err))
8753 +               goto out;
8754 +       ebrange = au_dbrange_test(dentry);
8755 +       if (!ebrange)
8756 +               ebrange = au_do_refresh_hdentry(dentry, parent);
8757 +
8758 +       if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8759 +               AuDebugOn(au_dbtop(dentry) < 0 && au_dbbot(dentry) >= 0);
8760 +               if (d_really_is_positive(dentry)) {
8761 +                       inode = d_inode(dentry);
8762 +                       err = au_refresh_hinode_self(inode);
8763 +               }
8764 +               au_dbg_verify_dinode(dentry);
8765 +               if (!err)
8766 +                       goto out_dgen; /* success */
8767 +               goto out;
8768 +       }
8769 +
8770 +       /* temporary dinfo */
8771 +       AuDbgDentry(dentry);
8772 +       err = -ENOMEM;
8773 +       tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8774 +       if (unlikely(!tmp))
8775 +               goto out;
8776 +       au_di_swap(tmp, dinfo);
8777 +       /* returns the number of positive dentries */
8778 +       /*
8779 +        * if current working dir is removed, it returns an error.
8780 +        * but the dentry is legal.
8781 +        */
8782 +       err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
8783 +       AuDbgDentry(dentry);
8784 +       au_di_swap(tmp, dinfo);
8785 +       if (err == -ENOENT)
8786 +               err = 0;
8787 +       if (err >= 0) {
8788 +               /* compare/refresh by dinfo */
8789 +               AuDbgDentry(dentry);
8790 +               err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8791 +               au_dbg_verify_dinode(dentry);
8792 +               AuTraceErr(err);
8793 +       }
8794 +       au_di_realloc(dinfo, nbr, /*may_shrink*/1); /* harmless if err */
8795 +       au_rw_write_unlock(&tmp->di_rwsem);
8796 +       au_di_free(tmp);
8797 +       if (unlikely(err))
8798 +               goto out;
8799 +
8800 +out_dgen:
8801 +       au_update_digen(dentry);
8802 +out:
8803 +       if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8804 +               AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8805 +               AuDbgDentry(dentry);
8806 +       }
8807 +       AuTraceErr(err);
8808 +       return err;
8809 +}
8810 +
8811 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8812 +                          struct dentry *dentry, aufs_bindex_t bindex)
8813 +{
8814 +       int err, valid;
8815 +
8816 +       err = 0;
8817 +       if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8818 +               goto out;
8819 +
8820 +       AuDbg("b%d\n", bindex);
8821 +       /*
8822 +        * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8823 +        * due to whiteout and branch permission.
8824 +        */
8825 +       flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8826 +                  | LOOKUP_FOLLOW | LOOKUP_EXCL);
8827 +       /* it may return tri-state */
8828 +       valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8829 +
8830 +       if (unlikely(valid < 0))
8831 +               err = valid;
8832 +       else if (!valid)
8833 +               err = -EINVAL;
8834 +
8835 +out:
8836 +       AuTraceErr(err);
8837 +       return err;
8838 +}
8839 +
8840 +/* todo: remove this */
8841 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8842 +                         unsigned int flags, int do_udba, int dirren)
8843 +{
8844 +       int err;
8845 +       umode_t mode, h_mode;
8846 +       aufs_bindex_t bindex, btail, btop, ibs, ibe;
8847 +       unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8848 +       struct inode *h_inode, *h_cached_inode;
8849 +       struct dentry *h_dentry;
8850 +       struct qstr *name, *h_name;
8851 +
8852 +       err = 0;
8853 +       plus = 0;
8854 +       mode = 0;
8855 +       ibs = -1;
8856 +       ibe = -1;
8857 +       unhashed = !!d_unhashed(dentry);
8858 +       is_root = !!IS_ROOT(dentry);
8859 +       name = &dentry->d_name;
8860 +       tmpfile = au_di(dentry)->di_tmpfile;
8861 +
8862 +       /*
8863 +        * Theoretically, REVAL test should be unnecessary in case of
8864 +        * {FS,I}NOTIFY.
8865 +        * But {fs,i}notify doesn't fire some necessary events,
8866 +        *      IN_ATTRIB for atime/nlink/pageio
8867 +        * Let's do REVAL test too.
8868 +        */
8869 +       if (do_udba && inode) {
8870 +               mode = (inode->i_mode & S_IFMT);
8871 +               plus = (inode->i_nlink > 0);
8872 +               ibs = au_ibtop(inode);
8873 +               ibe = au_ibbot(inode);
8874 +       }
8875 +
8876 +       btop = au_dbtop(dentry);
8877 +       btail = btop;
8878 +       if (inode && S_ISDIR(inode->i_mode))
8879 +               btail = au_dbtaildir(dentry);
8880 +       for (bindex = btop; bindex <= btail; bindex++) {
8881 +               h_dentry = au_h_dptr(dentry, bindex);
8882 +               if (!h_dentry)
8883 +                       continue;
8884 +
8885 +               AuDbg("b%d, %pd\n", bindex, h_dentry);
8886 +               h_nfs = !!au_test_nfs(h_dentry->d_sb);
8887 +               spin_lock(&h_dentry->d_lock);
8888 +               h_name = &h_dentry->d_name;
8889 +               if (unlikely(do_udba
8890 +                            && !is_root
8891 +                            && ((!h_nfs
8892 +                                 && (unhashed != !!d_unhashed(h_dentry)
8893 +                                     || (!tmpfile && !dirren
8894 +                                         && !au_qstreq(name, h_name))
8895 +                                         ))
8896 +                                || (h_nfs
8897 +                                    && !(flags & LOOKUP_OPEN)
8898 +                                    && (h_dentry->d_flags
8899 +                                        & DCACHE_NFSFS_RENAMED)))
8900 +                           )) {
8901 +                       int h_unhashed;
8902 +
8903 +                       h_unhashed = d_unhashed(h_dentry);
8904 +                       spin_unlock(&h_dentry->d_lock);
8905 +                       AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8906 +                             unhashed, h_unhashed, dentry, h_dentry);
8907 +                       goto err;
8908 +               }
8909 +               spin_unlock(&h_dentry->d_lock);
8910 +
8911 +               err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8912 +               if (unlikely(err))
8913 +                       /* do not goto err, to keep the errno */
8914 +                       break;
8915 +
8916 +               /* todo: plink too? */
8917 +               if (!do_udba)
8918 +                       continue;
8919 +
8920 +               /* UDBA tests */
8921 +               if (unlikely(!!inode != d_is_positive(h_dentry)))
8922 +                       goto err;
8923 +
8924 +               h_inode = NULL;
8925 +               if (d_is_positive(h_dentry))
8926 +                       h_inode = d_inode(h_dentry);
8927 +               h_plus = plus;
8928 +               h_mode = mode;
8929 +               h_cached_inode = h_inode;
8930 +               if (h_inode) {
8931 +                       h_mode = (h_inode->i_mode & S_IFMT);
8932 +                       h_plus = (h_inode->i_nlink > 0);
8933 +               }
8934 +               if (inode && ibs <= bindex && bindex <= ibe)
8935 +                       h_cached_inode = au_h_iptr(inode, bindex);
8936 +
8937 +               if (!h_nfs) {
8938 +                       if (unlikely(plus != h_plus && !tmpfile))
8939 +                               goto err;
8940 +               } else {
8941 +                       if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8942 +                                    && !is_root
8943 +                                    && !IS_ROOT(h_dentry)
8944 +                                    && unhashed != d_unhashed(h_dentry)))
8945 +                               goto err;
8946 +               }
8947 +               if (unlikely(mode != h_mode
8948 +                            || h_cached_inode != h_inode))
8949 +                       goto err;
8950 +               continue;
8951 +
8952 +err:
8953 +               err = -EINVAL;
8954 +               break;
8955 +       }
8956 +
8957 +       AuTraceErr(err);
8958 +       return err;
8959 +}
8960 +
8961 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8962 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8963 +{
8964 +       int err;
8965 +       struct dentry *parent;
8966 +
8967 +       if (!au_digen_test(dentry, sigen))
8968 +               return 0;
8969 +
8970 +       parent = dget_parent(dentry);
8971 +       di_read_lock_parent(parent, AuLock_IR);
8972 +       AuDebugOn(au_digen_test(parent, sigen));
8973 +       au_dbg_verify_gen(parent, sigen);
8974 +       err = au_refresh_dentry(dentry, parent);
8975 +       di_read_unlock(parent, AuLock_IR);
8976 +       dput(parent);
8977 +       AuTraceErr(err);
8978 +       return err;
8979 +}
8980 +
8981 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8982 +{
8983 +       int err;
8984 +       struct dentry *d, *parent;
8985 +
8986 +       if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8987 +               return simple_reval_dpath(dentry, sigen);
8988 +
8989 +       /* slow loop, keep it simple and stupid */
8990 +       /* cf: au_cpup_dirs() */
8991 +       err = 0;
8992 +       parent = NULL;
8993 +       while (au_digen_test(dentry, sigen)) {
8994 +               d = dentry;
8995 +               while (1) {
8996 +                       dput(parent);
8997 +                       parent = dget_parent(d);
8998 +                       if (!au_digen_test(parent, sigen))
8999 +                               break;
9000 +                       d = parent;
9001 +               }
9002 +
9003 +               if (d != dentry)
9004 +                       di_write_lock_child2(d);
9005 +
9006 +               /* someone might update our dentry while we were sleeping */
9007 +               if (au_digen_test(d, sigen)) {
9008 +                       /*
9009 +                        * todo: consolidate with simple_reval_dpath(),
9010 +                        * do_refresh() and au_reval_for_attr().
9011 +                        */
9012 +                       di_read_lock_parent(parent, AuLock_IR);
9013 +                       err = au_refresh_dentry(d, parent);
9014 +                       di_read_unlock(parent, AuLock_IR);
9015 +               }
9016 +
9017 +               if (d != dentry)
9018 +                       di_write_unlock(d);
9019 +               dput(parent);
9020 +               if (unlikely(err))
9021 +                       break;
9022 +       }
9023 +
9024 +       return err;
9025 +}
9026 +
9027 +/*
9028 + * if valid returns 1, otherwise 0.
9029 + */
9030 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
9031 +{
9032 +       int valid, err;
9033 +       unsigned int sigen;
9034 +       unsigned char do_udba, dirren;
9035 +       struct super_block *sb;
9036 +       struct inode *inode;
9037 +
9038 +       /* todo: support rcu-walk? */
9039 +       if (flags & LOOKUP_RCU)
9040 +               return -ECHILD;
9041 +
9042 +       valid = 0;
9043 +       if (unlikely(!au_di(dentry)))
9044 +               goto out;
9045 +
9046 +       valid = 1;
9047 +       sb = dentry->d_sb;
9048 +       /*
9049 +        * todo: very ugly
9050 +        * i_mutex of parent dir may be held,
9051 +        * but we should not return 'invalid' due to busy.
9052 +        */
9053 +       err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
9054 +       if (unlikely(err)) {
9055 +               valid = err;
9056 +               AuTraceErr(err);
9057 +               goto out;
9058 +       }
9059 +       inode = NULL;
9060 +       if (d_really_is_positive(dentry))
9061 +               inode = d_inode(dentry);
9062 +       if (unlikely(inode && au_is_bad_inode(inode))) {
9063 +               err = -EINVAL;
9064 +               AuTraceErr(err);
9065 +               goto out_dgrade;
9066 +       }
9067 +       if (unlikely(au_dbrange_test(dentry))) {
9068 +               err = -EINVAL;
9069 +               AuTraceErr(err);
9070 +               goto out_dgrade;
9071 +       }
9072 +
9073 +       sigen = au_sigen(sb);
9074 +       if (au_digen_test(dentry, sigen)) {
9075 +               AuDebugOn(IS_ROOT(dentry));
9076 +               err = au_reval_dpath(dentry, sigen);
9077 +               if (unlikely(err)) {
9078 +                       AuTraceErr(err);
9079 +                       goto out_dgrade;
9080 +               }
9081 +       }
9082 +       di_downgrade_lock(dentry, AuLock_IR);
9083 +
9084 +       err = -EINVAL;
9085 +       if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
9086 +           && inode
9087 +           && !(inode->i_state && I_LINKABLE)
9088 +           && (IS_DEADDIR(inode) || !inode->i_nlink)) {
9089 +               AuTraceErr(err);
9090 +               goto out_inval;
9091 +       }
9092 +
9093 +       do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
9094 +       if (do_udba && inode) {
9095 +               aufs_bindex_t btop = au_ibtop(inode);
9096 +               struct inode *h_inode;
9097 +
9098 +               if (btop >= 0) {
9099 +                       h_inode = au_h_iptr(inode, btop);
9100 +                       if (h_inode && au_test_higen(inode, h_inode)) {
9101 +                               AuTraceErr(err);
9102 +                               goto out_inval;
9103 +                       }
9104 +               }
9105 +       }
9106 +
9107 +       dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
9108 +       err = h_d_revalidate(dentry, inode, flags, do_udba, dirren);
9109 +       if (unlikely(!err && do_udba && au_dbtop(dentry) < 0)) {
9110 +               err = -EIO;
9111 +               AuDbg("both of real entry and whiteout found, %p, err %d\n",
9112 +                     dentry, err);
9113 +       }
9114 +       goto out_inval;
9115 +
9116 +out_dgrade:
9117 +       di_downgrade_lock(dentry, AuLock_IR);
9118 +out_inval:
9119 +       aufs_read_unlock(dentry, AuLock_IR);
9120 +       AuTraceErr(err);
9121 +       valid = !err;
9122 +out:
9123 +       if (!valid) {
9124 +               AuDbg("%pd invalid, %d\n", dentry, valid);
9125 +               d_drop(dentry);
9126 +       }
9127 +       return valid;
9128 +}
9129 +
9130 +static void aufs_d_release(struct dentry *dentry)
9131 +{
9132 +       if (au_di(dentry)) {
9133 +               au_di_fin(dentry);
9134 +               au_hn_di_reinit(dentry);
9135 +       }
9136 +}
9137 +
9138 +const struct dentry_operations aufs_dop = {
9139 +       .d_revalidate           = aufs_d_revalidate,
9140 +       .d_weak_revalidate      = aufs_d_revalidate,
9141 +       .d_release              = aufs_d_release
9142 +};
9143 +
9144 +/* aufs_dop without d_revalidate */
9145 +const struct dentry_operations aufs_dop_noreval = {
9146 +       .d_release              = aufs_d_release
9147 +};
9148 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
9149 --- /usr/share/empty/fs/aufs/dentry.h   1970-01-01 01:00:00.000000000 +0100
9150 +++ linux/fs/aufs/dentry.h      2022-12-17 09:21:34.796521861 +0100
9151 @@ -0,0 +1,270 @@
9152 +/* SPDX-License-Identifier: GPL-2.0 */
9153 +/*
9154 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9155 + *
9156 + * This program is free software; you can redistribute it and/or modify
9157 + * it under the terms of the GNU General Public License as published by
9158 + * the Free Software Foundation; either version 2 of the License, or
9159 + * (at your option) any later version.
9160 + *
9161 + * This program is distributed in the hope that it will be useful,
9162 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9163 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9164 + * GNU General Public License for more details.
9165 + *
9166 + * You should have received a copy of the GNU General Public License
9167 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
9168 + */
9169 +
9170 +/*
9171 + * lookup and dentry operations
9172 + */
9173 +
9174 +#ifndef __AUFS_DENTRY_H__
9175 +#define __AUFS_DENTRY_H__
9176 +
9177 +#ifdef __KERNEL__
9178 +
9179 +#include <linux/dcache.h>
9180 +#include "dirren.h"
9181 +#include "rwsem.h"
9182 +
9183 +struct au_hdentry {
9184 +       struct dentry           *hd_dentry;
9185 +       aufs_bindex_t           hd_id;
9186 +};
9187 +
9188 +struct au_dinfo {
9189 +       atomic_t                di_generation;
9190 +
9191 +       struct au_rwsem         di_rwsem;
9192 +       aufs_bindex_t           di_btop, di_bbot, di_bwh, di_bdiropq;
9193 +       unsigned char           di_tmpfile; /* to allow the different name */
9194 +       struct au_hdentry       *di_hdentry;
9195 +       struct file             *di_htmpfile;
9196 +       struct rcu_head         rcu;
9197 +} ____cacheline_aligned_in_smp;
9198 +
9199 +/* ---------------------------------------------------------------------- */
9200 +
9201 +/* flags for au_lkup_dentry() */
9202 +#define AuLkup_ALLOW_NEG       1
9203 +#define AuLkup_IGNORE_PERM     (1 << 1)
9204 +#define AuLkup_DIRREN          (1 << 2)
9205 +#define au_ftest_lkup(flags, name)     ((flags) & AuLkup_##name)
9206 +#define au_fset_lkup(flags, name) \
9207 +       do { (flags) |= AuLkup_##name; } while (0)
9208 +#define au_fclr_lkup(flags, name) \
9209 +       do { (flags) &= ~AuLkup_##name; } while (0)
9210 +
9211 +#ifndef CONFIG_AUFS_DIRREN
9212 +#undef AuLkup_DIRREN
9213 +#define AuLkup_DIRREN 0
9214 +#endif
9215 +
9216 +struct au_do_lookup_args {
9217 +       unsigned int            flags;
9218 +       mode_t                  type;
9219 +       struct qstr             whname, *name;
9220 +       struct au_dr_lookup     dirren;
9221 +};
9222 +
9223 +/* ---------------------------------------------------------------------- */
9224 +
9225 +/* dentry.c */
9226 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
9227 +struct au_branch;
9228 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
9229 +                              struct path *ppath);
9230 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
9231 +               struct dentry *h_parent, struct au_branch *br);
9232 +
9233 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
9234 +                  unsigned int flags);
9235 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
9236 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
9237 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
9238 +void au_refresh_dop(struct dentry *dentry, int force_reval);
9239 +
9240 +/* dinfo.c */
9241 +void au_di_init_once(void *_di);
9242 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
9243 +void au_di_free(struct au_dinfo *dinfo);
9244 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
9245 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
9246 +int au_di_init(struct dentry *dentry);
9247 +void au_di_fin(struct dentry *dentry);
9248 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink);
9249 +
9250 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
9251 +void di_read_unlock(struct dentry *d, int flags);
9252 +void di_downgrade_lock(struct dentry *d, int flags);
9253 +void di_write_lock(struct dentry *d, unsigned int lsc);
9254 +void di_write_unlock(struct dentry *d);
9255 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
9256 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
9257 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
9258 +
9259 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
9260 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
9261 +aufs_bindex_t au_dbtail(struct dentry *dentry);
9262 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
9263 +
9264 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9265 +                  struct dentry *h_dentry);
9266 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
9267 +int au_dbrange_test(struct dentry *dentry);
9268 +void au_update_digen(struct dentry *dentry);
9269 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
9270 +void au_update_dbtop(struct dentry *dentry);
9271 +void au_update_dbbot(struct dentry *dentry);
9272 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
9273 +
9274 +/* ---------------------------------------------------------------------- */
9275 +
9276 +static inline struct au_dinfo *au_di(struct dentry *dentry)
9277 +{
9278 +       return dentry->d_fsdata;
9279 +}
9280 +
9281 +/* ---------------------------------------------------------------------- */
9282 +
9283 +/* lock subclass for dinfo */
9284 +enum {
9285 +       AuLsc_DI_CHILD,         /* child first */
9286 +       AuLsc_DI_CHILD2,        /* rename(2), link(2), and cpup at hnotify */
9287 +       AuLsc_DI_CHILD3,        /* copyup dirs */
9288 +       AuLsc_DI_PARENT,
9289 +       AuLsc_DI_PARENT2,
9290 +       AuLsc_DI_PARENT3,
9291 +       AuLsc_DI_TMP            /* temp for replacing dinfo */
9292 +};
9293 +
9294 +/*
9295 + * di_read_lock_child, di_write_lock_child,
9296 + * di_read_lock_child2, di_write_lock_child2,
9297 + * di_read_lock_child3, di_write_lock_child3,
9298 + * di_read_lock_parent, di_write_lock_parent,
9299 + * di_read_lock_parent2, di_write_lock_parent2,
9300 + * di_read_lock_parent3, di_write_lock_parent3,
9301 + */
9302 +#define AuReadLockFunc(name, lsc) \
9303 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
9304 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
9305 +
9306 +#define AuWriteLockFunc(name, lsc) \
9307 +static inline void di_write_lock_##name(struct dentry *d) \
9308 +{ di_write_lock(d, AuLsc_DI_##lsc); }
9309 +
9310 +#define AuRWLockFuncs(name, lsc) \
9311 +       AuReadLockFunc(name, lsc) \
9312 +       AuWriteLockFunc(name, lsc)
9313 +
9314 +AuRWLockFuncs(child, CHILD);
9315 +AuRWLockFuncs(child2, CHILD2);
9316 +AuRWLockFuncs(child3, CHILD3);
9317 +AuRWLockFuncs(parent, PARENT);
9318 +AuRWLockFuncs(parent2, PARENT2);
9319 +AuRWLockFuncs(parent3, PARENT3);
9320 +
9321 +#undef AuReadLockFunc
9322 +#undef AuWriteLockFunc
9323 +#undef AuRWLockFuncs
9324 +
9325 +#define DiMustNoWaiters(d)     AuRwMustNoWaiters(&au_di(d)->di_rwsem)
9326 +#define DiMustAnyLock(d)       AuRwMustAnyLock(&au_di(d)->di_rwsem)
9327 +#define DiMustWriteLock(d)     AuRwMustWriteLock(&au_di(d)->di_rwsem)
9328 +
9329 +/* ---------------------------------------------------------------------- */
9330 +
9331 +/* todo: memory barrier? */
9332 +static inline unsigned int au_digen(struct dentry *d)
9333 +{
9334 +       return atomic_read(&au_di(d)->di_generation);
9335 +}
9336 +
9337 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
9338 +{
9339 +       hdentry->hd_dentry = NULL;
9340 +}
9341 +
9342 +static inline struct au_hdentry *au_hdentry(struct au_dinfo *di,
9343 +                                           aufs_bindex_t bindex)
9344 +{
9345 +       return di->di_hdentry + bindex;
9346 +}
9347 +
9348 +static inline void au_hdput(struct au_hdentry *hd)
9349 +{
9350 +       if (hd)
9351 +               dput(hd->hd_dentry);
9352 +}
9353 +
9354 +static inline aufs_bindex_t au_dbtop(struct dentry *dentry)
9355 +{
9356 +       DiMustAnyLock(dentry);
9357 +       return au_di(dentry)->di_btop;
9358 +}
9359 +
9360 +static inline aufs_bindex_t au_dbbot(struct dentry *dentry)
9361 +{
9362 +       DiMustAnyLock(dentry);
9363 +       return au_di(dentry)->di_bbot;
9364 +}
9365 +
9366 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
9367 +{
9368 +       DiMustAnyLock(dentry);
9369 +       return au_di(dentry)->di_bwh;
9370 +}
9371 +
9372 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
9373 +{
9374 +       DiMustAnyLock(dentry);
9375 +       return au_di(dentry)->di_bdiropq;
9376 +}
9377 +
9378 +/* todo: hard/soft set? */
9379 +static inline void au_set_dbtop(struct dentry *dentry, aufs_bindex_t bindex)
9380 +{
9381 +       DiMustWriteLock(dentry);
9382 +       au_di(dentry)->di_btop = bindex;
9383 +}
9384 +
9385 +static inline void au_set_dbbot(struct dentry *dentry, aufs_bindex_t bindex)
9386 +{
9387 +       DiMustWriteLock(dentry);
9388 +       au_di(dentry)->di_bbot = bindex;
9389 +}
9390 +
9391 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
9392 +{
9393 +       DiMustWriteLock(dentry);
9394 +       /* dbwh can be outside of btop - bbot range */
9395 +       au_di(dentry)->di_bwh = bindex;
9396 +}
9397 +
9398 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
9399 +{
9400 +       DiMustWriteLock(dentry);
9401 +       au_di(dentry)->di_bdiropq = bindex;
9402 +}
9403 +
9404 +/* ---------------------------------------------------------------------- */
9405 +
9406 +#ifdef CONFIG_AUFS_HNOTIFY
9407 +static inline void au_digen_dec(struct dentry *d)
9408 +{
9409 +       atomic_dec(&au_di(d)->di_generation);
9410 +}
9411 +
9412 +static inline void au_hn_di_reinit(struct dentry *dentry)
9413 +{
9414 +       dentry->d_fsdata = NULL;
9415 +}
9416 +#else
9417 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
9418 +#endif /* CONFIG_AUFS_HNOTIFY */
9419 +
9420 +#endif /* __KERNEL__ */
9421 +#endif /* __AUFS_DENTRY_H__ */
9422 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
9423 --- /usr/share/empty/fs/aufs/dinfo.c    1970-01-01 01:00:00.000000000 +0100
9424 +++ linux/fs/aufs/dinfo.c       2022-12-17 09:21:34.796521861 +0100
9425 @@ -0,0 +1,555 @@
9426 +// SPDX-License-Identifier: GPL-2.0
9427 +/*
9428 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9429 + *
9430 + * This program is free software; you can redistribute it and/or modify
9431 + * it under the terms of the GNU General Public License as published by
9432 + * the Free Software Foundation; either version 2 of the License, or
9433 + * (at your option) any later version.
9434 + *
9435 + * This program is distributed in the hope that it will be useful,
9436 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9437 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9438 + * GNU General Public License for more details.
9439 + *
9440 + * You should have received a copy of the GNU General Public License
9441 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
9442 + */
9443 +
9444 +/*
9445 + * dentry private data
9446 + */
9447 +
9448 +#include "aufs.h"
9449 +
9450 +void au_di_init_once(void *_dinfo)
9451 +{
9452 +       struct au_dinfo *dinfo = _dinfo;
9453 +
9454 +       au_rw_init(&dinfo->di_rwsem);
9455 +}
9456 +
9457 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
9458 +{
9459 +       struct au_dinfo *dinfo;
9460 +       int nbr, i;
9461 +
9462 +       dinfo = au_cache_alloc_dinfo();
9463 +       if (unlikely(!dinfo))
9464 +               goto out;
9465 +
9466 +       nbr = au_sbbot(sb) + 1;
9467 +       if (nbr <= 0)
9468 +               nbr = 1;
9469 +       dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
9470 +       if (dinfo->di_hdentry) {
9471 +               au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
9472 +               dinfo->di_btop = -1;
9473 +               dinfo->di_bbot = -1;
9474 +               dinfo->di_bwh = -1;
9475 +               dinfo->di_bdiropq = -1;
9476 +               dinfo->di_tmpfile = 0;
9477 +               for (i = 0; i < nbr; i++)
9478 +                       dinfo->di_hdentry[i].hd_id = -1;
9479 +               dinfo->di_htmpfile = NULL;
9480 +               goto out;
9481 +       }
9482 +
9483 +       au_cache_free_dinfo(dinfo);
9484 +       dinfo = NULL;
9485 +
9486 +out:
9487 +       return dinfo;
9488 +}
9489 +
9490 +void au_di_free(struct au_dinfo *dinfo)
9491 +{
9492 +       struct au_hdentry *p;
9493 +       aufs_bindex_t bbot, bindex;
9494 +
9495 +       /* dentry may not be revalidated */
9496 +       bindex = dinfo->di_btop;
9497 +       if (bindex >= 0) {
9498 +               bbot = dinfo->di_bbot;
9499 +               p = au_hdentry(dinfo, bindex);
9500 +               while (bindex++ <= bbot)
9501 +                       au_hdput(p++);
9502 +       }
9503 +       au_kfree_try_rcu(dinfo->di_hdentry);
9504 +       au_cache_free_dinfo(dinfo);
9505 +}
9506 +
9507 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
9508 +{
9509 +       struct au_hdentry *p;
9510 +       aufs_bindex_t bi;
9511 +
9512 +       AuRwMustWriteLock(&a->di_rwsem);
9513 +       AuRwMustWriteLock(&b->di_rwsem);
9514 +
9515 +#define DiSwap(v, name)                                \
9516 +       do {                                    \
9517 +               v = a->di_##name;               \
9518 +               a->di_##name = b->di_##name;    \
9519 +               b->di_##name = v;               \
9520 +       } while (0)
9521 +
9522 +       DiSwap(p, hdentry);
9523 +       DiSwap(bi, btop);
9524 +       DiSwap(bi, bbot);
9525 +       DiSwap(bi, bwh);
9526 +       DiSwap(bi, bdiropq);
9527 +       /* smp_mb(); */
9528 +
9529 +#undef DiSwap
9530 +}
9531 +
9532 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
9533 +{
9534 +       AuRwMustWriteLock(&dst->di_rwsem);
9535 +       AuRwMustWriteLock(&src->di_rwsem);
9536 +
9537 +       dst->di_btop = src->di_btop;
9538 +       dst->di_bbot = src->di_bbot;
9539 +       dst->di_bwh = src->di_bwh;
9540 +       dst->di_bdiropq = src->di_bdiropq;
9541 +       /* smp_mb(); */
9542 +}
9543 +
9544 +int au_di_init(struct dentry *dentry)
9545 +{
9546 +       int err;
9547 +       struct super_block *sb;
9548 +       struct au_dinfo *dinfo;
9549 +
9550 +       err = 0;
9551 +       sb = dentry->d_sb;
9552 +       dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
9553 +       if (dinfo) {
9554 +               atomic_set(&dinfo->di_generation, au_sigen(sb));
9555 +               /* smp_mb(); */ /* atomic_set */
9556 +               dentry->d_fsdata = dinfo;
9557 +       } else
9558 +               err = -ENOMEM;
9559 +
9560 +       return err;
9561 +}
9562 +
9563 +void au_di_fin(struct dentry *dentry)
9564 +{
9565 +       struct au_dinfo *dinfo;
9566 +
9567 +       dinfo = au_di(dentry);
9568 +       AuRwDestroy(&dinfo->di_rwsem);
9569 +       au_di_free(dinfo);
9570 +}
9571 +
9572 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink)
9573 +{
9574 +       int err, sz;
9575 +       struct au_hdentry *hdp;
9576 +
9577 +       AuRwMustWriteLock(&dinfo->di_rwsem);
9578 +
9579 +       err = -ENOMEM;
9580 +       sz = sizeof(*hdp) * (dinfo->di_bbot + 1);
9581 +       if (!sz)
9582 +               sz = sizeof(*hdp);
9583 +       hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS,
9584 +                          may_shrink);
9585 +       if (hdp) {
9586 +               dinfo->di_hdentry = hdp;
9587 +               err = 0;
9588 +       }
9589 +
9590 +       return err;
9591 +}
9592 +
9593 +/* ---------------------------------------------------------------------- */
9594 +
9595 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
9596 +{
9597 +       switch (lsc) {
9598 +       case AuLsc_DI_CHILD:
9599 +               ii_write_lock_child(inode);
9600 +               break;
9601 +       case AuLsc_DI_CHILD2:
9602 +               ii_write_lock_child2(inode);
9603 +               break;
9604 +       case AuLsc_DI_CHILD3:
9605 +               ii_write_lock_child3(inode);
9606 +               break;
9607 +       case AuLsc_DI_PARENT:
9608 +               ii_write_lock_parent(inode);
9609 +               break;
9610 +       case AuLsc_DI_PARENT2:
9611 +               ii_write_lock_parent2(inode);
9612 +               break;
9613 +       case AuLsc_DI_PARENT3:
9614 +               ii_write_lock_parent3(inode);
9615 +               break;
9616 +       default:
9617 +               BUG();
9618 +       }
9619 +}
9620 +
9621 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9622 +{
9623 +       switch (lsc) {
9624 +       case AuLsc_DI_CHILD:
9625 +               ii_read_lock_child(inode);
9626 +               break;
9627 +       case AuLsc_DI_CHILD2:
9628 +               ii_read_lock_child2(inode);
9629 +               break;
9630 +       case AuLsc_DI_CHILD3:
9631 +               ii_read_lock_child3(inode);
9632 +               break;
9633 +       case AuLsc_DI_PARENT:
9634 +               ii_read_lock_parent(inode);
9635 +               break;
9636 +       case AuLsc_DI_PARENT2:
9637 +               ii_read_lock_parent2(inode);
9638 +               break;
9639 +       case AuLsc_DI_PARENT3:
9640 +               ii_read_lock_parent3(inode);
9641 +               break;
9642 +       default:
9643 +               BUG();
9644 +       }
9645 +}
9646 +
9647 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9648 +{
9649 +       struct inode *inode;
9650 +
9651 +       au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9652 +       if (d_really_is_positive(d)) {
9653 +               inode = d_inode(d);
9654 +               if (au_ftest_lock(flags, IW))
9655 +                       do_ii_write_lock(inode, lsc);
9656 +               else if (au_ftest_lock(flags, IR))
9657 +                       do_ii_read_lock(inode, lsc);
9658 +       }
9659 +}
9660 +
9661 +void di_read_unlock(struct dentry *d, int flags)
9662 +{
9663 +       struct inode *inode;
9664 +
9665 +       if (d_really_is_positive(d)) {
9666 +               inode = d_inode(d);
9667 +               if (au_ftest_lock(flags, IW)) {
9668 +                       au_dbg_verify_dinode(d);
9669 +                       ii_write_unlock(inode);
9670 +               } else if (au_ftest_lock(flags, IR)) {
9671 +                       au_dbg_verify_dinode(d);
9672 +                       ii_read_unlock(inode);
9673 +               }
9674 +       }
9675 +       au_rw_read_unlock(&au_di(d)->di_rwsem);
9676 +}
9677 +
9678 +void di_downgrade_lock(struct dentry *d, int flags)
9679 +{
9680 +       if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9681 +               ii_downgrade_lock(d_inode(d));
9682 +       au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9683 +}
9684 +
9685 +void di_write_lock(struct dentry *d, unsigned int lsc)
9686 +{
9687 +       au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9688 +       if (d_really_is_positive(d))
9689 +               do_ii_write_lock(d_inode(d), lsc);
9690 +}
9691 +
9692 +void di_write_unlock(struct dentry *d)
9693 +{
9694 +       au_dbg_verify_dinode(d);
9695 +       if (d_really_is_positive(d))
9696 +               ii_write_unlock(d_inode(d));
9697 +       au_rw_write_unlock(&au_di(d)->di_rwsem);
9698 +}
9699 +
9700 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9701 +{
9702 +       AuDebugOn(d1 == d2
9703 +                 || d_inode(d1) == d_inode(d2)
9704 +                 || d1->d_sb != d2->d_sb);
9705 +
9706 +       if ((isdir && au_test_subdir(d1, d2))
9707 +           || d1 < d2) {
9708 +               di_write_lock_child(d1);
9709 +               di_write_lock_child2(d2);
9710 +       } else {
9711 +               di_write_lock_child(d2);
9712 +               di_write_lock_child2(d1);
9713 +       }
9714 +}
9715 +
9716 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9717 +{
9718 +       AuDebugOn(d1 == d2
9719 +                 || d_inode(d1) == d_inode(d2)
9720 +                 || d1->d_sb != d2->d_sb);
9721 +
9722 +       if ((isdir && au_test_subdir(d1, d2))
9723 +           || d1 < d2) {
9724 +               di_write_lock_parent(d1);
9725 +               di_write_lock_parent2(d2);
9726 +       } else {
9727 +               di_write_lock_parent(d2);
9728 +               di_write_lock_parent2(d1);
9729 +       }
9730 +}
9731 +
9732 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9733 +{
9734 +       di_write_unlock(d1);
9735 +       if (d_inode(d1) == d_inode(d2))
9736 +               au_rw_write_unlock(&au_di(d2)->di_rwsem);
9737 +       else
9738 +               di_write_unlock(d2);
9739 +}
9740 +
9741 +/* ---------------------------------------------------------------------- */
9742 +
9743 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9744 +{
9745 +       struct dentry *d;
9746 +
9747 +       DiMustAnyLock(dentry);
9748 +
9749 +       if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
9750 +               return NULL;
9751 +       AuDebugOn(bindex < 0);
9752 +       d = au_hdentry(au_di(dentry), bindex)->hd_dentry;
9753 +       AuDebugOn(d && au_dcount(d) <= 0);
9754 +       return d;
9755 +}
9756 +
9757 +/*
9758 + * extended version of au_h_dptr().
9759 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9760 + * error.
9761 + */
9762 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9763 +{
9764 +       struct dentry *h_dentry;
9765 +       struct inode *inode, *h_inode;
9766 +
9767 +       AuDebugOn(d_really_is_negative(dentry));
9768 +
9769 +       h_dentry = NULL;
9770 +       if (au_dbtop(dentry) <= bindex
9771 +           && bindex <= au_dbbot(dentry))
9772 +               h_dentry = au_h_dptr(dentry, bindex);
9773 +       if (h_dentry && !au_d_linkable(h_dentry)) {
9774 +               dget(h_dentry);
9775 +               goto out; /* success */
9776 +       }
9777 +
9778 +       inode = d_inode(dentry);
9779 +       AuDebugOn(bindex < au_ibtop(inode));
9780 +       AuDebugOn(au_ibbot(inode) < bindex);
9781 +       h_inode = au_h_iptr(inode, bindex);
9782 +       h_dentry = d_find_alias(h_inode);
9783 +       if (h_dentry) {
9784 +               if (!IS_ERR(h_dentry)) {
9785 +                       if (!au_d_linkable(h_dentry))
9786 +                               goto out; /* success */
9787 +                       dput(h_dentry);
9788 +               } else
9789 +                       goto out;
9790 +       }
9791 +
9792 +       if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9793 +               h_dentry = au_plink_lkup(inode, bindex);
9794 +               AuDebugOn(!h_dentry);
9795 +               if (!IS_ERR(h_dentry)) {
9796 +                       if (!au_d_hashed_positive(h_dentry))
9797 +                               goto out; /* success */
9798 +                       dput(h_dentry);
9799 +                       h_dentry = NULL;
9800 +               }
9801 +       }
9802 +
9803 +out:
9804 +       AuDbgDentry(h_dentry);
9805 +       return h_dentry;
9806 +}
9807 +
9808 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9809 +{
9810 +       aufs_bindex_t bbot, bwh;
9811 +
9812 +       bbot = au_dbbot(dentry);
9813 +       if (0 <= bbot) {
9814 +               bwh = au_dbwh(dentry);
9815 +               if (!bwh)
9816 +                       return bwh;
9817 +               if (0 < bwh && bwh < bbot)
9818 +                       return bwh - 1;
9819 +       }
9820 +       return bbot;
9821 +}
9822 +
9823 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9824 +{
9825 +       aufs_bindex_t bbot, bopq;
9826 +
9827 +       bbot = au_dbtail(dentry);
9828 +       if (0 <= bbot) {
9829 +               bopq = au_dbdiropq(dentry);
9830 +               if (0 <= bopq && bopq < bbot)
9831 +                       bbot = bopq;
9832 +       }
9833 +       return bbot;
9834 +}
9835 +
9836 +/* ---------------------------------------------------------------------- */
9837 +
9838 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9839 +                  struct dentry *h_dentry)
9840 +{
9841 +       struct au_dinfo *dinfo;
9842 +       struct au_hdentry *hd;
9843 +       struct au_branch *br;
9844 +
9845 +       DiMustWriteLock(dentry);
9846 +
9847 +       dinfo = au_di(dentry);
9848 +       hd = au_hdentry(dinfo, bindex);
9849 +       au_hdput(hd);
9850 +       hd->hd_dentry = h_dentry;
9851 +       if (h_dentry) {
9852 +               br = au_sbr(dentry->d_sb, bindex);
9853 +               hd->hd_id = br->br_id;
9854 +       }
9855 +}
9856 +
9857 +int au_dbrange_test(struct dentry *dentry)
9858 +{
9859 +       int err;
9860 +       aufs_bindex_t btop, bbot;
9861 +
9862 +       err = 0;
9863 +       btop = au_dbtop(dentry);
9864 +       bbot = au_dbbot(dentry);
9865 +       if (btop >= 0)
9866 +               AuDebugOn(bbot < 0 && btop > bbot);
9867 +       else {
9868 +               err = -EIO;
9869 +               AuDebugOn(bbot >= 0);
9870 +       }
9871 +
9872 +       return err;
9873 +}
9874 +
9875 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9876 +{
9877 +       int err;
9878 +
9879 +       err = 0;
9880 +       if (unlikely(au_digen(dentry) != sigen
9881 +                    || au_iigen_test(d_inode(dentry), sigen)))
9882 +               err = -EIO;
9883 +
9884 +       return err;
9885 +}
9886 +
9887 +void au_update_digen(struct dentry *dentry)
9888 +{
9889 +       atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9890 +       /* smp_mb(); */ /* atomic_set */
9891 +}
9892 +
9893 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9894 +{
9895 +       struct au_dinfo *dinfo;
9896 +       struct dentry *h_d;
9897 +       struct au_hdentry *hdp;
9898 +       aufs_bindex_t bindex, bbot;
9899 +
9900 +       DiMustWriteLock(dentry);
9901 +
9902 +       dinfo = au_di(dentry);
9903 +       if (!dinfo || dinfo->di_btop < 0)
9904 +               return;
9905 +
9906 +       if (do_put_zero) {
9907 +               bbot = dinfo->di_bbot;
9908 +               bindex = dinfo->di_btop;
9909 +               hdp = au_hdentry(dinfo, bindex);
9910 +               for (; bindex <= bbot; bindex++, hdp++) {
9911 +                       h_d = hdp->hd_dentry;
9912 +                       if (h_d && d_is_negative(h_d))
9913 +                               au_set_h_dptr(dentry, bindex, NULL);
9914 +               }
9915 +       }
9916 +
9917 +       dinfo->di_btop = 0;
9918 +       hdp = au_hdentry(dinfo, dinfo->di_btop);
9919 +       for (; dinfo->di_btop <= dinfo->di_bbot; dinfo->di_btop++, hdp++)
9920 +               if (hdp->hd_dentry)
9921 +                       break;
9922 +       if (dinfo->di_btop > dinfo->di_bbot) {
9923 +               dinfo->di_btop = -1;
9924 +               dinfo->di_bbot = -1;
9925 +               return;
9926 +       }
9927 +
9928 +       hdp = au_hdentry(dinfo, dinfo->di_bbot);
9929 +       for (; dinfo->di_bbot >= 0; dinfo->di_bbot--, hdp--)
9930 +               if (hdp->hd_dentry)
9931 +                       break;
9932 +       AuDebugOn(dinfo->di_btop > dinfo->di_bbot || dinfo->di_bbot < 0);
9933 +}
9934 +
9935 +void au_update_dbtop(struct dentry *dentry)
9936 +{
9937 +       aufs_bindex_t bindex, bbot;
9938 +       struct dentry *h_dentry;
9939 +
9940 +       bbot = au_dbbot(dentry);
9941 +       for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
9942 +               h_dentry = au_h_dptr(dentry, bindex);
9943 +               if (!h_dentry)
9944 +                       continue;
9945 +               if (d_is_positive(h_dentry)) {
9946 +                       au_set_dbtop(dentry, bindex);
9947 +                       return;
9948 +               }
9949 +               au_set_h_dptr(dentry, bindex, NULL);
9950 +       }
9951 +}
9952 +
9953 +void au_update_dbbot(struct dentry *dentry)
9954 +{
9955 +       aufs_bindex_t bindex, btop;
9956 +       struct dentry *h_dentry;
9957 +
9958 +       btop = au_dbtop(dentry);
9959 +       for (bindex = au_dbbot(dentry); bindex >= btop; bindex--) {
9960 +               h_dentry = au_h_dptr(dentry, bindex);
9961 +               if (!h_dentry)
9962 +                       continue;
9963 +               if (d_is_positive(h_dentry)) {
9964 +                       au_set_dbbot(dentry, bindex);
9965 +                       return;
9966 +               }
9967 +               au_set_h_dptr(dentry, bindex, NULL);
9968 +       }
9969 +}
9970 +
9971 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9972 +{
9973 +       aufs_bindex_t bindex, bbot;
9974 +
9975 +       bbot = au_dbbot(dentry);
9976 +       for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++)
9977 +               if (au_h_dptr(dentry, bindex) == h_dentry)
9978 +                       return bindex;
9979 +       return -1;
9980 +}
9981 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9982 --- /usr/share/empty/fs/aufs/dir.c      1970-01-01 01:00:00.000000000 +0100
9983 +++ linux/fs/aufs/dir.c 2022-12-17 09:21:34.796521861 +0100
9984 @@ -0,0 +1,765 @@
9985 +// SPDX-License-Identifier: GPL-2.0
9986 +/*
9987 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9988 + *
9989 + * This program is free software; you can redistribute it and/or modify
9990 + * it under the terms of the GNU General Public License as published by
9991 + * the Free Software Foundation; either version 2 of the License, or
9992 + * (at your option) any later version.
9993 + *
9994 + * This program is distributed in the hope that it will be useful,
9995 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9996 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9997 + * GNU General Public License for more details.
9998 + *
9999 + * You should have received a copy of the GNU General Public License
10000 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
10001 + */
10002 +
10003 +/*
10004 + * directory operations
10005 + */
10006 +
10007 +#include <linux/fs_stack.h>
10008 +#include <linux/iversion.h>
10009 +#include "aufs.h"
10010 +
10011 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
10012 +{
10013 +       unsigned int nlink;
10014 +
10015 +       AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10016 +
10017 +       nlink = dir->i_nlink;
10018 +       nlink += h_dir->i_nlink - 2;
10019 +       if (h_dir->i_nlink < 2)
10020 +               nlink += 2;
10021 +       smp_mb(); /* for i_nlink */
10022 +       /* 0 can happen in revaliding */
10023 +       set_nlink(dir, nlink);
10024 +}
10025 +
10026 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
10027 +{
10028 +       unsigned int nlink;
10029 +
10030 +       AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10031 +
10032 +       nlink = dir->i_nlink;
10033 +       nlink -= h_dir->i_nlink - 2;
10034 +       if (h_dir->i_nlink < 2)
10035 +               nlink -= 2;
10036 +       smp_mb(); /* for i_nlink */
10037 +       /* nlink == 0 means the branch-fs is broken */
10038 +       set_nlink(dir, nlink);
10039 +}
10040 +
10041 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
10042 +{
10043 +       loff_t sz;
10044 +       aufs_bindex_t bindex, bbot;
10045 +       struct file *h_file;
10046 +       struct dentry *h_dentry;
10047 +
10048 +       sz = 0;
10049 +       if (file) {
10050 +               AuDebugOn(!d_is_dir(file->f_path.dentry));
10051 +
10052 +               bbot = au_fbbot_dir(file);
10053 +               for (bindex = au_fbtop(file);
10054 +                    bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10055 +                    bindex++) {
10056 +                       h_file = au_hf_dir(file, bindex);
10057 +                       if (h_file && file_inode(h_file))
10058 +                               sz += vfsub_f_size_read(h_file);
10059 +               }
10060 +       } else {
10061 +               AuDebugOn(!dentry);
10062 +               AuDebugOn(!d_is_dir(dentry));
10063 +
10064 +               bbot = au_dbtaildir(dentry);
10065 +               for (bindex = au_dbtop(dentry);
10066 +                    bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10067 +                    bindex++) {
10068 +                       h_dentry = au_h_dptr(dentry, bindex);
10069 +                       if (h_dentry && d_is_positive(h_dentry))
10070 +                               sz += i_size_read(d_inode(h_dentry));
10071 +               }
10072 +       }
10073 +       if (sz < KMALLOC_MAX_SIZE)
10074 +               sz = roundup_pow_of_two(sz);
10075 +       if (sz > KMALLOC_MAX_SIZE)
10076 +               sz = KMALLOC_MAX_SIZE;
10077 +       else if (sz < NAME_MAX) {
10078 +               BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
10079 +               sz = AUFS_RDBLK_DEF;
10080 +       }
10081 +       return sz;
10082 +}
10083 +
10084 +struct au_dir_ts_arg {
10085 +       struct dentry *dentry;
10086 +       aufs_bindex_t brid;
10087 +};
10088 +
10089 +static void au_do_dir_ts(void *arg)
10090 +{
10091 +       struct au_dir_ts_arg *a = arg;
10092 +       struct au_dtime dt;
10093 +       struct path h_path;
10094 +       struct inode *dir, *h_dir;
10095 +       struct super_block *sb;
10096 +       struct au_branch *br;
10097 +       struct au_hinode *hdir;
10098 +       int err;
10099 +       aufs_bindex_t btop, bindex;
10100 +
10101 +       sb = a->dentry->d_sb;
10102 +       if (d_really_is_negative(a->dentry))
10103 +               goto out;
10104 +       /* no dir->i_mutex lock */
10105 +       aufs_read_lock(a->dentry, AuLock_DW); /* noflush */
10106 +
10107 +       dir = d_inode(a->dentry);
10108 +       btop = au_ibtop(dir);
10109 +       bindex = au_br_index(sb, a->brid);
10110 +       if (bindex < btop)
10111 +               goto out_unlock;
10112 +
10113 +       br = au_sbr(sb, bindex);
10114 +       h_path.dentry = au_h_dptr(a->dentry, bindex);
10115 +       if (!h_path.dentry)
10116 +               goto out_unlock;
10117 +       h_path.mnt = au_br_mnt(br);
10118 +       au_dtime_store(&dt, a->dentry, &h_path);
10119 +
10120 +       br = au_sbr(sb, btop);
10121 +       if (!au_br_writable(br->br_perm))
10122 +               goto out_unlock;
10123 +       h_path.dentry = au_h_dptr(a->dentry, btop);
10124 +       h_path.mnt = au_br_mnt(br);
10125 +       err = vfsub_mnt_want_write(h_path.mnt);
10126 +       if (err)
10127 +               goto out_unlock;
10128 +       hdir = au_hi(dir, btop);
10129 +       au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
10130 +       h_dir = au_h_iptr(dir, btop);
10131 +       if (h_dir->i_nlink
10132 +           && timespec64_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
10133 +               dt.dt_h_path = h_path;
10134 +               au_dtime_revert(&dt);
10135 +       }
10136 +       au_hn_inode_unlock(hdir);
10137 +       vfsub_mnt_drop_write(h_path.mnt);
10138 +       au_cpup_attr_timesizes(dir);
10139 +
10140 +out_unlock:
10141 +       aufs_read_unlock(a->dentry, AuLock_DW);
10142 +out:
10143 +       dput(a->dentry);
10144 +       au_nwt_done(&au_sbi(sb)->si_nowait);
10145 +       au_kfree_try_rcu(arg);
10146 +}
10147 +
10148 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
10149 +{
10150 +       int perm, wkq_err;
10151 +       aufs_bindex_t btop;
10152 +       struct au_dir_ts_arg *arg;
10153 +       struct dentry *dentry;
10154 +       struct super_block *sb;
10155 +
10156 +       IMustLock(dir);
10157 +
10158 +       dentry = d_find_any_alias(dir);
10159 +       AuDebugOn(!dentry);
10160 +       sb = dentry->d_sb;
10161 +       btop = au_ibtop(dir);
10162 +       if (btop == bindex) {
10163 +               au_cpup_attr_timesizes(dir);
10164 +               goto out;
10165 +       }
10166 +
10167 +       perm = au_sbr_perm(sb, btop);
10168 +       if (!au_br_writable(perm))
10169 +               goto out;
10170 +
10171 +       arg = kmalloc(sizeof(*arg), GFP_NOFS);
10172 +       if (!arg)
10173 +               goto out;
10174 +
10175 +       arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
10176 +       arg->brid = au_sbr_id(sb, bindex);
10177 +       wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
10178 +       if (unlikely(wkq_err)) {
10179 +               pr_err("wkq %d\n", wkq_err);
10180 +               dput(dentry);
10181 +               au_kfree_try_rcu(arg);
10182 +       }
10183 +
10184 +out:
10185 +       dput(dentry);
10186 +}
10187 +
10188 +/* ---------------------------------------------------------------------- */
10189 +
10190 +static int reopen_dir(struct file *file)
10191 +{
10192 +       int err;
10193 +       unsigned int flags;
10194 +       aufs_bindex_t bindex, btail, btop;
10195 +       struct dentry *dentry, *h_dentry;
10196 +       struct file *h_file;
10197 +
10198 +       /* open all lower dirs */
10199 +       dentry = file->f_path.dentry;
10200 +       btop = au_dbtop(dentry);
10201 +       for (bindex = au_fbtop(file); bindex < btop; bindex++)
10202 +               au_set_h_fptr(file, bindex, NULL);
10203 +       au_set_fbtop(file, btop);
10204 +
10205 +       btail = au_dbtaildir(dentry);
10206 +       for (bindex = au_fbbot_dir(file); btail < bindex; bindex--)
10207 +               au_set_h_fptr(file, bindex, NULL);
10208 +       au_set_fbbot_dir(file, btail);
10209 +
10210 +       flags = vfsub_file_flags(file);
10211 +       for (bindex = btop; bindex <= btail; bindex++) {
10212 +               h_dentry = au_h_dptr(dentry, bindex);
10213 +               if (!h_dentry)
10214 +                       continue;
10215 +               h_file = au_hf_dir(file, bindex);
10216 +               if (h_file)
10217 +                       continue;
10218 +
10219 +               h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10220 +               err = PTR_ERR(h_file);
10221 +               if (IS_ERR(h_file))
10222 +                       goto out; /* close all? */
10223 +               au_set_h_fptr(file, bindex, h_file);
10224 +       }
10225 +       au_update_figen(file);
10226 +       /* todo: necessary? */
10227 +       /* file->f_ra = h_file->f_ra; */
10228 +       err = 0;
10229 +
10230 +out:
10231 +       return err;
10232 +}
10233 +
10234 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
10235 +{
10236 +       int err;
10237 +       aufs_bindex_t bindex, btail;
10238 +       struct dentry *dentry, *h_dentry;
10239 +       struct vfsmount *mnt;
10240 +
10241 +       FiMustWriteLock(file);
10242 +       AuDebugOn(h_file);
10243 +
10244 +       err = 0;
10245 +       mnt = file->f_path.mnt;
10246 +       dentry = file->f_path.dentry;
10247 +       file->f_version = inode_query_iversion(d_inode(dentry));
10248 +       bindex = au_dbtop(dentry);
10249 +       au_set_fbtop(file, bindex);
10250 +       btail = au_dbtaildir(dentry);
10251 +       au_set_fbbot_dir(file, btail);
10252 +       for (; !err && bindex <= btail; bindex++) {
10253 +               h_dentry = au_h_dptr(dentry, bindex);
10254 +               if (!h_dentry)
10255 +                       continue;
10256 +
10257 +               err = vfsub_test_mntns(mnt, h_dentry->d_sb);
10258 +               if (unlikely(err))
10259 +                       break;
10260 +               h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10261 +               if (IS_ERR(h_file)) {
10262 +                       err = PTR_ERR(h_file);
10263 +                       break;
10264 +               }
10265 +               au_set_h_fptr(file, bindex, h_file);
10266 +       }
10267 +       au_update_figen(file);
10268 +       /* todo: necessary? */
10269 +       /* file->f_ra = h_file->f_ra; */
10270 +       if (!err)
10271 +               return 0; /* success */
10272 +
10273 +       /* close all */
10274 +       for (bindex = au_fbtop(file); bindex <= btail; bindex++)
10275 +               au_set_h_fptr(file, bindex, NULL);
10276 +       au_set_fbtop(file, -1);
10277 +       au_set_fbbot_dir(file, -1);
10278 +
10279 +       return err;
10280 +}
10281 +
10282 +static int aufs_open_dir(struct inode *inode __maybe_unused,
10283 +                        struct file *file)
10284 +{
10285 +       int err;
10286 +       struct super_block *sb;
10287 +       struct au_fidir *fidir;
10288 +
10289 +       err = -ENOMEM;
10290 +       sb = file->f_path.dentry->d_sb;
10291 +       si_read_lock(sb, AuLock_FLUSH);
10292 +       fidir = au_fidir_alloc(sb);
10293 +       if (fidir) {
10294 +               struct au_do_open_args args = {
10295 +                       .open   = do_open_dir,
10296 +                       .fidir  = fidir
10297 +               };
10298 +               err = au_do_open(file, &args);
10299 +               if (unlikely(err))
10300 +                       au_kfree_rcu(fidir);
10301 +       }
10302 +       si_read_unlock(sb);
10303 +       return err;
10304 +}
10305 +
10306 +static int aufs_release_dir(struct inode *inode __maybe_unused,
10307 +                           struct file *file)
10308 +{
10309 +       struct au_vdir *vdir_cache;
10310 +       struct au_finfo *finfo;
10311 +       struct au_fidir *fidir;
10312 +       struct au_hfile *hf;
10313 +       aufs_bindex_t bindex, bbot;
10314 +
10315 +       finfo = au_fi(file);
10316 +       fidir = finfo->fi_hdir;
10317 +       if (fidir) {
10318 +               au_hbl_del(&finfo->fi_hlist,
10319 +                          &au_sbi(file->f_path.dentry->d_sb)->si_files);
10320 +               vdir_cache = fidir->fd_vdir_cache; /* lock-free */
10321 +               if (vdir_cache)
10322 +                       au_vdir_free(vdir_cache);
10323 +
10324 +               bindex = finfo->fi_btop;
10325 +               if (bindex >= 0) {
10326 +                       hf = fidir->fd_hfile + bindex;
10327 +                       /*
10328 +                        * calls fput() instead of filp_close(),
10329 +                        * since no dnotify or lock for the lower file.
10330 +                        */
10331 +                       bbot = fidir->fd_bbot;
10332 +                       for (; bindex <= bbot; bindex++, hf++)
10333 +                               if (hf->hf_file)
10334 +                                       au_hfput(hf, /*execed*/0);
10335 +               }
10336 +               au_kfree_rcu(fidir);
10337 +               finfo->fi_hdir = NULL;
10338 +       }
10339 +       au_finfo_fin(file);
10340 +       return 0;
10341 +}
10342 +
10343 +/* ---------------------------------------------------------------------- */
10344 +
10345 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
10346 +{
10347 +       int err;
10348 +       aufs_bindex_t bindex, bbot;
10349 +       struct file *h_file;
10350 +
10351 +       err = 0;
10352 +       bbot = au_fbbot_dir(file);
10353 +       for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10354 +               h_file = au_hf_dir(file, bindex);
10355 +               if (h_file)
10356 +                       err = vfsub_flush(h_file, id);
10357 +       }
10358 +       return err;
10359 +}
10360 +
10361 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
10362 +{
10363 +       return au_do_flush(file, id, au_do_flush_dir);
10364 +}
10365 +
10366 +/* ---------------------------------------------------------------------- */
10367 +
10368 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
10369 +{
10370 +       int err;
10371 +       aufs_bindex_t bbot, bindex;
10372 +       struct inode *inode;
10373 +       struct super_block *sb;
10374 +
10375 +       err = 0;
10376 +       sb = dentry->d_sb;
10377 +       inode = d_inode(dentry);
10378 +       IMustLock(inode);
10379 +       bbot = au_dbbot(dentry);
10380 +       for (bindex = au_dbtop(dentry); !err && bindex <= bbot; bindex++) {
10381 +               struct path h_path;
10382 +
10383 +               if (au_test_ro(sb, bindex, inode))
10384 +                       continue;
10385 +               h_path.dentry = au_h_dptr(dentry, bindex);
10386 +               if (!h_path.dentry)
10387 +                       continue;
10388 +
10389 +               h_path.mnt = au_sbr_mnt(sb, bindex);
10390 +               err = vfsub_fsync(NULL, &h_path, datasync);
10391 +       }
10392 +
10393 +       return err;
10394 +}
10395 +
10396 +static int au_do_fsync_dir(struct file *file, int datasync)
10397 +{
10398 +       int err;
10399 +       aufs_bindex_t bbot, bindex;
10400 +       struct file *h_file;
10401 +       struct super_block *sb;
10402 +       struct inode *inode;
10403 +
10404 +       err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10405 +       if (unlikely(err))
10406 +               goto out;
10407 +
10408 +       inode = file_inode(file);
10409 +       sb = inode->i_sb;
10410 +       bbot = au_fbbot_dir(file);
10411 +       for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10412 +               h_file = au_hf_dir(file, bindex);
10413 +               if (!h_file || au_test_ro(sb, bindex, inode))
10414 +                       continue;
10415 +
10416 +               err = vfsub_fsync(h_file, &h_file->f_path, datasync);
10417 +       }
10418 +
10419 +out:
10420 +       return err;
10421 +}
10422 +
10423 +/*
10424 + * @file may be NULL
10425 + */
10426 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
10427 +                         int datasync)
10428 +{
10429 +       int err;
10430 +       struct dentry *dentry;
10431 +       struct inode *inode;
10432 +       struct super_block *sb;
10433 +
10434 +       err = 0;
10435 +       dentry = file->f_path.dentry;
10436 +       inode = d_inode(dentry);
10437 +       inode_lock(inode);
10438 +       sb = dentry->d_sb;
10439 +       si_noflush_read_lock(sb);
10440 +       if (file)
10441 +               err = au_do_fsync_dir(file, datasync);
10442 +       else {
10443 +               di_write_lock_child(dentry);
10444 +               err = au_do_fsync_dir_no_file(dentry, datasync);
10445 +       }
10446 +       au_cpup_attr_timesizes(inode);
10447 +       di_write_unlock(dentry);
10448 +       if (file)
10449 +               fi_write_unlock(file);
10450 +
10451 +       si_read_unlock(sb);
10452 +       inode_unlock(inode);
10453 +       return err;
10454 +}
10455 +
10456 +/* ---------------------------------------------------------------------- */
10457 +
10458 +static int aufs_iterate_shared(struct file *file, struct dir_context *ctx)
10459 +{
10460 +       int err;
10461 +       struct dentry *dentry;
10462 +       struct inode *inode, *h_inode;
10463 +       struct super_block *sb;
10464 +
10465 +       AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
10466 +
10467 +       dentry = file->f_path.dentry;
10468 +       inode = d_inode(dentry);
10469 +       IMustLock(inode);
10470 +
10471 +       sb = dentry->d_sb;
10472 +       si_read_lock(sb, AuLock_FLUSH);
10473 +       err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10474 +       if (unlikely(err))
10475 +               goto out;
10476 +       err = au_alive_dir(dentry);
10477 +       if (!err)
10478 +               err = au_vdir_init(file);
10479 +       di_downgrade_lock(dentry, AuLock_IR);
10480 +       if (unlikely(err))
10481 +               goto out_unlock;
10482 +
10483 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
10484 +       if (!au_test_nfsd()) {
10485 +               err = au_vdir_fill_de(file, ctx);
10486 +               fsstack_copy_attr_atime(inode, h_inode);
10487 +       } else {
10488 +               /*
10489 +                * nfsd filldir may call lookup_one_len(), vfs_getattr(),
10490 +                * encode_fh() and others.
10491 +                */
10492 +               atomic_inc(&h_inode->i_count);
10493 +               di_read_unlock(dentry, AuLock_IR);
10494 +               si_read_unlock(sb);
10495 +               err = au_vdir_fill_de(file, ctx);
10496 +               fsstack_copy_attr_atime(inode, h_inode);
10497 +               fi_write_unlock(file);
10498 +               iput(h_inode);
10499 +
10500 +               AuTraceErr(err);
10501 +               return err;
10502 +       }
10503 +
10504 +out_unlock:
10505 +       di_read_unlock(dentry, AuLock_IR);
10506 +       fi_write_unlock(file);
10507 +out:
10508 +       si_read_unlock(sb);
10509 +       return err;
10510 +}
10511 +
10512 +/* ---------------------------------------------------------------------- */
10513 +
10514 +#define AuTestEmpty_WHONLY     1
10515 +#define AuTestEmpty_CALLED     (1 << 1)
10516 +#define AuTestEmpty_SHWH       (1 << 2)
10517 +#define au_ftest_testempty(flags, name)        ((flags) & AuTestEmpty_##name)
10518 +#define au_fset_testempty(flags, name) \
10519 +       do { (flags) |= AuTestEmpty_##name; } while (0)
10520 +#define au_fclr_testempty(flags, name) \
10521 +       do { (flags) &= ~AuTestEmpty_##name; } while (0)
10522 +
10523 +#ifndef CONFIG_AUFS_SHWH
10524 +#undef AuTestEmpty_SHWH
10525 +#define AuTestEmpty_SHWH       0
10526 +#endif
10527 +
10528 +struct test_empty_arg {
10529 +       struct dir_context ctx;
10530 +       struct au_nhash *whlist;
10531 +       unsigned int flags;
10532 +       int err;
10533 +       aufs_bindex_t bindex;
10534 +};
10535 +
10536 +static bool test_empty_cb(struct dir_context *ctx, const char *__name,
10537 +                         int namelen, loff_t offset __maybe_unused, u64 ino,
10538 +                         unsigned int d_type)
10539 +{
10540 +       struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
10541 +                                                 ctx);
10542 +       char *name = (void *)__name;
10543 +
10544 +       arg->err = 0;
10545 +       au_fset_testempty(arg->flags, CALLED);
10546 +       /* smp_mb(); */
10547 +       if (name[0] == '.'
10548 +           && (namelen == 1 || (name[1] == '.' && namelen == 2)))
10549 +               goto out; /* success */
10550 +
10551 +       if (namelen <= AUFS_WH_PFX_LEN
10552 +           || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
10553 +               if (au_ftest_testempty(arg->flags, WHONLY)
10554 +                   && !au_nhash_test_known_wh(arg->whlist, name, namelen))
10555 +                       arg->err = -ENOTEMPTY;
10556 +               goto out;
10557 +       }
10558 +
10559 +       name += AUFS_WH_PFX_LEN;
10560 +       namelen -= AUFS_WH_PFX_LEN;
10561 +       if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
10562 +               arg->err = au_nhash_append_wh
10563 +                       (arg->whlist, name, namelen, ino, d_type, arg->bindex,
10564 +                        au_ftest_testempty(arg->flags, SHWH));
10565 +
10566 +out:
10567 +       /* smp_mb(); */
10568 +       AuTraceErr(arg->err);
10569 +       return !arg->err;
10570 +}
10571 +
10572 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10573 +{
10574 +       int err;
10575 +       struct file *h_file;
10576 +       struct au_branch *br;
10577 +
10578 +       h_file = au_h_open(dentry, arg->bindex,
10579 +                          O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
10580 +                          /*file*/NULL, /*force_wr*/0);
10581 +       err = PTR_ERR(h_file);
10582 +       if (IS_ERR(h_file))
10583 +               goto out;
10584 +
10585 +       err = 0;
10586 +       if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
10587 +           && !file_inode(h_file)->i_nlink)
10588 +               goto out_put;
10589 +
10590 +       do {
10591 +               arg->err = 0;
10592 +               au_fclr_testempty(arg->flags, CALLED);
10593 +               /* smp_mb(); */
10594 +               err = vfsub_iterate_dir(h_file, &arg->ctx);
10595 +               if (err >= 0)
10596 +                       err = arg->err;
10597 +       } while (!err && au_ftest_testempty(arg->flags, CALLED));
10598 +
10599 +out_put:
10600 +       fput(h_file);
10601 +       br = au_sbr(dentry->d_sb, arg->bindex);
10602 +       au_lcnt_dec(&br->br_nfiles);
10603 +out:
10604 +       return err;
10605 +}
10606 +
10607 +struct do_test_empty_args {
10608 +       int *errp;
10609 +       struct dentry *dentry;
10610 +       struct test_empty_arg *arg;
10611 +};
10612 +
10613 +static void call_do_test_empty(void *args)
10614 +{
10615 +       struct do_test_empty_args *a = args;
10616 +       *a->errp = do_test_empty(a->dentry, a->arg);
10617 +}
10618 +
10619 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10620 +{
10621 +       int err, wkq_err;
10622 +       struct dentry *h_dentry;
10623 +       struct inode *h_inode;
10624 +       struct user_namespace *h_userns;
10625 +
10626 +       h_userns = au_sbr_userns(dentry->d_sb, arg->bindex);
10627 +       h_dentry = au_h_dptr(dentry, arg->bindex);
10628 +       h_inode = d_inode(h_dentry);
10629 +       /* todo: i_mode changes anytime? */
10630 +       inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
10631 +       err = au_test_h_perm_sio(h_userns, h_inode, MAY_EXEC | MAY_READ);
10632 +       inode_unlock_shared(h_inode);
10633 +       if (!err)
10634 +               err = do_test_empty(dentry, arg);
10635 +       else {
10636 +               struct do_test_empty_args args = {
10637 +                       .errp   = &err,
10638 +                       .dentry = dentry,
10639 +                       .arg    = arg
10640 +               };
10641 +               unsigned int flags = arg->flags;
10642 +
10643 +               wkq_err = au_wkq_wait(call_do_test_empty, &args);
10644 +               if (unlikely(wkq_err))
10645 +                       err = wkq_err;
10646 +               arg->flags = flags;
10647 +       }
10648 +
10649 +       return err;
10650 +}
10651 +
10652 +int au_test_empty_lower(struct dentry *dentry)
10653 +{
10654 +       int err;
10655 +       unsigned int rdhash;
10656 +       aufs_bindex_t bindex, btop, btail;
10657 +       struct au_nhash whlist;
10658 +       struct test_empty_arg arg = {
10659 +               .ctx = {
10660 +                       .actor = test_empty_cb
10661 +               }
10662 +       };
10663 +       int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10664 +
10665 +       SiMustAnyLock(dentry->d_sb);
10666 +
10667 +       rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10668 +       if (!rdhash)
10669 +               rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10670 +       err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10671 +       if (unlikely(err))
10672 +               goto out;
10673 +
10674 +       arg.flags = 0;
10675 +       arg.whlist = &whlist;
10676 +       btop = au_dbtop(dentry);
10677 +       if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10678 +               au_fset_testempty(arg.flags, SHWH);
10679 +       test_empty = do_test_empty;
10680 +       if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10681 +               test_empty = sio_test_empty;
10682 +       arg.bindex = btop;
10683 +       err = test_empty(dentry, &arg);
10684 +       if (unlikely(err))
10685 +               goto out_whlist;
10686 +
10687 +       au_fset_testempty(arg.flags, WHONLY);
10688 +       btail = au_dbtaildir(dentry);
10689 +       for (bindex = btop + 1; !err && bindex <= btail; bindex++) {
10690 +               struct dentry *h_dentry;
10691 +
10692 +               h_dentry = au_h_dptr(dentry, bindex);
10693 +               if (h_dentry && d_is_positive(h_dentry)) {
10694 +                       arg.bindex = bindex;
10695 +                       err = test_empty(dentry, &arg);
10696 +               }
10697 +       }
10698 +
10699 +out_whlist:
10700 +       au_nhash_wh_free(&whlist);
10701 +out:
10702 +       return err;
10703 +}
10704 +
10705 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10706 +{
10707 +       int err;
10708 +       struct test_empty_arg arg = {
10709 +               .ctx = {
10710 +                       .actor = test_empty_cb
10711 +               }
10712 +       };
10713 +       aufs_bindex_t bindex, btail;
10714 +
10715 +       err = 0;
10716 +       arg.whlist = whlist;
10717 +       arg.flags = AuTestEmpty_WHONLY;
10718 +       if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10719 +               au_fset_testempty(arg.flags, SHWH);
10720 +       btail = au_dbtaildir(dentry);
10721 +       for (bindex = au_dbtop(dentry); !err && bindex <= btail; bindex++) {
10722 +               struct dentry *h_dentry;
10723 +
10724 +               h_dentry = au_h_dptr(dentry, bindex);
10725 +               if (h_dentry && d_is_positive(h_dentry)) {
10726 +                       arg.bindex = bindex;
10727 +                       err = sio_test_empty(dentry, &arg);
10728 +               }
10729 +       }
10730 +
10731 +       return err;
10732 +}
10733 +
10734 +/* ---------------------------------------------------------------------- */
10735 +
10736 +const struct file_operations aufs_dir_fop = {
10737 +       .owner          = THIS_MODULE,
10738 +       .llseek         = default_llseek,
10739 +       .read           = generic_read_dir,
10740 +       .iterate_shared = aufs_iterate_shared,
10741 +       .unlocked_ioctl = aufs_ioctl_dir,
10742 +#ifdef CONFIG_COMPAT
10743 +       .compat_ioctl   = aufs_compat_ioctl_dir,
10744 +#endif
10745 +       .open           = aufs_open_dir,
10746 +       .release        = aufs_release_dir,
10747 +       .flush          = aufs_flush_dir,
10748 +       .fsync          = aufs_fsync_dir
10749 +};
10750 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10751 --- /usr/share/empty/fs/aufs/dir.h      1970-01-01 01:00:00.000000000 +0100
10752 +++ linux/fs/aufs/dir.h 2022-11-05 23:02:18.962555950 +0100
10753 @@ -0,0 +1,134 @@
10754 +/* SPDX-License-Identifier: GPL-2.0 */
10755 +/*
10756 + * Copyright (C) 2005-2022 Junjiro R. Okajima
10757 + *
10758 + * This program is free software; you can redistribute it and/or modify
10759 + * it under the terms of the GNU General Public License as published by
10760 + * the Free Software Foundation; either version 2 of the License, or
10761 + * (at your option) any later version.
10762 + *
10763 + * This program is distributed in the hope that it will be useful,
10764 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10765 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
10766 + * GNU General Public License for more details.
10767 + *
10768 + * You should have received a copy of the GNU General Public License
10769 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
10770 + */
10771 +
10772 +/*
10773 + * directory operations
10774 + */
10775 +
10776 +#ifndef __AUFS_DIR_H__
10777 +#define __AUFS_DIR_H__
10778 +
10779 +#ifdef __KERNEL__
10780 +
10781 +#include <linux/fs.h>
10782 +
10783 +/* ---------------------------------------------------------------------- */
10784 +
10785 +/* need to be faster and smaller */
10786 +
10787 +struct au_nhash {
10788 +       unsigned int            nh_num;
10789 +       struct hlist_head       *nh_head;
10790 +};
10791 +
10792 +struct au_vdir_destr {
10793 +       unsigned char   len;
10794 +       unsigned char   name[];
10795 +} __packed;
10796 +
10797 +struct au_vdir_dehstr {
10798 +       struct hlist_node       hash;
10799 +       struct au_vdir_destr    *str;
10800 +       struct rcu_head         rcu;
10801 +} ____cacheline_aligned_in_smp;
10802 +
10803 +struct au_vdir_de {
10804 +       ino_t                   de_ino;
10805 +       unsigned char           de_type;
10806 +       /* caution: packed */
10807 +       struct au_vdir_destr    de_str;
10808 +} __packed;
10809 +
10810 +struct au_vdir_wh {
10811 +       struct hlist_node       wh_hash;
10812 +#ifdef CONFIG_AUFS_SHWH
10813 +       ino_t                   wh_ino;
10814 +       aufs_bindex_t           wh_bindex;
10815 +       unsigned char           wh_type;
10816 +#else
10817 +       aufs_bindex_t           wh_bindex;
10818 +#endif
10819 +       /* caution: packed */
10820 +       struct au_vdir_destr    wh_str;
10821 +} __packed;
10822 +
10823 +union au_vdir_deblk_p {
10824 +       unsigned char           *deblk;
10825 +       struct au_vdir_de       *de;
10826 +};
10827 +
10828 +struct au_vdir {
10829 +       unsigned char   **vd_deblk;
10830 +       unsigned long   vd_nblk;
10831 +       struct {
10832 +               unsigned long           ul;
10833 +               union au_vdir_deblk_p   p;
10834 +       } vd_last;
10835 +
10836 +       u64             vd_version;
10837 +       unsigned int    vd_deblk_sz;
10838 +       unsigned long   vd_jiffy;
10839 +       struct rcu_head rcu;
10840 +} ____cacheline_aligned_in_smp;
10841 +
10842 +/* ---------------------------------------------------------------------- */
10843 +
10844 +/* dir.c */
10845 +extern const struct file_operations aufs_dir_fop;
10846 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10847 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10848 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10849 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10850 +int au_test_empty_lower(struct dentry *dentry);
10851 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10852 +
10853 +/* vdir.c */
10854 +unsigned int au_rdhash_est(loff_t sz);
10855 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10856 +void au_nhash_wh_free(struct au_nhash *whlist);
10857 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10858 +                           int limit);
10859 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10860 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10861 +                      unsigned int d_type, aufs_bindex_t bindex,
10862 +                      unsigned char shwh);
10863 +void au_vdir_free(struct au_vdir *vdir);
10864 +int au_vdir_init(struct file *file);
10865 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10866 +
10867 +/* ioctl.c */
10868 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10869 +
10870 +#ifdef CONFIG_AUFS_RDU
10871 +/* rdu.c */
10872 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10873 +#ifdef CONFIG_COMPAT
10874 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10875 +                        unsigned long arg);
10876 +#endif
10877 +#else
10878 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10879 +       unsigned int cmd, unsigned long arg)
10880 +#ifdef CONFIG_COMPAT
10881 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10882 +       unsigned int cmd, unsigned long arg)
10883 +#endif
10884 +#endif
10885 +
10886 +#endif /* __KERNEL__ */
10887 +#endif /* __AUFS_DIR_H__ */
10888 diff -urN /usr/share/empty/fs/aufs/dirren.c linux/fs/aufs/dirren.c
10889 --- /usr/share/empty/fs/aufs/dirren.c   1970-01-01 01:00:00.000000000 +0100
10890 +++ linux/fs/aufs/dirren.c      2022-11-05 23:02:18.962555950 +0100
10891 @@ -0,0 +1,1315 @@
10892 +// SPDX-License-Identifier: GPL-2.0
10893 +/*
10894 + * Copyright (C) 2017-2022 Junjiro R. Okajima
10895 + *
10896 + * This program is free software; you can redistribute it and/or modify
10897 + * it under the terms of the GNU General Public License as published by
10898 + * the Free Software Foundation; either version 2 of the License, or
10899 + * (at your option) any later version.
10900 + *
10901 + * This program is distributed in the hope that it will be useful,
10902 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10903 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
10904 + * GNU General Public License for more details.
10905 + *
10906 + * You should have received a copy of the GNU General Public License
10907 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
10908 + */
10909 +
10910 +/*
10911 + * special handling in renaming a directory
10912 + * in order to support looking-up the before-renamed name on the lower readonly
10913 + * branches
10914 + */
10915 +
10916 +#include <linux/byteorder/generic.h>
10917 +#include "aufs.h"
10918 +
10919 +static void au_dr_hino_del(struct au_dr_br *dr, struct au_dr_hino *ent)
10920 +{
10921 +       int idx;
10922 +
10923 +       idx = au_dr_ihash(ent->dr_h_ino);
10924 +       au_hbl_del(&ent->dr_hnode, dr->dr_h_ino + idx);
10925 +}
10926 +
10927 +static int au_dr_hino_test_empty(struct au_dr_br *dr)
10928 +{
10929 +       int ret, i;
10930 +       struct hlist_bl_head *hbl;
10931 +
10932 +       ret = 1;
10933 +       for (i = 0; ret && i < AuDirren_NHASH; i++) {
10934 +               hbl = dr->dr_h_ino + i;
10935 +               hlist_bl_lock(hbl);
10936 +               ret &= hlist_bl_empty(hbl);
10937 +               hlist_bl_unlock(hbl);
10938 +       }
10939 +
10940 +       return ret;
10941 +}
10942 +
10943 +static struct au_dr_hino *au_dr_hino_find(struct au_dr_br *dr, ino_t ino)
10944 +{
10945 +       struct au_dr_hino *found, *ent;
10946 +       struct hlist_bl_head *hbl;
10947 +       struct hlist_bl_node *pos;
10948 +       int idx;
10949 +
10950 +       found = NULL;
10951 +       idx = au_dr_ihash(ino);
10952 +       hbl = dr->dr_h_ino + idx;
10953 +       hlist_bl_lock(hbl);
10954 +       hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10955 +               if (ent->dr_h_ino == ino) {
10956 +                       found = ent;
10957 +                       break;
10958 +               }
10959 +       hlist_bl_unlock(hbl);
10960 +
10961 +       return found;
10962 +}
10963 +
10964 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t ino,
10965 +                       struct au_dr_hino *add_ent)
10966 +{
10967 +       int found, idx;
10968 +       struct hlist_bl_head *hbl;
10969 +       struct hlist_bl_node *pos;
10970 +       struct au_dr_hino *ent;
10971 +
10972 +       found = 0;
10973 +       idx = au_dr_ihash(ino);
10974 +       hbl = dr->dr_h_ino + idx;
10975 +#if 0 /* debug print */
10976 +       {
10977 +               struct hlist_bl_node *tmp;
10978 +
10979 +               hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10980 +                       AuDbg("hi%llu\n", (unsigned long long)ent->dr_h_ino);
10981 +       }
10982 +#endif
10983 +       hlist_bl_lock(hbl);
10984 +       hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10985 +               if (ent->dr_h_ino == ino) {
10986 +                       found = 1;
10987 +                       break;
10988 +               }
10989 +       if (!found && add_ent)
10990 +               hlist_bl_add_head(&add_ent->dr_hnode, hbl);
10991 +       hlist_bl_unlock(hbl);
10992 +
10993 +       if (!found && add_ent)
10994 +               AuDbg("i%llu added\n", (unsigned long long)add_ent->dr_h_ino);
10995 +
10996 +       return found;
10997 +}
10998 +
10999 +void au_dr_hino_free(struct au_dr_br *dr)
11000 +{
11001 +       int i;
11002 +       struct hlist_bl_head *hbl;
11003 +       struct hlist_bl_node *pos, *tmp;
11004 +       struct au_dr_hino *ent;
11005 +
11006 +       /* SiMustWriteLock(sb); */
11007 +
11008 +       for (i = 0; i < AuDirren_NHASH; i++) {
11009 +               hbl = dr->dr_h_ino + i;
11010 +               /* no spinlock since sbinfo must be write-locked */
11011 +               hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
11012 +                       au_kfree_rcu(ent);
11013 +               INIT_HLIST_BL_HEAD(hbl);
11014 +       }
11015 +}
11016 +
11017 +/* returns the number of inodes or an error */
11018 +static int au_dr_hino_store(struct super_block *sb, struct au_branch *br,
11019 +                           struct file *hinofile)
11020 +{
11021 +       int err, i;
11022 +       ssize_t ssz;
11023 +       loff_t pos, oldsize;
11024 +       __be64 u64;
11025 +       struct inode *hinoinode;
11026 +       struct hlist_bl_head *hbl;
11027 +       struct hlist_bl_node *n1, *n2;
11028 +       struct au_dr_hino *ent;
11029 +
11030 +       SiMustWriteLock(sb);
11031 +       AuDebugOn(!au_br_writable(br->br_perm));
11032 +
11033 +       hinoinode = file_inode(hinofile);
11034 +       oldsize = i_size_read(hinoinode);
11035 +
11036 +       err = 0;
11037 +       pos = 0;
11038 +       hbl = br->br_dirren.dr_h_ino;
11039 +       for (i = 0; !err && i < AuDirren_NHASH; i++, hbl++) {
11040 +               /* no bit-lock since sbinfo must be write-locked */
11041 +               hlist_bl_for_each_entry_safe(ent, n1, n2, hbl, dr_hnode) {
11042 +                       AuDbg("hi%llu, %pD2\n",
11043 +                             (unsigned long long)ent->dr_h_ino, hinofile);
11044 +                       u64 = cpu_to_be64(ent->dr_h_ino);
11045 +                       ssz = vfsub_write_k(hinofile, &u64, sizeof(u64), &pos);
11046 +                       if (ssz == sizeof(u64))
11047 +                               continue;
11048 +
11049 +                       /* write error */
11050 +                       pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11051 +                       err = -ENOSPC;
11052 +                       if (ssz < 0)
11053 +                               err = ssz;
11054 +                       break;
11055 +               }
11056 +       }
11057 +       /* regardless the error */
11058 +       if (pos < oldsize) {
11059 +               err = vfsub_trunc(&hinofile->f_path, pos, /*attr*/0, hinofile);
11060 +               AuTraceErr(err);
11061 +       }
11062 +
11063 +       AuTraceErr(err);
11064 +       return err;
11065 +}
11066 +
11067 +static int au_dr_hino_load(struct au_dr_br *dr, struct file *hinofile)
11068 +{
11069 +       int err, hidx;
11070 +       ssize_t ssz;
11071 +       size_t sz, n;
11072 +       loff_t pos;
11073 +       uint64_t u64;
11074 +       struct au_dr_hino *ent;
11075 +       struct inode *hinoinode;
11076 +       struct hlist_bl_head *hbl;
11077 +
11078 +       err = 0;
11079 +       pos = 0;
11080 +       hbl = dr->dr_h_ino;
11081 +       hinoinode = file_inode(hinofile);
11082 +       sz = i_size_read(hinoinode);
11083 +       AuDebugOn(sz % sizeof(u64));
11084 +       n = sz / sizeof(u64);
11085 +       while (n--) {
11086 +               ssz = vfsub_read_k(hinofile, &u64, sizeof(u64), &pos);
11087 +               if (unlikely(ssz != sizeof(u64))) {
11088 +                       pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11089 +                       err = -EINVAL;
11090 +                       if (ssz < 0)
11091 +                               err = ssz;
11092 +                       goto out_free;
11093 +               }
11094 +
11095 +               ent = kmalloc(sizeof(*ent), GFP_NOFS);
11096 +               if (!ent) {
11097 +                       err = -ENOMEM;
11098 +                       AuTraceErr(err);
11099 +                       goto out_free;
11100 +               }
11101 +               ent->dr_h_ino = be64_to_cpu((__force __be64)u64);
11102 +               AuDbg("hi%llu, %pD2\n",
11103 +                     (unsigned long long)ent->dr_h_ino, hinofile);
11104 +               hidx = au_dr_ihash(ent->dr_h_ino);
11105 +               au_hbl_add(&ent->dr_hnode, hbl + hidx);
11106 +       }
11107 +       goto out; /* success */
11108 +
11109 +out_free:
11110 +       au_dr_hino_free(dr);
11111 +out:
11112 +       AuTraceErr(err);
11113 +       return err;
11114 +}
11115 +
11116 +/*
11117 + * @bindex/@br is a switch to distinguish whether suspending hnotify or not.
11118 + * @path is a switch to distinguish load and store.
11119 + */
11120 +static int au_dr_hino(struct super_block *sb, aufs_bindex_t bindex,
11121 +                     struct au_branch *br, const struct path *path)
11122 +{
11123 +       int err, flags;
11124 +       unsigned char load, suspend;
11125 +       struct file *hinofile;
11126 +       struct au_hinode *hdir;
11127 +       struct inode *dir, *delegated;
11128 +       struct path hinopath;
11129 +       struct qstr hinoname = QSTR_INIT(AUFS_WH_DR_BRHINO,
11130 +                                        sizeof(AUFS_WH_DR_BRHINO) - 1);
11131 +
11132 +       AuDebugOn(bindex < 0 && !br);
11133 +       AuDebugOn(bindex >= 0 && br);
11134 +
11135 +       err = -EINVAL;
11136 +       suspend = !br;
11137 +       if (suspend)
11138 +               br = au_sbr(sb, bindex);
11139 +       load = !!path;
11140 +       if (!load) {
11141 +               path = &br->br_path;
11142 +               AuDebugOn(!au_br_writable(br->br_perm));
11143 +               if (unlikely(!au_br_writable(br->br_perm)))
11144 +                       goto out;
11145 +       }
11146 +
11147 +       hdir = NULL;
11148 +       if (suspend) {
11149 +               dir = d_inode(sb->s_root);
11150 +               hdir = au_hinode(au_ii(dir), bindex);
11151 +               dir = hdir->hi_inode;
11152 +               au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11153 +       } else {
11154 +               dir = d_inode(path->dentry);
11155 +               inode_lock_nested(dir, AuLsc_I_CHILD);
11156 +       }
11157 +       hinopath.mnt = path->mnt;
11158 +       hinopath.dentry = vfsub_lkup_one(&hinoname, (struct path *)path);
11159 +       err = PTR_ERR(hinopath.dentry);
11160 +       if (IS_ERR(hinopath.dentry))
11161 +               goto out_unlock;
11162 +
11163 +       err = 0;
11164 +       flags = O_RDONLY;
11165 +       if (load) {
11166 +               if (d_is_negative(hinopath.dentry))
11167 +                       goto out_dput; /* success */
11168 +       } else {
11169 +               if (au_dr_hino_test_empty(&br->br_dirren)) {
11170 +                       if (d_is_positive(hinopath.dentry)) {
11171 +                               delegated = NULL;
11172 +                               err = vfsub_unlink(dir, &hinopath, &delegated,
11173 +                                                  /*force*/0);
11174 +                               AuTraceErr(err);
11175 +                               if (unlikely(err))
11176 +                                       pr_err("ignored err %d, %pd2\n",
11177 +                                              err, hinopath.dentry);
11178 +                               if (unlikely(err == -EWOULDBLOCK))
11179 +                                       iput(delegated);
11180 +                               err = 0;
11181 +                       }
11182 +                       goto out_dput;
11183 +               } else if (!d_is_positive(hinopath.dentry)) {
11184 +                       err = vfsub_create(dir, &hinopath, 0600,
11185 +                                          /*want_excl*/false);
11186 +                       AuTraceErr(err);
11187 +                       if (unlikely(err))
11188 +                               goto out_dput;
11189 +               }
11190 +               flags = O_WRONLY;
11191 +       }
11192 +       hinofile = vfsub_dentry_open(&hinopath, flags);
11193 +       if (suspend)
11194 +               au_hn_inode_unlock(hdir);
11195 +       else
11196 +               inode_unlock(dir);
11197 +       dput(hinopath.dentry);
11198 +       AuTraceErrPtr(hinofile);
11199 +       if (IS_ERR(hinofile)) {
11200 +               err = PTR_ERR(hinofile);
11201 +               goto out;
11202 +       }
11203 +
11204 +       if (load)
11205 +               err = au_dr_hino_load(&br->br_dirren, hinofile);
11206 +       else
11207 +               err = au_dr_hino_store(sb, br, hinofile);
11208 +       fput(hinofile);
11209 +       goto out;
11210 +
11211 +out_dput:
11212 +       dput(hinopath.dentry);
11213 +out_unlock:
11214 +       if (suspend)
11215 +               au_hn_inode_unlock(hdir);
11216 +       else
11217 +               inode_unlock(dir);
11218 +out:
11219 +       AuTraceErr(err);
11220 +       return err;
11221 +}
11222 +
11223 +/* ---------------------------------------------------------------------- */
11224 +
11225 +static int au_dr_brid_init(struct au_dr_brid *brid, const struct path *path)
11226 +{
11227 +       int err;
11228 +       struct kstatfs kstfs;
11229 +       dev_t dev;
11230 +       struct dentry *dentry;
11231 +       struct super_block *sb;
11232 +
11233 +       err = vfs_statfs((void *)path, &kstfs);
11234 +       AuTraceErr(err);
11235 +       if (unlikely(err))
11236 +               goto out;
11237 +
11238 +       /* todo: support for UUID */
11239 +
11240 +       if (kstfs.f_fsid.val[0] || kstfs.f_fsid.val[1]) {
11241 +               brid->type = AuBrid_FSID;
11242 +               brid->fsid = kstfs.f_fsid;
11243 +       } else {
11244 +               dentry = path->dentry;
11245 +               sb = dentry->d_sb;
11246 +               dev = sb->s_dev;
11247 +               if (dev) {
11248 +                       brid->type = AuBrid_DEV;
11249 +                       brid->dev = dev;
11250 +               }
11251 +       }
11252 +
11253 +out:
11254 +       return err;
11255 +}
11256 +
11257 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
11258 +                 const struct path *path)
11259 +{
11260 +       int err, i;
11261 +       struct au_dr_br *dr;
11262 +       struct hlist_bl_head *hbl;
11263 +
11264 +       dr = &br->br_dirren;
11265 +       hbl = dr->dr_h_ino;
11266 +       for (i = 0; i < AuDirren_NHASH; i++, hbl++)
11267 +               INIT_HLIST_BL_HEAD(hbl);
11268 +
11269 +       err = au_dr_brid_init(&dr->dr_brid, path);
11270 +       if (unlikely(err))
11271 +               goto out;
11272 +
11273 +       if (au_opt_test(au_mntflags(sb), DIRREN))
11274 +               err = au_dr_hino(sb, /*bindex*/-1, br, path);
11275 +
11276 +out:
11277 +       AuTraceErr(err);
11278 +       return err;
11279 +}
11280 +
11281 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br)
11282 +{
11283 +       int err;
11284 +
11285 +       err = 0;
11286 +       if (au_br_writable(br->br_perm))
11287 +               err = au_dr_hino(sb, /*bindex*/-1, br, /*path*/NULL);
11288 +       if (!err)
11289 +               au_dr_hino_free(&br->br_dirren);
11290 +
11291 +       return err;
11292 +}
11293 +
11294 +/* ---------------------------------------------------------------------- */
11295 +
11296 +static int au_brid_str(struct au_dr_brid *brid, struct inode *h_inode,
11297 +                      char *buf, size_t sz)
11298 +{
11299 +       int err;
11300 +       unsigned int major, minor;
11301 +       char *p;
11302 +
11303 +       p = buf;
11304 +       err = snprintf(p, sz, "%d_", brid->type);
11305 +       AuDebugOn(err > sz);
11306 +       p += err;
11307 +       sz -= err;
11308 +       switch (brid->type) {
11309 +       case AuBrid_Unset:
11310 +               return -EINVAL;
11311 +       case AuBrid_UUID:
11312 +               err = snprintf(p, sz, "%pU", brid->uuid.b);
11313 +               break;
11314 +       case AuBrid_FSID:
11315 +               err = snprintf(p, sz, "%08x-%08x",
11316 +                              brid->fsid.val[0], brid->fsid.val[1]);
11317 +               break;
11318 +       case AuBrid_DEV:
11319 +               major = MAJOR(brid->dev);
11320 +               minor = MINOR(brid->dev);
11321 +               if (major <= 0xff && minor <= 0xff)
11322 +                       err = snprintf(p, sz, "%02x%02x", major, minor);
11323 +               else
11324 +                       err = snprintf(p, sz, "%03x:%05x", major, minor);
11325 +               break;
11326 +       }
11327 +       AuDebugOn(err > sz);
11328 +       p += err;
11329 +       sz -= err;
11330 +       err = snprintf(p, sz, "_%llu", (unsigned long long)h_inode->i_ino);
11331 +       AuDebugOn(err > sz);
11332 +       p += err;
11333 +       sz -= err;
11334 +
11335 +       return p - buf;
11336 +}
11337 +
11338 +static int au_drinfo_name(struct au_branch *br, char *name, int len)
11339 +{
11340 +       int rlen;
11341 +       struct dentry *br_dentry;
11342 +       struct inode *br_inode;
11343 +
11344 +       br_dentry = au_br_dentry(br);
11345 +       br_inode = d_inode(br_dentry);
11346 +       rlen = au_brid_str(&br->br_dirren.dr_brid, br_inode, name, len);
11347 +       AuDebugOn(rlen >= AUFS_DIRREN_ENV_VAL_SZ);
11348 +       AuDebugOn(rlen > len);
11349 +
11350 +       return rlen;
11351 +}
11352 +
11353 +/* ---------------------------------------------------------------------- */
11354 +
11355 +/*
11356 + * from the given @h_dentry, construct drinfo at @*fdata.
11357 + * when the size of @*fdata is not enough, reallocate and return new @fdata and
11358 + * @allocated.
11359 + */
11360 +static int au_drinfo_construct(struct au_drinfo_fdata **fdata,
11361 +                              struct dentry *h_dentry,
11362 +                              unsigned char *allocated)
11363 +{
11364 +       int err, v;
11365 +       struct au_drinfo_fdata *f, *p;
11366 +       struct au_drinfo *drinfo;
11367 +       struct inode *h_inode;
11368 +       struct qstr *qname;
11369 +
11370 +       err = 0;
11371 +       f = *fdata;
11372 +       h_inode = d_inode(h_dentry);
11373 +       qname = &h_dentry->d_name;
11374 +       drinfo = &f->drinfo;
11375 +       drinfo->ino = (__force uint64_t)cpu_to_be64(h_inode->i_ino);
11376 +       drinfo->oldnamelen = qname->len;
11377 +       if (*allocated < sizeof(*f) + qname->len) {
11378 +               v = roundup_pow_of_two(*allocated + qname->len);
11379 +               p = au_krealloc(f, v, GFP_NOFS, /*may_shrink*/0);
11380 +               if (unlikely(!p)) {
11381 +                       err = -ENOMEM;
11382 +                       AuTraceErr(err);
11383 +                       goto out;
11384 +               }
11385 +               f = p;
11386 +               *fdata = f;
11387 +               *allocated = v;
11388 +               drinfo = &f->drinfo;
11389 +       }
11390 +       memcpy(drinfo->oldname, qname->name, qname->len);
11391 +       AuDbg("i%llu, %.*s\n",
11392 +             be64_to_cpu((__force __be64)drinfo->ino), drinfo->oldnamelen,
11393 +             drinfo->oldname);
11394 +
11395 +out:
11396 +       AuTraceErr(err);
11397 +       return err;
11398 +}
11399 +
11400 +/* callers have to free the return value */
11401 +static struct au_drinfo *au_drinfo_read_k(struct file *file, ino_t h_ino)
11402 +{
11403 +       struct au_drinfo *ret, *drinfo;
11404 +       struct au_drinfo_fdata fdata;
11405 +       int len;
11406 +       loff_t pos;
11407 +       ssize_t ssz;
11408 +
11409 +       ret = ERR_PTR(-EIO);
11410 +       pos = 0;
11411 +       ssz = vfsub_read_k(file, &fdata, sizeof(fdata), &pos);
11412 +       if (unlikely(ssz != sizeof(fdata))) {
11413 +               AuIOErr("ssz %zd, %u, %pD2\n",
11414 +                       ssz, (unsigned int)sizeof(fdata), file);
11415 +               goto out;
11416 +       }
11417 +
11418 +       fdata.magic = ntohl((__force __be32)fdata.magic);
11419 +       switch (fdata.magic) {
11420 +       case AUFS_DRINFO_MAGIC_V1:
11421 +               break;
11422 +       default:
11423 +               AuIOErr("magic-num 0x%x, 0x%x, %pD2\n",
11424 +                       fdata.magic, AUFS_DRINFO_MAGIC_V1, file);
11425 +               goto out;
11426 +       }
11427 +
11428 +       drinfo = &fdata.drinfo;
11429 +       len = drinfo->oldnamelen;
11430 +       if (!len) {
11431 +               AuIOErr("broken drinfo %pD2\n", file);
11432 +               goto out;
11433 +       }
11434 +
11435 +       ret = NULL;
11436 +       drinfo->ino = be64_to_cpu((__force __be64)drinfo->ino);
11437 +       if (unlikely(h_ino && drinfo->ino != h_ino)) {
11438 +               AuDbg("ignored i%llu, i%llu, %pD2\n",
11439 +                     (unsigned long long)drinfo->ino,
11440 +                     (unsigned long long)h_ino, file);
11441 +               goto out; /* success */
11442 +       }
11443 +
11444 +       ret = kmalloc(sizeof(*ret) + len, GFP_NOFS);
11445 +       if (unlikely(!ret)) {
11446 +               ret = ERR_PTR(-ENOMEM);
11447 +               AuTraceErrPtr(ret);
11448 +               goto out;
11449 +       }
11450 +
11451 +       *ret = *drinfo;
11452 +       ssz = vfsub_read_k(file, (void *)ret->oldname, len, &pos);
11453 +       if (unlikely(ssz != len)) {
11454 +               au_kfree_rcu(ret);
11455 +               ret = ERR_PTR(-EIO);
11456 +               AuIOErr("ssz %zd, %u, %pD2\n", ssz, len, file);
11457 +               goto out;
11458 +       }
11459 +
11460 +       AuDbg("oldname %.*s\n", ret->oldnamelen, ret->oldname);
11461 +
11462 +out:
11463 +       return ret;
11464 +}
11465 +
11466 +/* ---------------------------------------------------------------------- */
11467 +
11468 +/* in order to be revertible */
11469 +struct au_drinfo_rev_elm {
11470 +       int                     created;
11471 +       struct dentry           *info_dentry;
11472 +       struct au_drinfo        *info_last;
11473 +};
11474 +
11475 +struct au_drinfo_rev {
11476 +       unsigned char                   already;
11477 +       aufs_bindex_t                   nelm;
11478 +       struct au_drinfo_rev_elm        elm[];
11479 +};
11480 +
11481 +/* todo: isn't it too large? */
11482 +struct au_drinfo_store {
11483 +       struct path h_ppath;
11484 +       struct dentry *h_dentry;
11485 +       struct au_drinfo_fdata *fdata;
11486 +       char *infoname;                 /* inside of whname, just after PFX */
11487 +       char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ];
11488 +       aufs_bindex_t btgt, btail;
11489 +       unsigned char no_sio,
11490 +               allocated,              /* current size of *fdata */
11491 +               infonamelen,            /* room size for p */
11492 +               whnamelen,              /* length of the generated name */
11493 +               renameback;             /* renamed back */
11494 +};
11495 +
11496 +/* on rename(2) error, the caller should revert it using @elm */
11497 +static int au_drinfo_do_store(struct au_drinfo_store *w,
11498 +                             struct au_drinfo_rev_elm *elm)
11499 +{
11500 +       int err, len;
11501 +       ssize_t ssz;
11502 +       loff_t pos;
11503 +       struct path infopath = {
11504 +               .mnt = w->h_ppath.mnt
11505 +       };
11506 +       struct inode *h_dir, *h_inode, *delegated;
11507 +       struct file *infofile;
11508 +       struct qstr *qname;
11509 +
11510 +       AuDebugOn(elm
11511 +                 && memcmp(elm, page_address(ZERO_PAGE(0)), sizeof(*elm)));
11512 +
11513 +       infopath.dentry = vfsub_lookup_one_len(w->whname, &w->h_ppath,
11514 +                                              w->whnamelen);
11515 +       AuTraceErrPtr(infopath.dentry);
11516 +       if (IS_ERR(infopath.dentry)) {
11517 +               err = PTR_ERR(infopath.dentry);
11518 +               goto out;
11519 +       }
11520 +
11521 +       err = 0;
11522 +       h_dir = d_inode(w->h_ppath.dentry);
11523 +       if (elm && d_is_negative(infopath.dentry)) {
11524 +               err = vfsub_create(h_dir, &infopath, 0600, /*want_excl*/true);
11525 +               AuTraceErr(err);
11526 +               if (unlikely(err))
11527 +                       goto out_dput;
11528 +               elm->created = 1;
11529 +               elm->info_dentry = dget(infopath.dentry);
11530 +       }
11531 +
11532 +       infofile = vfsub_dentry_open(&infopath, O_RDWR);
11533 +       AuTraceErrPtr(infofile);
11534 +       if (IS_ERR(infofile)) {
11535 +               err = PTR_ERR(infofile);
11536 +               goto out_dput;
11537 +       }
11538 +
11539 +       h_inode = d_inode(infopath.dentry);
11540 +       if (elm && i_size_read(h_inode)) {
11541 +               h_inode = d_inode(w->h_dentry);
11542 +               elm->info_last = au_drinfo_read_k(infofile, h_inode->i_ino);
11543 +               AuTraceErrPtr(elm->info_last);
11544 +               if (IS_ERR(elm->info_last)) {
11545 +                       err = PTR_ERR(elm->info_last);
11546 +                       elm->info_last = NULL;
11547 +                       AuDebugOn(elm->info_dentry);
11548 +                       goto out_fput;
11549 +               }
11550 +       }
11551 +
11552 +       if (elm && w->renameback) {
11553 +               delegated = NULL;
11554 +               err = vfsub_unlink(h_dir, &infopath, &delegated, /*force*/0);
11555 +               AuTraceErr(err);
11556 +               if (unlikely(err == -EWOULDBLOCK))
11557 +                       iput(delegated);
11558 +               goto out_fput;
11559 +       }
11560 +
11561 +       pos = 0;
11562 +       qname = &w->h_dentry->d_name;
11563 +       len = sizeof(*w->fdata) + qname->len;
11564 +       if (!elm)
11565 +               len = sizeof(*w->fdata) + w->fdata->drinfo.oldnamelen;
11566 +       ssz = vfsub_write_k(infofile, w->fdata, len, &pos);
11567 +       if (ssz == len) {
11568 +               AuDbg("hi%llu, %.*s\n", w->fdata->drinfo.ino,
11569 +                     w->fdata->drinfo.oldnamelen, w->fdata->drinfo.oldname);
11570 +               goto out_fput; /* success */
11571 +       } else {
11572 +               err = -EIO;
11573 +               if (ssz < 0)
11574 +                       err = ssz;
11575 +               /* the caller should revert it using @elm */
11576 +       }
11577 +
11578 +out_fput:
11579 +       fput(infofile);
11580 +out_dput:
11581 +       dput(infopath.dentry);
11582 +out:
11583 +       AuTraceErr(err);
11584 +       return err;
11585 +}
11586 +
11587 +struct au_call_drinfo_do_store_args {
11588 +       int *errp;
11589 +       struct au_drinfo_store *w;
11590 +       struct au_drinfo_rev_elm *elm;
11591 +};
11592 +
11593 +static void au_call_drinfo_do_store(void *args)
11594 +{
11595 +       struct au_call_drinfo_do_store_args *a = args;
11596 +
11597 +       *a->errp = au_drinfo_do_store(a->w, a->elm);
11598 +}
11599 +
11600 +static int au_drinfo_store_sio(struct au_drinfo_store *w,
11601 +                              struct au_drinfo_rev_elm *elm)
11602 +{
11603 +       int err, wkq_err;
11604 +
11605 +       if (w->no_sio)
11606 +               err = au_drinfo_do_store(w, elm);
11607 +       else {
11608 +               struct au_call_drinfo_do_store_args a = {
11609 +                       .errp   = &err,
11610 +                       .w      = w,
11611 +                       .elm    = elm
11612 +               };
11613 +               wkq_err = au_wkq_wait(au_call_drinfo_do_store, &a);
11614 +               if (unlikely(wkq_err))
11615 +                       err = wkq_err;
11616 +       }
11617 +       AuTraceErr(err);
11618 +
11619 +       return err;
11620 +}
11621 +
11622 +static int au_drinfo_store_work_init(struct au_drinfo_store *w,
11623 +                                    aufs_bindex_t btgt)
11624 +{
11625 +       int err;
11626 +
11627 +       memset(w, 0, sizeof(*w));
11628 +       w->allocated = roundup_pow_of_two(sizeof(*w->fdata) + 40);
11629 +       strcpy(w->whname, AUFS_WH_DR_INFO_PFX);
11630 +       w->infoname = w->whname + sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11631 +       w->infonamelen = sizeof(w->whname) - sizeof(AUFS_WH_DR_INFO_PFX);
11632 +       w->btgt = btgt;
11633 +       w->no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
11634 +
11635 +       err = -ENOMEM;
11636 +       w->fdata = kcalloc(1, w->allocated, GFP_NOFS);
11637 +       if (unlikely(!w->fdata)) {
11638 +               AuTraceErr(err);
11639 +               goto out;
11640 +       }
11641 +       w->fdata->magic = (__force uint32_t)htonl(AUFS_DRINFO_MAGIC_V1);
11642 +       err = 0;
11643 +
11644 +out:
11645 +       return err;
11646 +}
11647 +
11648 +static void au_drinfo_store_work_fin(struct au_drinfo_store *w)
11649 +{
11650 +       au_kfree_rcu(w->fdata);
11651 +}
11652 +
11653 +static void au_drinfo_store_rev(struct au_drinfo_rev *rev,
11654 +                               struct au_drinfo_store *w)
11655 +{
11656 +       struct au_drinfo_rev_elm *elm;
11657 +       struct inode *h_dir, *delegated;
11658 +       int err, nelm;
11659 +       struct path infopath = {
11660 +               .mnt = w->h_ppath.mnt
11661 +       };
11662 +
11663 +       h_dir = d_inode(w->h_ppath.dentry);
11664 +       IMustLock(h_dir);
11665 +
11666 +       err = 0;
11667 +       elm = rev->elm;
11668 +       for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11669 +               AuDebugOn(elm->created && elm->info_last);
11670 +               if (elm->created) {
11671 +                       AuDbg("here\n");
11672 +                       delegated = NULL;
11673 +                       infopath.dentry = elm->info_dentry;
11674 +                       err = vfsub_unlink(h_dir, &infopath, &delegated,
11675 +                                          !w->no_sio);
11676 +                       AuTraceErr(err);
11677 +                       if (unlikely(err == -EWOULDBLOCK))
11678 +                               iput(delegated);
11679 +                       dput(elm->info_dentry);
11680 +               } else if (elm->info_last) {
11681 +                       AuDbg("here\n");
11682 +                       w->fdata->drinfo = *elm->info_last;
11683 +                       memcpy(w->fdata->drinfo.oldname,
11684 +                              elm->info_last->oldname,
11685 +                              elm->info_last->oldnamelen);
11686 +                       err = au_drinfo_store_sio(w, /*elm*/NULL);
11687 +                       au_kfree_rcu(elm->info_last);
11688 +               }
11689 +               if (unlikely(err))
11690 +                       AuIOErr("%d, %s\n", err, w->whname);
11691 +               /* go on even if err */
11692 +       }
11693 +}
11694 +
11695 +/* caller has to call au_dr_rename_fin() later */
11696 +static int au_drinfo_store(struct dentry *dentry, aufs_bindex_t btgt,
11697 +                          struct qstr *dst_name, void *_rev)
11698 +{
11699 +       int err, sz, nelm;
11700 +       aufs_bindex_t bindex, btail;
11701 +       struct au_drinfo_store work;
11702 +       struct au_drinfo_rev *rev, **p;
11703 +       struct au_drinfo_rev_elm *elm;
11704 +       struct super_block *sb;
11705 +       struct au_branch *br;
11706 +       struct au_hinode *hdir;
11707 +
11708 +       err = au_drinfo_store_work_init(&work, btgt);
11709 +       AuTraceErr(err);
11710 +       if (unlikely(err))
11711 +               goto out;
11712 +
11713 +       err = -ENOMEM;
11714 +       btail = au_dbtaildir(dentry);
11715 +       nelm = btail - btgt;
11716 +       sz = sizeof(*rev) + sizeof(*elm) * nelm;
11717 +       rev = kcalloc(1, sz, GFP_NOFS);
11718 +       if (unlikely(!rev)) {
11719 +               AuTraceErr(err);
11720 +               goto out_args;
11721 +       }
11722 +       rev->nelm = nelm;
11723 +       elm = rev->elm;
11724 +       p = _rev;
11725 +       *p = rev;
11726 +
11727 +       err = 0;
11728 +       sb = dentry->d_sb;
11729 +       work.h_ppath.dentry = au_h_dptr(dentry, btgt);
11730 +       work.h_ppath.mnt = au_sbr_mnt(sb, btgt);
11731 +       hdir = au_hi(d_inode(dentry), btgt);
11732 +       au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11733 +       for (bindex = btgt + 1; bindex <= btail; bindex++, elm++) {
11734 +               work.h_dentry = au_h_dptr(dentry, bindex);
11735 +               if (!work.h_dentry)
11736 +                       continue;
11737 +
11738 +               err = au_drinfo_construct(&work.fdata, work.h_dentry,
11739 +                                         &work.allocated);
11740 +               AuTraceErr(err);
11741 +               if (unlikely(err))
11742 +                       break;
11743 +
11744 +               work.renameback = au_qstreq(&work.h_dentry->d_name, dst_name);
11745 +               br = au_sbr(sb, bindex);
11746 +               work.whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11747 +               work.whnamelen += au_drinfo_name(br, work.infoname,
11748 +                                                work.infonamelen);
11749 +               AuDbg("whname %.*s, i%llu, %.*s\n",
11750 +                     work.whnamelen, work.whname,
11751 +                     be64_to_cpu((__force __be64)work.fdata->drinfo.ino),
11752 +                     work.fdata->drinfo.oldnamelen,
11753 +                     work.fdata->drinfo.oldname);
11754 +
11755 +               err = au_drinfo_store_sio(&work, elm);
11756 +               AuTraceErr(err);
11757 +               if (unlikely(err))
11758 +                       break;
11759 +       }
11760 +       if (unlikely(err)) {
11761 +               /* revert all drinfo */
11762 +               au_drinfo_store_rev(rev, &work);
11763 +               au_kfree_try_rcu(rev);
11764 +               *p = NULL;
11765 +       }
11766 +       au_hn_inode_unlock(hdir);
11767 +
11768 +out_args:
11769 +       au_drinfo_store_work_fin(&work);
11770 +out:
11771 +       return err;
11772 +}
11773 +
11774 +/* ---------------------------------------------------------------------- */
11775 +
11776 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
11777 +                struct qstr *dst_name, void *_rev)
11778 +{
11779 +       int err, already;
11780 +       ino_t ino;
11781 +       struct super_block *sb;
11782 +       struct au_branch *br;
11783 +       struct au_dr_br *dr;
11784 +       struct dentry *h_dentry;
11785 +       struct inode *h_inode;
11786 +       struct au_dr_hino *ent;
11787 +       struct au_drinfo_rev *rev, **p;
11788 +
11789 +       AuDbg("bindex %d\n", bindex);
11790 +
11791 +       err = -ENOMEM;
11792 +       ent = kmalloc(sizeof(*ent), GFP_NOFS);
11793 +       if (unlikely(!ent))
11794 +               goto out;
11795 +
11796 +       sb = src->d_sb;
11797 +       br = au_sbr(sb, bindex);
11798 +       dr = &br->br_dirren;
11799 +       h_dentry = au_h_dptr(src, bindex);
11800 +       h_inode = d_inode(h_dentry);
11801 +       ino = h_inode->i_ino;
11802 +       ent->dr_h_ino = ino;
11803 +       already = au_dr_hino_test_add(dr, ino, ent);
11804 +       AuDbg("b%d, hi%llu, already %d\n",
11805 +             bindex, (unsigned long long)ino, already);
11806 +
11807 +       err = au_drinfo_store(src, bindex, dst_name, _rev);
11808 +       AuTraceErr(err);
11809 +       if (!err) {
11810 +               p = _rev;
11811 +               rev = *p;
11812 +               rev->already = already;
11813 +               goto out; /* success */
11814 +       }
11815 +
11816 +       /* revert */
11817 +       if (!already)
11818 +               au_dr_hino_del(dr, ent);
11819 +       au_kfree_rcu(ent);
11820 +
11821 +out:
11822 +       AuTraceErr(err);
11823 +       return err;
11824 +}
11825 +
11826 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11827 +{
11828 +       struct au_drinfo_rev *rev;
11829 +       struct au_drinfo_rev_elm *elm;
11830 +       int nelm;
11831 +
11832 +       rev = _rev;
11833 +       elm = rev->elm;
11834 +       for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11835 +               dput(elm->info_dentry);
11836 +               au_kfree_rcu(elm->info_last);
11837 +       }
11838 +       au_kfree_try_rcu(rev);
11839 +}
11840 +
11841 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11842 +{
11843 +       int err;
11844 +       struct au_drinfo_store work;
11845 +       struct au_drinfo_rev *rev = _rev;
11846 +       struct super_block *sb;
11847 +       struct au_branch *br;
11848 +       struct inode *h_inode;
11849 +       struct au_dr_br *dr;
11850 +       struct au_dr_hino *ent;
11851 +
11852 +       err = au_drinfo_store_work_init(&work, btgt);
11853 +       if (unlikely(err))
11854 +               goto out;
11855 +
11856 +       sb = src->d_sb;
11857 +       br = au_sbr(sb, btgt);
11858 +       work.h_ppath.dentry = au_h_dptr(src, btgt);
11859 +       work.h_ppath.mnt = au_br_mnt(br);
11860 +       au_drinfo_store_rev(rev, &work);
11861 +       au_drinfo_store_work_fin(&work);
11862 +       if (rev->already)
11863 +               goto out;
11864 +
11865 +       dr = &br->br_dirren;
11866 +       h_inode = d_inode(work.h_ppath.dentry);
11867 +       ent = au_dr_hino_find(dr, h_inode->i_ino);
11868 +       BUG_ON(!ent);
11869 +       au_dr_hino_del(dr, ent);
11870 +       au_kfree_rcu(ent);
11871 +
11872 +out:
11873 +       au_kfree_try_rcu(rev);
11874 +       if (unlikely(err))
11875 +               pr_err("failed to remove dirren info\n");
11876 +}
11877 +
11878 +/* ---------------------------------------------------------------------- */
11879 +
11880 +static struct au_drinfo *au_drinfo_do_load(struct path *h_ppath,
11881 +                                          char *whname, int whnamelen,
11882 +                                          struct dentry **info_dentry)
11883 +{
11884 +       struct au_drinfo *drinfo;
11885 +       struct file *f;
11886 +       struct inode *h_dir;
11887 +       struct path infopath;
11888 +       int unlocked;
11889 +
11890 +       AuDbg("%pd/%.*s\n", h_ppath->dentry, whnamelen, whname);
11891 +
11892 +       *info_dentry = NULL;
11893 +       drinfo = NULL;
11894 +       unlocked = 0;
11895 +       h_dir = d_inode(h_ppath->dentry);
11896 +       inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
11897 +       infopath.dentry = vfsub_lookup_one_len(whname, h_ppath, whnamelen);
11898 +       if (IS_ERR(infopath.dentry)) {
11899 +               drinfo = (void *)infopath.dentry;
11900 +               goto out;
11901 +       }
11902 +
11903 +       if (d_is_negative(infopath.dentry))
11904 +               goto out_dput; /* success */
11905 +
11906 +       infopath.mnt = h_ppath->mnt;
11907 +       f = vfsub_dentry_open(&infopath, O_RDONLY);
11908 +       inode_unlock_shared(h_dir);
11909 +       unlocked = 1;
11910 +       if (IS_ERR(f)) {
11911 +               drinfo = (void *)f;
11912 +               goto out_dput;
11913 +       }
11914 +
11915 +       drinfo = au_drinfo_read_k(f, /*h_ino*/0);
11916 +       if (IS_ERR_OR_NULL(drinfo))
11917 +               goto out_fput;
11918 +
11919 +       AuDbg("oldname %.*s\n", drinfo->oldnamelen, drinfo->oldname);
11920 +       *info_dentry = dget(infopath.dentry); /* keep it alive */
11921 +
11922 +out_fput:
11923 +       fput(f);
11924 +out_dput:
11925 +       dput(infopath.dentry);
11926 +out:
11927 +       if (!unlocked)
11928 +               inode_unlock_shared(h_dir);
11929 +       AuTraceErrPtr(drinfo);
11930 +       return drinfo;
11931 +}
11932 +
11933 +struct au_drinfo_do_load_args {
11934 +       struct au_drinfo **drinfop;
11935 +       struct path *h_ppath;
11936 +       char *whname;
11937 +       int whnamelen;
11938 +       struct dentry **info_dentry;
11939 +};
11940 +
11941 +static void au_call_drinfo_do_load(void *args)
11942 +{
11943 +       struct au_drinfo_do_load_args *a = args;
11944 +
11945 +       *a->drinfop = au_drinfo_do_load(a->h_ppath, a->whname, a->whnamelen,
11946 +                                       a->info_dentry);
11947 +}
11948 +
11949 +struct au_drinfo_load {
11950 +       struct path h_ppath;
11951 +       struct qstr *qname;
11952 +       unsigned char no_sio;
11953 +
11954 +       aufs_bindex_t ninfo;
11955 +       struct au_drinfo **drinfo;
11956 +};
11957 +
11958 +static int au_drinfo_load(struct au_drinfo_load *w, aufs_bindex_t bindex,
11959 +                         struct au_branch *br)
11960 +{
11961 +       int err, wkq_err, whnamelen, e;
11962 +       char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ]
11963 +               = AUFS_WH_DR_INFO_PFX;
11964 +       struct au_drinfo *drinfo;
11965 +       struct qstr oldname;
11966 +       struct inode *h_dir, *delegated;
11967 +       struct dentry *info_dentry;
11968 +       struct path infopath;
11969 +
11970 +       whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11971 +       whnamelen += au_drinfo_name(br, whname + whnamelen,
11972 +                                   sizeof(whname) - whnamelen);
11973 +       if (w->no_sio)
11974 +               drinfo = au_drinfo_do_load(&w->h_ppath, whname, whnamelen,
11975 +                                          &info_dentry);
11976 +       else {
11977 +               struct au_drinfo_do_load_args args = {
11978 +                       .drinfop        = &drinfo,
11979 +                       .h_ppath        = &w->h_ppath,
11980 +                       .whname         = whname,
11981 +                       .whnamelen      = whnamelen,
11982 +                       .info_dentry    = &info_dentry
11983 +               };
11984 +               wkq_err = au_wkq_wait(au_call_drinfo_do_load, &args);
11985 +               if (unlikely(wkq_err))
11986 +                       drinfo = ERR_PTR(wkq_err);
11987 +       }
11988 +       err = PTR_ERR(drinfo);
11989 +       if (IS_ERR_OR_NULL(drinfo))
11990 +               goto out;
11991 +
11992 +       err = 0;
11993 +       oldname.len = drinfo->oldnamelen;
11994 +       oldname.name = drinfo->oldname;
11995 +       if (au_qstreq(w->qname, &oldname)) {
11996 +               /* the name is renamed back */
11997 +               au_kfree_rcu(drinfo);
11998 +               drinfo = NULL;
11999 +
12000 +               infopath.dentry = info_dentry;
12001 +               infopath.mnt = w->h_ppath.mnt;
12002 +               h_dir = d_inode(w->h_ppath.dentry);
12003 +               delegated = NULL;
12004 +               inode_lock_nested(h_dir, AuLsc_I_PARENT);
12005 +               e = vfsub_unlink(h_dir, &infopath, &delegated, !w->no_sio);
12006 +               inode_unlock(h_dir);
12007 +               if (unlikely(e))
12008 +                       AuIOErr("ignored %d, %pd2\n", e, &infopath.dentry);
12009 +               if (unlikely(e == -EWOULDBLOCK))
12010 +                       iput(delegated);
12011 +       }
12012 +       au_kfree_rcu(w->drinfo[bindex]);
12013 +       w->drinfo[bindex] = drinfo;
12014 +       dput(info_dentry);
12015 +
12016 +out:
12017 +       AuTraceErr(err);
12018 +       return err;
12019 +}
12020 +
12021 +/* ---------------------------------------------------------------------- */
12022 +
12023 +static void au_dr_lkup_free(struct au_drinfo **drinfo, int n)
12024 +{
12025 +       struct au_drinfo **p = drinfo;
12026 +
12027 +       while (n-- > 0)
12028 +               au_kfree_rcu(*drinfo++);
12029 +       au_kfree_try_rcu(p);
12030 +}
12031 +
12032 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12033 +              aufs_bindex_t btgt)
12034 +{
12035 +       int err, ninfo;
12036 +       struct au_drinfo_load w;
12037 +       aufs_bindex_t bindex, bbot;
12038 +       struct au_branch *br;
12039 +       struct inode *h_dir;
12040 +       struct au_dr_hino *ent;
12041 +       struct super_block *sb;
12042 +
12043 +       AuDbg("%.*s, name %.*s, whname %.*s, b%d\n",
12044 +             AuLNPair(&dentry->d_name), AuLNPair(&lkup->dirren.dr_name),
12045 +             AuLNPair(&lkup->whname), btgt);
12046 +
12047 +       sb = dentry->d_sb;
12048 +       bbot = au_sbbot(sb);
12049 +       w.ninfo = bbot + 1;
12050 +       if (!lkup->dirren.drinfo) {
12051 +               lkup->dirren.drinfo = kcalloc(w.ninfo,
12052 +                                             sizeof(*lkup->dirren.drinfo),
12053 +                                             GFP_NOFS);
12054 +               if (unlikely(!lkup->dirren.drinfo)) {
12055 +                       err = -ENOMEM;
12056 +                       goto out;
12057 +               }
12058 +               lkup->dirren.ninfo = w.ninfo;
12059 +       }
12060 +       w.drinfo = lkup->dirren.drinfo;
12061 +       w.no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
12062 +       w.h_ppath.dentry = au_h_dptr(dentry, btgt);
12063 +       AuDebugOn(!w.h_ppath.dentry);
12064 +       w.h_ppath.mnt = au_sbr_mnt(sb, btgt);
12065 +       w.qname = &dentry->d_name;
12066 +
12067 +       ninfo = 0;
12068 +       for (bindex = btgt + 1; bindex <= bbot; bindex++) {
12069 +               br = au_sbr(sb, bindex);
12070 +               err = au_drinfo_load(&w, bindex, br);
12071 +               if (unlikely(err))
12072 +                       goto out_free;
12073 +               if (w.drinfo[bindex])
12074 +                       ninfo++;
12075 +       }
12076 +       if (!ninfo) {
12077 +               br = au_sbr(sb, btgt);
12078 +               h_dir = d_inode(w.h_ppath.dentry);
12079 +               ent = au_dr_hino_find(&br->br_dirren, h_dir->i_ino);
12080 +               AuDebugOn(!ent);
12081 +               au_dr_hino_del(&br->br_dirren, ent);
12082 +               au_kfree_rcu(ent);
12083 +       }
12084 +       goto out; /* success */
12085 +
12086 +out_free:
12087 +       au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12088 +       lkup->dirren.ninfo = 0;
12089 +       lkup->dirren.drinfo = NULL;
12090 +out:
12091 +       AuTraceErr(err);
12092 +       return err;
12093 +}
12094 +
12095 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup)
12096 +{
12097 +       au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12098 +}
12099 +
12100 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt)
12101 +{
12102 +       int err;
12103 +       struct au_drinfo *drinfo;
12104 +
12105 +       err = 0;
12106 +       if (!lkup->dirren.drinfo)
12107 +               goto out;
12108 +       AuDebugOn(lkup->dirren.ninfo <= btgt);
12109 +       drinfo = lkup->dirren.drinfo[btgt];
12110 +       if (!drinfo)
12111 +               goto out;
12112 +
12113 +       au_kfree_try_rcu(lkup->whname.name);
12114 +       lkup->whname.name = NULL;
12115 +       lkup->dirren.dr_name.len = drinfo->oldnamelen;
12116 +       lkup->dirren.dr_name.name = drinfo->oldname;
12117 +       lkup->name = &lkup->dirren.dr_name;
12118 +       err = au_wh_name_alloc(&lkup->whname, lkup->name);
12119 +       if (!err)
12120 +               AuDbg("name %.*s, whname %.*s, b%d\n",
12121 +                     AuLNPair(lkup->name), AuLNPair(&lkup->whname),
12122 +                     btgt);
12123 +
12124 +out:
12125 +       AuTraceErr(err);
12126 +       return err;
12127 +}
12128 +
12129 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12130 +                    ino_t h_ino)
12131 +{
12132 +       int match;
12133 +       struct au_drinfo *drinfo;
12134 +
12135 +       match = 1;
12136 +       if (!lkup->dirren.drinfo)
12137 +               goto out;
12138 +       AuDebugOn(lkup->dirren.ninfo <= bindex);
12139 +       drinfo = lkup->dirren.drinfo[bindex];
12140 +       if (!drinfo)
12141 +               goto out;
12142 +
12143 +       match = (drinfo->ino == h_ino);
12144 +       AuDbg("match %d\n", match);
12145 +
12146 +out:
12147 +       return match;
12148 +}
12149 +
12150 +/* ---------------------------------------------------------------------- */
12151 +
12152 +int au_dr_opt_set(struct super_block *sb)
12153 +{
12154 +       int err;
12155 +       aufs_bindex_t bindex, bbot;
12156 +       struct au_branch *br;
12157 +
12158 +       err = 0;
12159 +       bbot = au_sbbot(sb);
12160 +       for (bindex = 0; !err && bindex <= bbot; bindex++) {
12161 +               br = au_sbr(sb, bindex);
12162 +               err = au_dr_hino(sb, bindex, /*br*/NULL, &br->br_path);
12163 +       }
12164 +
12165 +       return err;
12166 +}
12167 +
12168 +int au_dr_opt_flush(struct super_block *sb)
12169 +{
12170 +       int err;
12171 +       aufs_bindex_t bindex, bbot;
12172 +       struct au_branch *br;
12173 +
12174 +       err = 0;
12175 +       bbot = au_sbbot(sb);
12176 +       for (bindex = 0; !err && bindex <= bbot; bindex++) {
12177 +               br = au_sbr(sb, bindex);
12178 +               if (au_br_writable(br->br_perm))
12179 +                       err = au_dr_hino(sb, bindex, /*br*/NULL, /*path*/NULL);
12180 +       }
12181 +
12182 +       return err;
12183 +}
12184 +
12185 +int au_dr_opt_clr(struct super_block *sb, int no_flush)
12186 +{
12187 +       int err;
12188 +       aufs_bindex_t bindex, bbot;
12189 +       struct au_branch *br;
12190 +
12191 +       err = 0;
12192 +       if (!no_flush) {
12193 +               err = au_dr_opt_flush(sb);
12194 +               if (unlikely(err))
12195 +                       goto out;
12196 +       }
12197 +
12198 +       bbot = au_sbbot(sb);
12199 +       for (bindex = 0; bindex <= bbot; bindex++) {
12200 +               br = au_sbr(sb, bindex);
12201 +               au_dr_hino_free(&br->br_dirren);
12202 +       }
12203 +
12204 +out:
12205 +       return err;
12206 +}
12207 diff -urN /usr/share/empty/fs/aufs/dirren.h linux/fs/aufs/dirren.h
12208 --- /usr/share/empty/fs/aufs/dirren.h   1970-01-01 01:00:00.000000000 +0100
12209 +++ linux/fs/aufs/dirren.h      2022-11-05 23:02:18.962555950 +0100
12210 @@ -0,0 +1,140 @@
12211 +/* SPDX-License-Identifier: GPL-2.0 */
12212 +/*
12213 + * Copyright (C) 2017-2022 Junjiro R. Okajima
12214 + *
12215 + * This program is free software; you can redistribute it and/or modify
12216 + * it under the terms of the GNU General Public License as published by
12217 + * the Free Software Foundation; either version 2 of the License, or
12218 + * (at your option) any later version.
12219 + *
12220 + * This program is distributed in the hope that it will be useful,
12221 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12222 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12223 + * GNU General Public License for more details.
12224 + *
12225 + * You should have received a copy of the GNU General Public License
12226 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12227 + */
12228 +
12229 +/*
12230 + * renamed dir info
12231 + */
12232 +
12233 +#ifndef __AUFS_DIRREN_H__
12234 +#define __AUFS_DIRREN_H__
12235 +
12236 +#ifdef __KERNEL__
12237 +
12238 +#include <linux/dcache.h>
12239 +#include <linux/statfs.h>
12240 +#include <linux/uuid.h>
12241 +#include "hbl.h"
12242 +
12243 +#define AuDirren_NHASH 100
12244 +
12245 +#ifdef CONFIG_AUFS_DIRREN
12246 +enum au_brid_type {
12247 +       AuBrid_Unset,
12248 +       AuBrid_UUID,
12249 +       AuBrid_FSID,
12250 +       AuBrid_DEV
12251 +};
12252 +
12253 +struct au_dr_brid {
12254 +       enum au_brid_type       type;
12255 +       union {
12256 +               uuid_t  uuid;   /* unimplemented yet */
12257 +               fsid_t  fsid;
12258 +               dev_t   dev;
12259 +       };
12260 +};
12261 +
12262 +/* 20 is the max digits length of ulong 64 */
12263 +/* brid-type "_" uuid "_" inum */
12264 +#define AUFS_DIRREN_FNAME_SZ   (1 + 1 + UUID_STRING_LEN + 20)
12265 +#define AUFS_DIRREN_ENV_VAL_SZ (AUFS_DIRREN_FNAME_SZ + 1 + 20)
12266 +
12267 +struct au_dr_hino {
12268 +       struct hlist_bl_node    dr_hnode;
12269 +       ino_t                   dr_h_ino;
12270 +};
12271 +
12272 +struct au_dr_br {
12273 +       struct hlist_bl_head    dr_h_ino[AuDirren_NHASH];
12274 +       struct au_dr_brid       dr_brid;
12275 +};
12276 +
12277 +struct au_dr_lookup {
12278 +       /* dr_name is pointed by struct au_do_lookup_args.name */
12279 +       struct qstr             dr_name; /* subset of dr_info */
12280 +       aufs_bindex_t           ninfo;
12281 +       struct au_drinfo        **drinfo;
12282 +};
12283 +#else
12284 +struct au_dr_hino;
12285 +/* empty */
12286 +struct au_dr_br { };
12287 +struct au_dr_lookup { };
12288 +#endif
12289 +
12290 +/* ---------------------------------------------------------------------- */
12291 +
12292 +struct au_branch;
12293 +struct au_do_lookup_args;
12294 +struct au_hinode;
12295 +#ifdef CONFIG_AUFS_DIRREN
12296 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t h_ino,
12297 +                       struct au_dr_hino *add_ent);
12298 +void au_dr_hino_free(struct au_dr_br *dr);
12299 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
12300 +                 const struct path *path);
12301 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br);
12302 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
12303 +                struct qstr *dst_name, void *_rev);
12304 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *rev);
12305 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t bindex, void *rev);
12306 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12307 +              aufs_bindex_t bindex);
12308 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12309 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12310 +                    ino_t h_ino);
12311 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup);
12312 +int au_dr_opt_set(struct super_block *sb);
12313 +int au_dr_opt_flush(struct super_block *sb);
12314 +int au_dr_opt_clr(struct super_block *sb, int no_flush);
12315 +#else
12316 +AuStubInt0(au_dr_hino_test_add, struct au_dr_br *dr, ino_t h_ino,
12317 +          struct au_dr_hino *add_ent);
12318 +AuStubVoid(au_dr_hino_free, struct au_dr_br *dr);
12319 +AuStubInt0(au_dr_br_init, struct super_block *sb, struct au_branch *br,
12320 +          const struct path *path);
12321 +AuStubInt0(au_dr_br_fin, struct super_block *sb, struct au_branch *br);
12322 +AuStubInt0(au_dr_rename, struct dentry *src, aufs_bindex_t bindex,
12323 +          struct qstr *dst_name, void *_rev);
12324 +AuStubVoid(au_dr_rename_fin, struct dentry *src, aufs_bindex_t btgt, void *rev);
12325 +AuStubVoid(au_dr_rename_rev, struct dentry *src, aufs_bindex_t bindex,
12326 +          void *rev);
12327 +AuStubInt0(au_dr_lkup, struct au_do_lookup_args *lkup, struct dentry *dentry,
12328 +          aufs_bindex_t bindex);
12329 +AuStubInt0(au_dr_lkup_name, struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12330 +AuStubInt0(au_dr_lkup_h_ino, struct au_do_lookup_args *lkup,
12331 +          aufs_bindex_t bindex, ino_t h_ino);
12332 +AuStubVoid(au_dr_lkup_fin, struct au_do_lookup_args *lkup);
12333 +AuStubInt0(au_dr_opt_set, struct super_block *sb);
12334 +AuStubInt0(au_dr_opt_flush, struct super_block *sb);
12335 +AuStubInt0(au_dr_opt_clr, struct super_block *sb, int no_flush);
12336 +#endif
12337 +
12338 +/* ---------------------------------------------------------------------- */
12339 +
12340 +#ifdef CONFIG_AUFS_DIRREN
12341 +static inline int au_dr_ihash(ino_t h_ino)
12342 +{
12343 +       return h_ino % AuDirren_NHASH;
12344 +}
12345 +#else
12346 +AuStubInt0(au_dr_ihash, ino_t h_ino);
12347 +#endif
12348 +
12349 +#endif /* __KERNEL__ */
12350 +#endif /* __AUFS_DIRREN_H__ */
12351 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
12352 --- /usr/share/empty/fs/aufs/dynop.c    1970-01-01 01:00:00.000000000 +0100
12353 +++ linux/fs/aufs/dynop.c       2022-11-05 23:02:18.962555950 +0100
12354 @@ -0,0 +1,366 @@
12355 +// SPDX-License-Identifier: GPL-2.0
12356 +/*
12357 + * Copyright (C) 2010-2022 Junjiro R. Okajima
12358 + *
12359 + * This program is free software; you can redistribute it and/or modify
12360 + * it under the terms of the GNU General Public License as published by
12361 + * the Free Software Foundation; either version 2 of the License, or
12362 + * (at your option) any later version.
12363 + *
12364 + * This program is distributed in the hope that it will be useful,
12365 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12366 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12367 + * GNU General Public License for more details.
12368 + *
12369 + * You should have received a copy of the GNU General Public License
12370 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12371 + */
12372 +
12373 +/*
12374 + * dynamically customizable operations for regular files
12375 + */
12376 +
12377 +#include "aufs.h"
12378 +
12379 +#define DyPrSym(key)   AuDbgSym(key->dk_op.dy_hop)
12380 +
12381 +/*
12382 + * How large will these lists be?
12383 + * Usually just a few elements, 20-30 at most for each, I guess.
12384 + */
12385 +static struct hlist_bl_head dynop[AuDyLast];
12386 +
12387 +static struct au_dykey *dy_gfind_get(struct hlist_bl_head *hbl,
12388 +                                    const void *h_op)
12389 +{
12390 +       struct au_dykey *key, *tmp;
12391 +       struct hlist_bl_node *pos;
12392 +
12393 +       key = NULL;
12394 +       hlist_bl_lock(hbl);
12395 +       hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12396 +               if (tmp->dk_op.dy_hop == h_op) {
12397 +                       if (kref_get_unless_zero(&tmp->dk_kref))
12398 +                               key = tmp;
12399 +                       break;
12400 +               }
12401 +       hlist_bl_unlock(hbl);
12402 +
12403 +       return key;
12404 +}
12405 +
12406 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
12407 +{
12408 +       struct au_dykey **k, *found;
12409 +       const void *h_op = key->dk_op.dy_hop;
12410 +       int i;
12411 +
12412 +       found = NULL;
12413 +       k = br->br_dykey;
12414 +       for (i = 0; i < AuBrDynOp; i++)
12415 +               if (k[i]) {
12416 +                       if (k[i]->dk_op.dy_hop == h_op) {
12417 +                               found = k[i];
12418 +                               break;
12419 +                       }
12420 +               } else
12421 +                       break;
12422 +       if (!found) {
12423 +               spin_lock(&br->br_dykey_lock);
12424 +               for (; i < AuBrDynOp; i++)
12425 +                       if (k[i]) {
12426 +                               if (k[i]->dk_op.dy_hop == h_op) {
12427 +                                       found = k[i];
12428 +                                       break;
12429 +                               }
12430 +                       } else {
12431 +                               k[i] = key;
12432 +                               break;
12433 +                       }
12434 +               spin_unlock(&br->br_dykey_lock);
12435 +               BUG_ON(i == AuBrDynOp); /* expand the array */
12436 +       }
12437 +
12438 +       return found;
12439 +}
12440 +
12441 +/* kref_get() if @key is already added */
12442 +static struct au_dykey *dy_gadd(struct hlist_bl_head *hbl, struct au_dykey *key)
12443 +{
12444 +       struct au_dykey *tmp, *found;
12445 +       struct hlist_bl_node *pos;
12446 +       const void *h_op = key->dk_op.dy_hop;
12447 +
12448 +       found = NULL;
12449 +       hlist_bl_lock(hbl);
12450 +       hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12451 +               if (tmp->dk_op.dy_hop == h_op) {
12452 +                       if (kref_get_unless_zero(&tmp->dk_kref))
12453 +                               found = tmp;
12454 +                       break;
12455 +               }
12456 +       if (!found)
12457 +               hlist_bl_add_head(&key->dk_hnode, hbl);
12458 +       hlist_bl_unlock(hbl);
12459 +
12460 +       if (!found)
12461 +               DyPrSym(key);
12462 +       return found;
12463 +}
12464 +
12465 +static void dy_free_rcu(struct rcu_head *rcu)
12466 +{
12467 +       struct au_dykey *key;
12468 +
12469 +       key = container_of(rcu, struct au_dykey, dk_rcu);
12470 +       DyPrSym(key);
12471 +       kfree(key);
12472 +}
12473 +
12474 +static void dy_free(struct kref *kref)
12475 +{
12476 +       struct au_dykey *key;
12477 +       struct hlist_bl_head *hbl;
12478 +
12479 +       key = container_of(kref, struct au_dykey, dk_kref);
12480 +       hbl = dynop + key->dk_op.dy_type;
12481 +       au_hbl_del(&key->dk_hnode, hbl);
12482 +       call_rcu(&key->dk_rcu, dy_free_rcu);
12483 +}
12484 +
12485 +void au_dy_put(struct au_dykey *key)
12486 +{
12487 +       kref_put(&key->dk_kref, dy_free);
12488 +}
12489 +
12490 +/* ---------------------------------------------------------------------- */
12491 +
12492 +#define DyDbgSize(cnt, op)     AuDebugOn(cnt != sizeof(op)/sizeof(void *))
12493 +
12494 +#ifdef CONFIG_AUFS_DEBUG
12495 +#define DyDbgDeclare(cnt)      unsigned int cnt = 0
12496 +#define DyDbgInc(cnt)          do { cnt++; } while (0)
12497 +#else
12498 +#define DyDbgDeclare(cnt)      do {} while (0)
12499 +#define DyDbgInc(cnt)          do {} while (0)
12500 +#endif
12501 +
12502 +#define DySet(func, dst, src, h_op, h_sb) do {                         \
12503 +       DyDbgInc(cnt);                                                  \
12504 +       if (h_op->func) {                                               \
12505 +               if (src.func)                                           \
12506 +                       dst.func = src.func;                            \
12507 +               else                                                    \
12508 +                       AuDbg("%s %s\n", au_sbtype(h_sb), #func);       \
12509 +       }                                                               \
12510 +} while (0)
12511 +
12512 +#define DySetForce(func, dst, src) do {                \
12513 +       AuDebugOn(!src.func);                   \
12514 +       DyDbgInc(cnt);                          \
12515 +       dst.func = src.func;                    \
12516 +} while (0)
12517 +
12518 +#define DySetAop(func) \
12519 +       DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
12520 +#define DySetAopForce(func) \
12521 +       DySetForce(func, dyaop->da_op, aufs_aop)
12522 +
12523 +static void dy_aop(struct au_dykey *key, const void *h_op,
12524 +                  struct super_block *h_sb __maybe_unused)
12525 +{
12526 +       struct au_dyaop *dyaop = (void *)key;
12527 +       const struct address_space_operations *h_aop = h_op;
12528 +       DyDbgDeclare(cnt);
12529 +
12530 +       AuDbg("%s\n", au_sbtype(h_sb));
12531 +
12532 +       DySetAop(writepage);
12533 +       DySetAopForce(read_folio);      /* force */
12534 +       DySetAop(writepages);
12535 +       DySetAop(dirty_folio);
12536 +       DySetAop(invalidate_folio);
12537 +       DySetAop(readahead);
12538 +       DySetAop(write_begin);
12539 +       DySetAop(write_end);
12540 +       DySetAop(bmap);
12541 +       DySetAop(release_folio);
12542 +       DySetAop(free_folio);
12543 +       /* this one will be changed according to an aufs mount option */
12544 +       DySetAop(direct_IO);
12545 +       DySetAop(migrate_folio);
12546 +       DySetAop(launder_folio);
12547 +       DySetAop(is_partially_uptodate);
12548 +       DySetAop(is_dirty_writeback);
12549 +       DySetAop(error_remove_page);
12550 +       DySetAop(swap_activate);
12551 +       DySetAop(swap_deactivate);
12552 +       DySetAop(swap_rw);
12553 +
12554 +       DyDbgSize(cnt, *h_aop);
12555 +}
12556 +
12557 +/* ---------------------------------------------------------------------- */
12558 +
12559 +static void dy_bug(struct kref *kref)
12560 +{
12561 +       BUG();
12562 +}
12563 +
12564 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
12565 +{
12566 +       struct au_dykey *key, *old;
12567 +       struct hlist_bl_head *hbl;
12568 +       struct op {
12569 +               unsigned int sz;
12570 +               void (*set)(struct au_dykey *key, const void *h_op,
12571 +                           struct super_block *h_sb __maybe_unused);
12572 +       };
12573 +       static const struct op a[] = {
12574 +               [AuDy_AOP] = {
12575 +                       .sz     = sizeof(struct au_dyaop),
12576 +                       .set    = dy_aop
12577 +               }
12578 +       };
12579 +       const struct op *p;
12580 +
12581 +       hbl = dynop + op->dy_type;
12582 +       key = dy_gfind_get(hbl, op->dy_hop);
12583 +       if (key)
12584 +               goto out_add; /* success */
12585 +
12586 +       p = a + op->dy_type;
12587 +       key = kzalloc(p->sz, GFP_NOFS);
12588 +       if (unlikely(!key)) {
12589 +               key = ERR_PTR(-ENOMEM);
12590 +               goto out;
12591 +       }
12592 +
12593 +       key->dk_op.dy_hop = op->dy_hop;
12594 +       kref_init(&key->dk_kref);
12595 +       p->set(key, op->dy_hop, au_br_sb(br));
12596 +       old = dy_gadd(hbl, key);
12597 +       if (old) {
12598 +               au_kfree_rcu(key);
12599 +               key = old;
12600 +       }
12601 +
12602 +out_add:
12603 +       old = dy_bradd(br, key);
12604 +       if (old)
12605 +               /* its ref-count should never be zero here */
12606 +               kref_put(&key->dk_kref, dy_bug);
12607 +out:
12608 +       return key;
12609 +}
12610 +
12611 +/* ---------------------------------------------------------------------- */
12612 +/*
12613 + * Aufs prohibits O_DIRECT by default even if the branch supports it.
12614 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
12615 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
12616 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
12617 + * See the aufs manual in detail.
12618 + */
12619 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
12620 +{
12621 +       if (!do_dx)
12622 +               dyaop->da_op.direct_IO = NULL;
12623 +       else
12624 +               dyaop->da_op.direct_IO = aufs_aop.direct_IO;
12625 +}
12626 +
12627 +static struct au_dyaop *dy_aget(struct au_branch *br,
12628 +                               const struct address_space_operations *h_aop,
12629 +                               int do_dx)
12630 +{
12631 +       struct au_dyaop *dyaop;
12632 +       struct au_dynop op;
12633 +
12634 +       op.dy_type = AuDy_AOP;
12635 +       op.dy_haop = h_aop;
12636 +       dyaop = (void *)dy_get(&op, br);
12637 +       if (IS_ERR(dyaop))
12638 +               goto out;
12639 +       dy_adx(dyaop, do_dx);
12640 +
12641 +out:
12642 +       return dyaop;
12643 +}
12644 +
12645 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12646 +               struct inode *h_inode)
12647 +{
12648 +       int err, do_dx;
12649 +       struct super_block *sb;
12650 +       struct au_branch *br;
12651 +       struct au_dyaop *dyaop;
12652 +
12653 +       AuDebugOn(!S_ISREG(h_inode->i_mode));
12654 +       IiMustWriteLock(inode);
12655 +
12656 +       sb = inode->i_sb;
12657 +       br = au_sbr(sb, bindex);
12658 +       do_dx = !!au_opt_test(au_mntflags(sb), DIO);
12659 +       dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
12660 +       err = PTR_ERR(dyaop);
12661 +       if (IS_ERR(dyaop))
12662 +               /* unnecessary to call dy_fput() */
12663 +               goto out;
12664 +
12665 +       err = 0;
12666 +       inode->i_mapping->a_ops = &dyaop->da_op;
12667 +
12668 +out:
12669 +       return err;
12670 +}
12671 +
12672 +/*
12673 + * Is it safe to replace a_ops during the inode/file is in operation?
12674 + * Yes, I hope so.
12675 + */
12676 +int au_dy_irefresh(struct inode *inode)
12677 +{
12678 +       int err;
12679 +       aufs_bindex_t btop;
12680 +       struct inode *h_inode;
12681 +
12682 +       err = 0;
12683 +       if (S_ISREG(inode->i_mode)) {
12684 +               btop = au_ibtop(inode);
12685 +               h_inode = au_h_iptr(inode, btop);
12686 +               err = au_dy_iaop(inode, btop, h_inode);
12687 +       }
12688 +       return err;
12689 +}
12690 +
12691 +void au_dy_arefresh(int do_dx)
12692 +{
12693 +       struct hlist_bl_head *hbl;
12694 +       struct hlist_bl_node *pos;
12695 +       struct au_dykey *key;
12696 +
12697 +       hbl = dynop + AuDy_AOP;
12698 +       hlist_bl_lock(hbl);
12699 +       hlist_bl_for_each_entry(key, pos, hbl, dk_hnode)
12700 +               dy_adx((void *)key, do_dx);
12701 +       hlist_bl_unlock(hbl);
12702 +}
12703 +
12704 +/* ---------------------------------------------------------------------- */
12705 +
12706 +void __init au_dy_init(void)
12707 +{
12708 +       int i;
12709 +
12710 +       for (i = 0; i < AuDyLast; i++)
12711 +               INIT_HLIST_BL_HEAD(dynop + i);
12712 +}
12713 +
12714 +void au_dy_fin(void)
12715 +{
12716 +       int i;
12717 +
12718 +       for (i = 0; i < AuDyLast; i++)
12719 +               WARN_ON(!hlist_bl_empty(dynop + i));
12720 +}
12721 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
12722 --- /usr/share/empty/fs/aufs/dynop.h    1970-01-01 01:00:00.000000000 +0100
12723 +++ linux/fs/aufs/dynop.h       2022-11-05 23:02:18.962555950 +0100
12724 @@ -0,0 +1,77 @@
12725 +/* SPDX-License-Identifier: GPL-2.0 */
12726 +/*
12727 + * Copyright (C) 2010-2022 Junjiro R. Okajima
12728 + *
12729 + * This program is free software; you can redistribute it and/or modify
12730 + * it under the terms of the GNU General Public License as published by
12731 + * the Free Software Foundation; either version 2 of the License, or
12732 + * (at your option) any later version.
12733 + *
12734 + * This program is distributed in the hope that it will be useful,
12735 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12736 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12737 + * GNU General Public License for more details.
12738 + *
12739 + * You should have received a copy of the GNU General Public License
12740 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12741 + */
12742 +
12743 +/*
12744 + * dynamically customizable operations (for regular files only)
12745 + */
12746 +
12747 +#ifndef __AUFS_DYNOP_H__
12748 +#define __AUFS_DYNOP_H__
12749 +
12750 +#ifdef __KERNEL__
12751 +
12752 +#include <linux/fs.h>
12753 +#include <linux/kref.h>
12754 +
12755 +enum {AuDy_AOP, AuDyLast};
12756 +
12757 +struct au_dynop {
12758 +       int                                             dy_type;
12759 +       union {
12760 +               const void                              *dy_hop;
12761 +               const struct address_space_operations   *dy_haop;
12762 +       };
12763 +};
12764 +
12765 +struct au_dykey {
12766 +       union {
12767 +               struct hlist_bl_node    dk_hnode;
12768 +               struct rcu_head         dk_rcu;
12769 +       };
12770 +       struct au_dynop         dk_op;
12771 +
12772 +       /*
12773 +        * during I am in the branch local array, kref is gotten. when the
12774 +        * branch is removed, kref is put.
12775 +        */
12776 +       struct kref             dk_kref;
12777 +};
12778 +
12779 +/* stop unioning since their sizes are very different from each other */
12780 +struct au_dyaop {
12781 +       struct au_dykey                 da_key;
12782 +       struct address_space_operations da_op; /* not const */
12783 +};
12784 +/* make sure that 'struct au_dykey *' can be any type */
12785 +static_assert(!offsetof(struct au_dyaop, da_key));
12786 +
12787 +/* ---------------------------------------------------------------------- */
12788 +
12789 +/* dynop.c */
12790 +struct au_branch;
12791 +void au_dy_put(struct au_dykey *key);
12792 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12793 +               struct inode *h_inode);
12794 +int au_dy_irefresh(struct inode *inode);
12795 +void au_dy_arefresh(int do_dio);
12796 +
12797 +void __init au_dy_init(void);
12798 +void au_dy_fin(void);
12799 +
12800 +#endif /* __KERNEL__ */
12801 +#endif /* __AUFS_DYNOP_H__ */
12802 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
12803 --- /usr/share/empty/fs/aufs/export.c   1970-01-01 01:00:00.000000000 +0100
12804 +++ linux/fs/aufs/export.c      2022-12-17 09:21:34.796521861 +0100
12805 @@ -0,0 +1,830 @@
12806 +// SPDX-License-Identifier: GPL-2.0
12807 +/*
12808 + * Copyright (C) 2005-2022 Junjiro R. Okajima
12809 + *
12810 + * This program is free software; you can redistribute it and/or modify
12811 + * it under the terms of the GNU General Public License as published by
12812 + * the Free Software Foundation; either version 2 of the License, or
12813 + * (at your option) any later version.
12814 + *
12815 + * This program is distributed in the hope that it will be useful,
12816 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12817 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12818 + * GNU General Public License for more details.
12819 + *
12820 + * You should have received a copy of the GNU General Public License
12821 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12822 + */
12823 +
12824 +/*
12825 + * export via nfs
12826 + */
12827 +
12828 +#include <linux/exportfs.h>
12829 +#include <linux/fs_struct.h>
12830 +#include <linux/nsproxy.h>
12831 +#include <linux/random.h>
12832 +#include <linux/writeback.h>
12833 +#include "aufs.h"
12834 +
12835 +union conv {
12836 +#ifdef CONFIG_AUFS_INO_T_64
12837 +       __u32 a[2];
12838 +#else
12839 +       __u32 a[1];
12840 +#endif
12841 +       ino_t ino;
12842 +};
12843 +
12844 +static ino_t decode_ino(__u32 *a)
12845 +{
12846 +       union conv u;
12847 +
12848 +       BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
12849 +       u.a[0] = a[0];
12850 +#ifdef CONFIG_AUFS_INO_T_64
12851 +       u.a[1] = a[1];
12852 +#endif
12853 +       return u.ino;
12854 +}
12855 +
12856 +static void encode_ino(__u32 *a, ino_t ino)
12857 +{
12858 +       union conv u;
12859 +
12860 +       u.ino = ino;
12861 +       a[0] = u.a[0];
12862 +#ifdef CONFIG_AUFS_INO_T_64
12863 +       a[1] = u.a[1];
12864 +#endif
12865 +}
12866 +
12867 +/* NFS file handle */
12868 +enum {
12869 +       Fh_br_id,
12870 +       Fh_sigen,
12871 +#ifdef CONFIG_AUFS_INO_T_64
12872 +       /* support 64bit inode number */
12873 +       Fh_ino1,
12874 +       Fh_ino2,
12875 +       Fh_dir_ino1,
12876 +       Fh_dir_ino2,
12877 +#else
12878 +       Fh_ino1,
12879 +       Fh_dir_ino1,
12880 +#endif
12881 +       Fh_igen,
12882 +       Fh_h_type,
12883 +       Fh_tail,
12884 +
12885 +       Fh_ino = Fh_ino1,
12886 +       Fh_dir_ino = Fh_dir_ino1
12887 +};
12888 +
12889 +static int au_test_anon(struct dentry *dentry)
12890 +{
12891 +       /* note: read d_flags without d_lock */
12892 +       return !!(dentry->d_flags & DCACHE_DISCONNECTED);
12893 +}
12894 +
12895 +int au_test_nfsd(void)
12896 +{
12897 +       int ret;
12898 +       struct task_struct *tsk = current;
12899 +       char comm[sizeof(tsk->comm)];
12900 +
12901 +       ret = 0;
12902 +       if (tsk->flags & PF_KTHREAD) {
12903 +               get_task_comm(comm, tsk);
12904 +               ret = !strcmp(comm, "nfsd");
12905 +       }
12906 +
12907 +       return ret;
12908 +}
12909 +
12910 +/* ---------------------------------------------------------------------- */
12911 +/* inode generation external table */
12912 +
12913 +void au_xigen_inc(struct inode *inode)
12914 +{
12915 +       loff_t pos;
12916 +       ssize_t sz;
12917 +       __u32 igen;
12918 +       struct super_block *sb;
12919 +       struct au_sbinfo *sbinfo;
12920 +
12921 +       sb = inode->i_sb;
12922 +       AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
12923 +
12924 +       sbinfo = au_sbi(sb);
12925 +       pos = inode->i_ino;
12926 +       pos *= sizeof(igen);
12927 +       igen = inode->i_generation + 1;
12928 +       sz = xino_fwrite(sbinfo->si_xigen, &igen, sizeof(igen), &pos);
12929 +       if (sz == sizeof(igen))
12930 +               return; /* success */
12931 +
12932 +       if (unlikely(sz >= 0))
12933 +               AuIOErr("xigen error (%zd)\n", sz);
12934 +}
12935 +
12936 +int au_xigen_new(struct inode *inode)
12937 +{
12938 +       int err;
12939 +       loff_t pos;
12940 +       ssize_t sz;
12941 +       struct super_block *sb;
12942 +       struct au_sbinfo *sbinfo;
12943 +       struct file *file;
12944 +
12945 +       err = 0;
12946 +       /* todo: dirty, at mount time */
12947 +       if (inode->i_ino == AUFS_ROOT_INO)
12948 +               goto out;
12949 +       sb = inode->i_sb;
12950 +       SiMustAnyLock(sb);
12951 +       if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
12952 +               goto out;
12953 +
12954 +       err = -EFBIG;
12955 +       pos = inode->i_ino;
12956 +       if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
12957 +               AuIOErr1("too large i%lld\n", pos);
12958 +               goto out;
12959 +       }
12960 +       pos *= sizeof(inode->i_generation);
12961 +
12962 +       err = 0;
12963 +       sbinfo = au_sbi(sb);
12964 +       file = sbinfo->si_xigen;
12965 +       BUG_ON(!file);
12966 +
12967 +       if (vfsub_f_size_read(file)
12968 +           < pos + sizeof(inode->i_generation)) {
12969 +               inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
12970 +               sz = xino_fwrite(file, &inode->i_generation,
12971 +                                sizeof(inode->i_generation), &pos);
12972 +       } else
12973 +               sz = xino_fread(file, &inode->i_generation,
12974 +                               sizeof(inode->i_generation), &pos);
12975 +       if (sz == sizeof(inode->i_generation))
12976 +               goto out; /* success */
12977 +
12978 +       err = sz;
12979 +       if (unlikely(sz >= 0)) {
12980 +               err = -EIO;
12981 +               AuIOErr("xigen error (%zd)\n", sz);
12982 +       }
12983 +
12984 +out:
12985 +       return err;
12986 +}
12987 +
12988 +int au_xigen_set(struct super_block *sb, struct path *path)
12989 +{
12990 +       int err;
12991 +       struct au_sbinfo *sbinfo;
12992 +       struct file *file;
12993 +
12994 +       SiMustWriteLock(sb);
12995 +
12996 +       sbinfo = au_sbi(sb);
12997 +       file = au_xino_create2(sb, path, sbinfo->si_xigen);
12998 +       err = PTR_ERR(file);
12999 +       if (IS_ERR(file))
13000 +               goto out;
13001 +       err = 0;
13002 +       if (sbinfo->si_xigen)
13003 +               fput(sbinfo->si_xigen);
13004 +       sbinfo->si_xigen = file;
13005 +
13006 +out:
13007 +       AuTraceErr(err);
13008 +       return err;
13009 +}
13010 +
13011 +void au_xigen_clr(struct super_block *sb)
13012 +{
13013 +       struct au_sbinfo *sbinfo;
13014 +
13015 +       SiMustWriteLock(sb);
13016 +
13017 +       sbinfo = au_sbi(sb);
13018 +       if (sbinfo->si_xigen) {
13019 +               fput(sbinfo->si_xigen);
13020 +               sbinfo->si_xigen = NULL;
13021 +       }
13022 +}
13023 +
13024 +/* ---------------------------------------------------------------------- */
13025 +
13026 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
13027 +                                   ino_t dir_ino)
13028 +{
13029 +       struct dentry *dentry, *d;
13030 +       struct inode *inode;
13031 +       unsigned int sigen;
13032 +
13033 +       dentry = NULL;
13034 +       inode = ilookup(sb, ino);
13035 +       if (!inode)
13036 +               goto out;
13037 +
13038 +       dentry = ERR_PTR(-ESTALE);
13039 +       sigen = au_sigen(sb);
13040 +       if (unlikely(au_is_bad_inode(inode)
13041 +                    || IS_DEADDIR(inode)
13042 +                    || sigen != au_iigen(inode, NULL)))
13043 +               goto out_iput;
13044 +
13045 +       dentry = NULL;
13046 +       if (!dir_ino || S_ISDIR(inode->i_mode))
13047 +               dentry = d_find_alias(inode);
13048 +       else {
13049 +               spin_lock(&inode->i_lock);
13050 +               hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
13051 +                       spin_lock(&d->d_lock);
13052 +                       if (!au_test_anon(d)
13053 +                           && d_inode(d->d_parent)->i_ino == dir_ino) {
13054 +                               dentry = dget_dlock(d);
13055 +                               spin_unlock(&d->d_lock);
13056 +                               break;
13057 +                       }
13058 +                       spin_unlock(&d->d_lock);
13059 +               }
13060 +               spin_unlock(&inode->i_lock);
13061 +       }
13062 +       if (unlikely(dentry && au_digen_test(dentry, sigen))) {
13063 +               /* need to refresh */
13064 +               dput(dentry);
13065 +               dentry = NULL;
13066 +       }
13067 +
13068 +out_iput:
13069 +       iput(inode);
13070 +out:
13071 +       AuTraceErrPtr(dentry);
13072 +       return dentry;
13073 +}
13074 +
13075 +/* ---------------------------------------------------------------------- */
13076 +
13077 +/* todo: dirty? */
13078 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
13079 +
13080 +struct au_compare_mnt_args {
13081 +       /* input */
13082 +       struct super_block *sb;
13083 +
13084 +       /* output */
13085 +       struct vfsmount *mnt;
13086 +};
13087 +
13088 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
13089 +{
13090 +       struct au_compare_mnt_args *a = arg;
13091 +
13092 +       if (mnt->mnt_sb != a->sb)
13093 +               return 0;
13094 +       a->mnt = mntget(mnt);
13095 +       return 1;
13096 +}
13097 +
13098 +static struct vfsmount *au_mnt_get(struct super_block *sb)
13099 +{
13100 +       int err;
13101 +       struct path root;
13102 +       struct au_compare_mnt_args args = {
13103 +               .sb = sb
13104 +       };
13105 +
13106 +       get_fs_root(current->fs, &root);
13107 +       rcu_read_lock();
13108 +       err = iterate_mounts(au_compare_mnt, &args, root.mnt);
13109 +       rcu_read_unlock();
13110 +       path_put(&root);
13111 +       AuDebugOn(!err);
13112 +       AuDebugOn(!args.mnt);
13113 +       return args.mnt;
13114 +}
13115 +
13116 +struct au_nfsd_si_lock {
13117 +       unsigned int sigen;
13118 +       aufs_bindex_t bindex, br_id;
13119 +       unsigned char force_lock;
13120 +};
13121 +
13122 +static int si_nfsd_read_lock(struct super_block *sb,
13123 +                            struct au_nfsd_si_lock *nsi_lock)
13124 +{
13125 +       int err;
13126 +       aufs_bindex_t bindex;
13127 +
13128 +       si_read_lock(sb, AuLock_FLUSH);
13129 +
13130 +       /* branch id may be wrapped around */
13131 +       err = 0;
13132 +       bindex = au_br_index(sb, nsi_lock->br_id);
13133 +       if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
13134 +               goto out; /* success */
13135 +
13136 +       err = -ESTALE;
13137 +       bindex = -1;
13138 +       if (!nsi_lock->force_lock)
13139 +               si_read_unlock(sb);
13140 +
13141 +out:
13142 +       nsi_lock->bindex = bindex;
13143 +       return err;
13144 +}
13145 +
13146 +struct find_name_by_ino {
13147 +       struct dir_context ctx;
13148 +       int called, found;
13149 +       ino_t ino;
13150 +       char *name;
13151 +       int namelen;
13152 +};
13153 +
13154 +static bool
13155 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
13156 +                loff_t offset, u64 ino, unsigned int d_type)
13157 +{
13158 +       struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
13159 +                                                 ctx);
13160 +
13161 +       a->called++;
13162 +       if (a->ino != ino)
13163 +               return true;
13164 +
13165 +       memcpy(a->name, name, namelen);
13166 +       a->namelen = namelen;
13167 +       a->found = 1;
13168 +       return false;
13169 +}
13170 +
13171 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
13172 +                                    struct au_nfsd_si_lock *nsi_lock)
13173 +{
13174 +       struct dentry *dentry, *parent;
13175 +       struct file *file;
13176 +       struct inode *dir;
13177 +       struct find_name_by_ino arg = {
13178 +               .ctx = {
13179 +                       .actor = find_name_by_ino
13180 +               }
13181 +       };
13182 +       int err;
13183 +
13184 +       parent = path->dentry;
13185 +       if (nsi_lock)
13186 +               si_read_unlock(parent->d_sb);
13187 +       file = vfsub_dentry_open(path, au_dir_roflags);
13188 +       dentry = (void *)file;
13189 +       if (IS_ERR(file))
13190 +               goto out;
13191 +
13192 +       dentry = ERR_PTR(-ENOMEM);
13193 +       arg.name = (void *)__get_free_page(GFP_NOFS);
13194 +       if (unlikely(!arg.name))
13195 +               goto out_file;
13196 +       arg.ino = ino;
13197 +       arg.found = 0;
13198 +       do {
13199 +               arg.called = 0;
13200 +               /* smp_mb(); */
13201 +               err = vfsub_iterate_dir(file, &arg.ctx);
13202 +       } while (!err && !arg.found && arg.called);
13203 +       dentry = ERR_PTR(err);
13204 +       if (unlikely(err))
13205 +               goto out_name;
13206 +       /* instead of ENOENT */
13207 +       dentry = ERR_PTR(-ESTALE);
13208 +       if (!arg.found)
13209 +               goto out_name;
13210 +
13211 +       /* do not call vfsub_lkup_one() */
13212 +       dir = d_inode(parent);
13213 +       dentry = vfsub_lookup_one_len_unlocked(arg.name, path, arg.namelen);
13214 +       AuTraceErrPtr(dentry);
13215 +       if (IS_ERR(dentry))
13216 +               goto out_name;
13217 +       AuDebugOn(au_test_anon(dentry));
13218 +       if (unlikely(d_really_is_negative(dentry))) {
13219 +               dput(dentry);
13220 +               dentry = ERR_PTR(-ENOENT);
13221 +       }
13222 +
13223 +out_name:
13224 +       free_page((unsigned long)arg.name);
13225 +out_file:
13226 +       fput(file);
13227 +out:
13228 +       if (unlikely(nsi_lock
13229 +                    && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
13230 +               if (!IS_ERR(dentry)) {
13231 +                       dput(dentry);
13232 +                       dentry = ERR_PTR(-ESTALE);
13233 +               }
13234 +       AuTraceErrPtr(dentry);
13235 +       return dentry;
13236 +}
13237 +
13238 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
13239 +                                       ino_t dir_ino,
13240 +                                       struct au_nfsd_si_lock *nsi_lock)
13241 +{
13242 +       struct dentry *dentry;
13243 +       struct path path;
13244 +
13245 +       if (dir_ino != AUFS_ROOT_INO) {
13246 +               path.dentry = decode_by_ino(sb, dir_ino, 0);
13247 +               dentry = path.dentry;
13248 +               if (!path.dentry || IS_ERR(path.dentry))
13249 +                       goto out;
13250 +               AuDebugOn(au_test_anon(path.dentry));
13251 +       } else
13252 +               path.dentry = dget(sb->s_root);
13253 +
13254 +       path.mnt = au_mnt_get(sb);
13255 +       dentry = au_lkup_by_ino(&path, ino, nsi_lock);
13256 +       path_put(&path);
13257 +
13258 +out:
13259 +       AuTraceErrPtr(dentry);
13260 +       return dentry;
13261 +}
13262 +
13263 +/* ---------------------------------------------------------------------- */
13264 +
13265 +static int h_acceptable(void *expv, struct dentry *dentry)
13266 +{
13267 +       return 1;
13268 +}
13269 +
13270 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
13271 +                          char *buf, int len, struct super_block *sb)
13272 +{
13273 +       char *p;
13274 +       int n;
13275 +       struct path path;
13276 +
13277 +       p = d_path(h_rootpath, buf, len);
13278 +       if (IS_ERR(p))
13279 +               goto out;
13280 +       n = strlen(p);
13281 +
13282 +       path.mnt = h_rootpath->mnt;
13283 +       path.dentry = h_parent;
13284 +       p = d_path(&path, buf, len);
13285 +       if (IS_ERR(p))
13286 +               goto out;
13287 +       if (n != 1)
13288 +               p += n;
13289 +
13290 +       path.mnt = au_mnt_get(sb);
13291 +       path.dentry = sb->s_root;
13292 +       p = d_path(&path, buf, len - strlen(p));
13293 +       mntput(path.mnt);
13294 +       if (IS_ERR(p))
13295 +               goto out;
13296 +       if (n != 1)
13297 +               p[strlen(p)] = '/';
13298 +
13299 +out:
13300 +       AuTraceErrPtr(p);
13301 +       return p;
13302 +}
13303 +
13304 +static
13305 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
13306 +                             int fh_len, struct au_nfsd_si_lock *nsi_lock)
13307 +{
13308 +       struct dentry *dentry, *h_parent, *root;
13309 +       struct super_block *h_sb;
13310 +       char *pathname, *p;
13311 +       struct vfsmount *h_mnt;
13312 +       struct au_branch *br;
13313 +       int err;
13314 +       struct path path;
13315 +
13316 +       br = au_sbr(sb, nsi_lock->bindex);
13317 +       h_mnt = au_br_mnt(br);
13318 +       h_sb = h_mnt->mnt_sb;
13319 +       /* todo: call lower fh_to_dentry()? fh_to_parent()? */
13320 +       lockdep_off();
13321 +       h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
13322 +                                     fh_len - Fh_tail, fh[Fh_h_type],
13323 +                                     h_acceptable, /*context*/NULL);
13324 +       lockdep_on();
13325 +       dentry = h_parent;
13326 +       if (unlikely(!h_parent || IS_ERR(h_parent))) {
13327 +               AuWarn1("%s decode_fh failed, %ld\n",
13328 +                       au_sbtype(h_sb), PTR_ERR(h_parent));
13329 +               goto out;
13330 +       }
13331 +       dentry = NULL;
13332 +       if (unlikely(au_test_anon(h_parent))) {
13333 +               AuWarn1("%s decode_fh returned a disconnected dentry\n",
13334 +                       au_sbtype(h_sb));
13335 +               goto out_h_parent;
13336 +       }
13337 +
13338 +       dentry = ERR_PTR(-ENOMEM);
13339 +       pathname = (void *)__get_free_page(GFP_NOFS);
13340 +       if (unlikely(!pathname))
13341 +               goto out_h_parent;
13342 +
13343 +       root = sb->s_root;
13344 +       path.mnt = h_mnt;
13345 +       di_read_lock_parent(root, !AuLock_IR);
13346 +       path.dentry = au_h_dptr(root, nsi_lock->bindex);
13347 +       di_read_unlock(root, !AuLock_IR);
13348 +       p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
13349 +       dentry = (void *)p;
13350 +       if (IS_ERR(p))
13351 +               goto out_pathname;
13352 +
13353 +       si_read_unlock(sb);
13354 +       err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
13355 +       dentry = ERR_PTR(err);
13356 +       if (unlikely(err))
13357 +               goto out_relock;
13358 +
13359 +       dentry = ERR_PTR(-ENOENT);
13360 +       AuDebugOn(au_test_anon(path.dentry));
13361 +       if (unlikely(d_really_is_negative(path.dentry)))
13362 +               goto out_path;
13363 +
13364 +       if (ino != d_inode(path.dentry)->i_ino)
13365 +               dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
13366 +       else
13367 +               dentry = dget(path.dentry);
13368 +
13369 +out_path:
13370 +       path_put(&path);
13371 +out_relock:
13372 +       if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
13373 +               if (!IS_ERR(dentry)) {
13374 +                       dput(dentry);
13375 +                       dentry = ERR_PTR(-ESTALE);
13376 +               }
13377 +out_pathname:
13378 +       free_page((unsigned long)pathname);
13379 +out_h_parent:
13380 +       dput(h_parent);
13381 +out:
13382 +       AuTraceErrPtr(dentry);
13383 +       return dentry;
13384 +}
13385 +
13386 +/* ---------------------------------------------------------------------- */
13387 +
13388 +static struct dentry *
13389 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
13390 +                 int fh_type)
13391 +{
13392 +       struct dentry *dentry;
13393 +       __u32 *fh = fid->raw;
13394 +       struct au_branch *br;
13395 +       ino_t ino, dir_ino;
13396 +       struct au_nfsd_si_lock nsi_lock = {
13397 +               .force_lock     = 0
13398 +       };
13399 +
13400 +       dentry = ERR_PTR(-ESTALE);
13401 +       /* it should never happen, but the file handle is unreliable */
13402 +       if (unlikely(fh_len < Fh_tail))
13403 +               goto out;
13404 +       nsi_lock.sigen = fh[Fh_sigen];
13405 +       nsi_lock.br_id = fh[Fh_br_id];
13406 +
13407 +       /* branch id may be wrapped around */
13408 +       br = NULL;
13409 +       if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
13410 +               goto out;
13411 +       nsi_lock.force_lock = 1;
13412 +
13413 +       /* is this inode still cached? */
13414 +       ino = decode_ino(fh + Fh_ino);
13415 +       /* it should never happen */
13416 +       if (unlikely(ino == AUFS_ROOT_INO))
13417 +               goto out_unlock;
13418 +
13419 +       dir_ino = decode_ino(fh + Fh_dir_ino);
13420 +       dentry = decode_by_ino(sb, ino, dir_ino);
13421 +       if (IS_ERR(dentry))
13422 +               goto out_unlock;
13423 +       if (dentry)
13424 +               goto accept;
13425 +
13426 +       /* is the parent dir cached? */
13427 +       br = au_sbr(sb, nsi_lock.bindex);
13428 +       au_lcnt_inc(&br->br_nfiles);
13429 +       dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
13430 +       if (IS_ERR(dentry))
13431 +               goto out_unlock;
13432 +       if (dentry)
13433 +               goto accept;
13434 +
13435 +       /* lookup path */
13436 +       dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
13437 +       if (IS_ERR(dentry))
13438 +               goto out_unlock;
13439 +       if (unlikely(!dentry))
13440 +               /* todo?: make it ESTALE */
13441 +               goto out_unlock;
13442 +
13443 +accept:
13444 +       if (!au_digen_test(dentry, au_sigen(sb))
13445 +           && d_inode(dentry)->i_generation == fh[Fh_igen])
13446 +               goto out_unlock; /* success */
13447 +
13448 +       dput(dentry);
13449 +       dentry = ERR_PTR(-ESTALE);
13450 +out_unlock:
13451 +       if (br)
13452 +               au_lcnt_dec(&br->br_nfiles);
13453 +       si_read_unlock(sb);
13454 +out:
13455 +       AuTraceErrPtr(dentry);
13456 +       return dentry;
13457 +}
13458 +
13459 +#if 0 /* reserved for future use */
13460 +/* support subtreecheck option */
13461 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
13462 +                                       int fh_len, int fh_type)
13463 +{
13464 +       struct dentry *parent;
13465 +       __u32 *fh = fid->raw;
13466 +       ino_t dir_ino;
13467 +
13468 +       dir_ino = decode_ino(fh + Fh_dir_ino);
13469 +       parent = decode_by_ino(sb, dir_ino, 0);
13470 +       if (IS_ERR(parent))
13471 +               goto out;
13472 +       if (!parent)
13473 +               parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
13474 +                                       dir_ino, fh, fh_len);
13475 +
13476 +out:
13477 +       AuTraceErrPtr(parent);
13478 +       return parent;
13479 +}
13480 +#endif
13481 +
13482 +/* ---------------------------------------------------------------------- */
13483 +
13484 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
13485 +                         struct inode *dir)
13486 +{
13487 +       int err;
13488 +       aufs_bindex_t bindex;
13489 +       struct super_block *sb, *h_sb;
13490 +       struct dentry *dentry, *parent, *h_parent;
13491 +       struct inode *h_dir;
13492 +       struct au_branch *br;
13493 +
13494 +       err = -ENOSPC;
13495 +       if (unlikely(*max_len <= Fh_tail)) {
13496 +               AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
13497 +               goto out;
13498 +       }
13499 +
13500 +       err = FILEID_ROOT;
13501 +       if (inode->i_ino == AUFS_ROOT_INO) {
13502 +               AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
13503 +               goto out;
13504 +       }
13505 +
13506 +       h_parent = NULL;
13507 +       sb = inode->i_sb;
13508 +       err = si_read_lock(sb, AuLock_FLUSH);
13509 +       if (unlikely(err))
13510 +               goto out;
13511 +
13512 +#ifdef CONFIG_AUFS_DEBUG
13513 +       if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
13514 +               AuWarn1("NFS-exporting requires xino\n");
13515 +#endif
13516 +       err = -EIO;
13517 +       parent = NULL;
13518 +       ii_read_lock_child(inode);
13519 +       bindex = au_ibtop(inode);
13520 +       if (!dir) {
13521 +               dentry = d_find_any_alias(inode);
13522 +               if (unlikely(!dentry))
13523 +                       goto out_unlock;
13524 +               AuDebugOn(au_test_anon(dentry));
13525 +               parent = dget_parent(dentry);
13526 +               dput(dentry);
13527 +               if (unlikely(!parent))
13528 +                       goto out_unlock;
13529 +               if (d_really_is_positive(parent))
13530 +                       dir = d_inode(parent);
13531 +       }
13532 +
13533 +       ii_read_lock_parent(dir);
13534 +       h_dir = au_h_iptr(dir, bindex);
13535 +       ii_read_unlock(dir);
13536 +       if (unlikely(!h_dir))
13537 +               goto out_parent;
13538 +       h_parent = d_find_any_alias(h_dir);
13539 +       if (unlikely(!h_parent))
13540 +               goto out_hparent;
13541 +
13542 +       err = -EPERM;
13543 +       br = au_sbr(sb, bindex);
13544 +       h_sb = au_br_sb(br);
13545 +       if (unlikely(!h_sb->s_export_op)) {
13546 +               AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
13547 +               goto out_hparent;
13548 +       }
13549 +
13550 +       fh[Fh_br_id] = br->br_id;
13551 +       fh[Fh_sigen] = au_sigen(sb);
13552 +       encode_ino(fh + Fh_ino, inode->i_ino);
13553 +       encode_ino(fh + Fh_dir_ino, dir->i_ino);
13554 +       fh[Fh_igen] = inode->i_generation;
13555 +
13556 +       *max_len -= Fh_tail;
13557 +       fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
13558 +                                          max_len,
13559 +                                          /*connectable or subtreecheck*/0);
13560 +       err = fh[Fh_h_type];
13561 +       *max_len += Fh_tail;
13562 +       /* todo: macros? */
13563 +       if (err != FILEID_INVALID)
13564 +               err = 99;
13565 +       else
13566 +               AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
13567 +
13568 +out_hparent:
13569 +       dput(h_parent);
13570 +out_parent:
13571 +       dput(parent);
13572 +out_unlock:
13573 +       ii_read_unlock(inode);
13574 +       si_read_unlock(sb);
13575 +out:
13576 +       if (unlikely(err < 0))
13577 +               err = FILEID_INVALID;
13578 +       return err;
13579 +}
13580 +
13581 +/* ---------------------------------------------------------------------- */
13582 +
13583 +static int aufs_commit_metadata(struct inode *inode)
13584 +{
13585 +       int err;
13586 +       aufs_bindex_t bindex;
13587 +       struct super_block *sb;
13588 +       struct inode *h_inode;
13589 +       int (*f)(struct inode *inode);
13590 +
13591 +       sb = inode->i_sb;
13592 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13593 +       ii_write_lock_child(inode);
13594 +       bindex = au_ibtop(inode);
13595 +       AuDebugOn(bindex < 0);
13596 +       h_inode = au_h_iptr(inode, bindex);
13597 +
13598 +       f = h_inode->i_sb->s_export_op->commit_metadata;
13599 +       if (f)
13600 +               err = f(h_inode);
13601 +       else
13602 +               err = sync_inode_metadata(h_inode, /*wait*/1);
13603 +
13604 +       au_cpup_attr_timesizes(inode);
13605 +       ii_write_unlock(inode);
13606 +       si_read_unlock(sb);
13607 +       return err;
13608 +}
13609 +
13610 +/* ---------------------------------------------------------------------- */
13611 +
13612 +static struct export_operations aufs_export_op = {
13613 +       .fh_to_dentry           = aufs_fh_to_dentry,
13614 +       /* .fh_to_parent        = aufs_fh_to_parent, */
13615 +       .encode_fh              = aufs_encode_fh,
13616 +       .commit_metadata        = aufs_commit_metadata
13617 +};
13618 +
13619 +void au_export_init(struct super_block *sb)
13620 +{
13621 +       struct au_sbinfo *sbinfo;
13622 +       __u32 u;
13623 +
13624 +       BUILD_BUG_ON_MSG(IS_BUILTIN(CONFIG_AUFS_FS)
13625 +                        && IS_MODULE(CONFIG_EXPORTFS),
13626 +                        AUFS_NAME ": unsupported configuration "
13627 +                        "CONFIG_EXPORTFS=m and CONFIG_AUFS_FS=y");
13628 +
13629 +       sb->s_export_op = &aufs_export_op;
13630 +       sbinfo = au_sbi(sb);
13631 +       sbinfo->si_xigen = NULL;
13632 +       get_random_bytes(&u, sizeof(u));
13633 +       BUILD_BUG_ON(sizeof(u) != sizeof(int));
13634 +       atomic_set(&sbinfo->si_xigen_next, u);
13635 +}
13636 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
13637 --- /usr/share/empty/fs/aufs/fhsm.c     1970-01-01 01:00:00.000000000 +0100
13638 +++ linux/fs/aufs/fhsm.c        2022-11-05 23:02:18.962555950 +0100
13639 @@ -0,0 +1,426 @@
13640 +// SPDX-License-Identifier: GPL-2.0
13641 +/*
13642 + * Copyright (C) 2011-2022 Junjiro R. Okajima
13643 + *
13644 + * This program is free software; you can redistribute it and/or modify
13645 + * it under the terms of the GNU General Public License as published by
13646 + * the Free Software Foundation; either version 2 of the License, or
13647 + * (at your option) any later version.
13648 + *
13649 + * This program is distributed in the hope that it will be useful,
13650 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13651 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
13652 + * GNU General Public License for more details.
13653 + *
13654 + * You should have received a copy of the GNU General Public License
13655 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
13656 + */
13657 +
13658 +/*
13659 + * File-based Hierarchy Storage Management
13660 + */
13661 +
13662 +#include <linux/anon_inodes.h>
13663 +#include <linux/poll.h>
13664 +#include <linux/seq_file.h>
13665 +#include <linux/statfs.h>
13666 +#include "aufs.h"
13667 +
13668 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
13669 +{
13670 +       struct au_sbinfo *sbinfo;
13671 +       struct au_fhsm *fhsm;
13672 +
13673 +       SiMustAnyLock(sb);
13674 +
13675 +       sbinfo = au_sbi(sb);
13676 +       fhsm = &sbinfo->si_fhsm;
13677 +       AuDebugOn(!fhsm);
13678 +       return fhsm->fhsm_bottom;
13679 +}
13680 +
13681 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
13682 +{
13683 +       struct au_sbinfo *sbinfo;
13684 +       struct au_fhsm *fhsm;
13685 +
13686 +       SiMustWriteLock(sb);
13687 +
13688 +       sbinfo = au_sbi(sb);
13689 +       fhsm = &sbinfo->si_fhsm;
13690 +       AuDebugOn(!fhsm);
13691 +       fhsm->fhsm_bottom = bindex;
13692 +}
13693 +
13694 +/* ---------------------------------------------------------------------- */
13695 +
13696 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
13697 +{
13698 +       struct au_br_fhsm *bf;
13699 +
13700 +       bf = br->br_fhsm;
13701 +       MtxMustLock(&bf->bf_lock);
13702 +
13703 +       return !bf->bf_readable
13704 +               || time_after(jiffies,
13705 +                             bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
13706 +}
13707 +
13708 +/* ---------------------------------------------------------------------- */
13709 +
13710 +static void au_fhsm_notify(struct super_block *sb, int val)
13711 +{
13712 +       struct au_sbinfo *sbinfo;
13713 +       struct au_fhsm *fhsm;
13714 +
13715 +       SiMustAnyLock(sb);
13716 +
13717 +       sbinfo = au_sbi(sb);
13718 +       fhsm = &sbinfo->si_fhsm;
13719 +       if (au_fhsm_pid(fhsm)
13720 +           && atomic_read(&fhsm->fhsm_readable) != -1) {
13721 +               atomic_set(&fhsm->fhsm_readable, val);
13722 +               if (val)
13723 +                       wake_up(&fhsm->fhsm_wqh);
13724 +       }
13725 +}
13726 +
13727 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
13728 +                       struct aufs_stfs *rstfs, int do_lock, int do_notify)
13729 +{
13730 +       int err;
13731 +       struct au_branch *br;
13732 +       struct au_br_fhsm *bf;
13733 +
13734 +       br = au_sbr(sb, bindex);
13735 +       AuDebugOn(au_br_rdonly(br));
13736 +       bf = br->br_fhsm;
13737 +       AuDebugOn(!bf);
13738 +
13739 +       if (do_lock)
13740 +               mutex_lock(&bf->bf_lock);
13741 +       else
13742 +               MtxMustLock(&bf->bf_lock);
13743 +
13744 +       /* sb->s_root for NFS is unreliable */
13745 +       err = au_br_stfs(br, &bf->bf_stfs);
13746 +       if (unlikely(err)) {
13747 +               AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
13748 +               goto out;
13749 +       }
13750 +
13751 +       bf->bf_jiffy = jiffies;
13752 +       bf->bf_readable = 1;
13753 +       if (do_notify)
13754 +               au_fhsm_notify(sb, /*val*/1);
13755 +       if (rstfs)
13756 +               *rstfs = bf->bf_stfs;
13757 +
13758 +out:
13759 +       if (do_lock)
13760 +               mutex_unlock(&bf->bf_lock);
13761 +       au_fhsm_notify(sb, /*val*/1);
13762 +
13763 +       return err;
13764 +}
13765 +
13766 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
13767 +{
13768 +       int err;
13769 +       struct au_sbinfo *sbinfo;
13770 +       struct au_fhsm *fhsm;
13771 +       struct au_branch *br;
13772 +       struct au_br_fhsm *bf;
13773 +
13774 +       AuDbg("b%d, force %d\n", bindex, force);
13775 +       SiMustAnyLock(sb);
13776 +
13777 +       sbinfo = au_sbi(sb);
13778 +       fhsm = &sbinfo->si_fhsm;
13779 +       if (!au_ftest_si(sbinfo, FHSM)
13780 +           || fhsm->fhsm_bottom == bindex)
13781 +               return;
13782 +
13783 +       br = au_sbr(sb, bindex);
13784 +       bf = br->br_fhsm;
13785 +       AuDebugOn(!bf);
13786 +       mutex_lock(&bf->bf_lock);
13787 +       if (force
13788 +           || au_fhsm_pid(fhsm)
13789 +           || au_fhsm_test_jiffy(sbinfo, br))
13790 +               err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
13791 +                                 /*do_notify*/1);
13792 +       mutex_unlock(&bf->bf_lock);
13793 +}
13794 +
13795 +void au_fhsm_wrote_all(struct super_block *sb, int force)
13796 +{
13797 +       aufs_bindex_t bindex, bbot;
13798 +       struct au_branch *br;
13799 +
13800 +       /* exclude the bottom */
13801 +       bbot = au_fhsm_bottom(sb);
13802 +       for (bindex = 0; bindex < bbot; bindex++) {
13803 +               br = au_sbr(sb, bindex);
13804 +               if (au_br_fhsm(br->br_perm))
13805 +                       au_fhsm_wrote(sb, bindex, force);
13806 +       }
13807 +}
13808 +
13809 +/* ---------------------------------------------------------------------- */
13810 +
13811 +static __poll_t au_fhsm_poll(struct file *file, struct poll_table_struct *wait)
13812 +{
13813 +       __poll_t mask;
13814 +       struct au_sbinfo *sbinfo;
13815 +       struct au_fhsm *fhsm;
13816 +
13817 +       mask = 0;
13818 +       sbinfo = file->private_data;
13819 +       fhsm = &sbinfo->si_fhsm;
13820 +       poll_wait(file, &fhsm->fhsm_wqh, wait);
13821 +       if (atomic_read(&fhsm->fhsm_readable))
13822 +               mask = EPOLLIN /* | EPOLLRDNORM */;
13823 +
13824 +       if (!mask)
13825 +               AuDbg("mask 0x%x\n", mask);
13826 +       return mask;
13827 +}
13828 +
13829 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
13830 +                             struct aufs_stfs *stfs, __s16 brid)
13831 +{
13832 +       int err;
13833 +
13834 +       err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
13835 +       if (!err)
13836 +               err = __put_user(brid, &stbr->brid);
13837 +       if (unlikely(err))
13838 +               err = -EFAULT;
13839 +
13840 +       return err;
13841 +}
13842 +
13843 +static ssize_t au_fhsm_do_read(struct super_block *sb,
13844 +                              struct aufs_stbr __user *stbr, size_t count)
13845 +{
13846 +       ssize_t err;
13847 +       int nstbr;
13848 +       aufs_bindex_t bindex, bbot;
13849 +       struct au_branch *br;
13850 +       struct au_br_fhsm *bf;
13851 +
13852 +       /* except the bottom branch */
13853 +       err = 0;
13854 +       nstbr = 0;
13855 +       bbot = au_fhsm_bottom(sb);
13856 +       for (bindex = 0; !err && bindex < bbot; bindex++) {
13857 +               br = au_sbr(sb, bindex);
13858 +               if (!au_br_fhsm(br->br_perm))
13859 +                       continue;
13860 +
13861 +               bf = br->br_fhsm;
13862 +               mutex_lock(&bf->bf_lock);
13863 +               if (bf->bf_readable) {
13864 +                       err = -EFAULT;
13865 +                       if (count >= sizeof(*stbr))
13866 +                               err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
13867 +                                                         br->br_id);
13868 +                       if (!err) {
13869 +                               bf->bf_readable = 0;
13870 +                               count -= sizeof(*stbr);
13871 +                               nstbr++;
13872 +                       }
13873 +               }
13874 +               mutex_unlock(&bf->bf_lock);
13875 +       }
13876 +       if (!err)
13877 +               err = sizeof(*stbr) * nstbr;
13878 +
13879 +       return err;
13880 +}
13881 +
13882 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
13883 +                          loff_t *pos)
13884 +{
13885 +       ssize_t err;
13886 +       int readable;
13887 +       aufs_bindex_t nfhsm, bindex, bbot;
13888 +       struct au_sbinfo *sbinfo;
13889 +       struct au_fhsm *fhsm;
13890 +       struct au_branch *br;
13891 +       struct super_block *sb;
13892 +
13893 +       err = 0;
13894 +       sbinfo = file->private_data;
13895 +       fhsm = &sbinfo->si_fhsm;
13896 +need_data:
13897 +       spin_lock_irq(&fhsm->fhsm_wqh.lock);
13898 +       if (!atomic_read(&fhsm->fhsm_readable)) {
13899 +               if (vfsub_file_flags(file) & O_NONBLOCK)
13900 +                       err = -EAGAIN;
13901 +               else
13902 +                       err = wait_event_interruptible_locked_irq
13903 +                               (fhsm->fhsm_wqh,
13904 +                                atomic_read(&fhsm->fhsm_readable));
13905 +       }
13906 +       spin_unlock_irq(&fhsm->fhsm_wqh.lock);
13907 +       if (unlikely(err))
13908 +               goto out;
13909 +
13910 +       /* sb may already be dead */
13911 +       au_rw_read_lock(&sbinfo->si_rwsem);
13912 +       readable = atomic_read(&fhsm->fhsm_readable);
13913 +       if (readable > 0) {
13914 +               sb = sbinfo->si_sb;
13915 +               AuDebugOn(!sb);
13916 +               /* exclude the bottom branch */
13917 +               nfhsm = 0;
13918 +               bbot = au_fhsm_bottom(sb);
13919 +               for (bindex = 0; bindex < bbot; bindex++) {
13920 +                       br = au_sbr(sb, bindex);
13921 +                       if (au_br_fhsm(br->br_perm))
13922 +                               nfhsm++;
13923 +               }
13924 +               err = -EMSGSIZE;
13925 +               if (nfhsm * sizeof(struct aufs_stbr) <= count) {
13926 +                       atomic_set(&fhsm->fhsm_readable, 0);
13927 +                       err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
13928 +                                            count);
13929 +               }
13930 +       }
13931 +       au_rw_read_unlock(&sbinfo->si_rwsem);
13932 +       if (!readable)
13933 +               goto need_data;
13934 +
13935 +out:
13936 +       return err;
13937 +}
13938 +
13939 +static int au_fhsm_release(struct inode *inode, struct file *file)
13940 +{
13941 +       struct au_sbinfo *sbinfo;
13942 +       struct au_fhsm *fhsm;
13943 +
13944 +       /* sb may already be dead */
13945 +       sbinfo = file->private_data;
13946 +       fhsm = &sbinfo->si_fhsm;
13947 +       spin_lock(&fhsm->fhsm_spin);
13948 +       fhsm->fhsm_pid = 0;
13949 +       spin_unlock(&fhsm->fhsm_spin);
13950 +       kobject_put(&sbinfo->si_kobj);
13951 +
13952 +       return 0;
13953 +}
13954 +
13955 +static const struct file_operations au_fhsm_fops = {
13956 +       .owner          = THIS_MODULE,
13957 +       .llseek         = noop_llseek,
13958 +       .read           = au_fhsm_read,
13959 +       .poll           = au_fhsm_poll,
13960 +       .release        = au_fhsm_release
13961 +};
13962 +
13963 +int au_fhsm_fd(struct super_block *sb, int oflags)
13964 +{
13965 +       int err, fd;
13966 +       struct au_sbinfo *sbinfo;
13967 +       struct au_fhsm *fhsm;
13968 +
13969 +       err = -EPERM;
13970 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
13971 +               goto out;
13972 +
13973 +       err = -EINVAL;
13974 +       if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
13975 +               goto out;
13976 +
13977 +       err = 0;
13978 +       sbinfo = au_sbi(sb);
13979 +       fhsm = &sbinfo->si_fhsm;
13980 +       spin_lock(&fhsm->fhsm_spin);
13981 +       if (!fhsm->fhsm_pid)
13982 +               fhsm->fhsm_pid = current->pid;
13983 +       else
13984 +               err = -EBUSY;
13985 +       spin_unlock(&fhsm->fhsm_spin);
13986 +       if (unlikely(err))
13987 +               goto out;
13988 +
13989 +       oflags |= O_RDONLY;
13990 +       /* oflags |= FMODE_NONOTIFY; */
13991 +       fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
13992 +       err = fd;
13993 +       if (unlikely(fd < 0))
13994 +               goto out_pid;
13995 +
13996 +       /* succeed regardless 'fhsm' status */
13997 +       kobject_get(&sbinfo->si_kobj);
13998 +       si_noflush_read_lock(sb);
13999 +       if (au_ftest_si(sbinfo, FHSM))
14000 +               au_fhsm_wrote_all(sb, /*force*/0);
14001 +       si_read_unlock(sb);
14002 +       goto out; /* success */
14003 +
14004 +out_pid:
14005 +       spin_lock(&fhsm->fhsm_spin);
14006 +       fhsm->fhsm_pid = 0;
14007 +       spin_unlock(&fhsm->fhsm_spin);
14008 +out:
14009 +       AuTraceErr(err);
14010 +       return err;
14011 +}
14012 +
14013 +/* ---------------------------------------------------------------------- */
14014 +
14015 +int au_fhsm_br_alloc(struct au_branch *br)
14016 +{
14017 +       int err;
14018 +
14019 +       err = 0;
14020 +       br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
14021 +       if (br->br_fhsm)
14022 +               au_br_fhsm_init(br->br_fhsm);
14023 +       else
14024 +               err = -ENOMEM;
14025 +
14026 +       return err;
14027 +}
14028 +
14029 +/* ---------------------------------------------------------------------- */
14030 +
14031 +void au_fhsm_fin(struct super_block *sb)
14032 +{
14033 +       au_fhsm_notify(sb, /*val*/-1);
14034 +}
14035 +
14036 +void au_fhsm_init(struct au_sbinfo *sbinfo)
14037 +{
14038 +       struct au_fhsm *fhsm;
14039 +
14040 +       fhsm = &sbinfo->si_fhsm;
14041 +       spin_lock_init(&fhsm->fhsm_spin);
14042 +       init_waitqueue_head(&fhsm->fhsm_wqh);
14043 +       atomic_set(&fhsm->fhsm_readable, 0);
14044 +       fhsm->fhsm_expire
14045 +               = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
14046 +       fhsm->fhsm_bottom = -1;
14047 +}
14048 +
14049 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
14050 +{
14051 +       sbinfo->si_fhsm.fhsm_expire
14052 +               = msecs_to_jiffies(sec * MSEC_PER_SEC);
14053 +}
14054 +
14055 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
14056 +{
14057 +       unsigned int u;
14058 +
14059 +       if (!au_ftest_si(sbinfo, FHSM))
14060 +               return;
14061 +
14062 +       u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
14063 +       if (u != AUFS_FHSM_CACHE_DEF_SEC)
14064 +               seq_printf(seq, ",fhsm_sec=%u", u);
14065 +}
14066 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
14067 --- /usr/share/empty/fs/aufs/file.c     1970-01-01 01:00:00.000000000 +0100
14068 +++ linux/fs/aufs/file.c        2022-12-17 09:21:34.799855195 +0100
14069 @@ -0,0 +1,860 @@
14070 +// SPDX-License-Identifier: GPL-2.0
14071 +/*
14072 + * Copyright (C) 2005-2022 Junjiro R. Okajima
14073 + *
14074 + * This program is free software; you can redistribute it and/or modify
14075 + * it under the terms of the GNU General Public License as published by
14076 + * the Free Software Foundation; either version 2 of the License, or
14077 + * (at your option) any later version.
14078 + *
14079 + * This program is distributed in the hope that it will be useful,
14080 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14081 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14082 + * GNU General Public License for more details.
14083 + *
14084 + * You should have received a copy of the GNU General Public License
14085 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
14086 + */
14087 +
14088 +/*
14089 + * handling file/dir, and address_space operation
14090 + */
14091 +
14092 +#ifdef CONFIG_AUFS_DEBUG
14093 +#include <linux/migrate.h>
14094 +#endif
14095 +#include <linux/pagemap.h>
14096 +#include "aufs.h"
14097 +
14098 +/* drop flags for writing */
14099 +unsigned int au_file_roflags(unsigned int flags)
14100 +{
14101 +       flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
14102 +       flags |= O_RDONLY | O_NOATIME;
14103 +       return flags;
14104 +}
14105 +
14106 +/* common functions to regular file and dir */
14107 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14108 +                      struct file *file, int force_wr)
14109 +{
14110 +       struct file *h_file;
14111 +       struct dentry *h_dentry;
14112 +       struct inode *h_inode;
14113 +       struct super_block *sb;
14114 +       struct au_branch *br;
14115 +       struct path h_path;
14116 +       int err;
14117 +
14118 +       /* a race condition can happen between open and unlink/rmdir */
14119 +       h_file = ERR_PTR(-ENOENT);
14120 +       h_dentry = au_h_dptr(dentry, bindex);
14121 +       if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
14122 +               goto out;
14123 +       h_inode = d_inode(h_dentry);
14124 +       spin_lock(&h_dentry->d_lock);
14125 +       err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
14126 +               /* || !d_inode(dentry)->i_nlink */
14127 +               ;
14128 +       spin_unlock(&h_dentry->d_lock);
14129 +       if (unlikely(err))
14130 +               goto out;
14131 +
14132 +       sb = dentry->d_sb;
14133 +       br = au_sbr(sb, bindex);
14134 +       err = au_br_test_oflag(flags, br);
14135 +       h_file = ERR_PTR(err);
14136 +       if (unlikely(err))
14137 +               goto out;
14138 +
14139 +       /* drop flags for writing */
14140 +       if (au_test_ro(sb, bindex, d_inode(dentry))) {
14141 +               if (force_wr && !(flags & O_WRONLY))
14142 +                       force_wr = 0;
14143 +               flags = au_file_roflags(flags);
14144 +               if (force_wr) {
14145 +                       h_file = ERR_PTR(-EROFS);
14146 +                       flags = au_file_roflags(flags);
14147 +                       if (unlikely(vfsub_native_ro(h_inode)
14148 +                                    || IS_APPEND(h_inode)))
14149 +                               goto out;
14150 +                       flags &= ~O_ACCMODE;
14151 +                       flags |= O_WRONLY;
14152 +               }
14153 +       }
14154 +       flags &= ~O_CREAT;
14155 +       au_lcnt_inc(&br->br_nfiles);
14156 +       h_path.dentry = h_dentry;
14157 +       h_path.mnt = au_br_mnt(br);
14158 +       h_file = vfsub_dentry_open(&h_path, flags);
14159 +       if (IS_ERR(h_file))
14160 +               goto out_br;
14161 +
14162 +       if (flags & __FMODE_EXEC) {
14163 +               err = deny_write_access(h_file);
14164 +               if (unlikely(err)) {
14165 +                       fput(h_file);
14166 +                       h_file = ERR_PTR(err);
14167 +                       goto out_br;
14168 +               }
14169 +       }
14170 +       fsnotify_open(h_file);
14171 +       goto out; /* success */
14172 +
14173 +out_br:
14174 +       au_lcnt_dec(&br->br_nfiles);
14175 +out:
14176 +       return h_file;
14177 +}
14178 +
14179 +static int au_cmoo(struct dentry *dentry)
14180 +{
14181 +       int err, cmoo, matched;
14182 +       unsigned int udba;
14183 +       struct path h_path;
14184 +       struct au_pin pin;
14185 +       struct au_cp_generic cpg = {
14186 +               .dentry = dentry,
14187 +               .bdst   = -1,
14188 +               .bsrc   = -1,
14189 +               .len    = -1,
14190 +               .pin    = &pin,
14191 +               .flags  = AuCpup_DTIME | AuCpup_HOPEN
14192 +       };
14193 +       struct inode *delegated;
14194 +       struct super_block *sb;
14195 +       struct au_sbinfo *sbinfo;
14196 +       struct au_fhsm *fhsm;
14197 +       pid_t pid;
14198 +       struct au_branch *br;
14199 +       struct dentry *parent;
14200 +       struct au_hinode *hdir;
14201 +
14202 +       DiMustWriteLock(dentry);
14203 +       IiMustWriteLock(d_inode(dentry));
14204 +
14205 +       err = 0;
14206 +       if (IS_ROOT(dentry))
14207 +               goto out;
14208 +       cpg.bsrc = au_dbtop(dentry);
14209 +       if (!cpg.bsrc)
14210 +               goto out;
14211 +
14212 +       sb = dentry->d_sb;
14213 +       sbinfo = au_sbi(sb);
14214 +       fhsm = &sbinfo->si_fhsm;
14215 +       pid = au_fhsm_pid(fhsm);
14216 +       rcu_read_lock();
14217 +       matched = (pid
14218 +                  && (current->pid == pid
14219 +                      || rcu_dereference(current->real_parent)->pid == pid));
14220 +       rcu_read_unlock();
14221 +       if (matched)
14222 +               goto out;
14223 +
14224 +       br = au_sbr(sb, cpg.bsrc);
14225 +       cmoo = au_br_cmoo(br->br_perm);
14226 +       if (!cmoo)
14227 +               goto out;
14228 +       if (!d_is_reg(dentry))
14229 +               cmoo &= AuBrAttr_COO_ALL;
14230 +       if (!cmoo)
14231 +               goto out;
14232 +
14233 +       parent = dget_parent(dentry);
14234 +       di_write_lock_parent(parent);
14235 +       err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
14236 +       cpg.bdst = err;
14237 +       if (unlikely(err < 0)) {
14238 +               err = 0;        /* there is no upper writable branch */
14239 +               goto out_dgrade;
14240 +       }
14241 +       AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
14242 +
14243 +       /* do not respect the coo attrib for the target branch */
14244 +       err = au_cpup_dirs(dentry, cpg.bdst);
14245 +       if (unlikely(err))
14246 +               goto out_dgrade;
14247 +
14248 +       di_downgrade_lock(parent, AuLock_IR);
14249 +       udba = au_opt_udba(sb);
14250 +       err = au_pin(&pin, dentry, cpg.bdst, udba,
14251 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14252 +       if (unlikely(err))
14253 +               goto out_parent;
14254 +
14255 +       err = au_sio_cpup_simple(&cpg);
14256 +       au_unpin(&pin);
14257 +       if (unlikely(err))
14258 +               goto out_parent;
14259 +       if (!(cmoo & AuBrWAttr_MOO))
14260 +               goto out_parent; /* success */
14261 +
14262 +       err = au_pin(&pin, dentry, cpg.bsrc, udba,
14263 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14264 +       if (unlikely(err))
14265 +               goto out_parent;
14266 +
14267 +       h_path.mnt = au_br_mnt(br);
14268 +       h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
14269 +       hdir = au_hi(d_inode(parent), cpg.bsrc);
14270 +       delegated = NULL;
14271 +       err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
14272 +       au_unpin(&pin);
14273 +       /* todo: keep h_dentry or not? */
14274 +       if (unlikely(err == -EWOULDBLOCK)) {
14275 +               pr_warn("cannot retry for NFSv4 delegation"
14276 +                       " for an internal unlink\n");
14277 +               iput(delegated);
14278 +       }
14279 +       if (unlikely(err)) {
14280 +               pr_err("unlink %pd after coo failed (%d), ignored\n",
14281 +                      dentry, err);
14282 +               err = 0;
14283 +       }
14284 +       goto out_parent; /* success */
14285 +
14286 +out_dgrade:
14287 +       di_downgrade_lock(parent, AuLock_IR);
14288 +out_parent:
14289 +       di_read_unlock(parent, AuLock_IR);
14290 +       dput(parent);
14291 +out:
14292 +       AuTraceErr(err);
14293 +       return err;
14294 +}
14295 +
14296 +int au_do_open(struct file *file, struct au_do_open_args *args)
14297 +{
14298 +       int err, aopen = args->aopen;
14299 +       struct dentry *dentry;
14300 +       struct au_finfo *finfo;
14301 +
14302 +       if (!aopen)
14303 +               err = au_finfo_init(file, args->fidir);
14304 +       else {
14305 +               lockdep_off();
14306 +               err = au_finfo_init(file, args->fidir);
14307 +               lockdep_on();
14308 +       }
14309 +       if (unlikely(err))
14310 +               goto out;
14311 +
14312 +       dentry = file->f_path.dentry;
14313 +       AuDebugOn(IS_ERR_OR_NULL(dentry));
14314 +       di_write_lock_child(dentry);
14315 +       err = au_cmoo(dentry);
14316 +       if (!err) {
14317 +               if (!aopen) {
14318 +                       err = args->open(file, vfsub_file_flags(file),
14319 +                                        au_di(dentry)->di_htmpfile);
14320 +                       di_write_unlock(dentry);
14321 +               } else {
14322 +                       di_downgrade_lock(dentry, AuLock_IR);
14323 +                       lockdep_off();
14324 +                       err = args->open(file, vfsub_file_flags(file),
14325 +                                        args->h_file);
14326 +                       lockdep_on();
14327 +                       di_read_unlock(dentry, AuLock_IR);
14328 +               }
14329 +       }
14330 +
14331 +       finfo = au_fi(file);
14332 +       if (!err) {
14333 +               finfo->fi_file = file;
14334 +               au_hbl_add(&finfo->fi_hlist,
14335 +                          &au_sbi(file->f_path.dentry->d_sb)->si_files);
14336 +       }
14337 +       if (!aopen)
14338 +               fi_write_unlock(file);
14339 +       else {
14340 +               lockdep_off();
14341 +               fi_write_unlock(file);
14342 +               lockdep_on();
14343 +       }
14344 +       if (unlikely(err)) {
14345 +               finfo->fi_hdir = NULL;
14346 +               au_finfo_fin(file);
14347 +       }
14348 +
14349 +out:
14350 +       AuTraceErr(err);
14351 +       return err;
14352 +}
14353 +
14354 +int au_reopen_nondir(struct file *file)
14355 +{
14356 +       int err;
14357 +       aufs_bindex_t btop;
14358 +       struct dentry *dentry;
14359 +       struct au_branch *br;
14360 +       struct file *h_file, *h_file_tmp;
14361 +
14362 +       dentry = file->f_path.dentry;
14363 +       btop = au_dbtop(dentry);
14364 +       br = au_sbr(dentry->d_sb, btop);
14365 +       h_file_tmp = NULL;
14366 +       if (au_fbtop(file) == btop) {
14367 +               h_file = au_hf_top(file);
14368 +               if (file->f_mode == h_file->f_mode)
14369 +                       return 0; /* success */
14370 +               h_file_tmp = h_file;
14371 +               get_file(h_file_tmp);
14372 +               au_lcnt_inc(&br->br_nfiles);
14373 +               au_set_h_fptr(file, btop, NULL);
14374 +       }
14375 +       AuDebugOn(au_fi(file)->fi_hdir);
14376 +       /*
14377 +        * it can happen
14378 +        * file exists on both of rw and ro
14379 +        * open --> dbtop and fbtop are both 0
14380 +        * prepend a branch as rw, "rw" become ro
14381 +        * remove rw/file
14382 +        * delete the top branch, "rw" becomes rw again
14383 +        *      --> dbtop is 1, fbtop is still 0
14384 +        * write --> fbtop is 0 but dbtop is 1
14385 +        */
14386 +       /* AuDebugOn(au_fbtop(file) < btop); */
14387 +
14388 +       h_file = au_h_open(dentry, btop, vfsub_file_flags(file) & ~O_TRUNC,
14389 +                          file, /*force_wr*/0);
14390 +       err = PTR_ERR(h_file);
14391 +       if (IS_ERR(h_file)) {
14392 +               if (h_file_tmp) {
14393 +                       /* revert */
14394 +                       au_set_h_fptr(file, btop, h_file_tmp);
14395 +                       h_file_tmp = NULL;
14396 +               }
14397 +               goto out; /* todo: close all? */
14398 +       }
14399 +
14400 +       err = 0;
14401 +       au_set_fbtop(file, btop);
14402 +       au_set_h_fptr(file, btop, h_file);
14403 +       au_update_figen(file);
14404 +       /* todo: necessary? */
14405 +       /* file->f_ra = h_file->f_ra; */
14406 +
14407 +out:
14408 +       if (h_file_tmp) {
14409 +               fput(h_file_tmp);
14410 +               au_lcnt_dec(&br->br_nfiles);
14411 +       }
14412 +       return err;
14413 +}
14414 +
14415 +/* ---------------------------------------------------------------------- */
14416 +
14417 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
14418 +                       struct dentry *hi_wh)
14419 +{
14420 +       int err;
14421 +       aufs_bindex_t btop;
14422 +       struct au_dinfo *dinfo;
14423 +       struct dentry *h_dentry;
14424 +       struct au_hdentry *hdp;
14425 +
14426 +       dinfo = au_di(file->f_path.dentry);
14427 +       AuRwMustWriteLock(&dinfo->di_rwsem);
14428 +
14429 +       btop = dinfo->di_btop;
14430 +       dinfo->di_btop = btgt;
14431 +       hdp = au_hdentry(dinfo, btgt);
14432 +       h_dentry = hdp->hd_dentry;
14433 +       hdp->hd_dentry = hi_wh;
14434 +       err = au_reopen_nondir(file);
14435 +       hdp->hd_dentry = h_dentry;
14436 +       dinfo->di_btop = btop;
14437 +
14438 +       return err;
14439 +}
14440 +
14441 +static int au_ready_to_write_wh(struct file *file, loff_t len,
14442 +                               aufs_bindex_t bcpup, struct au_pin *pin)
14443 +{
14444 +       int err;
14445 +       struct inode *inode, *h_inode;
14446 +       struct dentry *h_dentry, *hi_wh;
14447 +       struct au_cp_generic cpg = {
14448 +               .dentry = file->f_path.dentry,
14449 +               .bdst   = bcpup,
14450 +               .bsrc   = -1,
14451 +               .len    = len,
14452 +               .pin    = pin
14453 +       };
14454 +
14455 +       au_update_dbtop(cpg.dentry);
14456 +       inode = d_inode(cpg.dentry);
14457 +       h_inode = NULL;
14458 +       if (au_dbtop(cpg.dentry) <= bcpup
14459 +           && au_dbbot(cpg.dentry) >= bcpup) {
14460 +               h_dentry = au_h_dptr(cpg.dentry, bcpup);
14461 +               if (h_dentry && d_is_positive(h_dentry))
14462 +                       h_inode = d_inode(h_dentry);
14463 +       }
14464 +       hi_wh = au_hi_wh(inode, bcpup);
14465 +       if (!hi_wh && !h_inode)
14466 +               err = au_sio_cpup_wh(&cpg, file);
14467 +       else
14468 +               /* already copied-up after unlink */
14469 +               err = au_reopen_wh(file, bcpup, hi_wh);
14470 +
14471 +       if (!err
14472 +           && (inode->i_nlink > 1
14473 +               || (inode->i_state & I_LINKABLE))
14474 +           && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
14475 +               au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
14476 +
14477 +       return err;
14478 +}
14479 +
14480 +/*
14481 + * prepare the @file for writing.
14482 + */
14483 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
14484 +{
14485 +       int err;
14486 +       aufs_bindex_t dbtop;
14487 +       struct dentry *parent;
14488 +       struct inode *inode;
14489 +       struct super_block *sb;
14490 +       struct file *h_file;
14491 +       struct au_cp_generic cpg = {
14492 +               .dentry = file->f_path.dentry,
14493 +               .bdst   = -1,
14494 +               .bsrc   = -1,
14495 +               .len    = len,
14496 +               .pin    = pin,
14497 +               .flags  = AuCpup_DTIME
14498 +       };
14499 +
14500 +       sb = cpg.dentry->d_sb;
14501 +       inode = d_inode(cpg.dentry);
14502 +       cpg.bsrc = au_fbtop(file);
14503 +       err = au_test_ro(sb, cpg.bsrc, inode);
14504 +       if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
14505 +               err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
14506 +                            /*flags*/0);
14507 +               goto out;
14508 +       }
14509 +
14510 +       /* need to cpup or reopen */
14511 +       parent = dget_parent(cpg.dentry);
14512 +       di_write_lock_parent(parent);
14513 +       err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14514 +       cpg.bdst = err;
14515 +       if (unlikely(err < 0))
14516 +               goto out_dgrade;
14517 +       err = 0;
14518 +
14519 +       if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
14520 +               err = au_cpup_dirs(cpg.dentry, cpg.bdst);
14521 +               if (unlikely(err))
14522 +                       goto out_dgrade;
14523 +       }
14524 +
14525 +       err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14526 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14527 +       if (unlikely(err))
14528 +               goto out_dgrade;
14529 +
14530 +       dbtop = au_dbtop(cpg.dentry);
14531 +       if (dbtop <= cpg.bdst)
14532 +               cpg.bsrc = cpg.bdst;
14533 +
14534 +       if (dbtop <= cpg.bdst           /* just reopen */
14535 +           || !d_unhashed(cpg.dentry)  /* copyup and reopen */
14536 +               ) {
14537 +               h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
14538 +               if (IS_ERR(h_file))
14539 +                       err = PTR_ERR(h_file);
14540 +               else {
14541 +                       di_downgrade_lock(parent, AuLock_IR);
14542 +                       if (dbtop > cpg.bdst)
14543 +                               err = au_sio_cpup_simple(&cpg);
14544 +                       if (!err)
14545 +                               err = au_reopen_nondir(file);
14546 +                       au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
14547 +               }
14548 +       } else {                        /* copyup as wh and reopen */
14549 +               /*
14550 +                * since writable hfsplus branch is not supported,
14551 +                * h_open_pre/post() are unnecessary.
14552 +                */
14553 +               err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
14554 +               di_downgrade_lock(parent, AuLock_IR);
14555 +       }
14556 +
14557 +       if (!err) {
14558 +               au_pin_set_parent_lflag(pin, /*lflag*/0);
14559 +               goto out_dput; /* success */
14560 +       }
14561 +       au_unpin(pin);
14562 +       goto out_unlock;
14563 +
14564 +out_dgrade:
14565 +       di_downgrade_lock(parent, AuLock_IR);
14566 +out_unlock:
14567 +       di_read_unlock(parent, AuLock_IR);
14568 +out_dput:
14569 +       dput(parent);
14570 +out:
14571 +       return err;
14572 +}
14573 +
14574 +/* ---------------------------------------------------------------------- */
14575 +
14576 +int au_do_flush(struct file *file, fl_owner_t id,
14577 +               int (*flush)(struct file *file, fl_owner_t id))
14578 +{
14579 +       int err;
14580 +       struct super_block *sb;
14581 +       struct inode *inode;
14582 +
14583 +       inode = file_inode(file);
14584 +       sb = inode->i_sb;
14585 +       si_noflush_read_lock(sb);
14586 +       fi_read_lock(file);
14587 +       ii_read_lock_child(inode);
14588 +
14589 +       err = flush(file, id);
14590 +       au_cpup_attr_timesizes(inode);
14591 +
14592 +       ii_read_unlock(inode);
14593 +       fi_read_unlock(file);
14594 +       si_read_unlock(sb);
14595 +       return err;
14596 +}
14597 +
14598 +/* ---------------------------------------------------------------------- */
14599 +
14600 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
14601 +{
14602 +       int err;
14603 +       struct au_pin pin;
14604 +       struct au_finfo *finfo;
14605 +       struct dentry *parent, *hi_wh;
14606 +       struct inode *inode;
14607 +       struct super_block *sb;
14608 +       struct au_cp_generic cpg = {
14609 +               .dentry = file->f_path.dentry,
14610 +               .bdst   = -1,
14611 +               .bsrc   = -1,
14612 +               .len    = -1,
14613 +               .pin    = &pin,
14614 +               .flags  = AuCpup_DTIME
14615 +       };
14616 +
14617 +       FiMustWriteLock(file);
14618 +
14619 +       err = 0;
14620 +       finfo = au_fi(file);
14621 +       sb = cpg.dentry->d_sb;
14622 +       inode = d_inode(cpg.dentry);
14623 +       cpg.bdst = au_ibtop(inode);
14624 +       if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
14625 +               goto out;
14626 +
14627 +       parent = dget_parent(cpg.dentry);
14628 +       if (au_test_ro(sb, cpg.bdst, inode)) {
14629 +               di_read_lock_parent(parent, !AuLock_IR);
14630 +               err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14631 +               cpg.bdst = err;
14632 +               di_read_unlock(parent, !AuLock_IR);
14633 +               if (unlikely(err < 0))
14634 +                       goto out_parent;
14635 +               err = 0;
14636 +       }
14637 +
14638 +       di_read_lock_parent(parent, AuLock_IR);
14639 +       hi_wh = au_hi_wh(inode, cpg.bdst);
14640 +       if (!S_ISDIR(inode->i_mode)
14641 +           && au_opt_test(au_mntflags(sb), PLINK)
14642 +           && au_plink_test(inode)
14643 +           && !d_unhashed(cpg.dentry)
14644 +           && cpg.bdst < au_dbtop(cpg.dentry)) {
14645 +               err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
14646 +               if (unlikely(err))
14647 +                       goto out_unlock;
14648 +
14649 +               /* always superio. */
14650 +               err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14651 +                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14652 +               if (!err) {
14653 +                       err = au_sio_cpup_simple(&cpg);
14654 +                       au_unpin(&pin);
14655 +               }
14656 +       } else if (hi_wh) {
14657 +               /* already copied-up after unlink */
14658 +               err = au_reopen_wh(file, cpg.bdst, hi_wh);
14659 +               *need_reopen = 0;
14660 +       }
14661 +
14662 +out_unlock:
14663 +       di_read_unlock(parent, AuLock_IR);
14664 +out_parent:
14665 +       dput(parent);
14666 +out:
14667 +       return err;
14668 +}
14669 +
14670 +static void au_do_refresh_dir(struct file *file)
14671 +{
14672 +       aufs_bindex_t bindex, bbot, new_bindex, brid;
14673 +       struct au_hfile *p, tmp, *q;
14674 +       struct au_finfo *finfo;
14675 +       struct super_block *sb;
14676 +       struct au_fidir *fidir;
14677 +
14678 +       FiMustWriteLock(file);
14679 +
14680 +       sb = file->f_path.dentry->d_sb;
14681 +       finfo = au_fi(file);
14682 +       fidir = finfo->fi_hdir;
14683 +       AuDebugOn(!fidir);
14684 +       p = fidir->fd_hfile + finfo->fi_btop;
14685 +       brid = p->hf_br->br_id;
14686 +       bbot = fidir->fd_bbot;
14687 +       for (bindex = finfo->fi_btop; bindex <= bbot; bindex++, p++) {
14688 +               if (!p->hf_file)
14689 +                       continue;
14690 +
14691 +               new_bindex = au_br_index(sb, p->hf_br->br_id);
14692 +               if (new_bindex == bindex)
14693 +                       continue;
14694 +               if (new_bindex < 0) {
14695 +                       au_set_h_fptr(file, bindex, NULL);
14696 +                       continue;
14697 +               }
14698 +
14699 +               /* swap two lower inode, and loop again */
14700 +               q = fidir->fd_hfile + new_bindex;
14701 +               tmp = *q;
14702 +               *q = *p;
14703 +               *p = tmp;
14704 +               if (tmp.hf_file) {
14705 +                       bindex--;
14706 +                       p--;
14707 +               }
14708 +       }
14709 +
14710 +       p = fidir->fd_hfile;
14711 +       if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
14712 +               bbot = au_sbbot(sb);
14713 +               for (finfo->fi_btop = 0; finfo->fi_btop <= bbot;
14714 +                    finfo->fi_btop++, p++)
14715 +                       if (p->hf_file) {
14716 +                               if (file_inode(p->hf_file))
14717 +                                       break;
14718 +                               au_hfput(p, /*execed*/0);
14719 +                       }
14720 +       } else {
14721 +               bbot = au_br_index(sb, brid);
14722 +               for (finfo->fi_btop = 0; finfo->fi_btop < bbot;
14723 +                    finfo->fi_btop++, p++)
14724 +                       if (p->hf_file)
14725 +                               au_hfput(p, /*execed*/0);
14726 +               bbot = au_sbbot(sb);
14727 +       }
14728 +
14729 +       p = fidir->fd_hfile + bbot;
14730 +       for (fidir->fd_bbot = bbot; fidir->fd_bbot >= finfo->fi_btop;
14731 +            fidir->fd_bbot--, p--)
14732 +               if (p->hf_file) {
14733 +                       if (file_inode(p->hf_file))
14734 +                               break;
14735 +                       au_hfput(p, /*execed*/0);
14736 +               }
14737 +       AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
14738 +}
14739 +
14740 +/*
14741 + * after branch manipulating, refresh the file.
14742 + */
14743 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
14744 +{
14745 +       int err, need_reopen, nbr;
14746 +       aufs_bindex_t bbot, bindex;
14747 +       struct dentry *dentry;
14748 +       struct super_block *sb;
14749 +       struct au_finfo *finfo;
14750 +       struct au_hfile *hfile;
14751 +
14752 +       dentry = file->f_path.dentry;
14753 +       sb = dentry->d_sb;
14754 +       nbr = au_sbbot(sb) + 1;
14755 +       finfo = au_fi(file);
14756 +       if (!finfo->fi_hdir) {
14757 +               hfile = &finfo->fi_htop;
14758 +               AuDebugOn(!hfile->hf_file);
14759 +               bindex = au_br_index(sb, hfile->hf_br->br_id);
14760 +               AuDebugOn(bindex < 0);
14761 +               if (bindex != finfo->fi_btop)
14762 +                       au_set_fbtop(file, bindex);
14763 +       } else {
14764 +               err = au_fidir_realloc(finfo, nbr, /*may_shrink*/0);
14765 +               if (unlikely(err))
14766 +                       goto out;
14767 +               au_do_refresh_dir(file);
14768 +       }
14769 +
14770 +       err = 0;
14771 +       need_reopen = 1;
14772 +       if (!au_test_mmapped(file))
14773 +               err = au_file_refresh_by_inode(file, &need_reopen);
14774 +       if (finfo->fi_hdir)
14775 +               /* harmless if err */
14776 +               au_fidir_realloc(finfo, nbr, /*may_shrink*/1);
14777 +       if (!err && need_reopen && !d_unlinked(dentry))
14778 +               err = reopen(file);
14779 +       if (!err) {
14780 +               au_update_figen(file);
14781 +               goto out; /* success */
14782 +       }
14783 +
14784 +       /* error, close all lower files */
14785 +       if (finfo->fi_hdir) {
14786 +               bbot = au_fbbot_dir(file);
14787 +               for (bindex = au_fbtop(file); bindex <= bbot; bindex++)
14788 +                       au_set_h_fptr(file, bindex, NULL);
14789 +       }
14790 +
14791 +out:
14792 +       return err;
14793 +}
14794 +
14795 +/* common function to regular file and dir */
14796 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14797 +                         int wlock, unsigned int fi_lsc)
14798 +{
14799 +       int err;
14800 +       unsigned int sigen, figen;
14801 +       aufs_bindex_t btop;
14802 +       unsigned char pseudo_link;
14803 +       struct dentry *dentry;
14804 +       struct inode *inode;
14805 +
14806 +       err = 0;
14807 +       dentry = file->f_path.dentry;
14808 +       inode = d_inode(dentry);
14809 +       sigen = au_sigen(dentry->d_sb);
14810 +       fi_write_lock_nested(file, fi_lsc);
14811 +       figen = au_figen(file);
14812 +       if (!fi_lsc)
14813 +               di_write_lock_child(dentry);
14814 +       else
14815 +               di_write_lock_child2(dentry);
14816 +       btop = au_dbtop(dentry);
14817 +       pseudo_link = (btop != au_ibtop(inode));
14818 +       if (sigen == figen && !pseudo_link && au_fbtop(file) == btop) {
14819 +               if (!wlock) {
14820 +                       di_downgrade_lock(dentry, AuLock_IR);
14821 +                       fi_downgrade_lock(file);
14822 +               }
14823 +               goto out; /* success */
14824 +       }
14825 +
14826 +       AuDbg("sigen %d, figen %d\n", sigen, figen);
14827 +       if (au_digen_test(dentry, sigen)) {
14828 +               err = au_reval_dpath(dentry, sigen);
14829 +               AuDebugOn(!err && au_digen_test(dentry, sigen));
14830 +       }
14831 +
14832 +       if (!err)
14833 +               err = refresh_file(file, reopen);
14834 +       if (!err) {
14835 +               if (!wlock) {
14836 +                       di_downgrade_lock(dentry, AuLock_IR);
14837 +                       fi_downgrade_lock(file);
14838 +               }
14839 +       } else {
14840 +               di_write_unlock(dentry);
14841 +               fi_write_unlock(file);
14842 +       }
14843 +
14844 +out:
14845 +       return err;
14846 +}
14847 +
14848 +/* ---------------------------------------------------------------------- */
14849 +
14850 +/* cf. aufs_nopage() */
14851 +/* for madvise(2) */
14852 +static int aufs_read_folio(struct file *file __maybe_unused, struct folio *folio)
14853 +{
14854 +       folio_unlock(folio);
14855 +       return 0;
14856 +}
14857 +
14858 +/* it will never be called, but necessary to support O_DIRECT */
14859 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
14860 +{ BUG(); return 0; }
14861 +
14862 +/* they will never be called. */
14863 +#ifdef CONFIG_AUFS_DEBUG
14864 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
14865 +                           loff_t pos, unsigned len,
14866 +                           struct page **pagep, void **fsdata)
14867 +{ AuUnsupport(); return 0; }
14868 +static int aufs_write_end(struct file *file, struct address_space *mapping,
14869 +                         loff_t pos, unsigned len, unsigned copied,
14870 +                         struct page *page, void *fsdata)
14871 +{ AuUnsupport(); return 0; }
14872 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
14873 +{ AuUnsupport(); return 0; }
14874 +
14875 +static bool aufs_dirty_folio(struct address_space *mapping, struct folio *folio)
14876 +{ AuUnsupport(); return true; }
14877 +static void aufs_invalidate_folio(struct folio *folio, size_t offset, size_t len)
14878 +{ AuUnsupport(); }
14879 +static bool aufs_release_folio(struct folio *folio, gfp_t gfp)
14880 +{ AuUnsupport(); return true; }
14881 +#if 0 /* called by memory compaction regardless file */
14882 +static int aufs_migrate_folio(struct address_space *mapping, struct folio *dst,
14883 +                             struct folio *src, enum migrate_mode mode)
14884 +{ AuUnsupport(); return 0; }
14885 +#endif
14886 +static int aufs_launder_folio(struct folio *folio)
14887 +{ AuUnsupport(); return 0; }
14888 +static bool aufs_is_partially_uptodate(struct folio *folio, size_t from,
14889 +                                     size_t count)
14890 +{ AuUnsupport(); return true; }
14891 +static void aufs_is_dirty_writeback(struct folio *folio, bool *dirty,
14892 +                                   bool *writeback)
14893 +{ AuUnsupport(); }
14894 +static int aufs_error_remove_page(struct address_space *mapping,
14895 +                                 struct page *page)
14896 +{ AuUnsupport(); return 0; }
14897 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
14898 +                             sector_t *span)
14899 +{ AuUnsupport(); return 0; }
14900 +static void aufs_swap_deactivate(struct file *file)
14901 +{ AuUnsupport(); }
14902 +static int aufs_swap_rw(struct kiocb *iocb, struct iov_iter *iter)
14903 +{ AuUnsupport(); return 0; }
14904 +#endif /* CONFIG_AUFS_DEBUG */
14905 +
14906 +const struct address_space_operations aufs_aop = {
14907 +       .read_folio             = aufs_read_folio,
14908 +       .direct_IO              = aufs_direct_IO,
14909 +#ifdef CONFIG_AUFS_DEBUG
14910 +       .writepage              = aufs_writepage,
14911 +       /* no writepages, because of writepage */
14912 +       .dirty_folio            = aufs_dirty_folio,
14913 +       /* no readpages, because of readpage */
14914 +       .write_begin            = aufs_write_begin,
14915 +       .write_end              = aufs_write_end,
14916 +       /* no bmap, no block device */
14917 +       .invalidate_folio       = aufs_invalidate_folio,
14918 +       .release_folio          = aufs_release_folio,
14919 +       /* is fallback_migrate_page ok? */
14920 +       /* .migrate_folio       = aufs_migrate_folio, */
14921 +       .launder_folio          = aufs_launder_folio,
14922 +       .is_partially_uptodate  = aufs_is_partially_uptodate,
14923 +       .is_dirty_writeback     = aufs_is_dirty_writeback,
14924 +       .error_remove_page      = aufs_error_remove_page,
14925 +       .swap_activate          = aufs_swap_activate,
14926 +       .swap_deactivate        = aufs_swap_deactivate,
14927 +       .swap_rw                = aufs_swap_rw
14928 +#endif /* CONFIG_AUFS_DEBUG */
14929 +};
14930 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
14931 --- /usr/share/empty/fs/aufs/file.h     1970-01-01 01:00:00.000000000 +0100
14932 +++ linux/fs/aufs/file.h        2022-11-05 23:02:18.965889284 +0100
14933 @@ -0,0 +1,342 @@
14934 +/* SPDX-License-Identifier: GPL-2.0 */
14935 +/*
14936 + * Copyright (C) 2005-2022 Junjiro R. Okajima
14937 + *
14938 + * This program is free software; you can redistribute it and/or modify
14939 + * it under the terms of the GNU General Public License as published by
14940 + * the Free Software Foundation; either version 2 of the License, or
14941 + * (at your option) any later version.
14942 + *
14943 + * This program is distributed in the hope that it will be useful,
14944 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14945 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14946 + * GNU General Public License for more details.
14947 + *
14948 + * You should have received a copy of the GNU General Public License
14949 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
14950 + */
14951 +
14952 +/*
14953 + * file operations
14954 + */
14955 +
14956 +#ifndef __AUFS_FILE_H__
14957 +#define __AUFS_FILE_H__
14958 +
14959 +#ifdef __KERNEL__
14960 +
14961 +#include <linux/file.h>
14962 +#include <linux/fs.h>
14963 +#include <linux/mm_types.h>
14964 +#include <linux/poll.h>
14965 +#include "rwsem.h"
14966 +
14967 +struct au_branch;
14968 +struct au_hfile {
14969 +       struct file             *hf_file;
14970 +       struct au_branch        *hf_br;
14971 +};
14972 +
14973 +struct au_vdir;
14974 +struct au_fidir {
14975 +       aufs_bindex_t           fd_bbot;
14976 +       aufs_bindex_t           fd_nent;
14977 +       struct au_vdir          *fd_vdir_cache;
14978 +       struct au_hfile         fd_hfile[];
14979 +};
14980 +
14981 +static inline int au_fidir_sz(int nent)
14982 +{
14983 +       AuDebugOn(nent < 0);
14984 +       return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
14985 +}
14986 +
14987 +struct au_finfo {
14988 +       atomic_t                fi_generation;
14989 +
14990 +       struct au_rwsem         fi_rwsem;
14991 +       aufs_bindex_t           fi_btop;
14992 +
14993 +       /* do not union them */
14994 +       struct {                                /* for non-dir */
14995 +               struct au_hfile                 fi_htop;
14996 +               atomic_t                        fi_mmapped;
14997 +       };
14998 +       struct au_fidir         *fi_hdir;       /* for dir only */
14999 +
15000 +       struct hlist_bl_node    fi_hlist;
15001 +       struct file             *fi_file;       /* very ugly */
15002 +       struct rcu_head         rcu;
15003 +} ____cacheline_aligned_in_smp;
15004 +
15005 +/* ---------------------------------------------------------------------- */
15006 +
15007 +/* file.c */
15008 +extern const struct address_space_operations aufs_aop;
15009 +unsigned int au_file_roflags(unsigned int flags);
15010 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
15011 +                      struct file *file, int force_wr);
15012 +struct au_do_open_args {
15013 +       int             aopen;
15014 +       int             (*open)(struct file *file, int flags,
15015 +                               struct file *h_file);
15016 +       struct au_fidir *fidir;
15017 +       struct file     *h_file;
15018 +};
15019 +int au_do_open(struct file *file, struct au_do_open_args *args);
15020 +int au_reopen_nondir(struct file *file);
15021 +struct au_pin;
15022 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
15023 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
15024 +                         int wlock, unsigned int fi_lsc);
15025 +int au_do_flush(struct file *file, fl_owner_t id,
15026 +               int (*flush)(struct file *file, fl_owner_t id));
15027 +
15028 +/* poll.c */
15029 +#ifdef CONFIG_AUFS_POLL
15030 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt);
15031 +#endif
15032 +
15033 +#ifdef CONFIG_AUFS_BR_HFSPLUS
15034 +/* hfsplus.c */
15035 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
15036 +                          int force_wr);
15037 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
15038 +                   struct file *h_file);
15039 +#else
15040 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
15041 +       aufs_bindex_t bindex, int force_wr)
15042 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
15043 +          struct file *h_file);
15044 +#endif
15045 +
15046 +/* f_op.c */
15047 +extern const struct file_operations aufs_file_fop;
15048 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
15049 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
15050 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc);
15051 +
15052 +/* finfo.c */
15053 +void au_hfput(struct au_hfile *hf, int execed);
15054 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
15055 +                  struct file *h_file);
15056 +
15057 +void au_update_figen(struct file *file);
15058 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
15059 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink);
15060 +
15061 +void au_fi_init_once(void *_fi);
15062 +void au_finfo_fin(struct file *file);
15063 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
15064 +
15065 +/* ioctl.c */
15066 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
15067 +#ifdef CONFIG_COMPAT
15068 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15069 +                          unsigned long arg);
15070 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15071 +                             unsigned long arg);
15072 +#endif
15073 +
15074 +/* ---------------------------------------------------------------------- */
15075 +
15076 +static inline struct au_finfo *au_fi(struct file *file)
15077 +{
15078 +       return file->private_data;
15079 +}
15080 +
15081 +/* ---------------------------------------------------------------------- */
15082 +
15083 +#define fi_read_lock(f)        au_rw_read_lock(&au_fi(f)->fi_rwsem)
15084 +#define fi_write_lock(f)       au_rw_write_lock(&au_fi(f)->fi_rwsem)
15085 +#define fi_read_trylock(f)     au_rw_read_trylock(&au_fi(f)->fi_rwsem)
15086 +#define fi_write_trylock(f)    au_rw_write_trylock(&au_fi(f)->fi_rwsem)
15087 +/*
15088 +#define fi_read_trylock_nested(f) \
15089 +       au_rw_read_trylock_nested(&au_fi(f)->fi_rwsem)
15090 +#define fi_write_trylock_nested(f) \
15091 +       au_rw_write_trylock_nested(&au_fi(f)->fi_rwsem)
15092 +*/
15093 +
15094 +#define fi_read_unlock(f)      au_rw_read_unlock(&au_fi(f)->fi_rwsem)
15095 +#define fi_write_unlock(f)     au_rw_write_unlock(&au_fi(f)->fi_rwsem)
15096 +#define fi_downgrade_lock(f)   au_rw_dgrade_lock(&au_fi(f)->fi_rwsem)
15097 +
15098 +/* lock subclass for finfo */
15099 +enum {
15100 +       AuLsc_FI_1,
15101 +       AuLsc_FI_2
15102 +};
15103 +
15104 +static inline void fi_read_lock_nested(struct file *f, unsigned int lsc)
15105 +{
15106 +       au_rw_read_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15107 +}
15108 +
15109 +static inline void fi_write_lock_nested(struct file *f, unsigned int lsc)
15110 +{
15111 +       au_rw_write_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15112 +}
15113 +
15114 +/*
15115 + * fi_read_lock_1, fi_write_lock_1,
15116 + * fi_read_lock_2, fi_write_lock_2
15117 + */
15118 +#define AuReadLockFunc(name) \
15119 +static inline void fi_read_lock_##name(struct file *f) \
15120 +{ fi_read_lock_nested(f, AuLsc_FI_##name); }
15121 +
15122 +#define AuWriteLockFunc(name) \
15123 +static inline void fi_write_lock_##name(struct file *f) \
15124 +{ fi_write_lock_nested(f, AuLsc_FI_##name); }
15125 +
15126 +#define AuRWLockFuncs(name) \
15127 +       AuReadLockFunc(name) \
15128 +       AuWriteLockFunc(name)
15129 +
15130 +AuRWLockFuncs(1);
15131 +AuRWLockFuncs(2);
15132 +
15133 +#undef AuReadLockFunc
15134 +#undef AuWriteLockFunc
15135 +#undef AuRWLockFuncs
15136 +
15137 +#define FiMustNoWaiters(f)     AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
15138 +#define FiMustAnyLock(f)       AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
15139 +#define FiMustWriteLock(f)     AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
15140 +
15141 +/* ---------------------------------------------------------------------- */
15142 +
15143 +/* todo: hard/soft set? */
15144 +static inline aufs_bindex_t au_fbtop(struct file *file)
15145 +{
15146 +       FiMustAnyLock(file);
15147 +       return au_fi(file)->fi_btop;
15148 +}
15149 +
15150 +static inline aufs_bindex_t au_fbbot_dir(struct file *file)
15151 +{
15152 +       FiMustAnyLock(file);
15153 +       AuDebugOn(!au_fi(file)->fi_hdir);
15154 +       return au_fi(file)->fi_hdir->fd_bbot;
15155 +}
15156 +
15157 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
15158 +{
15159 +       FiMustAnyLock(file);
15160 +       AuDebugOn(!au_fi(file)->fi_hdir);
15161 +       return au_fi(file)->fi_hdir->fd_vdir_cache;
15162 +}
15163 +
15164 +static inline void au_set_fbtop(struct file *file, aufs_bindex_t bindex)
15165 +{
15166 +       FiMustWriteLock(file);
15167 +       au_fi(file)->fi_btop = bindex;
15168 +}
15169 +
15170 +static inline void au_set_fbbot_dir(struct file *file, aufs_bindex_t bindex)
15171 +{
15172 +       FiMustWriteLock(file);
15173 +       AuDebugOn(!au_fi(file)->fi_hdir);
15174 +       au_fi(file)->fi_hdir->fd_bbot = bindex;
15175 +}
15176 +
15177 +static inline void au_set_fvdir_cache(struct file *file,
15178 +                                     struct au_vdir *vdir_cache)
15179 +{
15180 +       FiMustWriteLock(file);
15181 +       AuDebugOn(!au_fi(file)->fi_hdir);
15182 +       au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
15183 +}
15184 +
15185 +static inline struct file *au_hf_top(struct file *file)
15186 +{
15187 +       FiMustAnyLock(file);
15188 +       AuDebugOn(au_fi(file)->fi_hdir);
15189 +       return au_fi(file)->fi_htop.hf_file;
15190 +}
15191 +
15192 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
15193 +{
15194 +       FiMustAnyLock(file);
15195 +       AuDebugOn(!au_fi(file)->fi_hdir);
15196 +       return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
15197 +}
15198 +
15199 +/* todo: memory barrier? */
15200 +static inline unsigned int au_figen(struct file *f)
15201 +{
15202 +       return atomic_read(&au_fi(f)->fi_generation);
15203 +}
15204 +
15205 +static inline void au_set_mmapped(struct file *f)
15206 +{
15207 +       if (atomic_inc_return(&au_fi(f)->fi_mmapped))
15208 +               return;
15209 +       pr_warn("fi_mmapped wrapped around\n");
15210 +       while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
15211 +               ;
15212 +}
15213 +
15214 +static inline void au_unset_mmapped(struct file *f)
15215 +{
15216 +       atomic_dec(&au_fi(f)->fi_mmapped);
15217 +}
15218 +
15219 +static inline int au_test_mmapped(struct file *f)
15220 +{
15221 +       return atomic_read(&au_fi(f)->fi_mmapped);
15222 +}
15223 +
15224 +/* customize vma->vm_file */
15225 +
15226 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
15227 +                                      struct file *file)
15228 +{
15229 +       struct file *f;
15230 +
15231 +       f = vma->vm_file;
15232 +       get_file(file);
15233 +       vma->vm_file = file;
15234 +       fput(f);
15235 +}
15236 +
15237 +#ifdef CONFIG_MMU
15238 +#define AuDbgVmRegion(file, vma) do {} while (0)
15239 +
15240 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15241 +                                   struct file *file)
15242 +{
15243 +       au_do_vm_file_reset(vma, file);
15244 +}
15245 +#else
15246 +#define AuDbgVmRegion(file, vma) \
15247 +       AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
15248 +
15249 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15250 +                                   struct file *file)
15251 +{
15252 +       struct file *f;
15253 +
15254 +       au_do_vm_file_reset(vma, file);
15255 +       f = vma->vm_region->vm_file;
15256 +       get_file(file);
15257 +       vma->vm_region->vm_file = file;
15258 +       fput(f);
15259 +}
15260 +#endif /* CONFIG_MMU */
15261 +
15262 +/* handle vma->vm_prfile */
15263 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
15264 +                                   struct file *file)
15265 +{
15266 +       get_file(file);
15267 +       vma->vm_prfile = file;
15268 +#ifndef CONFIG_MMU
15269 +       get_file(file);
15270 +       vma->vm_region->vm_prfile = file;
15271 +#endif
15272 +}
15273 +
15274 +#endif /* __KERNEL__ */
15275 +#endif /* __AUFS_FILE_H__ */
15276 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
15277 --- /usr/share/empty/fs/aufs/finfo.c    1970-01-01 01:00:00.000000000 +0100
15278 +++ linux/fs/aufs/finfo.c       2022-11-05 23:02:18.965889284 +0100
15279 @@ -0,0 +1,149 @@
15280 +// SPDX-License-Identifier: GPL-2.0
15281 +/*
15282 + * Copyright (C) 2005-2022 Junjiro R. Okajima
15283 + *
15284 + * This program is free software; you can redistribute it and/or modify
15285 + * it under the terms of the GNU General Public License as published by
15286 + * the Free Software Foundation; either version 2 of the License, or
15287 + * (at your option) any later version.
15288 + *
15289 + * This program is distributed in the hope that it will be useful,
15290 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15291 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15292 + * GNU General Public License for more details.
15293 + *
15294 + * You should have received a copy of the GNU General Public License
15295 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
15296 + */
15297 +
15298 +/*
15299 + * file private data
15300 + */
15301 +
15302 +#include "aufs.h"
15303 +
15304 +void au_hfput(struct au_hfile *hf, int execed)
15305 +{
15306 +       if (execed)
15307 +               allow_write_access(hf->hf_file);
15308 +       fput(hf->hf_file);
15309 +       hf->hf_file = NULL;
15310 +       au_lcnt_dec(&hf->hf_br->br_nfiles);
15311 +       hf->hf_br = NULL;
15312 +}
15313 +
15314 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
15315 +{
15316 +       struct au_finfo *finfo = au_fi(file);
15317 +       struct au_hfile *hf;
15318 +       struct au_fidir *fidir;
15319 +
15320 +       fidir = finfo->fi_hdir;
15321 +       if (!fidir) {
15322 +               AuDebugOn(finfo->fi_btop != bindex);
15323 +               hf = &finfo->fi_htop;
15324 +       } else
15325 +               hf = fidir->fd_hfile + bindex;
15326 +
15327 +       if (hf && hf->hf_file)
15328 +               au_hfput(hf, vfsub_file_execed(file));
15329 +       if (val) {
15330 +               FiMustWriteLock(file);
15331 +               AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
15332 +               hf->hf_file = val;
15333 +               hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
15334 +       }
15335 +}
15336 +
15337 +void au_update_figen(struct file *file)
15338 +{
15339 +       atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
15340 +       /* smp_mb(); */ /* atomic_set */
15341 +}
15342 +
15343 +/* ---------------------------------------------------------------------- */
15344 +
15345 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
15346 +{
15347 +       struct au_fidir *fidir;
15348 +       int nbr;
15349 +
15350 +       nbr = au_sbbot(sb) + 1;
15351 +       if (nbr < 2)
15352 +               nbr = 2; /* initial allocate for 2 branches */
15353 +       fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
15354 +       if (fidir) {
15355 +               fidir->fd_bbot = -1;
15356 +               fidir->fd_nent = nbr;
15357 +       }
15358 +
15359 +       return fidir;
15360 +}
15361 +
15362 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink)
15363 +{
15364 +       int err;
15365 +       struct au_fidir *fidir, *p;
15366 +
15367 +       AuRwMustWriteLock(&finfo->fi_rwsem);
15368 +       fidir = finfo->fi_hdir;
15369 +       AuDebugOn(!fidir);
15370 +
15371 +       err = -ENOMEM;
15372 +       p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
15373 +                        GFP_NOFS, may_shrink);
15374 +       if (p) {
15375 +               p->fd_nent = nbr;
15376 +               finfo->fi_hdir = p;
15377 +               err = 0;
15378 +       }
15379 +
15380 +       return err;
15381 +}
15382 +
15383 +/* ---------------------------------------------------------------------- */
15384 +
15385 +void au_finfo_fin(struct file *file)
15386 +{
15387 +       struct au_finfo *finfo;
15388 +
15389 +       au_lcnt_dec(&au_sbi(file->f_path.dentry->d_sb)->si_nfiles);
15390 +
15391 +       finfo = au_fi(file);
15392 +       AuDebugOn(finfo->fi_hdir);
15393 +       AuRwDestroy(&finfo->fi_rwsem);
15394 +       au_cache_free_finfo(finfo);
15395 +}
15396 +
15397 +void au_fi_init_once(void *_finfo)
15398 +{
15399 +       struct au_finfo *finfo = _finfo;
15400 +
15401 +       au_rw_init(&finfo->fi_rwsem);
15402 +}
15403 +
15404 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
15405 +{
15406 +       int err;
15407 +       struct au_finfo *finfo;
15408 +       struct dentry *dentry;
15409 +
15410 +       err = -ENOMEM;
15411 +       dentry = file->f_path.dentry;
15412 +       finfo = au_cache_alloc_finfo();
15413 +       if (unlikely(!finfo))
15414 +               goto out;
15415 +
15416 +       err = 0;
15417 +       au_lcnt_inc(&au_sbi(dentry->d_sb)->si_nfiles);
15418 +       au_rw_write_lock(&finfo->fi_rwsem);
15419 +       finfo->fi_btop = -1;
15420 +       finfo->fi_hdir = fidir;
15421 +       atomic_set(&finfo->fi_generation, au_digen(dentry));
15422 +       /* smp_mb(); */ /* atomic_set */
15423 +
15424 +       file->private_data = finfo;
15425 +
15426 +out:
15427 +       return err;
15428 +}
15429 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
15430 --- /usr/share/empty/fs/aufs/f_op.c     1970-01-01 01:00:00.000000000 +0100
15431 +++ linux/fs/aufs/f_op.c        2023-02-20 21:05:51.959693785 +0100
15432 @@ -0,0 +1,780 @@
15433 +// SPDX-License-Identifier: GPL-2.0
15434 +/*
15435 + * Copyright (C) 2005-2022 Junjiro R. Okajima
15436 + *
15437 + * This program is free software; you can redistribute it and/or modify
15438 + * it under the terms of the GNU General Public License as published by
15439 + * the Free Software Foundation; either version 2 of the License, or
15440 + * (at your option) any later version.
15441 + *
15442 + * This program is distributed in the hope that it will be useful,
15443 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15444 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15445 + * GNU General Public License for more details.
15446 + *
15447 + * You should have received a copy of the GNU General Public License
15448 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
15449 + */
15450 +
15451 +/*
15452 + * file and vm operations
15453 + */
15454 +
15455 +#include <linux/aio.h>
15456 +#include <linux/fs_stack.h>
15457 +#include <linux/mman.h>
15458 +#include <linux/security.h>
15459 +#include "aufs.h"
15460 +
15461 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
15462 +{
15463 +       int err;
15464 +       aufs_bindex_t bindex;
15465 +       struct dentry *dentry, *h_dentry;
15466 +       struct au_finfo *finfo;
15467 +       struct inode *h_inode;
15468 +
15469 +       FiMustWriteLock(file);
15470 +
15471 +       err = 0;
15472 +       dentry = file->f_path.dentry;
15473 +       AuDebugOn(IS_ERR_OR_NULL(dentry));
15474 +       finfo = au_fi(file);
15475 +       memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
15476 +       atomic_set(&finfo->fi_mmapped, 0);
15477 +       bindex = au_dbtop(dentry);
15478 +       if (!h_file) {
15479 +               h_dentry = au_h_dptr(dentry, bindex);
15480 +               err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15481 +               if (unlikely(err))
15482 +                       goto out;
15483 +               h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
15484 +               if (IS_ERR(h_file)) {
15485 +                       err = PTR_ERR(h_file);
15486 +                       goto out;
15487 +               }
15488 +       } else {
15489 +               h_dentry = h_file->f_path.dentry;
15490 +               err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15491 +               if (unlikely(err))
15492 +                       goto out;
15493 +               /* br ref is already inc-ed */
15494 +       }
15495 +
15496 +       if (flags & __O_TMPFILE) {
15497 +               AuDebugOn(!h_file);
15498 +               AuDebugOn(h_file != au_di(dentry)->di_htmpfile);
15499 +               au_di(dentry)->di_htmpfile = NULL;
15500 +
15501 +               if (!(flags & O_EXCL)) {
15502 +                       h_inode = file_inode(h_file);
15503 +                       spin_lock(&h_inode->i_lock);
15504 +                       h_inode->i_state |= I_LINKABLE;
15505 +                       spin_unlock(&h_inode->i_lock);
15506 +               }
15507 +       }
15508 +       au_set_fbtop(file, bindex);
15509 +       au_set_h_fptr(file, bindex, h_file);
15510 +       au_update_figen(file);
15511 +       /* todo: necessary? */
15512 +       /* file->f_ra = h_file->f_ra; */
15513 +
15514 +out:
15515 +       return err;
15516 +}
15517 +
15518 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
15519 +                           struct file *file)
15520 +{
15521 +       int err;
15522 +       struct super_block *sb;
15523 +       struct au_do_open_args args = {
15524 +               .open   = au_do_open_nondir
15525 +       };
15526 +
15527 +       AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
15528 +             file, vfsub_file_flags(file), file->f_mode);
15529 +
15530 +       sb = file->f_path.dentry->d_sb;
15531 +       si_read_lock(sb, AuLock_FLUSH);
15532 +       err = au_do_open(file, &args);
15533 +       si_read_unlock(sb);
15534 +       return err;
15535 +}
15536 +
15537 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
15538 +{
15539 +       struct au_finfo *finfo;
15540 +       aufs_bindex_t bindex;
15541 +
15542 +       finfo = au_fi(file);
15543 +       au_hbl_del(&finfo->fi_hlist,
15544 +                  &au_sbi(file->f_path.dentry->d_sb)->si_files);
15545 +       bindex = finfo->fi_btop;
15546 +       if (bindex >= 0)
15547 +               au_set_h_fptr(file, bindex, NULL);
15548 +
15549 +       au_finfo_fin(file);
15550 +       return 0;
15551 +}
15552 +
15553 +/* ---------------------------------------------------------------------- */
15554 +
15555 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
15556 +{
15557 +       int err;
15558 +       struct file *h_file;
15559 +
15560 +       err = 0;
15561 +       h_file = au_hf_top(file);
15562 +       if (h_file)
15563 +               err = vfsub_flush(h_file, id);
15564 +       return err;
15565 +}
15566 +
15567 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
15568 +{
15569 +       return au_do_flush(file, id, au_do_flush_nondir);
15570 +}
15571 +
15572 +/* ---------------------------------------------------------------------- */
15573 +/*
15574 + * read and write functions acquire [fdi]_rwsem once, but release before
15575 + * mmap_sem. This is because to stop a race condition between mmap(2).
15576 + * Releasing these aufs-rwsem should be safe, no branch-management (by keeping
15577 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
15578 + * read functions after [fdi]_rwsem are released, but it should be harmless.
15579 + */
15580 +
15581 +/* Callers should call au_read_post() or fput() in the end */
15582 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc)
15583 +{
15584 +       struct file *h_file;
15585 +       int err;
15586 +
15587 +       err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0, lsc);
15588 +       if (!err) {
15589 +               di_read_unlock(file->f_path.dentry, AuLock_IR);
15590 +               h_file = au_hf_top(file);
15591 +               get_file(h_file);
15592 +               if (!keep_fi)
15593 +                       fi_read_unlock(file);
15594 +       } else
15595 +               h_file = ERR_PTR(err);
15596 +
15597 +       return h_file;
15598 +}
15599 +
15600 +static void au_read_post(struct inode *inode, struct file *h_file)
15601 +{
15602 +       /* update without lock, I don't think it a problem */
15603 +       fsstack_copy_attr_atime(inode, file_inode(h_file));
15604 +       fput(h_file);
15605 +}
15606 +
15607 +struct au_write_pre {
15608 +       /* input */
15609 +       unsigned int lsc;
15610 +
15611 +       /* output */
15612 +       blkcnt_t blks;
15613 +       aufs_bindex_t btop;
15614 +};
15615 +
15616 +/*
15617 + * return with iinfo is write-locked
15618 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
15619 + * end
15620 + */
15621 +static struct file *au_write_pre(struct file *file, int do_ready,
15622 +                                struct au_write_pre *wpre)
15623 +{
15624 +       struct file *h_file;
15625 +       struct dentry *dentry;
15626 +       int err;
15627 +       unsigned int lsc;
15628 +       struct au_pin pin;
15629 +
15630 +       lsc = 0;
15631 +       if (wpre)
15632 +               lsc = wpre->lsc;
15633 +       err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1, lsc);
15634 +       h_file = ERR_PTR(err);
15635 +       if (unlikely(err))
15636 +               goto out;
15637 +
15638 +       dentry = file->f_path.dentry;
15639 +       if (do_ready) {
15640 +               err = au_ready_to_write(file, -1, &pin);
15641 +               if (unlikely(err)) {
15642 +                       h_file = ERR_PTR(err);
15643 +                       di_write_unlock(dentry);
15644 +                       goto out_fi;
15645 +               }
15646 +       }
15647 +
15648 +       di_downgrade_lock(dentry, /*flags*/0);
15649 +       if (wpre)
15650 +               wpre->btop = au_fbtop(file);
15651 +       h_file = au_hf_top(file);
15652 +       get_file(h_file);
15653 +       if (wpre)
15654 +               wpre->blks = file_inode(h_file)->i_blocks;
15655 +       if (do_ready)
15656 +               au_unpin(&pin);
15657 +       di_read_unlock(dentry, /*flags*/0);
15658 +
15659 +out_fi:
15660 +       fi_write_unlock(file);
15661 +out:
15662 +       return h_file;
15663 +}
15664 +
15665 +static void au_write_post(struct inode *inode, struct file *h_file,
15666 +                         struct au_write_pre *wpre, ssize_t written)
15667 +{
15668 +       struct inode *h_inode;
15669 +
15670 +       au_cpup_attr_timesizes(inode);
15671 +       AuDebugOn(au_ibtop(inode) != wpre->btop);
15672 +       h_inode = file_inode(h_file);
15673 +       inode->i_mode = h_inode->i_mode;
15674 +       ii_write_unlock(inode);
15675 +       /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
15676 +       if (written > 0)
15677 +               au_fhsm_wrote(inode->i_sb, wpre->btop,
15678 +                             /*force*/h_inode->i_blocks > wpre->blks);
15679 +       fput(h_file);
15680 +}
15681 +
15682 +/*
15683 + * todo: very ugly
15684 + * it locks both of i_mutex and si_rwsem for read in safe.
15685 + * if the plink maintenance mode continues forever (that is the problem),
15686 + * may loop forever.
15687 + */
15688 +static void au_mtx_and_read_lock(struct inode *inode)
15689 +{
15690 +       int err;
15691 +       struct super_block *sb = inode->i_sb;
15692 +
15693 +       while (1) {
15694 +               inode_lock(inode);
15695 +               err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
15696 +               if (!err)
15697 +                       break;
15698 +               inode_unlock(inode);
15699 +               si_read_lock(sb, AuLock_NOPLMW);
15700 +               si_read_unlock(sb);
15701 +       }
15702 +}
15703 +
15704 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
15705 +                         struct iov_iter *iov_iter)
15706 +{
15707 +       ssize_t err;
15708 +       struct file *file;
15709 +       ssize_t (*iter)(struct kiocb *, struct iov_iter *);
15710 +
15711 +       err = security_file_permission(h_file, rw);
15712 +       if (unlikely(err))
15713 +               goto out;
15714 +
15715 +       err = -ENOSYS;  /* the branch doesn't have its ->(read|write)_iter() */
15716 +       iter = NULL;
15717 +       if (rw == MAY_READ)
15718 +               iter = h_file->f_op->read_iter;
15719 +       else if (rw == MAY_WRITE)
15720 +               iter = h_file->f_op->write_iter;
15721 +
15722 +       file = kio->ki_filp;
15723 +       kio->ki_filp = h_file;
15724 +       if (iter) {
15725 +               lockdep_off();
15726 +               err = iter(kio, iov_iter);
15727 +               lockdep_on();
15728 +       } else
15729 +               /* currently there is no such fs */
15730 +               WARN_ON_ONCE(1);
15731 +       kio->ki_filp = file;
15732 +
15733 +out:
15734 +       return err;
15735 +}
15736 +
15737 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15738 +{
15739 +       ssize_t err;
15740 +       struct file *file, *h_file;
15741 +       struct inode *inode;
15742 +       struct super_block *sb;
15743 +
15744 +       file = kio->ki_filp;
15745 +       inode = file_inode(file);
15746 +       sb = inode->i_sb;
15747 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15748 +
15749 +       h_file = au_read_pre(file, /*keep_fi*/1, /*lsc*/0);
15750 +       err = PTR_ERR(h_file);
15751 +       if (IS_ERR(h_file))
15752 +               goto out;
15753 +
15754 +       if (au_test_loopback_kthread()) {
15755 +               au_warn_loopback(h_file->f_path.dentry->d_sb);
15756 +               if (file->f_mapping != h_file->f_mapping) {
15757 +                       file->f_mapping = h_file->f_mapping;
15758 +                       smp_mb(); /* unnecessary? */
15759 +               }
15760 +       }
15761 +       fi_read_unlock(file);
15762 +
15763 +       err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
15764 +       /* todo: necessary? */
15765 +       /* file->f_ra = h_file->f_ra; */
15766 +       au_read_post(inode, h_file);
15767 +
15768 +out:
15769 +       si_read_unlock(sb);
15770 +       return err;
15771 +}
15772 +
15773 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15774 +{
15775 +       ssize_t err;
15776 +       struct au_write_pre wpre;
15777 +       struct inode *inode;
15778 +       struct file *file, *h_file;
15779 +
15780 +       file = kio->ki_filp;
15781 +       inode = file_inode(file);
15782 +       au_mtx_and_read_lock(inode);
15783 +
15784 +       wpre.lsc = 0;
15785 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15786 +       err = PTR_ERR(h_file);
15787 +       if (IS_ERR(h_file))
15788 +               goto out;
15789 +
15790 +       err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
15791 +       au_write_post(inode, h_file, &wpre, err);
15792 +
15793 +out:
15794 +       si_read_unlock(inode->i_sb);
15795 +       inode_unlock(inode);
15796 +       return err;
15797 +}
15798 +
15799 +/*
15800 + * We may be able to remove aufs_splice_{read,write}() since almost all FSes
15801 + * don't have their own .splice_{read,write} implimentations, and they use
15802 + * generic_file_splice_read() and iter_file_splice_write() who can act like the
15803 + * simple converters to f_op->iter_read() and ->iter_write().
15804 + * But we keep our own implementations because some non-mainlined FSes may have
15805 + * their own .splice_{read,write} implimentations and aufs doesn't want to take
15806 + * away an opportunity to co-work with aufs from them.
15807 + */
15808 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
15809 +                               struct pipe_inode_info *pipe, size_t len,
15810 +                               unsigned int flags)
15811 +{
15812 +       ssize_t err;
15813 +       struct file *h_file;
15814 +       struct inode *inode;
15815 +       struct super_block *sb;
15816 +
15817 +       inode = file_inode(file);
15818 +       sb = inode->i_sb;
15819 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15820 +
15821 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15822 +       err = PTR_ERR(h_file);
15823 +       if (IS_ERR(h_file))
15824 +               goto out;
15825 +
15826 +       err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
15827 +       /* todo: necessary? */
15828 +       /* file->f_ra = h_file->f_ra; */
15829 +       au_read_post(inode, h_file);
15830 +
15831 +out:
15832 +       si_read_unlock(sb);
15833 +       return err;
15834 +}
15835 +
15836 +static ssize_t
15837 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
15838 +                 size_t len, unsigned int flags)
15839 +{
15840 +       ssize_t err;
15841 +       struct au_write_pre wpre;
15842 +       struct inode *inode;
15843 +       struct file *h_file;
15844 +
15845 +       inode = file_inode(file);
15846 +       au_mtx_and_read_lock(inode);
15847 +
15848 +       wpre.lsc = 0;
15849 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15850 +       err = PTR_ERR(h_file);
15851 +       if (IS_ERR(h_file))
15852 +               goto out;
15853 +
15854 +       err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
15855 +       au_write_post(inode, h_file, &wpre, err);
15856 +
15857 +out:
15858 +       si_read_unlock(inode->i_sb);
15859 +       inode_unlock(inode);
15860 +       return err;
15861 +}
15862 +
15863 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
15864 +                          loff_t len)
15865 +{
15866 +       long err;
15867 +       struct au_write_pre wpre;
15868 +       struct inode *inode;
15869 +       struct file *h_file;
15870 +
15871 +       inode = file_inode(file);
15872 +       au_mtx_and_read_lock(inode);
15873 +
15874 +       wpre.lsc = 0;
15875 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15876 +       err = PTR_ERR(h_file);
15877 +       if (IS_ERR(h_file))
15878 +               goto out;
15879 +
15880 +       lockdep_off();
15881 +       err = vfs_fallocate(h_file, mode, offset, len);
15882 +       lockdep_on();
15883 +       /*
15884 +        * we don't need to call file_modifed() here since au_write_post()
15885 +        * is equivalent and copies-up all timestamps and permission bits.
15886 +        */
15887 +       au_write_post(inode, h_file, &wpre, /*written*/1);
15888 +
15889 +out:
15890 +       si_read_unlock(inode->i_sb);
15891 +       inode_unlock(inode);
15892 +       return err;
15893 +}
15894 +
15895 +static ssize_t aufs_copy_file_range(struct file *src, loff_t src_pos,
15896 +                                   struct file *dst, loff_t dst_pos,
15897 +                                   size_t len, unsigned int flags)
15898 +{
15899 +       ssize_t err;
15900 +       struct au_write_pre wpre;
15901 +       enum { SRC, DST };
15902 +       struct {
15903 +               struct inode *inode;
15904 +               struct file *h_file;
15905 +               struct super_block *h_sb;
15906 +       } a[2];
15907 +#define a_src  a[SRC]
15908 +#define a_dst  a[DST]
15909 +
15910 +       err = -EINVAL;
15911 +       a_src.inode = file_inode(src);
15912 +       if (unlikely(!S_ISREG(a_src.inode->i_mode)))
15913 +               goto out;
15914 +       a_dst.inode = file_inode(dst);
15915 +       if (unlikely(!S_ISREG(a_dst.inode->i_mode)))
15916 +               goto out;
15917 +
15918 +       au_mtx_and_read_lock(a_dst.inode);
15919 +       /*
15920 +        * in order to match the order in di_write_lock2_{child,parent}(),
15921 +        * use f_path.dentry for this comparison.
15922 +        */
15923 +       if (src->f_path.dentry < dst->f_path.dentry) {
15924 +               a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_1);
15925 +               err = PTR_ERR(a_src.h_file);
15926 +               if (IS_ERR(a_src.h_file))
15927 +                       goto out_si;
15928 +
15929 +               wpre.lsc = AuLsc_FI_2;
15930 +               a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15931 +               err = PTR_ERR(a_dst.h_file);
15932 +               if (IS_ERR(a_dst.h_file)) {
15933 +                       au_read_post(a_src.inode, a_src.h_file);
15934 +                       goto out_si;
15935 +               }
15936 +       } else {
15937 +               wpre.lsc = AuLsc_FI_1;
15938 +               a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15939 +               err = PTR_ERR(a_dst.h_file);
15940 +               if (IS_ERR(a_dst.h_file))
15941 +                       goto out_si;
15942 +
15943 +               a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_2);
15944 +               err = PTR_ERR(a_src.h_file);
15945 +               if (IS_ERR(a_src.h_file)) {
15946 +                       au_write_post(a_dst.inode, a_dst.h_file, &wpre,
15947 +                                     /*written*/0);
15948 +                       goto out_si;
15949 +               }
15950 +       }
15951 +
15952 +       err = -EXDEV;
15953 +       a_src.h_sb = file_inode(a_src.h_file)->i_sb;
15954 +       a_dst.h_sb = file_inode(a_dst.h_file)->i_sb;
15955 +       if (unlikely(a_src.h_sb != a_dst.h_sb)) {
15956 +               AuDbgFile(src);
15957 +               AuDbgFile(dst);
15958 +               goto out_file;
15959 +       }
15960 +
15961 +       err = vfsub_copy_file_range(a_src.h_file, src_pos, a_dst.h_file,
15962 +                                   dst_pos, len, flags);
15963 +
15964 +out_file:
15965 +       au_write_post(a_dst.inode, a_dst.h_file, &wpre, err);
15966 +       fi_read_unlock(src);
15967 +       au_read_post(a_src.inode, a_src.h_file);
15968 +out_si:
15969 +       si_read_unlock(a_dst.inode->i_sb);
15970 +       inode_unlock(a_dst.inode);
15971 +out:
15972 +       return err;
15973 +#undef a_src
15974 +#undef a_dst
15975 +}
15976 +
15977 +/* ---------------------------------------------------------------------- */
15978 +
15979 +/*
15980 + * The locking order around current->mmap_sem.
15981 + * - in most and regular cases
15982 + *   file I/O syscall -- aufs_read() or something
15983 + *     -- si_rwsem for read -- mmap_sem
15984 + *     (Note that [fdi]i_rwsem are released before mmap_sem).
15985 + * - in mmap case
15986 + *   mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
15987 + * This AB-BA order is definitely bad, but is not a problem since "si_rwsem for
15988 + * read" allows multiple processes to acquire it and [fdi]i_rwsem are not held
15989 + * in file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
15990 + * It means that when aufs acquires si_rwsem for write, the process should never
15991 + * acquire mmap_sem.
15992 + *
15993 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
15994 + * problem either since any directory is not able to be mmap-ed.
15995 + * The similar scenario is applied to aufs_readlink() too.
15996 + */
15997 +
15998 +#if 0 /* stop calling security_file_mmap() */
15999 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
16000 +#define AuConv_VM_PROT(f, b)   _calc_vm_trans(f, VM_##b, PROT_##b)
16001 +
16002 +static unsigned long au_arch_prot_conv(unsigned long flags)
16003 +{
16004 +       /* currently ppc64 only */
16005 +#ifdef CONFIG_PPC64
16006 +       /* cf. linux/arch/powerpc/include/asm/mman.h */
16007 +       AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
16008 +       return AuConv_VM_PROT(flags, SAO);
16009 +#else
16010 +       AuDebugOn(arch_calc_vm_prot_bits(-1));
16011 +       return 0;
16012 +#endif
16013 +}
16014 +
16015 +static unsigned long au_prot_conv(unsigned long flags)
16016 +{
16017 +       return AuConv_VM_PROT(flags, READ)
16018 +               | AuConv_VM_PROT(flags, WRITE)
16019 +               | AuConv_VM_PROT(flags, EXEC)
16020 +               | au_arch_prot_conv(flags);
16021 +}
16022 +
16023 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
16024 +#define AuConv_VM_MAP(f, b)    _calc_vm_trans(f, VM_##b, MAP_##b)
16025 +
16026 +static unsigned long au_flag_conv(unsigned long flags)
16027 +{
16028 +       return AuConv_VM_MAP(flags, GROWSDOWN)
16029 +               | AuConv_VM_MAP(flags, DENYWRITE)
16030 +               | AuConv_VM_MAP(flags, LOCKED);
16031 +}
16032 +#endif
16033 +
16034 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
16035 +{
16036 +       int err;
16037 +       const unsigned char wlock
16038 +               = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
16039 +       struct super_block *sb;
16040 +       struct file *h_file;
16041 +       struct inode *inode;
16042 +
16043 +       AuDbgVmRegion(file, vma);
16044 +
16045 +       inode = file_inode(file);
16046 +       sb = inode->i_sb;
16047 +       lockdep_off();
16048 +       si_read_lock(sb, AuLock_NOPLMW);
16049 +
16050 +       h_file = au_write_pre(file, wlock, /*wpre*/NULL);
16051 +       lockdep_on();
16052 +       err = PTR_ERR(h_file);
16053 +       if (IS_ERR(h_file))
16054 +               goto out;
16055 +
16056 +       err = 0;
16057 +       au_set_mmapped(file);
16058 +       au_vm_file_reset(vma, h_file);
16059 +       /*
16060 +        * we cannot call security_mmap_file() here since it may acquire
16061 +        * mmap_sem or i_mutex.
16062 +        *
16063 +        * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
16064 +        *                       au_flag_conv(vma->vm_flags));
16065 +        */
16066 +       if (!err)
16067 +               err = call_mmap(h_file, vma);
16068 +       if (!err) {
16069 +               au_vm_prfile_set(vma, file);
16070 +               fsstack_copy_attr_atime(inode, file_inode(h_file));
16071 +               goto out_fput; /* success */
16072 +       }
16073 +       au_unset_mmapped(file);
16074 +       au_vm_file_reset(vma, file);
16075 +
16076 +out_fput:
16077 +       lockdep_off();
16078 +       ii_write_unlock(inode);
16079 +       lockdep_on();
16080 +       fput(h_file);
16081 +out:
16082 +       lockdep_off();
16083 +       si_read_unlock(sb);
16084 +       lockdep_on();
16085 +       AuTraceErr(err);
16086 +       return err;
16087 +}
16088 +
16089 +/* ---------------------------------------------------------------------- */
16090 +
16091 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
16092 +                            int datasync)
16093 +{
16094 +       int err;
16095 +       struct au_write_pre wpre;
16096 +       struct inode *inode;
16097 +       struct file *h_file;
16098 +
16099 +       err = 0; /* -EBADF; */ /* posix? */
16100 +       if (unlikely(!(file->f_mode & FMODE_WRITE)))
16101 +               goto out;
16102 +
16103 +       inode = file_inode(file);
16104 +       au_mtx_and_read_lock(inode);
16105 +
16106 +       wpre.lsc = 0;
16107 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
16108 +       err = PTR_ERR(h_file);
16109 +       if (IS_ERR(h_file))
16110 +               goto out_unlock;
16111 +
16112 +       err = vfsub_fsync(h_file, &h_file->f_path, datasync);
16113 +       au_write_post(inode, h_file, &wpre, /*written*/0);
16114 +
16115 +out_unlock:
16116 +       si_read_unlock(inode->i_sb);
16117 +       inode_unlock(inode);
16118 +out:
16119 +       return err;
16120 +}
16121 +
16122 +static int aufs_fasync(int fd, struct file *file, int flag)
16123 +{
16124 +       int err;
16125 +       struct file *h_file;
16126 +       struct super_block *sb;
16127 +
16128 +       sb = file->f_path.dentry->d_sb;
16129 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16130 +
16131 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16132 +       err = PTR_ERR(h_file);
16133 +       if (IS_ERR(h_file))
16134 +               goto out;
16135 +
16136 +       if (h_file->f_op->fasync)
16137 +               err = h_file->f_op->fasync(fd, h_file, flag);
16138 +       fput(h_file); /* instead of au_read_post() */
16139 +
16140 +out:
16141 +       si_read_unlock(sb);
16142 +       return err;
16143 +}
16144 +
16145 +static int aufs_setfl(struct file *file, unsigned long arg)
16146 +{
16147 +       int err;
16148 +       struct file *h_file;
16149 +       struct super_block *sb;
16150 +
16151 +       sb = file->f_path.dentry->d_sb;
16152 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16153 +
16154 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16155 +       err = PTR_ERR(h_file);
16156 +       if (IS_ERR(h_file))
16157 +               goto out;
16158 +
16159 +       /* stop calling h_file->fasync */
16160 +       arg |= vfsub_file_flags(file) & FASYNC;
16161 +       err = setfl(/*unused fd*/-1, h_file, arg);
16162 +       fput(h_file); /* instead of au_read_post() */
16163 +
16164 +out:
16165 +       si_read_unlock(sb);
16166 +       return err;
16167 +}
16168 +
16169 +/* ---------------------------------------------------------------------- */
16170 +
16171 +/* no one supports this operation, currently */
16172 +#if 0 /* reserved for future use */
16173 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
16174 +                            size_t len, loff_t *pos, int more)
16175 +{
16176 +}
16177 +#endif
16178 +
16179 +/* ---------------------------------------------------------------------- */
16180 +
16181 +const struct file_operations aufs_file_fop = {
16182 +       .owner          = THIS_MODULE,
16183 +
16184 +       .llseek         = default_llseek,
16185 +
16186 +       .read_iter      = aufs_read_iter,
16187 +       .write_iter     = aufs_write_iter,
16188 +
16189 +#ifdef CONFIG_AUFS_POLL
16190 +       .poll           = aufs_poll,
16191 +#endif
16192 +       .unlocked_ioctl = aufs_ioctl_nondir,
16193 +#ifdef CONFIG_COMPAT
16194 +       .compat_ioctl   = aufs_compat_ioctl_nondir,
16195 +#endif
16196 +       .mmap           = aufs_mmap,
16197 +       .open           = aufs_open_nondir,
16198 +       .flush          = aufs_flush_nondir,
16199 +       .release        = aufs_release_nondir,
16200 +       .fsync          = aufs_fsync_nondir,
16201 +       .fasync         = aufs_fasync,
16202 +       /* .sendpage    = aufs_sendpage, */
16203 +       .setfl          = aufs_setfl,
16204 +       .splice_write   = aufs_splice_write,
16205 +       .splice_read    = aufs_splice_read,
16206 +#if 0 /* reserved for future use */
16207 +       .aio_splice_write = aufs_aio_splice_write,
16208 +       .aio_splice_read  = aufs_aio_splice_read,
16209 +#endif
16210 +       .fallocate      = aufs_fallocate,
16211 +       .copy_file_range = aufs_copy_file_range
16212 +};
16213 diff -urN /usr/share/empty/fs/aufs/fsctx.c linux/fs/aufs/fsctx.c
16214 --- /usr/share/empty/fs/aufs/fsctx.c    1970-01-01 01:00:00.000000000 +0100
16215 +++ linux/fs/aufs/fsctx.c       2022-11-05 23:02:18.965889284 +0100
16216 @@ -0,0 +1,1242 @@
16217 +// SPDX-License-Identifier: GPL-2.0
16218 +/*
16219 + * Copyright (C) 2022 Junjiro R. Okajima
16220 + *
16221 + * This program is free software; you can redistribute it and/or modify
16222 + * it under the terms of the GNU General Public License as published by
16223 + * the Free Software Foundation; either version 2 of the License, or
16224 + * (at your option) any later version.
16225 + *
16226 + * This program is distributed in the hope that it will be useful,
16227 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16228 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16229 + * GNU General Public License for more details.
16230 + *
16231 + * You should have received a copy of the GNU General Public License
16232 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
16233 + */
16234 +
16235 +/*
16236 + * fs context, aka new mount api
16237 + */
16238 +
16239 +#include <linux/fs_context.h>
16240 +#include "aufs.h"
16241 +
16242 +struct au_fsctx_opts {
16243 +       aufs_bindex_t bindex;
16244 +       unsigned char skipped;
16245 +       struct au_opt *opt, *opt_tail;
16246 +       struct super_block *sb;
16247 +       struct au_sbinfo *sbinfo;
16248 +       struct au_opts opts;
16249 +};
16250 +
16251 +/* stop extra interpretation of errno in mount(8), and strange error messages */
16252 +static int cvt_err(int err)
16253 +{
16254 +       AuTraceErr(err);
16255 +
16256 +       switch (err) {
16257 +       case -ENOENT:
16258 +       case -ENOTDIR:
16259 +       case -EEXIST:
16260 +       case -EIO:
16261 +               err = -EINVAL;
16262 +       }
16263 +       return err;
16264 +}
16265 +
16266 +static int au_fsctx_reconfigure(struct fs_context *fc)
16267 +{
16268 +       int err, do_dx;
16269 +       unsigned int mntflags;
16270 +       struct dentry *root;
16271 +       struct super_block *sb;
16272 +       struct inode *inode;
16273 +       struct au_fsctx_opts *a = fc->fs_private;
16274 +
16275 +       AuDbg("fc %p\n", fc);
16276 +
16277 +       root = fc->root;
16278 +       sb = root->d_sb;
16279 +       err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16280 +       if (!err) {
16281 +               di_write_lock_child(root);
16282 +               err = au_opts_verify(sb, fc->sb_flags, /*pending*/0);
16283 +               aufs_write_unlock(root);
16284 +       }
16285 +
16286 +       inode = d_inode(root);
16287 +       inode_lock(inode);
16288 +       err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16289 +       if (unlikely(err))
16290 +               goto out;
16291 +       di_write_lock_child(root);
16292 +
16293 +       /* au_opts_remount() may return an error */
16294 +       err = au_opts_remount(sb, &a->opts);
16295 +
16296 +       if (au_ftest_opts(a->opts.flags, REFRESH))
16297 +               au_remount_refresh(sb, au_ftest_opts(a->opts.flags,
16298 +                                                    REFRESH_IDOP));
16299 +
16300 +       if (au_ftest_opts(a->opts.flags, REFRESH_DYAOP)) {
16301 +               mntflags = au_mntflags(sb);
16302 +               do_dx = !!au_opt_test(mntflags, DIO);
16303 +               au_dy_arefresh(do_dx);
16304 +       }
16305 +
16306 +       au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
16307 +       aufs_write_unlock(root);
16308 +
16309 +out:
16310 +       inode_unlock(inode);
16311 +       err = cvt_err(err);
16312 +       AuTraceErr(err);
16313 +
16314 +       return err;
16315 +}
16316 +
16317 +/* ---------------------------------------------------------------------- */
16318 +
16319 +static int au_fsctx_fill_super(struct super_block *sb, struct fs_context *fc)
16320 +{
16321 +       int err;
16322 +       struct au_fsctx_opts *a = fc->fs_private;
16323 +       struct au_sbinfo *sbinfo = a->sbinfo;
16324 +       struct dentry *root;
16325 +       struct inode *inode;
16326 +
16327 +       sbinfo->si_sb = sb;
16328 +       sb->s_fs_info = sbinfo;
16329 +       kobject_get(&sbinfo->si_kobj);
16330 +
16331 +       __si_write_lock(sb);
16332 +       si_pid_set(sb);
16333 +       au_sbilist_add(sb);
16334 +
16335 +       /* all timestamps always follow the ones on the branch */
16336 +       sb->s_flags |= SB_NOATIME | SB_NODIRATIME;
16337 +       sb->s_flags |= SB_I_VERSION; /* do we really need this? */
16338 +       sb->s_op = &aufs_sop;
16339 +       sb->s_d_op = &aufs_dop;
16340 +       sb->s_magic = AUFS_SUPER_MAGIC;
16341 +       sb->s_maxbytes = 0;
16342 +       sb->s_stack_depth = 1;
16343 +       au_export_init(sb);
16344 +       au_xattr_init(sb);
16345 +
16346 +       err = au_alloc_root(sb);
16347 +       if (unlikely(err)) {
16348 +               si_write_unlock(sb);
16349 +               goto out;
16350 +       }
16351 +       root = sb->s_root;
16352 +       inode = d_inode(root);
16353 +       ii_write_lock_parent(inode);
16354 +       aufs_write_unlock(root);
16355 +
16356 +       /* lock vfs_inode first, then aufs. */
16357 +       inode_lock(inode);
16358 +       aufs_write_lock(root);
16359 +       err = au_opts_mount(sb, &a->opts);
16360 +       AuTraceErr(err);
16361 +       if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
16362 +               sb->s_d_op = &aufs_dop_noreval;
16363 +               /* infofc(fc, "%ps", sb->s_d_op); */
16364 +               pr_info("%ps\n", sb->s_d_op);
16365 +               au_refresh_dop(root, /*force_reval*/0);
16366 +               sbinfo->si_iop_array = aufs_iop_nogetattr;
16367 +               au_refresh_iop(inode, /*force_getattr*/0);
16368 +       }
16369 +       aufs_write_unlock(root);
16370 +       inode_unlock(inode);
16371 +       if (!err)
16372 +               goto out; /* success */
16373 +
16374 +       dput(root);
16375 +       sb->s_root = NULL;
16376 +
16377 +out:
16378 +       if (unlikely(err))
16379 +               kobject_put(&sbinfo->si_kobj);
16380 +       AuTraceErr(err);
16381 +       err = cvt_err(err);
16382 +       AuTraceErr(err);
16383 +       return err;
16384 +}
16385 +
16386 +static int au_fsctx_get_tree(struct fs_context *fc)
16387 +{
16388 +       int err;
16389 +
16390 +       AuDbg("fc %p\n", fc);
16391 +       err = get_tree_nodev(fc, au_fsctx_fill_super);
16392 +
16393 +       AuTraceErr(err);
16394 +       return err;
16395 +}
16396 +
16397 +/* ---------------------------------------------------------------------- */
16398 +
16399 +static void au_fsctx_dump(struct au_opts *opts)
16400 +{
16401 +#ifdef CONFIG_AUFS_DEBUG
16402 +       /* reduce stack space */
16403 +       union {
16404 +               struct au_opt_add *add;
16405 +               struct au_opt_del *del;
16406 +               struct au_opt_mod *mod;
16407 +               struct au_opt_xino *xino;
16408 +               struct au_opt_xino_itrunc *xino_itrunc;
16409 +               struct au_opt_wbr_create *create;
16410 +       } u;
16411 +       struct au_opt *opt;
16412 +
16413 +       opt = opts->opt;
16414 +       while (opt->type != Opt_tail) {
16415 +               switch (opt->type) {
16416 +               case Opt_add:
16417 +                       u.add = &opt->add;
16418 +                       AuDbg("add {b%d, %s, 0x%x, %p}\n",
16419 +                                 u.add->bindex, u.add->pathname, u.add->perm,
16420 +                                 u.add->path.dentry);
16421 +                       break;
16422 +               case Opt_del:
16423 +                       fallthrough;
16424 +               case Opt_idel:
16425 +                       u.del = &opt->del;
16426 +                       AuDbg("del {%s, %p}\n",
16427 +                             u.del->pathname, u.del->h_path.dentry);
16428 +                       break;
16429 +               case Opt_mod:
16430 +                       fallthrough;
16431 +               case Opt_imod:
16432 +                       u.mod = &opt->mod;
16433 +                       AuDbg("mod {%s, 0x%x, %p}\n",
16434 +                                 u.mod->path, u.mod->perm, u.mod->h_root);
16435 +                       break;
16436 +               case Opt_append:
16437 +                       u.add = &opt->add;
16438 +                       AuDbg("append {b%d, %s, 0x%x, %p}\n",
16439 +                                 u.add->bindex, u.add->pathname, u.add->perm,
16440 +                                 u.add->path.dentry);
16441 +                       break;
16442 +               case Opt_prepend:
16443 +                       u.add = &opt->add;
16444 +                       AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
16445 +                                 u.add->bindex, u.add->pathname, u.add->perm,
16446 +                                 u.add->path.dentry);
16447 +                       break;
16448 +
16449 +               case Opt_dirwh:
16450 +                       AuDbg("dirwh %d\n", opt->dirwh);
16451 +                       break;
16452 +               case Opt_rdcache:
16453 +                       AuDbg("rdcache %d\n", opt->rdcache);
16454 +                       break;
16455 +               case Opt_rdblk:
16456 +                       AuDbg("rdblk %d\n", opt->rdblk);
16457 +                       break;
16458 +               case Opt_rdhash:
16459 +                       AuDbg("rdhash %u\n", opt->rdhash);
16460 +                       break;
16461 +
16462 +               case Opt_xino:
16463 +                       u.xino = &opt->xino;
16464 +                       AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
16465 +                       break;
16466 +
16467 +#define au_fsctx_TF(name)                                        \
16468 +                       case Opt_##name:                          \
16469 +                               if (opt->tf)                      \
16470 +                                       AuLabel(name);            \
16471 +                               else                              \
16472 +                                       AuLabel(no##name);        \
16473 +                               break;
16474 +
16475 +               /* simple true/false flag */
16476 +               au_fsctx_TF(trunc_xino);
16477 +               au_fsctx_TF(trunc_xib);
16478 +               au_fsctx_TF(dirperm1);
16479 +               au_fsctx_TF(plink);
16480 +               au_fsctx_TF(shwh);
16481 +               au_fsctx_TF(dio);
16482 +               au_fsctx_TF(warn_perm);
16483 +               au_fsctx_TF(verbose);
16484 +               au_fsctx_TF(sum);
16485 +               au_fsctx_TF(dirren);
16486 +               au_fsctx_TF(acl);
16487 +#undef au_fsctx_TF
16488 +
16489 +               case Opt_trunc_xino_path:
16490 +                       fallthrough;
16491 +               case Opt_itrunc_xino:
16492 +                       u.xino_itrunc = &opt->xino_itrunc;
16493 +                       AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
16494 +                       break;
16495 +               case Opt_noxino:
16496 +                       AuLabel(noxino);
16497 +                       break;
16498 +
16499 +               case Opt_list_plink:
16500 +                       AuLabel(list_plink);
16501 +                       break;
16502 +               case Opt_udba:
16503 +                       AuDbg("udba %d, %s\n",
16504 +                                 opt->udba, au_optstr_udba(opt->udba));
16505 +                       break;
16506 +               case Opt_diropq_a:
16507 +                       AuLabel(diropq_a);
16508 +                       break;
16509 +               case Opt_diropq_w:
16510 +                       AuLabel(diropq_w);
16511 +                       break;
16512 +               case Opt_wsum:
16513 +                       AuLabel(wsum);
16514 +                       break;
16515 +               case Opt_wbr_create:
16516 +                       u.create = &opt->wbr_create;
16517 +                       AuDbg("create %d, %s\n", u.create->wbr_create,
16518 +                                 au_optstr_wbr_create(u.create->wbr_create));
16519 +                       switch (u.create->wbr_create) {
16520 +                       case AuWbrCreate_MFSV:
16521 +                               fallthrough;
16522 +                       case AuWbrCreate_PMFSV:
16523 +                               AuDbg("%d sec\n", u.create->mfs_second);
16524 +                               break;
16525 +                       case AuWbrCreate_MFSRR:
16526 +                               fallthrough;
16527 +                       case AuWbrCreate_TDMFS:
16528 +                               AuDbg("%llu watermark\n",
16529 +                                         u.create->mfsrr_watermark);
16530 +                               break;
16531 +                       case AuWbrCreate_MFSRRV:
16532 +                               fallthrough;
16533 +                       case AuWbrCreate_TDMFSV:
16534 +                               fallthrough;
16535 +                       case AuWbrCreate_PMFSRRV:
16536 +                               AuDbg("%llu watermark, %d sec\n",
16537 +                                         u.create->mfsrr_watermark,
16538 +                                         u.create->mfs_second);
16539 +                               break;
16540 +                       }
16541 +                       break;
16542 +               case Opt_wbr_copyup:
16543 +                       AuDbg("copyup %d, %s\n", opt->wbr_copyup,
16544 +                                 au_optstr_wbr_copyup(opt->wbr_copyup));
16545 +                       break;
16546 +               case Opt_fhsm_sec:
16547 +                       AuDbg("fhsm_sec %u\n", opt->fhsm_second);
16548 +                       break;
16549 +
16550 +               default:
16551 +                       AuDbg("type %d\n", opt->type);
16552 +                       BUG();
16553 +               }
16554 +               opt++;
16555 +       }
16556 +#endif
16557 +}
16558 +
16559 +/* ---------------------------------------------------------------------- */
16560 +
16561 +/*
16562 + * For conditionally compiled mount options.
16563 + * Instead of fsparam_flag_no(), use this macro to distinguish ignore_silent.
16564 + */
16565 +#define au_ignore_flag(name, action)           \
16566 +       fsparam_flag(name, action),             \
16567 +       fsparam_flag("no" name, Opt_ignore_silent)
16568 +
16569 +const struct fs_parameter_spec aufs_fsctx_paramspec[] = {
16570 +       fsparam_string("br", Opt_br),
16571 +
16572 +       /* "add=%d:%s" or "ins=%d:%s" */
16573 +       fsparam_string("add", Opt_add),
16574 +       fsparam_string("ins", Opt_add),
16575 +       fsparam_path("append", Opt_append),
16576 +       fsparam_path("prepend", Opt_prepend),
16577 +
16578 +       fsparam_path("del", Opt_del),
16579 +       /* fsparam_s32("idel", Opt_idel), */
16580 +       fsparam_path("mod", Opt_mod),
16581 +       /* fsparam_string("imod", Opt_imod), */
16582 +
16583 +       fsparam_s32("dirwh", Opt_dirwh),
16584 +
16585 +       fsparam_path("xino", Opt_xino),
16586 +       fsparam_flag("noxino", Opt_noxino),
16587 +       fsparam_flag_no("trunc_xino", Opt_trunc_xino),
16588 +       /* "trunc_xino_v=%d:%d" */
16589 +       /* fsparam_string("trunc_xino_v", Opt_trunc_xino_v), */
16590 +       fsparam_path("trunc_xino", Opt_trunc_xino_path),
16591 +       fsparam_s32("itrunc_xino", Opt_itrunc_xino),
16592 +       /* fsparam_path("zxino", Opt_zxino), */
16593 +       fsparam_flag_no("trunc_xib", Opt_trunc_xib),
16594 +
16595 +#ifdef CONFIG_PROC_FS
16596 +       fsparam_flag_no("plink", Opt_plink),
16597 +#else
16598 +       au_ignore_flag("plink", Opt_ignore),
16599 +#endif
16600 +
16601 +#ifdef CONFIG_AUFS_DEBUG
16602 +       fsparam_flag("list_plink", Opt_list_plink),
16603 +#endif
16604 +
16605 +       fsparam_string("udba", Opt_udba),
16606 +
16607 +       fsparam_flag_no("dio", Opt_dio),
16608 +
16609 +#ifdef CONFIG_AUFS_DIRREN
16610 +       fsparam_flag_no("dirren", Opt_dirren),
16611 +#else
16612 +       au_ignore_flag("dirren", Opt_ignore),
16613 +#endif
16614 +
16615 +#ifdef CONFIG_AUFS_FHSM
16616 +       fsparam_s32("fhsm_sec", Opt_fhsm_sec),
16617 +#else
16618 +       fsparam_s32("fhsm_sec", Opt_ignore),
16619 +#endif
16620 +
16621 +       /* always | a | whiteouted | w */
16622 +       fsparam_string("diropq", Opt_diropq),
16623 +
16624 +       fsparam_flag_no("warn_perm", Opt_warn_perm),
16625 +
16626 +#ifdef CONFIG_AUFS_SHWH
16627 +       fsparam_flag_no("shwh", Opt_shwh),
16628 +#else
16629 +       au_ignore_flag("shwh", Opt_err),
16630 +#endif
16631 +
16632 +       fsparam_flag_no("dirperm1", Opt_dirperm1),
16633 +
16634 +       fsparam_flag_no("verbose", Opt_verbose),
16635 +       fsparam_flag("v", Opt_verbose),
16636 +       fsparam_flag("quiet", Opt_noverbose),
16637 +       fsparam_flag("q", Opt_noverbose),
16638 +       /* user-space may handle this */
16639 +       fsparam_flag("silent", Opt_noverbose),
16640 +
16641 +       fsparam_flag_no("sum", Opt_sum),
16642 +       fsparam_flag("wsum", Opt_wsum),
16643 +
16644 +       fsparam_s32("rdcache", Opt_rdcache),
16645 +       /* "def" or s32 */
16646 +       fsparam_string("rdblk", Opt_rdblk),
16647 +       /* "def" or s32 */
16648 +       fsparam_string("rdhash", Opt_rdhash),
16649 +
16650 +       fsparam_string("create", Opt_wbr_create),
16651 +       fsparam_string("create_policy", Opt_wbr_create),
16652 +       fsparam_string("cpup", Opt_wbr_copyup),
16653 +       fsparam_string("copyup", Opt_wbr_copyup),
16654 +       fsparam_string("copyup_policy", Opt_wbr_copyup),
16655 +
16656 +       /* generic VFS flag */
16657 +#ifdef CONFIG_FS_POSIX_ACL
16658 +       fsparam_flag_no("acl", Opt_acl),
16659 +#else
16660 +       au_ignore_flag("acl", Opt_ignore),
16661 +#endif
16662 +
16663 +       /* internal use for the scripts */
16664 +       fsparam_string("si", Opt_ignore_silent),
16665 +
16666 +       /* obsoleted, keep them temporary */
16667 +       fsparam_flag("nodlgt", Opt_ignore_silent),
16668 +       fsparam_flag("clean_plink", Opt_ignore),
16669 +       fsparam_string("dirs", Opt_br),
16670 +       fsparam_u32("debug", Opt_ignore),
16671 +       /* "whiteout" or "all" */
16672 +       fsparam_string("delete", Opt_ignore),
16673 +       fsparam_string("imap", Opt_ignore),
16674 +
16675 +       /* temporary workaround, due to old mount(8)? */
16676 +       fsparam_flag("relatime", Opt_ignore_silent),
16677 +
16678 +       {}
16679 +};
16680 +
16681 +static int au_fsctx_parse_do_add(struct fs_context *fc, struct au_opt *opt,
16682 +                                char *brspec, size_t speclen,
16683 +                                aufs_bindex_t bindex)
16684 +{
16685 +       int err;
16686 +       char *p;
16687 +
16688 +       AuDbg("brspec %s\n", brspec);
16689 +
16690 +       err = -ENOMEM;
16691 +       if (!speclen)
16692 +               speclen = strlen(brspec);
16693 +       /* will be freed by au_fsctx_free() */
16694 +       p = kmemdup_nul(brspec, speclen, GFP_NOFS);
16695 +       if (unlikely(!p)) {
16696 +               errorfc(fc, "failed in %s", brspec);
16697 +               goto out;
16698 +       }
16699 +       err = au_opt_add(opt, p, fc->sb_flags, bindex);
16700 +
16701 +out:
16702 +       AuTraceErr(err);
16703 +       return err;
16704 +}
16705 +
16706 +static int au_fsctx_parse_br(struct fs_context *fc, char *brspec)
16707 +{
16708 +       int err;
16709 +       char *p;
16710 +       struct au_fsctx_opts *a = fc->fs_private;
16711 +       struct au_opt *opt = a->opt;
16712 +       aufs_bindex_t bindex = a->bindex;
16713 +
16714 +       AuDbg("brspec %s\n", brspec);
16715 +
16716 +       err = -EINVAL;
16717 +       while ((p = strsep(&brspec, ":")) && *p) {
16718 +               err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, bindex);
16719 +               AuTraceErr(err);
16720 +               if (unlikely(err))
16721 +                       break;
16722 +               bindex++;
16723 +               opt++;
16724 +               if (unlikely(opt > a->opt_tail)) {
16725 +                       err = -E2BIG;
16726 +                       bindex--;
16727 +                       opt--;
16728 +                       break;
16729 +               }
16730 +               opt->type = Opt_tail;
16731 +               a->skipped = 1;
16732 +       }
16733 +       a->bindex = bindex;
16734 +       a->opt = opt;
16735 +
16736 +       AuTraceErr(err);
16737 +       return err;
16738 +}
16739 +
16740 +static int au_fsctx_parse_add(struct fs_context *fc, char *addspec)
16741 +{
16742 +       int err, n;
16743 +       char *p;
16744 +       struct au_fsctx_opts *a = fc->fs_private;
16745 +       struct au_opt *opt = a->opt;
16746 +
16747 +       err = -EINVAL;
16748 +       p = strchr(addspec, ':');
16749 +       if (unlikely(!p)) {
16750 +               errorfc(fc, "bad arg in %s", addspec);
16751 +               goto out;
16752 +       }
16753 +       *p++ = '\0';
16754 +       err = kstrtoint(addspec, 0, &n);
16755 +       if (unlikely(err)) {
16756 +               errorfc(fc, "bad integer in %s", addspec);
16757 +               goto out;
16758 +       }
16759 +       AuDbg("n %d\n", n);
16760 +       err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, n);
16761 +
16762 +out:
16763 +       AuTraceErr(err);
16764 +       return err;
16765 +}
16766 +
16767 +static int au_fsctx_parse_del(struct fs_context *fc, struct au_opt_del *del,
16768 +                             struct fs_parameter *param)
16769 +{
16770 +       int err;
16771 +
16772 +       err = -ENOMEM;
16773 +       /* will be freed by au_fsctx_free() */
16774 +       del->pathname = kmemdup_nul(param->string, param->size, GFP_NOFS);
16775 +       if (unlikely(!del->pathname))
16776 +               goto out;
16777 +       AuDbg("del %s\n", del->pathname);
16778 +       err = vfsub_kern_path(del->pathname, AuOpt_LkupDirFlags, &del->h_path);
16779 +       if (unlikely(err))
16780 +               errorfc(fc, "lookup failed %s (%d)", del->pathname, err);
16781 +
16782 +out:
16783 +       AuTraceErr(err);
16784 +       return err;
16785 +}
16786 +
16787 +#if 0 /* reserved for future use */
16788 +static int au_fsctx_parse_idel(struct fs_context *fc, struct au_opt_del *del,
16789 +                              aufs_bindex_t bindex)
16790 +{
16791 +       int err;
16792 +       struct super_block *sb;
16793 +       struct dentry *root;
16794 +       struct au_fsctx_opts *a = fc->fs_private;
16795 +
16796 +       sb = a->sb;
16797 +       AuDebugOn(!sb);
16798 +
16799 +       err = -EINVAL;
16800 +       root = sb->s_root;
16801 +       aufs_read_lock(root, AuLock_FLUSH);
16802 +       if (bindex < 0 || au_sbbot(sb) < bindex) {
16803 +               errorfc(fc, "out of bounds, %d", bindex);
16804 +               goto out;
16805 +       }
16806 +
16807 +       err = 0;
16808 +       del->h_path.dentry = dget(au_h_dptr(root, bindex));
16809 +       del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
16810 +
16811 +out:
16812 +       aufs_read_unlock(root, !AuLock_IR);
16813 +       AuTraceErr(err);
16814 +       return err;
16815 +}
16816 +#endif
16817 +
16818 +static int au_fsctx_parse_mod(struct fs_context *fc, struct au_opt_mod *mod,
16819 +                             struct fs_parameter *param)
16820 +{
16821 +       int err;
16822 +       struct path path;
16823 +       char *p;
16824 +
16825 +       err = -ENOMEM;
16826 +       /* will be freed by au_fsctx_free() */
16827 +       mod->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16828 +       if (unlikely(!mod->path))
16829 +               goto out;
16830 +
16831 +       err = -EINVAL;
16832 +       p = strchr(mod->path, '=');
16833 +       if (unlikely(!p)) {
16834 +               errorfc(fc, "no permission %s", mod->path);
16835 +               goto out;
16836 +       }
16837 +
16838 +       *p++ = 0;
16839 +       err = vfsub_kern_path(mod->path, AuOpt_LkupDirFlags, &path);
16840 +       if (unlikely(err)) {
16841 +               errorfc(fc, "lookup failed %s (%d)", mod->path, err);
16842 +               goto out;
16843 +       }
16844 +
16845 +       mod->perm = au_br_perm_val(p);
16846 +       AuDbg("mod path %s, perm 0x%x, %s", mod->path, mod->perm, p);
16847 +       mod->h_root = dget(path.dentry);
16848 +       path_put(&path);
16849 +
16850 +out:
16851 +       AuTraceErr(err);
16852 +       return err;
16853 +}
16854 +
16855 +#if 0 /* reserved for future use */
16856 +static int au_fsctx_parse_imod(struct fs_context *fc, struct au_opt_mod *mod,
16857 +                              char *ibrspec)
16858 +{
16859 +       int err, n;
16860 +       char *p;
16861 +       struct super_block *sb;
16862 +       struct dentry *root;
16863 +       struct au_fsctx_opts *a = fc->fs_private;
16864 +
16865 +       sb = a->sb;
16866 +       AuDebugOn(!sb);
16867 +
16868 +       err = -EINVAL;
16869 +       p = strchr(ibrspec, ':');
16870 +       if (unlikely(!p)) {
16871 +               errorfc(fc, "no index, %s", ibrspec);
16872 +               goto out;
16873 +       }
16874 +       *p++ = '\0';
16875 +       err = kstrtoint(ibrspec, 0, &n);
16876 +       if (unlikely(err)) {
16877 +               errorfc(fc, "bad integer in %s", ibrspec);
16878 +               goto out;
16879 +       }
16880 +       AuDbg("n %d\n", n);
16881 +
16882 +       root = sb->s_root;
16883 +       aufs_read_lock(root, AuLock_FLUSH);
16884 +       if (n < 0 || au_sbbot(sb) < n) {
16885 +               errorfc(fc, "out of bounds, %d", bindex);
16886 +               goto out_root;
16887 +       }
16888 +
16889 +       err = 0;
16890 +       mod->perm = au_br_perm_val(p);
16891 +       AuDbg("mod path %s, perm 0x%x, %s\n",
16892 +             mod->path, mod->perm, p);
16893 +       mod->h_root = dget(au_h_dptr(root, bindex));
16894 +
16895 +out_root:
16896 +       aufs_read_unlock(root, !AuLock_IR);
16897 +out:
16898 +       AuTraceErr(err);
16899 +       return err;
16900 +}
16901 +#endif
16902 +
16903 +static int au_fsctx_parse_xino(struct fs_context *fc,
16904 +                              struct au_opt_xino *xino,
16905 +                              struct fs_parameter *param)
16906 +{
16907 +       int err;
16908 +       struct au_fsctx_opts *a = fc->fs_private;
16909 +
16910 +       err = -ENOMEM;
16911 +       /* will be freed by au_opts_free() */
16912 +       xino->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16913 +       if (unlikely(!xino->path))
16914 +               goto out;
16915 +       AuDbg("path %s\n", xino->path);
16916 +
16917 +       xino->file = au_xino_create(a->sb, xino->path, /*silent*/0,
16918 +                                   /*wbrtop*/0);
16919 +       err = PTR_ERR(xino->file);
16920 +       if (IS_ERR(xino->file)) {
16921 +               xino->file = NULL;
16922 +               goto out;
16923 +       }
16924 +
16925 +       err = 0;
16926 +       if (unlikely(a->sb && xino->file->f_path.dentry->d_sb == a->sb)) {
16927 +               err = -EINVAL;
16928 +               errorfc(fc, "%s must be outside", xino->path);
16929 +       }
16930 +
16931 +out:
16932 +       AuTraceErr(err);
16933 +       return err;
16934 +}
16935 +
16936 +static
16937 +int au_fsctx_parse_xino_itrunc_path(struct fs_context *fc,
16938 +                                   struct au_opt_xino_itrunc *xino_itrunc,
16939 +                                   char *pathname)
16940 +{
16941 +       int err;
16942 +       aufs_bindex_t bbot, bindex;
16943 +       struct path path;
16944 +       struct dentry *root;
16945 +       struct au_fsctx_opts *a = fc->fs_private;
16946 +
16947 +       AuDebugOn(!a->sb);
16948 +
16949 +       err = vfsub_kern_path(pathname, AuOpt_LkupDirFlags, &path);
16950 +       if (unlikely(err)) {
16951 +               errorfc(fc, "lookup failed %s (%d)", pathname, err);
16952 +               goto out;
16953 +       }
16954 +
16955 +       xino_itrunc->bindex = -1;
16956 +       root = a->sb->s_root;
16957 +       aufs_read_lock(root, AuLock_FLUSH);
16958 +       bbot = au_sbbot(a->sb);
16959 +       for (bindex = 0; bindex <= bbot; bindex++) {
16960 +               if (au_h_dptr(root, bindex) == path.dentry) {
16961 +                       xino_itrunc->bindex = bindex;
16962 +                       break;
16963 +               }
16964 +       }
16965 +       aufs_read_unlock(root, !AuLock_IR);
16966 +       path_put(&path);
16967 +
16968 +       if (unlikely(xino_itrunc->bindex < 0)) {
16969 +               err = -EINVAL;
16970 +               errorfc(fc, "no such branch %s", pathname);
16971 +       }
16972 +
16973 +out:
16974 +       AuTraceErr(err);
16975 +       return err;
16976 +}
16977 +
16978 +static int au_fsctx_parse_xino_itrunc(struct fs_context *fc,
16979 +                                     struct au_opt_xino_itrunc *xino_itrunc,
16980 +                                     unsigned int bindex)
16981 +{
16982 +       int err;
16983 +       aufs_bindex_t bbot;
16984 +       struct super_block *sb;
16985 +       struct au_fsctx_opts *a = fc->fs_private;
16986 +
16987 +       sb = a->sb;
16988 +       AuDebugOn(!sb);
16989 +
16990 +       err = 0;
16991 +       si_noflush_read_lock(sb);
16992 +       bbot = au_sbbot(sb);
16993 +       si_read_unlock(sb);
16994 +       if (bindex <= bbot)
16995 +               xino_itrunc->bindex = bindex;
16996 +       else {
16997 +               err = -EINVAL;
16998 +               errorfc(fc, "out of bounds, %u", bindex);
16999 +       }
17000 +
17001 +       AuTraceErr(err);
17002 +       return err;
17003 +}
17004 +
17005 +static int au_fsctx_parse_param(struct fs_context *fc, struct fs_parameter *param)
17006 +{
17007 +       int err, token;
17008 +       struct fs_parse_result result;
17009 +       struct au_fsctx_opts *a = fc->fs_private;
17010 +       struct au_opt *opt = a->opt;
17011 +
17012 +       AuDbg("fc %p, param {key %s, string %s}\n",
17013 +             fc, param->key, param->string);
17014 +       err = fs_parse(fc, aufs_fsctx_paramspec, param, &result);
17015 +       if (unlikely(err < 0))
17016 +               goto out;
17017 +       token = err;
17018 +       AuDbg("token %d, res{negated %d, uint64 %llu}\n",
17019 +             token, result.negated, result.uint_64);
17020 +
17021 +       err = -EINVAL;
17022 +       a->skipped = 0;
17023 +       switch (token) {
17024 +       case Opt_br:
17025 +               err = au_fsctx_parse_br(fc, param->string);
17026 +               break;
17027 +       case Opt_add:
17028 +               err = au_fsctx_parse_add(fc, param->string);
17029 +               break;
17030 +       case Opt_append:
17031 +               err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17032 +                                           /*dummy bindex*/1);
17033 +               break;
17034 +       case Opt_prepend:
17035 +               err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17036 +                                           /*bindex*/0);
17037 +               break;
17038 +
17039 +       case Opt_del:
17040 +               err = au_fsctx_parse_del(fc, &opt->del, param);
17041 +               break;
17042 +#if 0 /* reserved for future use */
17043 +       case Opt_idel:
17044 +               if (!a->sb) {
17045 +                       err = 0;
17046 +                       a->skipped = 1;
17047 +                       break;
17048 +               }
17049 +               del->pathname = "(indexed)";
17050 +               err = au_opts_parse_idel(fc, &opt->del, result.uint_32);
17051 +               break;
17052 +#endif
17053 +
17054 +       case Opt_mod:
17055 +               err = au_fsctx_parse_mod(fc, &opt->mod, param);
17056 +               break;
17057 +#ifdef IMOD /* reserved for future use */
17058 +       case Opt_imod:
17059 +               if (!a->sb) {
17060 +                       err = 0;
17061 +                       a->skipped = 1;
17062 +                       break;
17063 +               }
17064 +               u.mod->path = "(indexed)";
17065 +               err = au_opts_parse_imod(fc, &opt->mod, param->string);
17066 +               break;
17067 +#endif
17068 +
17069 +       case Opt_xino:
17070 +               err = au_fsctx_parse_xino(fc, &opt->xino, param);
17071 +               break;
17072 +       case Opt_trunc_xino_path:
17073 +               if (!a->sb) {
17074 +                       errorfc(fc, "no such branch %s", param->string);
17075 +                       break;
17076 +               }
17077 +               err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17078 +                                                     param->string);
17079 +               break;
17080 +#if 0
17081 +       case Opt_trunc_xino_v:
17082 +               if (!a->sb) {
17083 +                       err = 0;
17084 +                       a->skipped = 1;
17085 +                       break;
17086 +               }
17087 +               err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17088 +                                                     param->string);
17089 +               break;
17090 +#endif
17091 +       case Opt_itrunc_xino:
17092 +               if (!a->sb) {
17093 +                       errorfc(fc, "out of bounds %s", param->string);
17094 +                       break;
17095 +               }
17096 +               err = au_fsctx_parse_xino_itrunc(fc, &opt->xino_itrunc,
17097 +                                                result.int_32);
17098 +               break;
17099 +
17100 +       case Opt_dirwh:
17101 +               err = 0;
17102 +               opt->dirwh = result.int_32;
17103 +               break;
17104 +
17105 +       case Opt_rdcache:
17106 +               if (unlikely(result.int_32 > AUFS_RDCACHE_MAX)) {
17107 +                       errorfc(fc, "rdcache must be smaller than %d",
17108 +                               AUFS_RDCACHE_MAX);
17109 +                       break;
17110 +               }
17111 +               err = 0;
17112 +               opt->rdcache = result.int_32;
17113 +               break;
17114 +
17115 +       case Opt_rdblk:
17116 +               err = 0;
17117 +               opt->rdblk = AUFS_RDBLK_DEF;
17118 +               if (!strcmp(param->string, "def"))
17119 +                       break;
17120 +
17121 +               err = kstrtoint(param->string, 0, &result.int_32);
17122 +               if (unlikely(err)) {
17123 +                       errorfc(fc, "bad value in %s", param->key);
17124 +                       break;
17125 +               }
17126 +               err = -EINVAL;
17127 +               if (unlikely(result.int_32 < 0
17128 +                            || result.int_32 > KMALLOC_MAX_SIZE)) {
17129 +                       errorfc(fc, "bad value in %s", param->key);
17130 +                       break;
17131 +               }
17132 +               if (unlikely(result.int_32 && result.int_32 < NAME_MAX)) {
17133 +                       errorfc(fc, "rdblk must be larger than %d", NAME_MAX);
17134 +                       break;
17135 +               }
17136 +               err = 0;
17137 +               opt->rdblk = result.int_32;
17138 +               break;
17139 +
17140 +       case Opt_rdhash:
17141 +               err = 0;
17142 +               opt->rdhash = AUFS_RDHASH_DEF;
17143 +               if (!strcmp(param->string, "def"))
17144 +                       break;
17145 +
17146 +               err = kstrtoint(param->string, 0, &result.int_32);
17147 +               if (unlikely(err)) {
17148 +                       errorfc(fc, "bad value in %s", param->key);
17149 +                       break;
17150 +               }
17151 +               /* how about zero? */
17152 +               if (result.int_32 < 0
17153 +                   || result.int_32 * sizeof(struct hlist_head)
17154 +                   > KMALLOC_MAX_SIZE) {
17155 +                       err = -EINVAL;
17156 +                       errorfc(fc, "bad integer in %s", param->key);
17157 +                       break;
17158 +               }
17159 +               opt->rdhash = result.int_32;
17160 +               break;
17161 +
17162 +       case Opt_diropq:
17163 +               /*
17164 +                * As other options, fs/aufs/opts.c can handle these strings by
17165 +                * match_token().  But "diropq=" is deprecated now and will
17166 +                * never have other value.  So simple strcmp() is enough here.
17167 +                */
17168 +               if (!strcmp(param->string, "a") ||
17169 +                   !strcmp(param->string, "always")) {
17170 +                       err = 0;
17171 +                       opt->type = Opt_diropq_a;
17172 +               } else if (!strcmp(param->string, "w") ||
17173 +                          !strcmp(param->string, "whiteouted")) {
17174 +                       err = 0;
17175 +                       opt->type = Opt_diropq_w;
17176 +               } else
17177 +                       errorfc(fc, "unknown value %s", param->string);
17178 +               break;
17179 +
17180 +       case Opt_udba:
17181 +               opt->udba = au_udba_val(param->string);
17182 +               if (opt->udba >= 0)
17183 +                       err = 0;
17184 +               else
17185 +                       errorf(fc, "wrong value, %s", param->string);
17186 +               break;
17187 +
17188 +       case Opt_wbr_create:
17189 +               opt->wbr_create.wbr_create
17190 +                       = au_wbr_create_val(param->string, &opt->wbr_create);
17191 +               if (opt->wbr_create.wbr_create >= 0)
17192 +                       err = 0;
17193 +               else
17194 +                       errorf(fc, "wrong value, %s", param->key);
17195 +               break;
17196 +
17197 +       case Opt_wbr_copyup:
17198 +               opt->wbr_copyup = au_wbr_copyup_val(param->string);
17199 +               if (opt->wbr_copyup >= 0)
17200 +                       err = 0;
17201 +               else
17202 +                       errorfc(fc, "wrong value, %s", param->key);
17203 +               break;
17204 +
17205 +       case Opt_fhsm_sec:
17206 +               if (unlikely(result.int_32 < 0)) {
17207 +                       errorfc(fc, "bad integer in %s\n", param->key);
17208 +                       break;
17209 +               }
17210 +               err = 0;
17211 +               if (sysaufs_brs)
17212 +                       opt->fhsm_second = result.int_32;
17213 +               else
17214 +                       warnfc(fc, "ignored %s %s", param->key, param->string);
17215 +               break;
17216 +
17217 +       /* simple true/false flag */
17218 +#define au_fsctx_TF(name)                              \
17219 +               case Opt_##name:                        \
17220 +                       err = 0;                        \
17221 +                       opt->tf = !result.negated;      \
17222 +                       break
17223 +       au_fsctx_TF(trunc_xino);
17224 +       au_fsctx_TF(trunc_xib);
17225 +       au_fsctx_TF(dirperm1);
17226 +       au_fsctx_TF(plink);
17227 +       au_fsctx_TF(shwh);
17228 +       au_fsctx_TF(dio);
17229 +       au_fsctx_TF(warn_perm);
17230 +       au_fsctx_TF(verbose);
17231 +       au_fsctx_TF(sum);
17232 +       au_fsctx_TF(dirren);
17233 +       au_fsctx_TF(acl);
17234 +#undef au_fsctx_TF
17235 +
17236 +       case Opt_noverbose:
17237 +               err = 0;
17238 +               opt->type = Opt_verbose;
17239 +               opt->tf = false;
17240 +               break;
17241 +
17242 +       case Opt_noxino:
17243 +               fallthrough;
17244 +       case Opt_list_plink:
17245 +               fallthrough;
17246 +       case Opt_wsum:
17247 +               err = 0;
17248 +               break;
17249 +
17250 +       case Opt_ignore:
17251 +               warnfc(fc, "ignored %s", param->key);
17252 +               fallthrough;
17253 +       case Opt_ignore_silent:
17254 +               a->skipped = 1;
17255 +               err = 0;
17256 +               break;
17257 +       default:
17258 +               a->skipped = 1;
17259 +               err = -ENOPARAM;
17260 +               break;
17261 +       }
17262 +       if (unlikely(err))
17263 +               goto out;
17264 +       if (a->skipped)
17265 +               goto out;
17266 +
17267 +       switch (token) {
17268 +       case Opt_br:
17269 +               fallthrough;
17270 +       case Opt_noverbose:
17271 +               fallthrough;
17272 +       case Opt_diropq:
17273 +               break;
17274 +       default:
17275 +               opt->type = token;
17276 +               break;
17277 +       }
17278 +       opt++;
17279 +       if (unlikely(opt > a->opt_tail)) {
17280 +               err = -E2BIG;
17281 +               opt--;
17282 +       }
17283 +       opt->type = Opt_tail;
17284 +       a->opt = opt;
17285 +
17286 +out:
17287 +       return err;
17288 +}
17289 +
17290 +/*
17291 + * these options accept both 'name=val' and 'name:val' form.
17292 + * some accept optional '=' in its value.
17293 + * eg. br:/br1=rw:/br2=ro and br=/br1=rw:/br2=ro
17294 + */
17295 +static inline unsigned int is_colonopt(char *str)
17296 +{
17297 +#define do_test(name)                                  \
17298 +       if (!strncmp(str, name ":", sizeof(name)))      \
17299 +               return sizeof(name) - 1
17300 +       do_test("br");
17301 +       do_test("add");
17302 +       do_test("ins");
17303 +       do_test("append");
17304 +       do_test("prepend");
17305 +       do_test("del");
17306 +       /* do_test("idel"); */
17307 +       do_test("mod");
17308 +       /* do_test("imod"); */
17309 +#undef do_test
17310 +
17311 +       return 0;
17312 +}
17313 +
17314 +static int au_fsctx_parse_monolithic(struct fs_context *fc, void *data)
17315 +{
17316 +       int err;
17317 +       unsigned int u;
17318 +       char *str;
17319 +       struct au_fsctx_opts *a = fc->fs_private;
17320 +
17321 +       str = data;
17322 +       AuDbg("str %s\n", str);
17323 +       while (str) {
17324 +               u = is_colonopt(str);
17325 +               if (u)
17326 +                       str[u] = '=';
17327 +               str = strchr(str, ',');
17328 +               if (!str)
17329 +                       break;
17330 +               str++;
17331 +       }
17332 +       str = data;
17333 +       AuDbg("str %s\n", str);
17334 +
17335 +       err = generic_parse_monolithic(fc, str);
17336 +       AuTraceErr(err);
17337 +       au_fsctx_dump(&a->opts);
17338 +
17339 +       return err;
17340 +}
17341 +
17342 +/* ---------------------------------------------------------------------- */
17343 +
17344 +static void au_fsctx_opts_free(struct au_opts *opts)
17345 +{
17346 +       struct au_opt *opt;
17347 +
17348 +       opt = opts->opt;
17349 +       while (opt->type != Opt_tail) {
17350 +               switch (opt->type) {
17351 +               case Opt_add:
17352 +                       fallthrough;
17353 +               case Opt_append:
17354 +                       fallthrough;
17355 +               case Opt_prepend:
17356 +                       kfree(opt->add.pathname);
17357 +                       path_put(&opt->add.path);
17358 +                       break;
17359 +               case Opt_del:
17360 +                       kfree(opt->del.pathname);
17361 +                       fallthrough;
17362 +               case Opt_idel:
17363 +                       path_put(&opt->del.h_path);
17364 +                       break;
17365 +               case Opt_mod:
17366 +                       kfree(opt->mod.path);
17367 +                       fallthrough;
17368 +               case Opt_imod:
17369 +                       dput(opt->mod.h_root);
17370 +                       break;
17371 +               case Opt_xino:
17372 +                       kfree(opt->xino.path);
17373 +                       fput(opt->xino.file);
17374 +                       break;
17375 +               }
17376 +               opt++;
17377 +       }
17378 +}
17379 +
17380 +static void au_fsctx_free(struct fs_context *fc)
17381 +{
17382 +       struct au_fsctx_opts *a = fc->fs_private;
17383 +
17384 +       /* fs_type=%p, root=%pD */
17385 +       AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17386 +             fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17387 +
17388 +       kobject_put(&a->sbinfo->si_kobj);
17389 +       au_fsctx_opts_free(&a->opts);
17390 +       free_page((unsigned long)a->opts.opt);
17391 +       au_kfree_rcu(a);
17392 +}
17393 +
17394 +static const struct fs_context_operations au_fsctx_ops = {
17395 +       .free                   = au_fsctx_free,
17396 +       .parse_param            = au_fsctx_parse_param,
17397 +       .parse_monolithic       = au_fsctx_parse_monolithic,
17398 +       .get_tree               = au_fsctx_get_tree,
17399 +       .reconfigure            = au_fsctx_reconfigure
17400 +       /*
17401 +        * nfs4 requires ->dup()? No.
17402 +        * I don't know what is this ->dup() for.
17403 +        */
17404 +};
17405 +
17406 +int aufs_fsctx_init(struct fs_context *fc)
17407 +{
17408 +       int err;
17409 +       struct au_fsctx_opts *a;
17410 +
17411 +       /* fs_type=%p, root=%pD */
17412 +       AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17413 +             fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17414 +
17415 +       /* they will be freed by au_fsctx_free() */
17416 +       err = -ENOMEM;
17417 +       a = kzalloc(sizeof(*a), GFP_NOFS);
17418 +       if (unlikely(!a))
17419 +               goto out;
17420 +       a->bindex = 0;
17421 +       a->opts.opt = (void *)__get_free_page(GFP_NOFS);
17422 +       if (unlikely(!a->opts.opt))
17423 +               goto out_a;
17424 +       a->opt = a->opts.opt;
17425 +       a->opt->type = Opt_tail;
17426 +       a->opts.max_opt = PAGE_SIZE / sizeof(*a->opts.opt);
17427 +       a->opt_tail = a->opt + a->opts.max_opt - 1;
17428 +       a->opts.sb_flags = fc->sb_flags;
17429 +
17430 +       a->sb = NULL;
17431 +       if (fc->root) {
17432 +               AuDebugOn(fc->purpose != FS_CONTEXT_FOR_RECONFIGURE);
17433 +               a->opts.flags = AuOpts_REMOUNT;
17434 +               a->sb = fc->root->d_sb;
17435 +               a->sbinfo = au_sbi(a->sb);
17436 +               kobject_get(&a->sbinfo->si_kobj);
17437 +       } else {
17438 +               a->sbinfo = au_si_alloc(a->sb);
17439 +               AuDebugOn(!a->sbinfo);
17440 +               err = PTR_ERR(a->sbinfo);
17441 +               if (IS_ERR(a->sbinfo))
17442 +                       goto out_opt;
17443 +               au_rw_write_unlock(&a->sbinfo->si_rwsem);
17444 +       }
17445 +
17446 +       err = 0;
17447 +       fc->fs_private = a;
17448 +       fc->ops = &au_fsctx_ops;
17449 +       goto out; /* success */
17450 +
17451 +out_opt:
17452 +       free_page((unsigned long)a->opts.opt);
17453 +out_a:
17454 +       au_kfree_rcu(a);
17455 +out:
17456 +       AuTraceErr(err);
17457 +       return err;
17458 +}
17459 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
17460 --- /usr/share/empty/fs/aufs/fstype.h   1970-01-01 01:00:00.000000000 +0100
17461 +++ linux/fs/aufs/fstype.h      2022-11-05 23:02:18.965889284 +0100
17462 @@ -0,0 +1,401 @@
17463 +/* SPDX-License-Identifier: GPL-2.0 */
17464 +/*
17465 + * Copyright (C) 2005-2022 Junjiro R. Okajima
17466 + *
17467 + * This program is free software; you can redistribute it and/or modify
17468 + * it under the terms of the GNU General Public License as published by
17469 + * the Free Software Foundation; either version 2 of the License, or
17470 + * (at your option) any later version.
17471 + *
17472 + * This program is distributed in the hope that it will be useful,
17473 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17474 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17475 + * GNU General Public License for more details.
17476 + *
17477 + * You should have received a copy of the GNU General Public License
17478 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
17479 + */
17480 +
17481 +/*
17482 + * judging filesystem type
17483 + */
17484 +
17485 +#ifndef __AUFS_FSTYPE_H__
17486 +#define __AUFS_FSTYPE_H__
17487 +
17488 +#ifdef __KERNEL__
17489 +
17490 +#include <linux/fs.h>
17491 +#include <linux/magic.h>
17492 +#include <linux/nfs_fs.h>
17493 +#include <linux/romfs_fs.h>
17494 +
17495 +static inline int au_test_aufs(struct super_block *sb)
17496 +{
17497 +       return sb->s_magic == AUFS_SUPER_MAGIC;
17498 +}
17499 +
17500 +static inline const char *au_sbtype(struct super_block *sb)
17501 +{
17502 +       return sb->s_type->name;
17503 +}
17504 +
17505 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
17506 +{
17507 +#if IS_ENABLED(CONFIG_ISO9660_FS)
17508 +       return sb->s_magic == ISOFS_SUPER_MAGIC;
17509 +#else
17510 +       return 0;
17511 +#endif
17512 +}
17513 +
17514 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
17515 +{
17516 +#if IS_ENABLED(CONFIG_ROMFS_FS)
17517 +       return sb->s_magic == ROMFS_MAGIC;
17518 +#else
17519 +       return 0;
17520 +#endif
17521 +}
17522 +
17523 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
17524 +{
17525 +#if IS_ENABLED(CONFIG_CRAMFS)
17526 +       return sb->s_magic == CRAMFS_MAGIC;
17527 +#endif
17528 +       return 0;
17529 +}
17530 +
17531 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
17532 +{
17533 +#if IS_ENABLED(CONFIG_NFS_FS)
17534 +       return sb->s_magic == NFS_SUPER_MAGIC;
17535 +#else
17536 +       return 0;
17537 +#endif
17538 +}
17539 +
17540 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
17541 +{
17542 +#if IS_ENABLED(CONFIG_FUSE_FS)
17543 +       return sb->s_magic == FUSE_SUPER_MAGIC;
17544 +#else
17545 +       return 0;
17546 +#endif
17547 +}
17548 +
17549 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
17550 +{
17551 +#if IS_ENABLED(CONFIG_XFS_FS)
17552 +       return sb->s_magic == XFS_SB_MAGIC;
17553 +#else
17554 +       return 0;
17555 +#endif
17556 +}
17557 +
17558 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
17559 +{
17560 +#ifdef CONFIG_TMPFS
17561 +       return sb->s_magic == TMPFS_MAGIC;
17562 +#else
17563 +       return 0;
17564 +#endif
17565 +}
17566 +
17567 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
17568 +{
17569 +#if IS_ENABLED(CONFIG_ECRYPT_FS)
17570 +       return !strcmp(au_sbtype(sb), "ecryptfs");
17571 +#else
17572 +       return 0;
17573 +#endif
17574 +}
17575 +
17576 +static inline int au_test_ramfs(struct super_block *sb)
17577 +{
17578 +       return sb->s_magic == RAMFS_MAGIC;
17579 +}
17580 +
17581 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
17582 +{
17583 +#if IS_ENABLED(CONFIG_UBIFS_FS)
17584 +       return sb->s_magic == UBIFS_SUPER_MAGIC;
17585 +#else
17586 +       return 0;
17587 +#endif
17588 +}
17589 +
17590 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
17591 +{
17592 +#ifdef CONFIG_PROC_FS
17593 +       return sb->s_magic == PROC_SUPER_MAGIC;
17594 +#else
17595 +       return 0;
17596 +#endif
17597 +}
17598 +
17599 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
17600 +{
17601 +#ifdef CONFIG_SYSFS
17602 +       return sb->s_magic == SYSFS_MAGIC;
17603 +#else
17604 +       return 0;
17605 +#endif
17606 +}
17607 +
17608 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
17609 +{
17610 +#if IS_ENABLED(CONFIG_CONFIGFS_FS)
17611 +       return sb->s_magic == CONFIGFS_MAGIC;
17612 +#else
17613 +       return 0;
17614 +#endif
17615 +}
17616 +
17617 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
17618 +{
17619 +#if IS_ENABLED(CONFIG_MINIX_FS)
17620 +       return sb->s_magic == MINIX3_SUPER_MAGIC
17621 +               || sb->s_magic == MINIX2_SUPER_MAGIC
17622 +               || sb->s_magic == MINIX2_SUPER_MAGIC2
17623 +               || sb->s_magic == MINIX_SUPER_MAGIC
17624 +               || sb->s_magic == MINIX_SUPER_MAGIC2;
17625 +#else
17626 +       return 0;
17627 +#endif
17628 +}
17629 +
17630 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
17631 +{
17632 +#if IS_ENABLED(CONFIG_FAT_FS)
17633 +       return sb->s_magic == MSDOS_SUPER_MAGIC;
17634 +#else
17635 +       return 0;
17636 +#endif
17637 +}
17638 +
17639 +static inline int au_test_msdos(struct super_block *sb)
17640 +{
17641 +       return au_test_fat(sb);
17642 +}
17643 +
17644 +static inline int au_test_vfat(struct super_block *sb)
17645 +{
17646 +       return au_test_fat(sb);
17647 +}
17648 +
17649 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
17650 +{
17651 +#ifdef CONFIG_SECURITYFS
17652 +       return sb->s_magic == SECURITYFS_MAGIC;
17653 +#else
17654 +       return 0;
17655 +#endif
17656 +}
17657 +
17658 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
17659 +{
17660 +#if IS_ENABLED(CONFIG_SQUASHFS)
17661 +       return sb->s_magic == SQUASHFS_MAGIC;
17662 +#else
17663 +       return 0;
17664 +#endif
17665 +}
17666 +
17667 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
17668 +{
17669 +#if IS_ENABLED(CONFIG_BTRFS_FS)
17670 +       return sb->s_magic == BTRFS_SUPER_MAGIC;
17671 +#else
17672 +       return 0;
17673 +#endif
17674 +}
17675 +
17676 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
17677 +{
17678 +#if IS_ENABLED(CONFIG_XENFS)
17679 +       return sb->s_magic == XENFS_SUPER_MAGIC;
17680 +#else
17681 +       return 0;
17682 +#endif
17683 +}
17684 +
17685 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
17686 +{
17687 +#ifdef CONFIG_DEBUG_FS
17688 +       return sb->s_magic == DEBUGFS_MAGIC;
17689 +#else
17690 +       return 0;
17691 +#endif
17692 +}
17693 +
17694 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
17695 +{
17696 +#if IS_ENABLED(CONFIG_NILFS)
17697 +       return sb->s_magic == NILFS_SUPER_MAGIC;
17698 +#else
17699 +       return 0;
17700 +#endif
17701 +}
17702 +
17703 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
17704 +{
17705 +#if IS_ENABLED(CONFIG_HFSPLUS_FS)
17706 +       return sb->s_magic == HFSPLUS_SUPER_MAGIC;
17707 +#else
17708 +       return 0;
17709 +#endif
17710 +}
17711 +
17712 +/* ---------------------------------------------------------------------- */
17713 +/*
17714 + * they can't be an aufs branch.
17715 + */
17716 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
17717 +{
17718 +       return
17719 +#ifndef CONFIG_AUFS_BR_RAMFS
17720 +               au_test_ramfs(sb) ||
17721 +#endif
17722 +               au_test_procfs(sb)
17723 +               || au_test_sysfs(sb)
17724 +               || au_test_configfs(sb)
17725 +               || au_test_debugfs(sb)
17726 +               || au_test_securityfs(sb)
17727 +               || au_test_xenfs(sb)
17728 +               || au_test_ecryptfs(sb)
17729 +               /* || !strcmp(au_sbtype(sb), "unionfs") */
17730 +               || au_test_aufs(sb); /* will be supported in next version */
17731 +}
17732 +
17733 +static inline int au_test_fs_remote(struct super_block *sb)
17734 +{
17735 +       return !au_test_tmpfs(sb)
17736 +#ifdef CONFIG_AUFS_BR_RAMFS
17737 +               && !au_test_ramfs(sb)
17738 +#endif
17739 +               && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
17740 +}
17741 +
17742 +/* ---------------------------------------------------------------------- */
17743 +
17744 +/*
17745 + * Note: these functions (below) are created after reading ->getattr() in all
17746 + * filesystems under linux/fs. it means we have to do so in every update...
17747 + */
17748 +
17749 +/*
17750 + * some filesystems require getattr to refresh the inode attributes before
17751 + * referencing.
17752 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
17753 + * and leave the work for d_revalidate()
17754 + */
17755 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
17756 +{
17757 +       return au_test_nfs(sb)
17758 +               || au_test_fuse(sb)
17759 +               /* || au_test_btrfs(sb) */      /* untested */
17760 +               ;
17761 +}
17762 +
17763 +/*
17764 + * filesystems which don't maintain i_size or i_blocks.
17765 + */
17766 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
17767 +{
17768 +       return au_test_xfs(sb)
17769 +               || au_test_btrfs(sb)
17770 +               || au_test_ubifs(sb)
17771 +               || au_test_hfsplus(sb)  /* maintained, but incorrect */
17772 +               /* || au_test_minix(sb) */      /* untested */
17773 +               ;
17774 +}
17775 +
17776 +/*
17777 + * filesystems which don't store the correct value in some of their inode
17778 + * attributes.
17779 + */
17780 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
17781 +{
17782 +       return au_test_fs_bad_iattr_size(sb)
17783 +               || au_test_fat(sb)
17784 +               || au_test_msdos(sb)
17785 +               || au_test_vfat(sb);
17786 +}
17787 +
17788 +/* they don't check i_nlink in link(2) */
17789 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
17790 +{
17791 +       return au_test_tmpfs(sb)
17792 +#ifdef CONFIG_AUFS_BR_RAMFS
17793 +               || au_test_ramfs(sb)
17794 +#endif
17795 +               || au_test_ubifs(sb)
17796 +               || au_test_hfsplus(sb);
17797 +}
17798 +
17799 +/*
17800 + * filesystems which sets S_NOATIME and S_NOCMTIME.
17801 + */
17802 +static inline int au_test_fs_notime(struct super_block *sb)
17803 +{
17804 +       return au_test_nfs(sb)
17805 +               || au_test_fuse(sb)
17806 +               || au_test_ubifs(sb)
17807 +               ;
17808 +}
17809 +
17810 +/* temporary support for i#1 in cramfs */
17811 +static inline int au_test_fs_unique_ino(struct inode *inode)
17812 +{
17813 +       if (au_test_cramfs(inode->i_sb))
17814 +               return inode->i_ino != 1;
17815 +       return 1;
17816 +}
17817 +
17818 +/* ---------------------------------------------------------------------- */
17819 +
17820 +/*
17821 + * the filesystem where the xino files placed must support i/o after unlink and
17822 + * maintain i_size and i_blocks.
17823 + */
17824 +static inline int au_test_fs_bad_xino(struct super_block *sb)
17825 +{
17826 +       return au_test_fs_remote(sb)
17827 +               || au_test_fs_bad_iattr_size(sb)
17828 +               /* don't want unnecessary work for xino */
17829 +               || au_test_aufs(sb)
17830 +               || au_test_ecryptfs(sb)
17831 +               || au_test_nilfs(sb);
17832 +}
17833 +
17834 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
17835 +{
17836 +       return au_test_tmpfs(sb)
17837 +               || au_test_ramfs(sb);
17838 +}
17839 +
17840 +/*
17841 + * test if the @sb is real-readonly.
17842 + */
17843 +static inline int au_test_fs_rr(struct super_block *sb)
17844 +{
17845 +       return au_test_squashfs(sb)
17846 +               || au_test_iso9660(sb)
17847 +               || au_test_cramfs(sb)
17848 +               || au_test_romfs(sb);
17849 +}
17850 +
17851 +/*
17852 + * test if the @inode is nfs with 'noacl' option
17853 + * NFS always sets SB_POSIXACL regardless its mount option 'noacl.'
17854 + */
17855 +static inline int au_test_nfs_noacl(struct inode *inode)
17856 +{
17857 +       return au_test_nfs(inode->i_sb)
17858 +               /* && IS_POSIXACL(inode) */
17859 +               && !nfs_server_capable(inode, NFS_CAP_ACLS);
17860 +}
17861 +
17862 +#endif /* __KERNEL__ */
17863 +#endif /* __AUFS_FSTYPE_H__ */
17864 diff -urN /usr/share/empty/fs/aufs/hbl.h linux/fs/aufs/hbl.h
17865 --- /usr/share/empty/fs/aufs/hbl.h      1970-01-01 01:00:00.000000000 +0100
17866 +++ linux/fs/aufs/hbl.h 2022-11-05 23:02:18.965889284 +0100
17867 @@ -0,0 +1,65 @@
17868 +/* SPDX-License-Identifier: GPL-2.0 */
17869 +/*
17870 + * Copyright (C) 2017-2022 Junjiro R. Okajima
17871 + *
17872 + * This program is free software; you can redistribute it and/or modify
17873 + * it under the terms of the GNU General Public License as published by
17874 + * the Free Software Foundation; either version 2 of the License, or
17875 + * (at your option) any later version.
17876 + *
17877 + * This program is distributed in the hope that it will be useful,
17878 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17879 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17880 + * GNU General Public License for more details.
17881 + *
17882 + * You should have received a copy of the GNU General Public License
17883 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
17884 + */
17885 +
17886 +/*
17887 + * helpers for hlist_bl.h
17888 + */
17889 +
17890 +#ifndef __AUFS_HBL_H__
17891 +#define __AUFS_HBL_H__
17892 +
17893 +#ifdef __KERNEL__
17894 +
17895 +#include <linux/list_bl.h>
17896 +
17897 +static inline void au_hbl_add(struct hlist_bl_node *node,
17898 +                             struct hlist_bl_head *hbl)
17899 +{
17900 +       hlist_bl_lock(hbl);
17901 +       hlist_bl_add_head(node, hbl);
17902 +       hlist_bl_unlock(hbl);
17903 +}
17904 +
17905 +static inline void au_hbl_del(struct hlist_bl_node *node,
17906 +                             struct hlist_bl_head *hbl)
17907 +{
17908 +       hlist_bl_lock(hbl);
17909 +       hlist_bl_del(node);
17910 +       hlist_bl_unlock(hbl);
17911 +}
17912 +
17913 +#define au_hbl_for_each(pos, head)                                     \
17914 +       for (pos = hlist_bl_first(head);                                \
17915 +            pos;                                                       \
17916 +            pos = pos->next)
17917 +
17918 +static inline unsigned long au_hbl_count(struct hlist_bl_head *hbl)
17919 +{
17920 +       unsigned long cnt;
17921 +       struct hlist_bl_node *pos;
17922 +
17923 +       cnt = 0;
17924 +       hlist_bl_lock(hbl);
17925 +       au_hbl_for_each(pos, hbl)
17926 +               cnt++;
17927 +       hlist_bl_unlock(hbl);
17928 +       return cnt;
17929 +}
17930 +
17931 +#endif /* __KERNEL__ */
17932 +#endif /* __AUFS_HBL_H__ */
17933 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
17934 --- /usr/share/empty/fs/aufs/hfsnotify.c        1970-01-01 01:00:00.000000000 +0100
17935 +++ linux/fs/aufs/hfsnotify.c   2022-11-05 23:02:18.965889284 +0100
17936 @@ -0,0 +1,290 @@
17937 +// SPDX-License-Identifier: GPL-2.0
17938 +/*
17939 + * Copyright (C) 2005-2022 Junjiro R. Okajima
17940 + *
17941 + * This program is free software; you can redistribute it and/or modify
17942 + * it under the terms of the GNU General Public License as published by
17943 + * the Free Software Foundation; either version 2 of the License, or
17944 + * (at your option) any later version.
17945 + *
17946 + * This program is distributed in the hope that it will be useful,
17947 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17948 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17949 + * GNU General Public License for more details.
17950 + *
17951 + * You should have received a copy of the GNU General Public License
17952 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
17953 + */
17954 +
17955 +/*
17956 + * fsnotify for the lower directories
17957 + */
17958 +
17959 +#include "aufs.h"
17960 +
17961 +/* FS_IN_IGNORED is unnecessary */
17962 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
17963 +                                | FS_CREATE | FS_EVENT_ON_CHILD);
17964 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
17965 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
17966 +
17967 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
17968 +{
17969 +       struct au_hnotify *hn = container_of(mark, struct au_hnotify,
17970 +                                            hn_mark);
17971 +       /* AuDbg("here\n"); */
17972 +       au_cache_free_hnotify(hn);
17973 +       smp_mb__before_atomic(); /* for atomic64_dec */
17974 +       if (atomic64_dec_and_test(&au_hfsn_ifree))
17975 +               wake_up(&au_hfsn_wq);
17976 +}
17977 +
17978 +static int au_hfsn_alloc(struct au_hinode *hinode)
17979 +{
17980 +       int err;
17981 +       struct au_hnotify *hn;
17982 +       struct super_block *sb;
17983 +       struct au_branch *br;
17984 +       struct fsnotify_mark *mark;
17985 +       aufs_bindex_t bindex;
17986 +
17987 +       hn = hinode->hi_notify;
17988 +       sb = hn->hn_aufs_inode->i_sb;
17989 +       bindex = au_br_index(sb, hinode->hi_id);
17990 +       br = au_sbr(sb, bindex);
17991 +       AuDebugOn(!br->br_hfsn);
17992 +
17993 +       mark = &hn->hn_mark;
17994 +       fsnotify_init_mark(mark, br->br_hfsn->hfsn_group);
17995 +       mark->mask = AuHfsnMask;
17996 +       /*
17997 +        * by udba rename or rmdir, aufs assign a new inode to the known
17998 +        * h_inode, so specify 1 to allow dups.
17999 +        */
18000 +       lockdep_off();
18001 +       err = fsnotify_add_inode_mark(mark, hinode->hi_inode, /*allow_dups*/1);
18002 +       lockdep_on();
18003 +
18004 +       return err;
18005 +}
18006 +
18007 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
18008 +{
18009 +       struct fsnotify_mark *mark;
18010 +       unsigned long long ull;
18011 +       struct fsnotify_group *group;
18012 +
18013 +       ull = atomic64_inc_return(&au_hfsn_ifree);
18014 +       BUG_ON(!ull);
18015 +
18016 +       mark = &hn->hn_mark;
18017 +       spin_lock(&mark->lock);
18018 +       group = mark->group;
18019 +       fsnotify_get_group(group);
18020 +       spin_unlock(&mark->lock);
18021 +       lockdep_off();
18022 +       fsnotify_destroy_mark(mark, group);
18023 +       fsnotify_put_mark(mark);
18024 +       fsnotify_put_group(group);
18025 +       lockdep_on();
18026 +
18027 +       /* free hn by myself */
18028 +       return 0;
18029 +}
18030 +
18031 +/* ---------------------------------------------------------------------- */
18032 +
18033 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
18034 +{
18035 +       struct fsnotify_mark *mark;
18036 +
18037 +       mark = &hinode->hi_notify->hn_mark;
18038 +       spin_lock(&mark->lock);
18039 +       if (do_set) {
18040 +               AuDebugOn(mark->mask & AuHfsnMask);
18041 +               mark->mask |= AuHfsnMask;
18042 +       } else {
18043 +               AuDebugOn(!(mark->mask & AuHfsnMask));
18044 +               mark->mask &= ~AuHfsnMask;
18045 +       }
18046 +       spin_unlock(&mark->lock);
18047 +       /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
18048 +}
18049 +
18050 +/* ---------------------------------------------------------------------- */
18051 +
18052 +/* #define AuDbgHnotify */
18053 +#ifdef AuDbgHnotify
18054 +static char *au_hfsn_name(u32 mask)
18055 +{
18056 +#ifdef CONFIG_AUFS_DEBUG
18057 +#define test_ret(flag)                         \
18058 +       do {                                    \
18059 +               if (mask & flag)                \
18060 +                       return #flag;           \
18061 +       } while (0)
18062 +       test_ret(FS_ACCESS);
18063 +       test_ret(FS_MODIFY);
18064 +       test_ret(FS_ATTRIB);
18065 +       test_ret(FS_CLOSE_WRITE);
18066 +       test_ret(FS_CLOSE_NOWRITE);
18067 +       test_ret(FS_OPEN);
18068 +       test_ret(FS_MOVED_FROM);
18069 +       test_ret(FS_MOVED_TO);
18070 +       test_ret(FS_CREATE);
18071 +       test_ret(FS_DELETE);
18072 +       test_ret(FS_DELETE_SELF);
18073 +       test_ret(FS_MOVE_SELF);
18074 +       test_ret(FS_UNMOUNT);
18075 +       test_ret(FS_Q_OVERFLOW);
18076 +       test_ret(FS_IN_IGNORED);
18077 +       test_ret(FS_ISDIR);
18078 +       test_ret(FS_IN_ONESHOT);
18079 +       test_ret(FS_EVENT_ON_CHILD);
18080 +       return "";
18081 +#undef test_ret
18082 +#else
18083 +       return "??";
18084 +#endif
18085 +}
18086 +#endif
18087 +
18088 +/* ---------------------------------------------------------------------- */
18089 +
18090 +static void au_hfsn_free_group(struct fsnotify_group *group)
18091 +{
18092 +       struct au_br_hfsnotify *hfsn = group->private;
18093 +
18094 +       /* AuDbg("here\n"); */
18095 +       au_kfree_try_rcu(hfsn);
18096 +}
18097 +
18098 +static int au_hfsn_handle_event(struct fsnotify_group *group,
18099 +                               u32 mask, const void *data, int data_type,
18100 +                               struct inode *dir,
18101 +                               const struct qstr *file_name, u32 cookie,
18102 +                               struct fsnotify_iter_info *iter_info)
18103 +{
18104 +       int err;
18105 +       struct au_hnotify *hnotify;
18106 +       struct inode *h_dir, *h_inode;
18107 +       struct fsnotify_mark *inode_mark;
18108 +
18109 +       AuDebugOn(!(data_type == FSNOTIFY_EVENT_INODE
18110 +                   || data_type == FSNOTIFY_EVENT_DENTRY));
18111 +
18112 +       err = 0;
18113 +       /* if FS_UNMOUNT happens, there must be another bug */
18114 +       AuDebugOn(mask & FS_UNMOUNT);
18115 +       if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
18116 +               goto out;
18117 +
18118 +       h_dir = dir;
18119 +       h_inode = NULL;
18120 +#ifdef AuDbgHnotify
18121 +       au_debug_on();
18122 +       if (1 || file_name.len != sizeof(AUFS_XINO_FNAME) - 1
18123 +           || strncmp(file_name.name, AUFS_XINO_FNAME, file_name.len)) {
18124 +               AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
18125 +                     h_dir->i_ino, mask, au_hfsn_name(mask),
18126 +                     AuLNPair(file_name), h_inode ? h_inode->i_ino : 0);
18127 +               /* WARN_ON(1); */
18128 +       }
18129 +       au_debug_off();
18130 +#endif
18131 +
18132 +       inode_mark = fsnotify_iter_inode_mark(iter_info);
18133 +       AuDebugOn(!inode_mark);
18134 +       hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
18135 +       err = au_hnotify(h_dir, hnotify, mask, file_name, h_inode);
18136 +
18137 +out:
18138 +       return err;
18139 +}
18140 +
18141 +static struct fsnotify_ops au_hfsn_ops = {
18142 +       .handle_event           = au_hfsn_handle_event,
18143 +       .free_group_priv        = au_hfsn_free_group,
18144 +       .free_mark              = au_hfsn_free_mark
18145 +};
18146 +
18147 +/* ---------------------------------------------------------------------- */
18148 +
18149 +static void au_hfsn_fin_br(struct au_branch *br)
18150 +{
18151 +       struct au_br_hfsnotify *hfsn;
18152 +
18153 +       hfsn = br->br_hfsn;
18154 +       if (hfsn) {
18155 +               lockdep_off();
18156 +               fsnotify_put_group(hfsn->hfsn_group);
18157 +               lockdep_on();
18158 +       }
18159 +}
18160 +
18161 +static int au_hfsn_init_br(struct au_branch *br, int perm)
18162 +{
18163 +       int err;
18164 +       struct fsnotify_group *group;
18165 +       struct au_br_hfsnotify *hfsn;
18166 +
18167 +       err = 0;
18168 +       br->br_hfsn = NULL;
18169 +       if (!au_br_hnotifyable(perm))
18170 +               goto out;
18171 +
18172 +       err = -ENOMEM;
18173 +       hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
18174 +       if (unlikely(!hfsn))
18175 +               goto out;
18176 +
18177 +       err = 0;
18178 +       group = fsnotify_alloc_group(&au_hfsn_ops,
18179 +                                    /*flags - not for userspace*/0);
18180 +       if (IS_ERR(group)) {
18181 +               err = PTR_ERR(group);
18182 +               pr_err("fsnotify_alloc_group() failed, %d\n", err);
18183 +               goto out_hfsn;
18184 +       }
18185 +
18186 +       group->private = hfsn;
18187 +       hfsn->hfsn_group = group;
18188 +       br->br_hfsn = hfsn;
18189 +       goto out; /* success */
18190 +
18191 +out_hfsn:
18192 +       au_kfree_try_rcu(hfsn);
18193 +out:
18194 +       return err;
18195 +}
18196 +
18197 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
18198 +{
18199 +       int err;
18200 +
18201 +       err = 0;
18202 +       if (!br->br_hfsn)
18203 +               err = au_hfsn_init_br(br, perm);
18204 +
18205 +       return err;
18206 +}
18207 +
18208 +/* ---------------------------------------------------------------------- */
18209 +
18210 +static void au_hfsn_fin(void)
18211 +{
18212 +       AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
18213 +       wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
18214 +}
18215 +
18216 +const struct au_hnotify_op au_hnotify_op = {
18217 +       .ctl            = au_hfsn_ctl,
18218 +       .alloc          = au_hfsn_alloc,
18219 +       .free           = au_hfsn_free,
18220 +
18221 +       .fin            = au_hfsn_fin,
18222 +
18223 +       .reset_br       = au_hfsn_reset_br,
18224 +       .fin_br         = au_hfsn_fin_br,
18225 +       .init_br        = au_hfsn_init_br
18226 +};
18227 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
18228 --- /usr/share/empty/fs/aufs/hfsplus.c  1970-01-01 01:00:00.000000000 +0100
18229 +++ linux/fs/aufs/hfsplus.c     2022-11-05 23:02:18.965889284 +0100
18230 @@ -0,0 +1,60 @@
18231 +// SPDX-License-Identifier: GPL-2.0
18232 +/*
18233 + * Copyright (C) 2010-2022 Junjiro R. Okajima
18234 + *
18235 + * This program is free software; you can redistribute it and/or modify
18236 + * it under the terms of the GNU General Public License as published by
18237 + * the Free Software Foundation; either version 2 of the License, or
18238 + * (at your option) any later version.
18239 + *
18240 + * This program is distributed in the hope that it will be useful,
18241 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18242 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18243 + * GNU General Public License for more details.
18244 + *
18245 + * You should have received a copy of the GNU General Public License
18246 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18247 + */
18248 +
18249 +/*
18250 + * special support for filesystems which acquires an inode mutex
18251 + * at final closing a file, eg, hfsplus.
18252 + *
18253 + * This trick is very simple and stupid, just to open the file before really
18254 + * necessary open to tell hfsplus that this is not the final closing.
18255 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
18256 + * and au_h_open_post() after releasing it.
18257 + */
18258 +
18259 +#include "aufs.h"
18260 +
18261 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
18262 +                          int force_wr)
18263 +{
18264 +       struct file *h_file;
18265 +       struct dentry *h_dentry;
18266 +
18267 +       h_dentry = au_h_dptr(dentry, bindex);
18268 +       AuDebugOn(!h_dentry);
18269 +       AuDebugOn(d_is_negative(h_dentry));
18270 +
18271 +       h_file = NULL;
18272 +       if (au_test_hfsplus(h_dentry->d_sb)
18273 +           && d_is_reg(h_dentry))
18274 +               h_file = au_h_open(dentry, bindex,
18275 +                                  O_RDONLY | O_NOATIME | O_LARGEFILE,
18276 +                                  /*file*/NULL, force_wr);
18277 +       return h_file;
18278 +}
18279 +
18280 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
18281 +                   struct file *h_file)
18282 +{
18283 +       struct au_branch *br;
18284 +
18285 +       if (h_file) {
18286 +               fput(h_file);
18287 +               br = au_sbr(dentry->d_sb, bindex);
18288 +               au_lcnt_dec(&br->br_nfiles);
18289 +       }
18290 +}
18291 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
18292 --- /usr/share/empty/fs/aufs/hnotify.c  1970-01-01 01:00:00.000000000 +0100
18293 +++ linux/fs/aufs/hnotify.c     2022-11-05 23:02:18.965889284 +0100
18294 @@ -0,0 +1,715 @@
18295 +// SPDX-License-Identifier: GPL-2.0
18296 +/*
18297 + * Copyright (C) 2005-2022 Junjiro R. Okajima
18298 + *
18299 + * This program is free software; you can redistribute it and/or modify
18300 + * it under the terms of the GNU General Public License as published by
18301 + * the Free Software Foundation; either version 2 of the License, or
18302 + * (at your option) any later version.
18303 + *
18304 + * This program is distributed in the hope that it will be useful,
18305 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18306 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18307 + * GNU General Public License for more details.
18308 + *
18309 + * You should have received a copy of the GNU General Public License
18310 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18311 + */
18312 +
18313 +/*
18314 + * abstraction to notify the direct changes on lower directories
18315 + */
18316 +
18317 +/* #include <linux/iversion.h> */
18318 +#include "aufs.h"
18319 +
18320 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
18321 +{
18322 +       int err;
18323 +       struct au_hnotify *hn;
18324 +
18325 +       err = -ENOMEM;
18326 +       hn = au_cache_alloc_hnotify();
18327 +       if (hn) {
18328 +               hn->hn_aufs_inode = inode;
18329 +               hinode->hi_notify = hn;
18330 +               err = au_hnotify_op.alloc(hinode);
18331 +               AuTraceErr(err);
18332 +               if (unlikely(err)) {
18333 +                       hinode->hi_notify = NULL;
18334 +                       au_cache_free_hnotify(hn);
18335 +                       /*
18336 +                        * The upper dir was removed by udba, but the same named
18337 +                        * dir left. In this case, aufs assigns a new inode
18338 +                        * number and set the monitor again.
18339 +                        * For the lower dir, the old monitor is still left.
18340 +                        */
18341 +                       if (err == -EEXIST)
18342 +                               err = 0;
18343 +               }
18344 +       }
18345 +
18346 +       AuTraceErr(err);
18347 +       return err;
18348 +}
18349 +
18350 +void au_hn_free(struct au_hinode *hinode)
18351 +{
18352 +       struct au_hnotify *hn;
18353 +
18354 +       hn = hinode->hi_notify;
18355 +       if (hn) {
18356 +               hinode->hi_notify = NULL;
18357 +               if (au_hnotify_op.free(hinode, hn))
18358 +                       au_cache_free_hnotify(hn);
18359 +       }
18360 +}
18361 +
18362 +/* ---------------------------------------------------------------------- */
18363 +
18364 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
18365 +{
18366 +       if (hinode->hi_notify)
18367 +               au_hnotify_op.ctl(hinode, do_set);
18368 +}
18369 +
18370 +void au_hn_reset(struct inode *inode, unsigned int flags)
18371 +{
18372 +       aufs_bindex_t bindex, bbot;
18373 +       struct inode *hi;
18374 +       struct dentry *iwhdentry;
18375 +
18376 +       bbot = au_ibbot(inode);
18377 +       for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
18378 +               hi = au_h_iptr(inode, bindex);
18379 +               if (!hi)
18380 +                       continue;
18381 +
18382 +               /* inode_lock_nested(hi, AuLsc_I_CHILD); */
18383 +               iwhdentry = au_hi_wh(inode, bindex);
18384 +               if (iwhdentry)
18385 +                       dget(iwhdentry);
18386 +               au_igrab(hi);
18387 +               au_set_h_iptr(inode, bindex, NULL, 0);
18388 +               au_set_h_iptr(inode, bindex, au_igrab(hi),
18389 +                             flags & ~AuHi_XINO);
18390 +               iput(hi);
18391 +               dput(iwhdentry);
18392 +               /* inode_unlock(hi); */
18393 +       }
18394 +}
18395 +
18396 +/* ---------------------------------------------------------------------- */
18397 +
18398 +static int hn_xino(struct inode *inode, struct inode *h_inode)
18399 +{
18400 +       int err;
18401 +       aufs_bindex_t bindex, bbot, bfound, btop;
18402 +       struct inode *h_i;
18403 +
18404 +       err = 0;
18405 +       if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18406 +               pr_warn("branch root dir was changed\n");
18407 +               goto out;
18408 +       }
18409 +
18410 +       bfound = -1;
18411 +       bbot = au_ibbot(inode);
18412 +       btop = au_ibtop(inode);
18413 +#if 0 /* reserved for future use */
18414 +       if (bindex == bbot) {
18415 +               /* keep this ino in rename case */
18416 +               goto out;
18417 +       }
18418 +#endif
18419 +       for (bindex = btop; bindex <= bbot; bindex++)
18420 +               if (au_h_iptr(inode, bindex) == h_inode) {
18421 +                       bfound = bindex;
18422 +                       break;
18423 +               }
18424 +       if (bfound < 0)
18425 +               goto out;
18426 +
18427 +       for (bindex = btop; bindex <= bbot; bindex++) {
18428 +               h_i = au_h_iptr(inode, bindex);
18429 +               if (!h_i)
18430 +                       continue;
18431 +
18432 +               err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
18433 +               /* ignore this error */
18434 +               /* bad action? */
18435 +       }
18436 +
18437 +       /* children inode number will be broken */
18438 +
18439 +out:
18440 +       AuTraceErr(err);
18441 +       return err;
18442 +}
18443 +
18444 +static int hn_gen_tree(struct dentry *dentry)
18445 +{
18446 +       int err, i, j, ndentry;
18447 +       struct au_dcsub_pages dpages;
18448 +       struct au_dpage *dpage;
18449 +       struct dentry **dentries;
18450 +
18451 +       err = au_dpages_init(&dpages, GFP_NOFS);
18452 +       if (unlikely(err))
18453 +               goto out;
18454 +       err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
18455 +       if (unlikely(err))
18456 +               goto out_dpages;
18457 +
18458 +       for (i = 0; i < dpages.ndpage; i++) {
18459 +               dpage = dpages.dpages + i;
18460 +               dentries = dpage->dentries;
18461 +               ndentry = dpage->ndentry;
18462 +               for (j = 0; j < ndentry; j++) {
18463 +                       struct dentry *d;
18464 +
18465 +                       d = dentries[j];
18466 +                       if (IS_ROOT(d))
18467 +                               continue;
18468 +
18469 +                       au_digen_dec(d);
18470 +                       if (d_really_is_positive(d))
18471 +                               /* todo: reset children xino?
18472 +                                  cached children only? */
18473 +                               au_iigen_dec(d_inode(d));
18474 +               }
18475 +       }
18476 +
18477 +out_dpages:
18478 +       au_dpages_free(&dpages);
18479 +out:
18480 +       return err;
18481 +}
18482 +
18483 +/*
18484 + * return 0 if processed.
18485 + */
18486 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
18487 +                          const unsigned int isdir)
18488 +{
18489 +       int err;
18490 +       struct dentry *d;
18491 +       struct qstr *dname;
18492 +
18493 +       err = 1;
18494 +       if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18495 +               pr_warn("branch root dir was changed\n");
18496 +               err = 0;
18497 +               goto out;
18498 +       }
18499 +
18500 +       if (!isdir) {
18501 +               AuDebugOn(!name);
18502 +               au_iigen_dec(inode);
18503 +               spin_lock(&inode->i_lock);
18504 +               hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
18505 +                       spin_lock(&d->d_lock);
18506 +                       dname = &d->d_name;
18507 +                       if (dname->len != nlen
18508 +                           && memcmp(dname->name, name, nlen)) {
18509 +                               spin_unlock(&d->d_lock);
18510 +                               continue;
18511 +                       }
18512 +                       err = 0;
18513 +                       au_digen_dec(d);
18514 +                       spin_unlock(&d->d_lock);
18515 +                       break;
18516 +               }
18517 +               spin_unlock(&inode->i_lock);
18518 +       } else {
18519 +               au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
18520 +               d = d_find_any_alias(inode);
18521 +               if (!d) {
18522 +                       au_iigen_dec(inode);
18523 +                       goto out;
18524 +               }
18525 +
18526 +               spin_lock(&d->d_lock);
18527 +               dname = &d->d_name;
18528 +               if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
18529 +                       spin_unlock(&d->d_lock);
18530 +                       err = hn_gen_tree(d);
18531 +                       spin_lock(&d->d_lock);
18532 +               }
18533 +               spin_unlock(&d->d_lock);
18534 +               dput(d);
18535 +       }
18536 +
18537 +out:
18538 +       AuTraceErr(err);
18539 +       return err;
18540 +}
18541 +
18542 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
18543 +{
18544 +       int err;
18545 +
18546 +       if (IS_ROOT(dentry)) {
18547 +               pr_warn("branch root dir was changed\n");
18548 +               return 0;
18549 +       }
18550 +
18551 +       err = 0;
18552 +       if (!isdir) {
18553 +               au_digen_dec(dentry);
18554 +               if (d_really_is_positive(dentry))
18555 +                       au_iigen_dec(d_inode(dentry));
18556 +       } else {
18557 +               au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
18558 +               if (d_really_is_positive(dentry))
18559 +                       err = hn_gen_tree(dentry);
18560 +       }
18561 +
18562 +       AuTraceErr(err);
18563 +       return err;
18564 +}
18565 +
18566 +/* ---------------------------------------------------------------------- */
18567 +
18568 +/* hnotify job flags */
18569 +#define AuHnJob_XINO0          1
18570 +#define AuHnJob_GEN            (1 << 1)
18571 +#define AuHnJob_DIRENT         (1 << 2)
18572 +#define AuHnJob_ISDIR          (1 << 3)
18573 +#define AuHnJob_TRYXINO0       (1 << 4)
18574 +#define AuHnJob_MNTPNT         (1 << 5)
18575 +#define au_ftest_hnjob(flags, name)    ((flags) & AuHnJob_##name)
18576 +#define au_fset_hnjob(flags, name) \
18577 +       do { (flags) |= AuHnJob_##name; } while (0)
18578 +#define au_fclr_hnjob(flags, name) \
18579 +       do { (flags) &= ~AuHnJob_##name; } while (0)
18580 +
18581 +enum {
18582 +       AuHn_CHILD,
18583 +       AuHn_PARENT,
18584 +       AuHnLast
18585 +};
18586 +
18587 +struct au_hnotify_args {
18588 +       struct inode *h_dir, *dir, *h_child_inode;
18589 +       u32 mask;
18590 +       unsigned int flags[AuHnLast];
18591 +       unsigned int h_child_nlen;
18592 +       char h_child_name[];
18593 +};
18594 +
18595 +struct hn_job_args {
18596 +       unsigned int flags;
18597 +       struct inode *inode, *h_inode, *dir, *h_dir;
18598 +       struct dentry *dentry;
18599 +       char *h_name;
18600 +       int h_nlen;
18601 +};
18602 +
18603 +static int hn_job(struct hn_job_args *a)
18604 +{
18605 +       const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
18606 +       int e;
18607 +
18608 +       /* reset xino */
18609 +       if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
18610 +               hn_xino(a->inode, a->h_inode); /* ignore this error */
18611 +
18612 +       if (au_ftest_hnjob(a->flags, TRYXINO0)
18613 +           && a->inode
18614 +           && a->h_inode) {
18615 +               inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
18616 +               if (!a->h_inode->i_nlink
18617 +                   && !(a->h_inode->i_state & I_LINKABLE))
18618 +                       hn_xino(a->inode, a->h_inode); /* ignore this error */
18619 +               inode_unlock_shared(a->h_inode);
18620 +       }
18621 +
18622 +       /* make the generation obsolete */
18623 +       if (au_ftest_hnjob(a->flags, GEN)) {
18624 +               e = -1;
18625 +               if (a->inode)
18626 +                       e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
18627 +                                             isdir);
18628 +               if (e && a->dentry)
18629 +                       hn_gen_by_name(a->dentry, isdir);
18630 +               /* ignore this error */
18631 +       }
18632 +
18633 +       /* make dir entries obsolete */
18634 +       if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
18635 +               struct au_vdir *vdir;
18636 +
18637 +               vdir = au_ivdir(a->inode);
18638 +               if (vdir)
18639 +                       vdir->vd_jiffy = 0;
18640 +               /* IMustLock(a->inode); */
18641 +               /* inode_inc_iversion(a->inode); */
18642 +       }
18643 +
18644 +       /* can do nothing but warn */
18645 +       if (au_ftest_hnjob(a->flags, MNTPNT)
18646 +           && a->dentry
18647 +           && d_mountpoint(a->dentry))
18648 +               pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
18649 +
18650 +       return 0;
18651 +}
18652 +
18653 +/* ---------------------------------------------------------------------- */
18654 +
18655 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
18656 +                                          struct inode *dir)
18657 +{
18658 +       struct dentry *dentry, *d, *parent;
18659 +       struct qstr *dname;
18660 +
18661 +       parent = d_find_any_alias(dir);
18662 +       if (!parent)
18663 +               return NULL;
18664 +
18665 +       dentry = NULL;
18666 +       spin_lock(&parent->d_lock);
18667 +       list_for_each_entry(d, &parent->d_subdirs, d_child) {
18668 +               /* AuDbg("%pd\n", d); */
18669 +               spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
18670 +               dname = &d->d_name;
18671 +               if (dname->len != nlen || memcmp(dname->name, name, nlen))
18672 +                       goto cont_unlock;
18673 +               if (au_di(d))
18674 +                       au_digen_dec(d);
18675 +               else
18676 +                       goto cont_unlock;
18677 +               if (au_dcount(d) > 0) {
18678 +                       dentry = dget_dlock(d);
18679 +                       spin_unlock(&d->d_lock);
18680 +                       break;
18681 +               }
18682 +
18683 +cont_unlock:
18684 +               spin_unlock(&d->d_lock);
18685 +       }
18686 +       spin_unlock(&parent->d_lock);
18687 +       dput(parent);
18688 +
18689 +       if (dentry)
18690 +               di_write_lock_child(dentry);
18691 +
18692 +       return dentry;
18693 +}
18694 +
18695 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
18696 +                                        aufs_bindex_t bindex, ino_t h_ino)
18697 +{
18698 +       struct inode *inode;
18699 +       ino_t ino;
18700 +       int err;
18701 +
18702 +       inode = NULL;
18703 +       err = au_xino_read(sb, bindex, h_ino, &ino);
18704 +       if (!err && ino)
18705 +               inode = ilookup(sb, ino);
18706 +       if (!inode)
18707 +               goto out;
18708 +
18709 +       if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18710 +               pr_warn("wrong root branch\n");
18711 +               iput(inode);
18712 +               inode = NULL;
18713 +               goto out;
18714 +       }
18715 +
18716 +       ii_write_lock_child(inode);
18717 +
18718 +out:
18719 +       return inode;
18720 +}
18721 +
18722 +static void au_hn_bh(void *_args)
18723 +{
18724 +       struct au_hnotify_args *a = _args;
18725 +       struct super_block *sb;
18726 +       aufs_bindex_t bindex, bbot, bfound;
18727 +       unsigned char xino, try_iput;
18728 +       int err;
18729 +       struct inode *inode;
18730 +       ino_t h_ino;
18731 +       struct hn_job_args args;
18732 +       struct dentry *dentry;
18733 +       struct au_sbinfo *sbinfo;
18734 +
18735 +       AuDebugOn(!_args);
18736 +       AuDebugOn(!a->h_dir);
18737 +       AuDebugOn(!a->dir);
18738 +       AuDebugOn(!a->mask);
18739 +       AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
18740 +             a->mask, a->dir->i_ino, a->h_dir->i_ino,
18741 +             a->h_child_inode ? a->h_child_inode->i_ino : 0);
18742 +
18743 +       inode = NULL;
18744 +       dentry = NULL;
18745 +       /*
18746 +        * do not lock a->dir->i_mutex here
18747 +        * because of d_revalidate() may cause a deadlock.
18748 +        */
18749 +       sb = a->dir->i_sb;
18750 +       AuDebugOn(!sb);
18751 +       sbinfo = au_sbi(sb);
18752 +       AuDebugOn(!sbinfo);
18753 +       si_write_lock(sb, AuLock_NOPLMW);
18754 +
18755 +       if (au_opt_test(sbinfo->si_mntflags, DIRREN))
18756 +               switch (a->mask & FS_EVENTS_POSS_ON_CHILD) {
18757 +               case FS_MOVED_FROM:
18758 +               case FS_MOVED_TO:
18759 +                       AuWarn1("DIRREN with UDBA may not work correctly "
18760 +                               "for the direct rename(2)\n");
18761 +               }
18762 +
18763 +       ii_read_lock_parent(a->dir);
18764 +       bfound = -1;
18765 +       bbot = au_ibbot(a->dir);
18766 +       for (bindex = au_ibtop(a->dir); bindex <= bbot; bindex++)
18767 +               if (au_h_iptr(a->dir, bindex) == a->h_dir) {
18768 +                       bfound = bindex;
18769 +                       break;
18770 +               }
18771 +       ii_read_unlock(a->dir);
18772 +       if (unlikely(bfound < 0))
18773 +               goto out;
18774 +
18775 +       xino = !!au_opt_test(au_mntflags(sb), XINO);
18776 +       h_ino = 0;
18777 +       if (a->h_child_inode)
18778 +               h_ino = a->h_child_inode->i_ino;
18779 +
18780 +       if (a->h_child_nlen
18781 +           && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
18782 +               || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
18783 +               dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
18784 +                                             a->dir);
18785 +       try_iput = 0;
18786 +       if (dentry && d_really_is_positive(dentry))
18787 +               inode = d_inode(dentry);
18788 +       if (xino && !inode && h_ino
18789 +           && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
18790 +               || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
18791 +               || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
18792 +               inode = lookup_wlock_by_ino(sb, bfound, h_ino);
18793 +               try_iput = 1;
18794 +       }
18795 +
18796 +       args.flags = a->flags[AuHn_CHILD];
18797 +       args.dentry = dentry;
18798 +       args.inode = inode;
18799 +       args.h_inode = a->h_child_inode;
18800 +       args.dir = a->dir;
18801 +       args.h_dir = a->h_dir;
18802 +       args.h_name = a->h_child_name;
18803 +       args.h_nlen = a->h_child_nlen;
18804 +       err = hn_job(&args);
18805 +       if (dentry) {
18806 +               if (au_di(dentry))
18807 +                       di_write_unlock(dentry);
18808 +               dput(dentry);
18809 +       }
18810 +       if (inode && try_iput) {
18811 +               ii_write_unlock(inode);
18812 +               iput(inode);
18813 +       }
18814 +
18815 +       ii_write_lock_parent(a->dir);
18816 +       args.flags = a->flags[AuHn_PARENT];
18817 +       args.dentry = NULL;
18818 +       args.inode = a->dir;
18819 +       args.h_inode = a->h_dir;
18820 +       args.dir = NULL;
18821 +       args.h_dir = NULL;
18822 +       args.h_name = NULL;
18823 +       args.h_nlen = 0;
18824 +       err = hn_job(&args);
18825 +       ii_write_unlock(a->dir);
18826 +
18827 +out:
18828 +       iput(a->h_child_inode);
18829 +       iput(a->h_dir);
18830 +       iput(a->dir);
18831 +       si_write_unlock(sb);
18832 +       au_nwt_done(&sbinfo->si_nowait);
18833 +       au_kfree_rcu(a);
18834 +}
18835 +
18836 +/* ---------------------------------------------------------------------- */
18837 +
18838 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
18839 +              const struct qstr *h_child_qstr, struct inode *h_child_inode)
18840 +{
18841 +       int err, len;
18842 +       unsigned int flags[AuHnLast], f;
18843 +       unsigned char isdir, isroot, wh;
18844 +       struct inode *dir;
18845 +       struct au_hnotify_args *args;
18846 +       char *p, *h_child_name;
18847 +
18848 +       err = 0;
18849 +       AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
18850 +       dir = igrab(hnotify->hn_aufs_inode);
18851 +       if (!dir)
18852 +               goto out;
18853 +
18854 +       isroot = (dir->i_ino == AUFS_ROOT_INO);
18855 +       wh = 0;
18856 +       h_child_name = (void *)h_child_qstr->name;
18857 +       len = h_child_qstr->len;
18858 +       if (h_child_name) {
18859 +               if (len > AUFS_WH_PFX_LEN
18860 +                   && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
18861 +                       h_child_name += AUFS_WH_PFX_LEN;
18862 +                       len -= AUFS_WH_PFX_LEN;
18863 +                       wh = 1;
18864 +               }
18865 +       }
18866 +
18867 +       isdir = 0;
18868 +       if (h_child_inode)
18869 +               isdir = !!S_ISDIR(h_child_inode->i_mode);
18870 +       flags[AuHn_PARENT] = AuHnJob_ISDIR;
18871 +       flags[AuHn_CHILD] = 0;
18872 +       if (isdir)
18873 +               flags[AuHn_CHILD] = AuHnJob_ISDIR;
18874 +       au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
18875 +       au_fset_hnjob(flags[AuHn_CHILD], GEN);
18876 +       switch (mask & ALL_FSNOTIFY_DIRENT_EVENTS) {
18877 +       case FS_MOVED_FROM:
18878 +       case FS_MOVED_TO:
18879 +               au_fset_hnjob(flags[AuHn_CHILD], XINO0);
18880 +               au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18881 +               fallthrough;
18882 +       case FS_CREATE:
18883 +               AuDebugOn(!h_child_name);
18884 +               break;
18885 +
18886 +       case FS_DELETE:
18887 +               /*
18888 +                * aufs never be able to get this child inode.
18889 +                * revalidation should be in d_revalidate()
18890 +                * by checking i_nlink, i_generation or d_unhashed().
18891 +                */
18892 +               AuDebugOn(!h_child_name);
18893 +               au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
18894 +               au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18895 +               break;
18896 +
18897 +       default:
18898 +               AuDebugOn(1);
18899 +       }
18900 +
18901 +       if (wh)
18902 +               h_child_inode = NULL;
18903 +
18904 +       err = -ENOMEM;
18905 +       /* iput() and kfree() will be called in au_hnotify() */
18906 +       args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
18907 +       if (unlikely(!args)) {
18908 +               AuErr1("no memory\n");
18909 +               iput(dir);
18910 +               goto out;
18911 +       }
18912 +       args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
18913 +       args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
18914 +       args->mask = mask;
18915 +       args->dir = dir;
18916 +       args->h_dir = igrab(h_dir);
18917 +       if (h_child_inode)
18918 +               h_child_inode = igrab(h_child_inode); /* can be NULL */
18919 +       args->h_child_inode = h_child_inode;
18920 +       args->h_child_nlen = len;
18921 +       if (len) {
18922 +               p = (void *)args;
18923 +               p += sizeof(*args);
18924 +               memcpy(p, h_child_name, len);
18925 +               p[len] = 0;
18926 +       }
18927 +
18928 +       /* NFS fires the event for silly-renamed one from kworker */
18929 +       f = 0;
18930 +       if (!dir->i_nlink
18931 +           || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
18932 +               f = AuWkq_NEST;
18933 +       err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
18934 +       if (unlikely(err)) {
18935 +               pr_err("wkq %d\n", err);
18936 +               iput(args->h_child_inode);
18937 +               iput(args->h_dir);
18938 +               iput(args->dir);
18939 +               au_kfree_rcu(args);
18940 +       }
18941 +
18942 +out:
18943 +       return err;
18944 +}
18945 +
18946 +/* ---------------------------------------------------------------------- */
18947 +
18948 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
18949 +{
18950 +       int err;
18951 +
18952 +       AuDebugOn(!(udba & AuOptMask_UDBA));
18953 +
18954 +       err = 0;
18955 +       if (au_hnotify_op.reset_br)
18956 +               err = au_hnotify_op.reset_br(udba, br, perm);
18957 +
18958 +       return err;
18959 +}
18960 +
18961 +int au_hnotify_init_br(struct au_branch *br, int perm)
18962 +{
18963 +       int err;
18964 +
18965 +       err = 0;
18966 +       if (au_hnotify_op.init_br)
18967 +               err = au_hnotify_op.init_br(br, perm);
18968 +
18969 +       return err;
18970 +}
18971 +
18972 +void au_hnotify_fin_br(struct au_branch *br)
18973 +{
18974 +       if (au_hnotify_op.fin_br)
18975 +               au_hnotify_op.fin_br(br);
18976 +}
18977 +
18978 +static void au_hn_destroy_cache(void)
18979 +{
18980 +       kmem_cache_destroy(au_cache[AuCache_HNOTIFY]);
18981 +       au_cache[AuCache_HNOTIFY] = NULL;
18982 +}
18983 +
18984 +int __init au_hnotify_init(void)
18985 +{
18986 +       int err;
18987 +
18988 +       err = -ENOMEM;
18989 +       au_cache[AuCache_HNOTIFY] = AuCache(au_hnotify);
18990 +       if (au_cache[AuCache_HNOTIFY]) {
18991 +               err = 0;
18992 +               if (au_hnotify_op.init)
18993 +                       err = au_hnotify_op.init();
18994 +               if (unlikely(err))
18995 +                       au_hn_destroy_cache();
18996 +       }
18997 +       AuTraceErr(err);
18998 +       return err;
18999 +}
19000 +
19001 +void au_hnotify_fin(void)
19002 +{
19003 +       if (au_hnotify_op.fin)
19004 +               au_hnotify_op.fin();
19005 +
19006 +       /* cf. au_cache_fin() */
19007 +       if (au_cache[AuCache_HNOTIFY])
19008 +               au_hn_destroy_cache();
19009 +}
19010 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
19011 --- /usr/share/empty/fs/aufs/iinfo.c    1970-01-01 01:00:00.000000000 +0100
19012 +++ linux/fs/aufs/iinfo.c       2022-11-05 23:02:18.965889284 +0100
19013 @@ -0,0 +1,286 @@
19014 +// SPDX-License-Identifier: GPL-2.0
19015 +/*
19016 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19017 + *
19018 + * This program is free software; you can redistribute it and/or modify
19019 + * it under the terms of the GNU General Public License as published by
19020 + * the Free Software Foundation; either version 2 of the License, or
19021 + * (at your option) any later version.
19022 + *
19023 + * This program is distributed in the hope that it will be useful,
19024 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19025 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19026 + * GNU General Public License for more details.
19027 + *
19028 + * You should have received a copy of the GNU General Public License
19029 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
19030 + */
19031 +
19032 +/*
19033 + * inode private data
19034 + */
19035 +
19036 +#include "aufs.h"
19037 +
19038 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
19039 +{
19040 +       struct inode *h_inode;
19041 +       struct au_hinode *hinode;
19042 +
19043 +       IiMustAnyLock(inode);
19044 +
19045 +       hinode = au_hinode(au_ii(inode), bindex);
19046 +       h_inode = hinode->hi_inode;
19047 +       AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19048 +       return h_inode;
19049 +}
19050 +
19051 +/* todo: hard/soft set? */
19052 +void au_hiput(struct au_hinode *hinode)
19053 +{
19054 +       au_hn_free(hinode);
19055 +       dput(hinode->hi_whdentry);
19056 +       iput(hinode->hi_inode);
19057 +}
19058 +
19059 +unsigned int au_hi_flags(struct inode *inode, int isdir)
19060 +{
19061 +       unsigned int flags;
19062 +       const unsigned int mnt_flags = au_mntflags(inode->i_sb);
19063 +
19064 +       flags = 0;
19065 +       if (au_opt_test(mnt_flags, XINO))
19066 +               au_fset_hi(flags, XINO);
19067 +       if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
19068 +               au_fset_hi(flags, HNOTIFY);
19069 +       return flags;
19070 +}
19071 +
19072 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
19073 +                  struct inode *h_inode, unsigned int flags)
19074 +{
19075 +       struct au_hinode *hinode;
19076 +       struct inode *hi;
19077 +       struct au_iinfo *iinfo = au_ii(inode);
19078 +
19079 +       IiMustWriteLock(inode);
19080 +
19081 +       hinode = au_hinode(iinfo, bindex);
19082 +       hi = hinode->hi_inode;
19083 +       AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19084 +
19085 +       if (hi)
19086 +               au_hiput(hinode);
19087 +       hinode->hi_inode = h_inode;
19088 +       if (h_inode) {
19089 +               int err;
19090 +               struct super_block *sb = inode->i_sb;
19091 +               struct au_branch *br;
19092 +
19093 +               AuDebugOn(inode->i_mode
19094 +                         && (h_inode->i_mode & S_IFMT)
19095 +                         != (inode->i_mode & S_IFMT));
19096 +               if (bindex == iinfo->ii_btop)
19097 +                       au_cpup_igen(inode, h_inode);
19098 +               br = au_sbr(sb, bindex);
19099 +               hinode->hi_id = br->br_id;
19100 +               if (au_ftest_hi(flags, XINO)) {
19101 +                       err = au_xino_write(sb, bindex, h_inode->i_ino,
19102 +                                           inode->i_ino);
19103 +                       if (unlikely(err))
19104 +                               AuIOErr1("failed au_xino_write() %d\n", err);
19105 +               }
19106 +
19107 +               if (au_ftest_hi(flags, HNOTIFY)
19108 +                   && au_br_hnotifyable(br->br_perm)) {
19109 +                       err = au_hn_alloc(hinode, inode);
19110 +                       if (unlikely(err))
19111 +                               AuIOErr1("au_hn_alloc() %d\n", err);
19112 +               }
19113 +       }
19114 +}
19115 +
19116 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
19117 +                 struct dentry *h_wh)
19118 +{
19119 +       struct au_hinode *hinode;
19120 +
19121 +       IiMustWriteLock(inode);
19122 +
19123 +       hinode = au_hinode(au_ii(inode), bindex);
19124 +       AuDebugOn(hinode->hi_whdentry);
19125 +       hinode->hi_whdentry = h_wh;
19126 +}
19127 +
19128 +void au_update_iigen(struct inode *inode, int half)
19129 +{
19130 +       struct au_iinfo *iinfo;
19131 +       struct au_iigen *iigen;
19132 +       unsigned int sigen;
19133 +
19134 +       sigen = au_sigen(inode->i_sb);
19135 +       iinfo = au_ii(inode);
19136 +       iigen = &iinfo->ii_generation;
19137 +       spin_lock(&iigen->ig_spin);
19138 +       iigen->ig_generation = sigen;
19139 +       if (half)
19140 +               au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
19141 +       else
19142 +               au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
19143 +       spin_unlock(&iigen->ig_spin);
19144 +}
19145 +
19146 +/* it may be called at remount time, too */
19147 +void au_update_ibrange(struct inode *inode, int do_put_zero)
19148 +{
19149 +       struct au_iinfo *iinfo;
19150 +       aufs_bindex_t bindex, bbot;
19151 +
19152 +       AuDebugOn(au_is_bad_inode(inode));
19153 +       IiMustWriteLock(inode);
19154 +
19155 +       iinfo = au_ii(inode);
19156 +       if (do_put_zero && iinfo->ii_btop >= 0) {
19157 +               for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19158 +                    bindex++) {
19159 +                       struct inode *h_i;
19160 +
19161 +                       h_i = au_hinode(iinfo, bindex)->hi_inode;
19162 +                       if (h_i
19163 +                           && !h_i->i_nlink
19164 +                           && !(h_i->i_state & I_LINKABLE))
19165 +                               au_set_h_iptr(inode, bindex, NULL, 0);
19166 +               }
19167 +       }
19168 +
19169 +       iinfo->ii_btop = -1;
19170 +       iinfo->ii_bbot = -1;
19171 +       bbot = au_sbbot(inode->i_sb);
19172 +       for (bindex = 0; bindex <= bbot; bindex++)
19173 +               if (au_hinode(iinfo, bindex)->hi_inode) {
19174 +                       iinfo->ii_btop = bindex;
19175 +                       break;
19176 +               }
19177 +       if (iinfo->ii_btop >= 0)
19178 +               for (bindex = bbot; bindex >= iinfo->ii_btop; bindex--)
19179 +                       if (au_hinode(iinfo, bindex)->hi_inode) {
19180 +                               iinfo->ii_bbot = bindex;
19181 +                               break;
19182 +                       }
19183 +       AuDebugOn(iinfo->ii_btop > iinfo->ii_bbot);
19184 +}
19185 +
19186 +/* ---------------------------------------------------------------------- */
19187 +
19188 +void au_icntnr_init_once(void *_c)
19189 +{
19190 +       struct au_icntnr *c = _c;
19191 +       struct au_iinfo *iinfo = &c->iinfo;
19192 +
19193 +       spin_lock_init(&iinfo->ii_generation.ig_spin);
19194 +       au_rw_init(&iinfo->ii_rwsem);
19195 +       inode_init_once(&c->vfs_inode);
19196 +}
19197 +
19198 +void au_hinode_init(struct au_hinode *hinode)
19199 +{
19200 +       hinode->hi_inode = NULL;
19201 +       hinode->hi_id = -1;
19202 +       au_hn_init(hinode);
19203 +       hinode->hi_whdentry = NULL;
19204 +}
19205 +
19206 +int au_iinfo_init(struct inode *inode)
19207 +{
19208 +       struct au_iinfo *iinfo;
19209 +       struct super_block *sb;
19210 +       struct au_hinode *hi;
19211 +       int nbr, i;
19212 +
19213 +       sb = inode->i_sb;
19214 +       iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19215 +       nbr = au_sbbot(sb) + 1;
19216 +       if (unlikely(nbr <= 0))
19217 +               nbr = 1;
19218 +       hi = kmalloc_array(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
19219 +       if (hi) {
19220 +               au_lcnt_inc(&au_sbi(sb)->si_ninodes);
19221 +
19222 +               iinfo->ii_hinode = hi;
19223 +               for (i = 0; i < nbr; i++, hi++)
19224 +                       au_hinode_init(hi);
19225 +
19226 +               iinfo->ii_generation.ig_generation = au_sigen(sb);
19227 +               iinfo->ii_btop = -1;
19228 +               iinfo->ii_bbot = -1;
19229 +               iinfo->ii_vdir = NULL;
19230 +               return 0;
19231 +       }
19232 +       return -ENOMEM;
19233 +}
19234 +
19235 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink)
19236 +{
19237 +       int err, i;
19238 +       struct au_hinode *hip;
19239 +
19240 +       AuRwMustWriteLock(&iinfo->ii_rwsem);
19241 +
19242 +       err = -ENOMEM;
19243 +       hip = au_krealloc(iinfo->ii_hinode, sizeof(*hip) * nbr, GFP_NOFS,
19244 +                         may_shrink);
19245 +       if (hip) {
19246 +               iinfo->ii_hinode = hip;
19247 +               i = iinfo->ii_bbot + 1;
19248 +               hip += i;
19249 +               for (; i < nbr; i++, hip++)
19250 +                       au_hinode_init(hip);
19251 +               err = 0;
19252 +       }
19253 +
19254 +       return err;
19255 +}
19256 +
19257 +void au_iinfo_fin(struct inode *inode)
19258 +{
19259 +       struct au_iinfo *iinfo;
19260 +       struct au_hinode *hi;
19261 +       struct super_block *sb;
19262 +       aufs_bindex_t bindex, bbot;
19263 +       const unsigned char unlinked = !inode->i_nlink;
19264 +
19265 +       AuDebugOn(au_is_bad_inode(inode));
19266 +
19267 +       sb = inode->i_sb;
19268 +       au_lcnt_dec(&au_sbi(sb)->si_ninodes);
19269 +       if (si_pid_test(sb))
19270 +               au_xino_delete_inode(inode, unlinked);
19271 +       else {
19272 +               /*
19273 +                * it is safe to hide the dependency between sbinfo and
19274 +                * sb->s_umount.
19275 +                */
19276 +               lockdep_off();
19277 +               si_noflush_read_lock(sb);
19278 +               au_xino_delete_inode(inode, unlinked);
19279 +               si_read_unlock(sb);
19280 +               lockdep_on();
19281 +       }
19282 +
19283 +       iinfo = au_ii(inode);
19284 +       if (iinfo->ii_vdir)
19285 +               au_vdir_free(iinfo->ii_vdir);
19286 +
19287 +       bindex = iinfo->ii_btop;
19288 +       if (bindex >= 0) {
19289 +               hi = au_hinode(iinfo, bindex);
19290 +               bbot = iinfo->ii_bbot;
19291 +               while (bindex++ <= bbot) {
19292 +                       if (hi->hi_inode)
19293 +                               au_hiput(hi);
19294 +                       hi++;
19295 +               }
19296 +       }
19297 +       au_kfree_rcu(iinfo->ii_hinode);
19298 +       AuRwDestroy(&iinfo->ii_rwsem);
19299 +}
19300 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
19301 --- /usr/share/empty/fs/aufs/inode.c    1970-01-01 01:00:00.000000000 +0100
19302 +++ linux/fs/aufs/inode.c       2023-02-20 21:05:51.959693785 +0100
19303 @@ -0,0 +1,531 @@
19304 +// SPDX-License-Identifier: GPL-2.0
19305 +/*
19306 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19307 + *
19308 + * This program is free software; you can redistribute it and/or modify
19309 + * it under the terms of the GNU General Public License as published by
19310 + * the Free Software Foundation; either version 2 of the License, or
19311 + * (at your option) any later version.
19312 + *
19313 + * This program is distributed in the hope that it will be useful,
19314 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19315 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19316 + * GNU General Public License for more details.
19317 + *
19318 + * You should have received a copy of the GNU General Public License
19319 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
19320 + */
19321 +
19322 +/*
19323 + * inode functions
19324 + */
19325 +
19326 +#include <linux/iversion.h>
19327 +#include "aufs.h"
19328 +
19329 +struct inode *au_igrab(struct inode *inode)
19330 +{
19331 +       if (inode) {
19332 +               AuDebugOn(!atomic_read(&inode->i_count));
19333 +               ihold(inode);
19334 +       }
19335 +       return inode;
19336 +}
19337 +
19338 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
19339 +{
19340 +       au_cpup_attr_all(inode, /*force*/0);
19341 +       au_update_iigen(inode, /*half*/1);
19342 +       if (do_version)
19343 +               inode_inc_iversion(inode);
19344 +}
19345 +
19346 +static int au_ii_refresh(struct inode *inode, int *update)
19347 +{
19348 +       int err, e, nbr;
19349 +       umode_t type;
19350 +       aufs_bindex_t bindex, new_bindex;
19351 +       struct super_block *sb;
19352 +       struct au_iinfo *iinfo;
19353 +       struct au_hinode *p, *q, tmp;
19354 +
19355 +       AuDebugOn(au_is_bad_inode(inode));
19356 +       IiMustWriteLock(inode);
19357 +
19358 +       *update = 0;
19359 +       sb = inode->i_sb;
19360 +       nbr = au_sbbot(sb) + 1;
19361 +       type = inode->i_mode & S_IFMT;
19362 +       iinfo = au_ii(inode);
19363 +       err = au_hinode_realloc(iinfo, nbr, /*may_shrink*/0);
19364 +       if (unlikely(err))
19365 +               goto out;
19366 +
19367 +       AuDebugOn(iinfo->ii_btop < 0);
19368 +       p = au_hinode(iinfo, iinfo->ii_btop);
19369 +       for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19370 +            bindex++, p++) {
19371 +               if (!p->hi_inode)
19372 +                       continue;
19373 +
19374 +               AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
19375 +               new_bindex = au_br_index(sb, p->hi_id);
19376 +               if (new_bindex == bindex)
19377 +                       continue;
19378 +
19379 +               if (new_bindex < 0) {
19380 +                       *update = 1;
19381 +                       au_hiput(p);
19382 +                       p->hi_inode = NULL;
19383 +                       continue;
19384 +               }
19385 +
19386 +               if (new_bindex < iinfo->ii_btop)
19387 +                       iinfo->ii_btop = new_bindex;
19388 +               if (iinfo->ii_bbot < new_bindex)
19389 +                       iinfo->ii_bbot = new_bindex;
19390 +               /* swap two lower inode, and loop again */
19391 +               q = au_hinode(iinfo, new_bindex);
19392 +               tmp = *q;
19393 +               *q = *p;
19394 +               *p = tmp;
19395 +               if (tmp.hi_inode) {
19396 +                       bindex--;
19397 +                       p--;
19398 +               }
19399 +       }
19400 +       au_update_ibrange(inode, /*do_put_zero*/0);
19401 +       au_hinode_realloc(iinfo, nbr, /*may_shrink*/1); /* harmless if err */
19402 +       e = au_dy_irefresh(inode);
19403 +       if (unlikely(e && !err))
19404 +               err = e;
19405 +
19406 +out:
19407 +       AuTraceErr(err);
19408 +       return err;
19409 +}
19410 +
19411 +void au_refresh_iop(struct inode *inode, int force_getattr)
19412 +{
19413 +       int type;
19414 +       struct au_sbinfo *sbi = au_sbi(inode->i_sb);
19415 +       const struct inode_operations *iop
19416 +               = force_getattr ? aufs_iop : sbi->si_iop_array;
19417 +
19418 +       if (inode->i_op == iop)
19419 +               return;
19420 +
19421 +       switch (inode->i_mode & S_IFMT) {
19422 +       case S_IFDIR:
19423 +               type = AuIop_DIR;
19424 +               break;
19425 +       case S_IFLNK:
19426 +               type = AuIop_SYMLINK;
19427 +               break;
19428 +       default:
19429 +               type = AuIop_OTHER;
19430 +               break;
19431 +       }
19432 +
19433 +       inode->i_op = iop + type;
19434 +       /* unnecessary smp_wmb() */
19435 +}
19436 +
19437 +int au_refresh_hinode_self(struct inode *inode)
19438 +{
19439 +       int err, update;
19440 +
19441 +       err = au_ii_refresh(inode, &update);
19442 +       if (!err)
19443 +               au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
19444 +
19445 +       AuTraceErr(err);
19446 +       return err;
19447 +}
19448 +
19449 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
19450 +{
19451 +       int err, e, update;
19452 +       unsigned int flags;
19453 +       umode_t mode;
19454 +       aufs_bindex_t bindex, bbot;
19455 +       unsigned char isdir;
19456 +       struct au_hinode *p;
19457 +       struct au_iinfo *iinfo;
19458 +
19459 +       err = au_ii_refresh(inode, &update);
19460 +       if (unlikely(err))
19461 +               goto out;
19462 +
19463 +       update = 0;
19464 +       iinfo = au_ii(inode);
19465 +       p = au_hinode(iinfo, iinfo->ii_btop);
19466 +       mode = (inode->i_mode & S_IFMT);
19467 +       isdir = S_ISDIR(mode);
19468 +       flags = au_hi_flags(inode, isdir);
19469 +       bbot = au_dbbot(dentry);
19470 +       for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
19471 +               struct inode *h_i, *h_inode;
19472 +               struct dentry *h_d;
19473 +
19474 +               h_d = au_h_dptr(dentry, bindex);
19475 +               if (!h_d || d_is_negative(h_d))
19476 +                       continue;
19477 +
19478 +               h_inode = d_inode(h_d);
19479 +               AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
19480 +               if (iinfo->ii_btop <= bindex && bindex <= iinfo->ii_bbot) {
19481 +                       h_i = au_h_iptr(inode, bindex);
19482 +                       if (h_i) {
19483 +                               if (h_i == h_inode)
19484 +                                       continue;
19485 +                               err = -EIO;
19486 +                               break;
19487 +                       }
19488 +               }
19489 +               if (bindex < iinfo->ii_btop)
19490 +                       iinfo->ii_btop = bindex;
19491 +               if (iinfo->ii_bbot < bindex)
19492 +                       iinfo->ii_bbot = bindex;
19493 +               au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
19494 +               update = 1;
19495 +       }
19496 +       au_update_ibrange(inode, /*do_put_zero*/0);
19497 +       e = au_dy_irefresh(inode);
19498 +       if (unlikely(e && !err))
19499 +               err = e;
19500 +       if (!err)
19501 +               au_refresh_hinode_attr(inode, update && isdir);
19502 +
19503 +out:
19504 +       AuTraceErr(err);
19505 +       return err;
19506 +}
19507 +
19508 +static int set_inode(struct inode *inode, struct dentry *dentry)
19509 +{
19510 +       int err;
19511 +       unsigned int flags;
19512 +       umode_t mode;
19513 +       aufs_bindex_t bindex, btop, btail;
19514 +       unsigned char isdir;
19515 +       struct dentry *h_dentry;
19516 +       struct inode *h_inode;
19517 +       struct au_iinfo *iinfo;
19518 +       const struct inode_operations *iop;
19519 +
19520 +       IiMustWriteLock(inode);
19521 +
19522 +       err = 0;
19523 +       isdir = 0;
19524 +       iop = au_sbi(inode->i_sb)->si_iop_array;
19525 +       btop = au_dbtop(dentry);
19526 +       h_dentry = au_h_dptr(dentry, btop);
19527 +       h_inode = d_inode(h_dentry);
19528 +       mode = h_inode->i_mode;
19529 +       switch (mode & S_IFMT) {
19530 +       case S_IFREG:
19531 +               btail = au_dbtail(dentry);
19532 +               inode->i_op = iop + AuIop_OTHER;
19533 +               inode->i_fop = &aufs_file_fop;
19534 +               err = au_dy_iaop(inode, btop, h_inode);
19535 +               if (unlikely(err))
19536 +                       goto out;
19537 +               break;
19538 +       case S_IFDIR:
19539 +               isdir = 1;
19540 +               btail = au_dbtaildir(dentry);
19541 +               inode->i_op = iop + AuIop_DIR;
19542 +               inode->i_fop = &aufs_dir_fop;
19543 +               break;
19544 +       case S_IFLNK:
19545 +               btail = au_dbtail(dentry);
19546 +               inode->i_op = iop + AuIop_SYMLINK;
19547 +               break;
19548 +       case S_IFBLK:
19549 +       case S_IFCHR:
19550 +       case S_IFIFO:
19551 +       case S_IFSOCK:
19552 +               btail = au_dbtail(dentry);
19553 +               inode->i_op = iop + AuIop_OTHER;
19554 +               init_special_inode(inode, mode, h_inode->i_rdev);
19555 +               break;
19556 +       default:
19557 +               AuIOErr("Unknown file type 0%o\n", mode);
19558 +               err = -EIO;
19559 +               goto out;
19560 +       }
19561 +
19562 +       /* do not set hnotify for whiteouted dirs (SHWH mode) */
19563 +       flags = au_hi_flags(inode, isdir);
19564 +       if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
19565 +           && au_ftest_hi(flags, HNOTIFY)
19566 +           && dentry->d_name.len > AUFS_WH_PFX_LEN
19567 +           && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
19568 +               au_fclr_hi(flags, HNOTIFY);
19569 +       iinfo = au_ii(inode);
19570 +       iinfo->ii_btop = btop;
19571 +       iinfo->ii_bbot = btail;
19572 +       for (bindex = btop; bindex <= btail; bindex++) {
19573 +               h_dentry = au_h_dptr(dentry, bindex);
19574 +               if (h_dentry)
19575 +                       au_set_h_iptr(inode, bindex,
19576 +                                     au_igrab(d_inode(h_dentry)), flags);
19577 +       }
19578 +       au_cpup_attr_all(inode, /*force*/1);
19579 +       /*
19580 +        * to force calling aufs_get_inode_acl() every time,
19581 +        * do not call cache_no_acl() for aufs inode.
19582 +        */
19583 +
19584 +out:
19585 +       return err;
19586 +}
19587 +
19588 +/*
19589 + * successful returns with iinfo write_locked
19590 + * minus: errno
19591 + * zero: success, matched
19592 + * plus: no error, but unmatched
19593 + */
19594 +static int reval_inode(struct inode *inode, struct dentry *dentry)
19595 +{
19596 +       int err;
19597 +       unsigned int gen, igflags;
19598 +       aufs_bindex_t bindex, bbot;
19599 +       struct inode *h_inode, *h_dinode;
19600 +       struct dentry *h_dentry;
19601 +
19602 +       /*
19603 +        * before this function, if aufs got any iinfo lock, it must be only
19604 +        * one, the parent dir.
19605 +        * it can happen by UDBA and the obsoleted inode number.
19606 +        */
19607 +       err = -EIO;
19608 +       if (unlikely(inode->i_ino == parent_ino(dentry)))
19609 +               goto out;
19610 +
19611 +       err = 1;
19612 +       ii_write_lock_new_child(inode);
19613 +       h_dentry = au_h_dptr(dentry, au_dbtop(dentry));
19614 +       h_dinode = d_inode(h_dentry);
19615 +       bbot = au_ibbot(inode);
19616 +       for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
19617 +               h_inode = au_h_iptr(inode, bindex);
19618 +               if (!h_inode || h_inode != h_dinode)
19619 +                       continue;
19620 +
19621 +               err = 0;
19622 +               gen = au_iigen(inode, &igflags);
19623 +               if (gen == au_digen(dentry)
19624 +                   && !au_ig_ftest(igflags, HALF_REFRESHED))
19625 +                       break;
19626 +
19627 +               /* fully refresh inode using dentry */
19628 +               err = au_refresh_hinode(inode, dentry);
19629 +               if (!err)
19630 +                       au_update_iigen(inode, /*half*/0);
19631 +               break;
19632 +       }
19633 +
19634 +       if (unlikely(err))
19635 +               ii_write_unlock(inode);
19636 +out:
19637 +       return err;
19638 +}
19639 +
19640 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19641 +          unsigned int d_type, ino_t *ino)
19642 +{
19643 +       int err, idx;
19644 +       const int isnondir = d_type != DT_DIR;
19645 +
19646 +       /* prevent hardlinked inode number from race condition */
19647 +       if (isnondir) {
19648 +               err = au_xinondir_enter(sb, bindex, h_ino, &idx);
19649 +               if (unlikely(err))
19650 +                       goto out;
19651 +       }
19652 +
19653 +       err = au_xino_read(sb, bindex, h_ino, ino);
19654 +       if (unlikely(err))
19655 +               goto out_xinondir;
19656 +
19657 +       if (!*ino) {
19658 +               err = -EIO;
19659 +               *ino = au_xino_new_ino(sb);
19660 +               if (unlikely(!*ino))
19661 +                       goto out_xinondir;
19662 +               err = au_xino_write(sb, bindex, h_ino, *ino);
19663 +               if (unlikely(err))
19664 +                       goto out_xinondir;
19665 +       }
19666 +
19667 +out_xinondir:
19668 +       if (isnondir && idx >= 0)
19669 +               au_xinondir_leave(sb, bindex, h_ino, idx);
19670 +out:
19671 +       return err;
19672 +}
19673 +
19674 +/* successful returns with iinfo write_locked */
19675 +/* todo: return with unlocked? */
19676 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
19677 +{
19678 +       struct inode *inode, *h_inode;
19679 +       struct dentry *h_dentry;
19680 +       struct super_block *sb;
19681 +       ino_t h_ino, ino;
19682 +       int err, idx, hlinked;
19683 +       aufs_bindex_t btop;
19684 +
19685 +       sb = dentry->d_sb;
19686 +       btop = au_dbtop(dentry);
19687 +       h_dentry = au_h_dptr(dentry, btop);
19688 +       h_inode = d_inode(h_dentry);
19689 +       h_ino = h_inode->i_ino;
19690 +       hlinked = !d_is_dir(h_dentry) && h_inode->i_nlink > 1;
19691 +
19692 +new_ino:
19693 +       /*
19694 +        * stop 'race'-ing between hardlinks under different
19695 +        * parents.
19696 +        */
19697 +       if (hlinked) {
19698 +               err = au_xinondir_enter(sb, btop, h_ino, &idx);
19699 +               inode = ERR_PTR(err);
19700 +               if (unlikely(err))
19701 +                       goto out;
19702 +       }
19703 +
19704 +       err = au_xino_read(sb, btop, h_ino, &ino);
19705 +       inode = ERR_PTR(err);
19706 +       if (unlikely(err))
19707 +               goto out_xinondir;
19708 +
19709 +       if (!ino) {
19710 +               ino = au_xino_new_ino(sb);
19711 +               if (unlikely(!ino)) {
19712 +                       inode = ERR_PTR(-EIO);
19713 +                       goto out_xinondir;
19714 +               }
19715 +       }
19716 +
19717 +       AuDbg("i%lu\n", (unsigned long)ino);
19718 +       inode = au_iget_locked(sb, ino);
19719 +       err = PTR_ERR(inode);
19720 +       if (IS_ERR(inode))
19721 +               goto out_xinondir;
19722 +
19723 +       AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
19724 +       if (inode->i_state & I_NEW) {
19725 +               ii_write_lock_new_child(inode);
19726 +               err = set_inode(inode, dentry);
19727 +               if (!err) {
19728 +                       unlock_new_inode(inode);
19729 +                       goto out_xinondir; /* success */
19730 +               }
19731 +
19732 +               /*
19733 +                * iget_failed() calls iput(), but we need to call
19734 +                * ii_write_unlock() after iget_failed(). so dirty hack for
19735 +                * i_count.
19736 +                */
19737 +               atomic_inc(&inode->i_count);
19738 +               iget_failed(inode);
19739 +               ii_write_unlock(inode);
19740 +               au_xino_write(sb, btop, h_ino, /*ino*/0);
19741 +               /* ignore this error */
19742 +               goto out_iput;
19743 +       } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
19744 +               /*
19745 +                * horrible race condition between lookup, readdir and copyup
19746 +                * (or something).
19747 +                */
19748 +               if (hlinked && idx >= 0)
19749 +                       au_xinondir_leave(sb, btop, h_ino, idx);
19750 +               err = reval_inode(inode, dentry);
19751 +               if (unlikely(err < 0)) {
19752 +                       hlinked = 0;
19753 +                       goto out_iput;
19754 +               }
19755 +               if (!err)
19756 +                       goto out; /* success */
19757 +               else if (hlinked && idx >= 0) {
19758 +                       err = au_xinondir_enter(sb, btop, h_ino, &idx);
19759 +                       if (unlikely(err)) {
19760 +                               iput(inode);
19761 +                               inode = ERR_PTR(err);
19762 +                               goto out;
19763 +                       }
19764 +               }
19765 +       }
19766 +
19767 +       if (unlikely(au_test_fs_unique_ino(h_inode)))
19768 +               AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
19769 +                       " b%d, %s, %pd, hi%lu, i%lu.\n",
19770 +                       btop, au_sbtype(h_dentry->d_sb), dentry,
19771 +                       (unsigned long)h_ino, (unsigned long)ino);
19772 +       ino = 0;
19773 +       err = au_xino_write(sb, btop, h_ino, /*ino*/0);
19774 +       if (!err) {
19775 +               iput(inode);
19776 +               if (hlinked && idx >= 0)
19777 +                       au_xinondir_leave(sb, btop, h_ino, idx);
19778 +               goto new_ino;
19779 +       }
19780 +
19781 +out_iput:
19782 +       iput(inode);
19783 +       inode = ERR_PTR(err);
19784 +out_xinondir:
19785 +       if (hlinked && idx >= 0)
19786 +               au_xinondir_leave(sb, btop, h_ino, idx);
19787 +out:
19788 +       return inode;
19789 +}
19790 +
19791 +/* ---------------------------------------------------------------------- */
19792 +
19793 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19794 +              struct inode *inode)
19795 +{
19796 +       int err;
19797 +       struct inode *hi;
19798 +
19799 +       err = au_br_rdonly(au_sbr(sb, bindex));
19800 +
19801 +       /* pseudo-link after flushed may happen out of bounds */
19802 +       if (!err
19803 +           && inode
19804 +           && au_ibtop(inode) <= bindex
19805 +           && bindex <= au_ibbot(inode)) {
19806 +               /*
19807 +                * permission check is unnecessary since vfsub routine
19808 +                * will be called later
19809 +                */
19810 +               hi = au_h_iptr(inode, bindex);
19811 +               if (hi)
19812 +                       err = IS_IMMUTABLE(hi) ? -EROFS : 0;
19813 +       }
19814 +
19815 +       return err;
19816 +}
19817 +
19818 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19819 +                  int mask)
19820 +{
19821 +       if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
19822 +               return 0;
19823 +       return inode_permission(h_userns, h_inode, mask);
19824 +}
19825 +
19826 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19827 +                      int mask)
19828 +{
19829 +       if (au_test_nfs(h_inode->i_sb)
19830 +           && (mask & MAY_WRITE)
19831 +           && S_ISDIR(h_inode->i_mode))
19832 +               mask |= MAY_READ; /* force permission check */
19833 +       return au_test_h_perm(h_userns, h_inode, mask);
19834 +}
19835 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
19836 --- /usr/share/empty/fs/aufs/inode.h    1970-01-01 01:00:00.000000000 +0100
19837 +++ linux/fs/aufs/inode.h       2023-02-20 21:05:51.959693785 +0100
19838 @@ -0,0 +1,707 @@
19839 +/* SPDX-License-Identifier: GPL-2.0 */
19840 +/*
19841 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19842 + *
19843 + * This program is free software; you can redistribute it and/or modify
19844 + * it under the terms of the GNU General Public License as published by
19845 + * the Free Software Foundation; either version 2 of the License, or
19846 + * (at your option) any later version.
19847 + *
19848 + * This program is distributed in the hope that it will be useful,
19849 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19850 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19851 + * GNU General Public License for more details.
19852 + *
19853 + * You should have received a copy of the GNU General Public License
19854 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
19855 + */
19856 +
19857 +/*
19858 + * inode operations
19859 + */
19860 +
19861 +#ifndef __AUFS_INODE_H__
19862 +#define __AUFS_INODE_H__
19863 +
19864 +#ifdef __KERNEL__
19865 +
19866 +#include <linux/fsnotify.h>
19867 +#include "rwsem.h"
19868 +
19869 +struct vfsmount;
19870 +
19871 +struct au_hnotify {
19872 +#ifdef CONFIG_AUFS_HNOTIFY
19873 +#ifdef CONFIG_AUFS_HFSNOTIFY
19874 +       /* never use fsnotify_add_vfsmount_mark() */
19875 +       struct fsnotify_mark            hn_mark;
19876 +#endif
19877 +       struct inode            *hn_aufs_inode; /* no get/put */
19878 +       struct rcu_head         rcu;
19879 +#endif
19880 +} ____cacheline_aligned_in_smp;
19881 +
19882 +struct au_hinode {
19883 +       struct inode            *hi_inode;
19884 +       aufs_bindex_t           hi_id;
19885 +#ifdef CONFIG_AUFS_HNOTIFY
19886 +       struct au_hnotify       *hi_notify;
19887 +#endif
19888 +
19889 +       /* reference to the copied-up whiteout with get/put */
19890 +       struct dentry           *hi_whdentry;
19891 +};
19892 +
19893 +/* ig_flags */
19894 +#define AuIG_HALF_REFRESHED            1
19895 +#define au_ig_ftest(flags, name)       ((flags) & AuIG_##name)
19896 +#define au_ig_fset(flags, name) \
19897 +       do { (flags) |= AuIG_##name; } while (0)
19898 +#define au_ig_fclr(flags, name) \
19899 +       do { (flags) &= ~AuIG_##name; } while (0)
19900 +
19901 +struct au_iigen {
19902 +       spinlock_t      ig_spin;
19903 +       __u32           ig_generation, ig_flags;
19904 +};
19905 +
19906 +struct au_vdir;
19907 +struct au_iinfo {
19908 +       struct au_iigen         ii_generation;
19909 +       struct super_block      *ii_hsb1;       /* no get/put */
19910 +
19911 +       struct au_rwsem         ii_rwsem;
19912 +       aufs_bindex_t           ii_btop, ii_bbot;
19913 +       __u32                   ii_higen;
19914 +       struct au_hinode        *ii_hinode;
19915 +       struct au_vdir          *ii_vdir;
19916 +};
19917 +
19918 +struct au_icntnr {
19919 +       struct au_iinfo         iinfo;
19920 +       struct inode            vfs_inode;
19921 +       struct hlist_bl_node    plink;
19922 +       struct rcu_head         rcu;
19923 +} ____cacheline_aligned_in_smp;
19924 +
19925 +/* au_pin flags */
19926 +#define AuPin_DI_LOCKED                1
19927 +#define AuPin_MNT_WRITE                (1 << 1)
19928 +#define au_ftest_pin(flags, name)      ((flags) & AuPin_##name)
19929 +#define au_fset_pin(flags, name) \
19930 +       do { (flags) |= AuPin_##name; } while (0)
19931 +#define au_fclr_pin(flags, name) \
19932 +       do { (flags) &= ~AuPin_##name; } while (0)
19933 +
19934 +struct au_pin {
19935 +       /* input */
19936 +       struct dentry *dentry;
19937 +       unsigned int udba;
19938 +       unsigned char lsc_di, lsc_hi, flags;
19939 +       aufs_bindex_t bindex;
19940 +
19941 +       /* output */
19942 +       struct dentry *parent;
19943 +       struct au_hinode *hdir;
19944 +       struct vfsmount *h_mnt;
19945 +
19946 +       /* temporary unlock/relock for copyup */
19947 +       struct dentry *h_dentry, *h_parent;
19948 +       struct au_branch *br;
19949 +       struct task_struct *task;
19950 +};
19951 +
19952 +void au_pin_hdir_unlock(struct au_pin *p);
19953 +int au_pin_hdir_lock(struct au_pin *p);
19954 +int au_pin_hdir_relock(struct au_pin *p);
19955 +void au_pin_hdir_acquire_nest(struct au_pin *p);
19956 +void au_pin_hdir_release(struct au_pin *p);
19957 +
19958 +/* ---------------------------------------------------------------------- */
19959 +
19960 +static inline struct au_iinfo *au_ii(struct inode *inode)
19961 +{
19962 +       BUG_ON(is_bad_inode(inode));
19963 +       return &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19964 +}
19965 +
19966 +/* ---------------------------------------------------------------------- */
19967 +
19968 +/* inode.c */
19969 +struct inode *au_igrab(struct inode *inode);
19970 +void au_refresh_iop(struct inode *inode, int force_getattr);
19971 +int au_refresh_hinode_self(struct inode *inode);
19972 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
19973 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19974 +          unsigned int d_type, ino_t *ino);
19975 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
19976 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19977 +              struct inode *inode);
19978 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19979 +                  int mask);
19980 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19981 +                      int mask);
19982 +
19983 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
19984 +                           ino_t h_ino, unsigned int d_type, ino_t *ino)
19985 +{
19986 +#ifdef CONFIG_AUFS_SHWH
19987 +       return au_ino(sb, bindex, h_ino, d_type, ino);
19988 +#else
19989 +       return 0;
19990 +#endif
19991 +}
19992 +
19993 +/* i_op.c */
19994 +enum {
19995 +       AuIop_SYMLINK,
19996 +       AuIop_DIR,
19997 +       AuIop_OTHER,
19998 +       AuIop_Last
19999 +};
20000 +extern struct inode_operations aufs_iop[AuIop_Last], /* not const */
20001 +       aufs_iop_nogetattr[AuIop_Last];
20002 +
20003 +/* au_wr_dir flags */
20004 +#define AuWrDir_ADD_ENTRY      1
20005 +#define AuWrDir_ISDIR          (1 << 1)
20006 +#define AuWrDir_TMPFILE                (1 << 2)
20007 +#define au_ftest_wrdir(flags, name)    ((flags) & AuWrDir_##name)
20008 +#define au_fset_wrdir(flags, name) \
20009 +       do { (flags) |= AuWrDir_##name; } while (0)
20010 +#define au_fclr_wrdir(flags, name) \
20011 +       do { (flags) &= ~AuWrDir_##name; } while (0)
20012 +
20013 +struct au_wr_dir_args {
20014 +       aufs_bindex_t force_btgt;
20015 +       unsigned char flags;
20016 +};
20017 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
20018 +             struct au_wr_dir_args *args);
20019 +
20020 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
20021 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
20022 +                aufs_bindex_t bindex, int lsc_di, int lsc_hi,
20023 +                unsigned int udba, unsigned char flags);
20024 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
20025 +          unsigned int udba, unsigned char flags) __must_check;
20026 +int au_do_pin(struct au_pin *pin) __must_check;
20027 +void au_unpin(struct au_pin *pin);
20028 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
20029 +
20030 +#define AuIcpup_DID_CPUP       1
20031 +#define au_ftest_icpup(flags, name)    ((flags) & AuIcpup_##name)
20032 +#define au_fset_icpup(flags, name) \
20033 +       do { (flags) |= AuIcpup_##name; } while (0)
20034 +#define au_fclr_icpup(flags, name) \
20035 +       do { (flags) &= ~AuIcpup_##name; } while (0)
20036 +
20037 +struct au_icpup_args {
20038 +       unsigned char flags;
20039 +       unsigned char pin_flags;
20040 +       aufs_bindex_t btgt;
20041 +       unsigned int udba;
20042 +       struct au_pin pin;
20043 +       struct path h_path;
20044 +       struct inode *h_inode;
20045 +};
20046 +
20047 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
20048 +                    struct au_icpup_args *a);
20049 +
20050 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
20051 +                     struct path *h_path, int locked);
20052 +
20053 +/* i_op_add.c */
20054 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20055 +              struct dentry *h_parent, int isdir);
20056 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
20057 +              struct dentry *dentry, umode_t mode, dev_t dev);
20058 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
20059 +                struct dentry *dentry, const char *symname);
20060 +int aufs_create(struct user_namespace *userns, struct inode *dir,
20061 +               struct dentry *dentry, umode_t mode, bool want_excl);
20062 +struct vfsub_aopen_args;
20063 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
20064 +                      struct vfsub_aopen_args *args);
20065 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
20066 +                struct file *file, umode_t mode);
20067 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
20068 +             struct dentry *dentry);
20069 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
20070 +              struct dentry *dentry, umode_t mode);
20071 +
20072 +/* i_op_del.c */
20073 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
20074 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
20075 +              struct dentry *h_parent, int isdir);
20076 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
20077 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
20078 +
20079 +/* i_op_ren.c */
20080 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
20081 +int aufs_rename(struct user_namespace *userns,
20082 +               struct inode *_src_dir, struct dentry *_src_dentry,
20083 +               struct inode *_dst_dir, struct dentry *_dst_dentry,
20084 +               unsigned int _flags);
20085 +
20086 +/* iinfo.c */
20087 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
20088 +void au_hiput(struct au_hinode *hinode);
20089 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
20090 +                 struct dentry *h_wh);
20091 +unsigned int au_hi_flags(struct inode *inode, int isdir);
20092 +
20093 +/* hinode flags */
20094 +#define AuHi_XINO      1
20095 +#define AuHi_HNOTIFY   (1 << 1)
20096 +#define au_ftest_hi(flags, name)       ((flags) & AuHi_##name)
20097 +#define au_fset_hi(flags, name) \
20098 +       do { (flags) |= AuHi_##name; } while (0)
20099 +#define au_fclr_hi(flags, name) \
20100 +       do { (flags) &= ~AuHi_##name; } while (0)
20101 +
20102 +#ifndef CONFIG_AUFS_HNOTIFY
20103 +#undef AuHi_HNOTIFY
20104 +#define AuHi_HNOTIFY   0
20105 +#endif
20106 +
20107 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
20108 +                  struct inode *h_inode, unsigned int flags);
20109 +
20110 +void au_update_iigen(struct inode *inode, int half);
20111 +void au_update_ibrange(struct inode *inode, int do_put_zero);
20112 +
20113 +void au_icntnr_init_once(void *_c);
20114 +void au_hinode_init(struct au_hinode *hinode);
20115 +int au_iinfo_init(struct inode *inode);
20116 +void au_iinfo_fin(struct inode *inode);
20117 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink);
20118 +
20119 +#ifdef CONFIG_PROC_FS
20120 +/* plink.c */
20121 +int au_plink_maint(struct super_block *sb, int flags);
20122 +struct au_sbinfo;
20123 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
20124 +int au_plink_maint_enter(struct super_block *sb);
20125 +#ifdef CONFIG_AUFS_DEBUG
20126 +void au_plink_list(struct super_block *sb);
20127 +#else
20128 +AuStubVoid(au_plink_list, struct super_block *sb)
20129 +#endif
20130 +int au_plink_test(struct inode *inode);
20131 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
20132 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
20133 +                    struct dentry *h_dentry);
20134 +void au_plink_put(struct super_block *sb, int verbose);
20135 +void au_plink_clean(struct super_block *sb, int verbose);
20136 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
20137 +#else
20138 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
20139 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
20140 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
20141 +AuStubVoid(au_plink_list, struct super_block *sb);
20142 +AuStubInt0(au_plink_test, struct inode *inode);
20143 +AuStub(struct dentry *, au_plink_lkup, return NULL,
20144 +       struct inode *inode, aufs_bindex_t bindex);
20145 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
20146 +          struct dentry *h_dentry);
20147 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
20148 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
20149 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
20150 +#endif /* CONFIG_PROC_FS */
20151 +
20152 +#ifdef CONFIG_AUFS_XATTR
20153 +/* xattr.c */
20154 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
20155 +                 unsigned int verbose);
20156 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
20157 +void au_xattr_init(struct super_block *sb);
20158 +#else
20159 +AuStubInt0(au_cpup_xattr, struct path *h_dst, struct path *h_src,
20160 +          int ignore_flags, unsigned int verbose);
20161 +AuStubVoid(au_xattr_init, struct super_block *sb);
20162 +#endif
20163 +
20164 +#ifdef CONFIG_FS_POSIX_ACL
20165 +struct posix_acl *aufs_get_inode_acl(struct inode *inode, int type, bool rcu);
20166 +struct posix_acl *aufs_get_acl(struct user_namespace *userns,
20167 +                              struct dentry *dentry, int type);
20168 +int aufs_set_acl(struct user_namespace *userns, struct dentry *dentry,
20169 +                struct posix_acl *acl, int type);
20170 +#endif
20171 +
20172 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
20173 +enum {
20174 +       AU_XATTR_SET,
20175 +       AU_ACL_SET
20176 +};
20177 +
20178 +struct au_sxattr {
20179 +       int type;
20180 +       union {
20181 +               struct {
20182 +                       const char      *name;
20183 +                       const void      *value;
20184 +                       size_t          size;
20185 +                       int             flags;
20186 +               } set;
20187 +               struct {
20188 +                       struct posix_acl *acl;
20189 +                       int             type;
20190 +               } acl_set;
20191 +       } u;
20192 +};
20193 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
20194 +                 struct au_sxattr *arg);
20195 +#endif
20196 +
20197 +/* ---------------------------------------------------------------------- */
20198 +
20199 +/* lock subclass for iinfo */
20200 +enum {
20201 +       AuLsc_II_CHILD,         /* child first */
20202 +       AuLsc_II_CHILD2,        /* rename(2), link(2), and cpup at hnotify */
20203 +       AuLsc_II_CHILD3,        /* copyup dirs */
20204 +       AuLsc_II_PARENT,        /* see AuLsc_I_PARENT in vfsub.h */
20205 +       AuLsc_II_PARENT2,
20206 +       AuLsc_II_PARENT3,       /* copyup dirs */
20207 +       AuLsc_II_NEW_CHILD
20208 +};
20209 +
20210 +/*
20211 + * ii_read_lock_child, ii_write_lock_child,
20212 + * ii_read_lock_child2, ii_write_lock_child2,
20213 + * ii_read_lock_child3, ii_write_lock_child3,
20214 + * ii_read_lock_parent, ii_write_lock_parent,
20215 + * ii_read_lock_parent2, ii_write_lock_parent2,
20216 + * ii_read_lock_parent3, ii_write_lock_parent3,
20217 + * ii_read_lock_new_child, ii_write_lock_new_child,
20218 + */
20219 +#define AuReadLockFunc(name, lsc) \
20220 +static inline void ii_read_lock_##name(struct inode *i) \
20221 +{ \
20222 +       au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20223 +}
20224 +
20225 +#define AuWriteLockFunc(name, lsc) \
20226 +static inline void ii_write_lock_##name(struct inode *i) \
20227 +{ \
20228 +       au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20229 +}
20230 +
20231 +#define AuRWLockFuncs(name, lsc) \
20232 +       AuReadLockFunc(name, lsc) \
20233 +       AuWriteLockFunc(name, lsc)
20234 +
20235 +AuRWLockFuncs(child, CHILD);
20236 +AuRWLockFuncs(child2, CHILD2);
20237 +AuRWLockFuncs(child3, CHILD3);
20238 +AuRWLockFuncs(parent, PARENT);
20239 +AuRWLockFuncs(parent2, PARENT2);
20240 +AuRWLockFuncs(parent3, PARENT3);
20241 +AuRWLockFuncs(new_child, NEW_CHILD);
20242 +
20243 +#undef AuReadLockFunc
20244 +#undef AuWriteLockFunc
20245 +#undef AuRWLockFuncs
20246 +
20247 +#define ii_read_unlock(i)      au_rw_read_unlock(&au_ii(i)->ii_rwsem)
20248 +#define ii_write_unlock(i)     au_rw_write_unlock(&au_ii(i)->ii_rwsem)
20249 +#define ii_downgrade_lock(i)   au_rw_dgrade_lock(&au_ii(i)->ii_rwsem)
20250 +
20251 +#define IiMustNoWaiters(i)     AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
20252 +#define IiMustAnyLock(i)       AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
20253 +#define IiMustWriteLock(i)     AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
20254 +
20255 +/* ---------------------------------------------------------------------- */
20256 +
20257 +static inline void au_icntnr_init(struct au_icntnr *c)
20258 +{
20259 +#ifdef CONFIG_AUFS_DEBUG
20260 +       c->vfs_inode.i_mode = 0;
20261 +#endif
20262 +}
20263 +
20264 +static inline unsigned int au_iigen(struct inode *inode, unsigned int *igflags)
20265 +{
20266 +       unsigned int gen;
20267 +       struct au_iinfo *iinfo;
20268 +       struct au_iigen *iigen;
20269 +
20270 +       iinfo = au_ii(inode);
20271 +       iigen = &iinfo->ii_generation;
20272 +       spin_lock(&iigen->ig_spin);
20273 +       if (igflags)
20274 +               *igflags = iigen->ig_flags;
20275 +       gen = iigen->ig_generation;
20276 +       spin_unlock(&iigen->ig_spin);
20277 +
20278 +       return gen;
20279 +}
20280 +
20281 +/* tiny test for inode number */
20282 +/* tmpfs generation is too rough */
20283 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
20284 +{
20285 +       struct au_iinfo *iinfo;
20286 +
20287 +       iinfo = au_ii(inode);
20288 +       AuRwMustAnyLock(&iinfo->ii_rwsem);
20289 +       return !(iinfo->ii_hsb1 == h_inode->i_sb
20290 +                && iinfo->ii_higen == h_inode->i_generation);
20291 +}
20292 +
20293 +static inline void au_iigen_dec(struct inode *inode)
20294 +{
20295 +       struct au_iinfo *iinfo;
20296 +       struct au_iigen *iigen;
20297 +
20298 +       iinfo = au_ii(inode);
20299 +       iigen = &iinfo->ii_generation;
20300 +       spin_lock(&iigen->ig_spin);
20301 +       iigen->ig_generation--;
20302 +       spin_unlock(&iigen->ig_spin);
20303 +}
20304 +
20305 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
20306 +{
20307 +       int err;
20308 +
20309 +       err = 0;
20310 +       if (unlikely(inode && au_iigen(inode, NULL) != sigen))
20311 +               err = -EIO;
20312 +
20313 +       return err;
20314 +}
20315 +
20316 +/* ---------------------------------------------------------------------- */
20317 +
20318 +static inline struct au_hinode *au_hinode(struct au_iinfo *iinfo,
20319 +                                         aufs_bindex_t bindex)
20320 +{
20321 +       return iinfo->ii_hinode + bindex;
20322 +}
20323 +
20324 +static inline int au_is_bad_inode(struct inode *inode)
20325 +{
20326 +       return !!(is_bad_inode(inode) || !au_hinode(au_ii(inode), 0));
20327 +}
20328 +
20329 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
20330 +                                       aufs_bindex_t bindex)
20331 +{
20332 +       IiMustAnyLock(inode);
20333 +       return au_hinode(au_ii(inode), bindex)->hi_id;
20334 +}
20335 +
20336 +static inline aufs_bindex_t au_ibtop(struct inode *inode)
20337 +{
20338 +       IiMustAnyLock(inode);
20339 +       return au_ii(inode)->ii_btop;
20340 +}
20341 +
20342 +static inline aufs_bindex_t au_ibbot(struct inode *inode)
20343 +{
20344 +       IiMustAnyLock(inode);
20345 +       return au_ii(inode)->ii_bbot;
20346 +}
20347 +
20348 +static inline struct au_vdir *au_ivdir(struct inode *inode)
20349 +{
20350 +       IiMustAnyLock(inode);
20351 +       return au_ii(inode)->ii_vdir;
20352 +}
20353 +
20354 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
20355 +{
20356 +       IiMustAnyLock(inode);
20357 +       return au_hinode(au_ii(inode), bindex)->hi_whdentry;
20358 +}
20359 +
20360 +static inline void au_set_ibtop(struct inode *inode, aufs_bindex_t bindex)
20361 +{
20362 +       IiMustWriteLock(inode);
20363 +       au_ii(inode)->ii_btop = bindex;
20364 +}
20365 +
20366 +static inline void au_set_ibbot(struct inode *inode, aufs_bindex_t bindex)
20367 +{
20368 +       IiMustWriteLock(inode);
20369 +       au_ii(inode)->ii_bbot = bindex;
20370 +}
20371 +
20372 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
20373 +{
20374 +       IiMustWriteLock(inode);
20375 +       au_ii(inode)->ii_vdir = vdir;
20376 +}
20377 +
20378 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
20379 +{
20380 +       IiMustAnyLock(inode);
20381 +       return au_hinode(au_ii(inode), bindex);
20382 +}
20383 +
20384 +/* ---------------------------------------------------------------------- */
20385 +
20386 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
20387 +{
20388 +       if (pin)
20389 +               return pin->parent;
20390 +       return NULL;
20391 +}
20392 +
20393 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
20394 +{
20395 +       if (pin && pin->hdir)
20396 +               return pin->hdir->hi_inode;
20397 +       return NULL;
20398 +}
20399 +
20400 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
20401 +{
20402 +       if (pin)
20403 +               return pin->hdir;
20404 +       return NULL;
20405 +}
20406 +
20407 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
20408 +{
20409 +       if (pin)
20410 +               pin->dentry = dentry;
20411 +}
20412 +
20413 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
20414 +                                          unsigned char lflag)
20415 +{
20416 +       if (pin) {
20417 +               if (lflag)
20418 +                       au_fset_pin(pin->flags, DI_LOCKED);
20419 +               else
20420 +                       au_fclr_pin(pin->flags, DI_LOCKED);
20421 +       }
20422 +}
20423 +
20424 +#if 0 /* reserved */
20425 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
20426 +{
20427 +       if (pin) {
20428 +               dput(pin->parent);
20429 +               pin->parent = dget(parent);
20430 +       }
20431 +}
20432 +#endif
20433 +
20434 +/* ---------------------------------------------------------------------- */
20435 +
20436 +struct au_branch;
20437 +#ifdef CONFIG_AUFS_HNOTIFY
20438 +struct au_hnotify_op {
20439 +       void (*ctl)(struct au_hinode *hinode, int do_set);
20440 +       int (*alloc)(struct au_hinode *hinode);
20441 +
20442 +       /*
20443 +        * if it returns true, the caller should free hinode->hi_notify,
20444 +        * otherwise ->free() frees it.
20445 +        */
20446 +       int (*free)(struct au_hinode *hinode,
20447 +                   struct au_hnotify *hn) __must_check;
20448 +
20449 +       void (*fin)(void);
20450 +       int (*init)(void);
20451 +
20452 +       int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
20453 +       void (*fin_br)(struct au_branch *br);
20454 +       int (*init_br)(struct au_branch *br, int perm);
20455 +};
20456 +
20457 +/* hnotify.c */
20458 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
20459 +void au_hn_free(struct au_hinode *hinode);
20460 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
20461 +void au_hn_reset(struct inode *inode, unsigned int flags);
20462 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
20463 +              const struct qstr *h_child_qstr, struct inode *h_child_inode);
20464 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
20465 +int au_hnotify_init_br(struct au_branch *br, int perm);
20466 +void au_hnotify_fin_br(struct au_branch *br);
20467 +int __init au_hnotify_init(void);
20468 +void au_hnotify_fin(void);
20469 +
20470 +/* hfsnotify.c */
20471 +extern const struct au_hnotify_op au_hnotify_op;
20472 +
20473 +static inline
20474 +void au_hn_init(struct au_hinode *hinode)
20475 +{
20476 +       hinode->hi_notify = NULL;
20477 +}
20478 +
20479 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
20480 +{
20481 +       return hinode->hi_notify;
20482 +}
20483 +
20484 +#else
20485 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
20486 +       struct au_hinode *hinode __maybe_unused,
20487 +       struct inode *inode __maybe_unused)
20488 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
20489 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
20490 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
20491 +          int do_set __maybe_unused)
20492 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
20493 +          unsigned int flags __maybe_unused)
20494 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
20495 +          struct au_branch *br __maybe_unused,
20496 +          int perm __maybe_unused)
20497 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
20498 +          int perm __maybe_unused)
20499 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
20500 +AuStubInt0(__init au_hnotify_init, void)
20501 +AuStubVoid(au_hnotify_fin, void)
20502 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
20503 +#endif /* CONFIG_AUFS_HNOTIFY */
20504 +
20505 +static inline void au_hn_suspend(struct au_hinode *hdir)
20506 +{
20507 +       au_hn_ctl(hdir, /*do_set*/0);
20508 +}
20509 +
20510 +static inline void au_hn_resume(struct au_hinode *hdir)
20511 +{
20512 +       au_hn_ctl(hdir, /*do_set*/1);
20513 +}
20514 +
20515 +static inline void au_hn_inode_lock(struct au_hinode *hdir)
20516 +{
20517 +       inode_lock(hdir->hi_inode);
20518 +       au_hn_suspend(hdir);
20519 +}
20520 +
20521 +static inline void au_hn_inode_lock_nested(struct au_hinode *hdir,
20522 +                                         unsigned int sc __maybe_unused)
20523 +{
20524 +       inode_lock_nested(hdir->hi_inode, sc);
20525 +       au_hn_suspend(hdir);
20526 +}
20527 +
20528 +#if 0 /* unused */
20529 +#include "vfsub.h"
20530 +static inline void au_hn_inode_lock_shared_nested(struct au_hinode *hdir,
20531 +                                                 unsigned int sc)
20532 +{
20533 +       inode_lock_shared_nested(hdir->hi_inode, sc);
20534 +       au_hn_suspend(hdir);
20535 +}
20536 +#endif
20537 +
20538 +static inline void au_hn_inode_unlock(struct au_hinode *hdir)
20539 +{
20540 +       au_hn_resume(hdir);
20541 +       inode_unlock(hdir->hi_inode);
20542 +}
20543 +
20544 +#endif /* __KERNEL__ */
20545 +#endif /* __AUFS_INODE_H__ */
20546 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
20547 --- /usr/share/empty/fs/aufs/ioctl.c    1970-01-01 01:00:00.000000000 +0100
20548 +++ linux/fs/aufs/ioctl.c       2022-11-05 23:02:18.965889284 +0100
20549 @@ -0,0 +1,220 @@
20550 +// SPDX-License-Identifier: GPL-2.0
20551 +/*
20552 + * Copyright (C) 2005-2022 Junjiro R. Okajima
20553 + *
20554 + * This program is free software; you can redistribute it and/or modify
20555 + * it under the terms of the GNU General Public License as published by
20556 + * the Free Software Foundation; either version 2 of the License, or
20557 + * (at your option) any later version.
20558 + *
20559 + * This program is distributed in the hope that it will be useful,
20560 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20561 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20562 + * GNU General Public License for more details.
20563 + *
20564 + * You should have received a copy of the GNU General Public License
20565 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
20566 + */
20567 +
20568 +/*
20569 + * ioctl
20570 + * plink-management and readdir in userspace.
20571 + * assist the pathconf(3) wrapper library.
20572 + * move-down
20573 + * File-based Hierarchical Storage Management.
20574 + */
20575 +
20576 +#include <linux/compat.h>
20577 +#include <linux/file.h>
20578 +#include "aufs.h"
20579 +
20580 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
20581 +{
20582 +       int err, fd;
20583 +       aufs_bindex_t wbi, bindex, bbot;
20584 +       struct file *h_file;
20585 +       struct super_block *sb;
20586 +       struct dentry *root;
20587 +       struct au_branch *br;
20588 +       struct aufs_wbr_fd wbrfd = {
20589 +               .oflags = au_dir_roflags,
20590 +               .brid   = -1
20591 +       };
20592 +       const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
20593 +               | O_NOATIME | O_CLOEXEC;
20594 +
20595 +       AuDebugOn(wbrfd.oflags & ~valid);
20596 +
20597 +       if (arg) {
20598 +               err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
20599 +               if (unlikely(err)) {
20600 +                       err = -EFAULT;
20601 +                       goto out;
20602 +               }
20603 +
20604 +               err = -EINVAL;
20605 +               AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
20606 +               wbrfd.oflags |= au_dir_roflags;
20607 +               AuDbg("0%o\n", wbrfd.oflags);
20608 +               if (unlikely(wbrfd.oflags & ~valid))
20609 +                       goto out;
20610 +       }
20611 +
20612 +       fd = get_unused_fd_flags(0);
20613 +       err = fd;
20614 +       if (unlikely(fd < 0))
20615 +               goto out;
20616 +
20617 +       h_file = ERR_PTR(-EINVAL);
20618 +       wbi = 0;
20619 +       br = NULL;
20620 +       sb = path->dentry->d_sb;
20621 +       root = sb->s_root;
20622 +       aufs_read_lock(root, AuLock_IR);
20623 +       bbot = au_sbbot(sb);
20624 +       if (wbrfd.brid >= 0) {
20625 +               wbi = au_br_index(sb, wbrfd.brid);
20626 +               if (unlikely(wbi < 0 || wbi > bbot))
20627 +                       goto out_unlock;
20628 +       }
20629 +
20630 +       h_file = ERR_PTR(-ENOENT);
20631 +       br = au_sbr(sb, wbi);
20632 +       if (!au_br_writable(br->br_perm)) {
20633 +               if (arg)
20634 +                       goto out_unlock;
20635 +
20636 +               bindex = wbi + 1;
20637 +               wbi = -1;
20638 +               for (; bindex <= bbot; bindex++) {
20639 +                       br = au_sbr(sb, bindex);
20640 +                       if (au_br_writable(br->br_perm)) {
20641 +                               wbi = bindex;
20642 +                               br = au_sbr(sb, wbi);
20643 +                               break;
20644 +                       }
20645 +               }
20646 +       }
20647 +       AuDbg("wbi %d\n", wbi);
20648 +       if (wbi >= 0)
20649 +               h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
20650 +                                  /*force_wr*/0);
20651 +
20652 +out_unlock:
20653 +       aufs_read_unlock(root, AuLock_IR);
20654 +       err = PTR_ERR(h_file);
20655 +       if (IS_ERR(h_file))
20656 +               goto out_fd;
20657 +
20658 +       au_lcnt_dec(&br->br_nfiles); /* cf. au_h_open() */
20659 +       fd_install(fd, h_file);
20660 +       err = fd;
20661 +       goto out; /* success */
20662 +
20663 +out_fd:
20664 +       put_unused_fd(fd);
20665 +out:
20666 +       AuTraceErr(err);
20667 +       return err;
20668 +}
20669 +
20670 +/* ---------------------------------------------------------------------- */
20671 +
20672 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
20673 +{
20674 +       long err;
20675 +       struct dentry *dentry;
20676 +
20677 +       switch (cmd) {
20678 +       case AUFS_CTL_RDU:
20679 +       case AUFS_CTL_RDU_INO:
20680 +               err = au_rdu_ioctl(file, cmd, arg);
20681 +               break;
20682 +
20683 +       case AUFS_CTL_WBR_FD:
20684 +               err = au_wbr_fd(&file->f_path, (void __user *)arg);
20685 +               break;
20686 +
20687 +       case AUFS_CTL_IBUSY:
20688 +               err = au_ibusy_ioctl(file, arg);
20689 +               break;
20690 +
20691 +       case AUFS_CTL_BRINFO:
20692 +               err = au_brinfo_ioctl(file, arg);
20693 +               break;
20694 +
20695 +       case AUFS_CTL_FHSM_FD:
20696 +               dentry = file->f_path.dentry;
20697 +               if (IS_ROOT(dentry))
20698 +                       err = au_fhsm_fd(dentry->d_sb, arg);
20699 +               else
20700 +                       err = -ENOTTY;
20701 +               break;
20702 +
20703 +       default:
20704 +               /* do not call the lower */
20705 +               AuDbg("0x%x\n", cmd);
20706 +               err = -ENOTTY;
20707 +       }
20708 +
20709 +       AuTraceErr(err);
20710 +       return err;
20711 +}
20712 +
20713 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
20714 +{
20715 +       long err;
20716 +
20717 +       switch (cmd) {
20718 +       case AUFS_CTL_MVDOWN:
20719 +               err = au_mvdown(file->f_path.dentry, (void __user *)arg);
20720 +               break;
20721 +
20722 +       case AUFS_CTL_WBR_FD:
20723 +               err = au_wbr_fd(&file->f_path, (void __user *)arg);
20724 +               break;
20725 +
20726 +       default:
20727 +               /* do not call the lower */
20728 +               AuDbg("0x%x\n", cmd);
20729 +               err = -ENOTTY;
20730 +       }
20731 +
20732 +       AuTraceErr(err);
20733 +       return err;
20734 +}
20735 +
20736 +#ifdef CONFIG_COMPAT
20737 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
20738 +                          unsigned long arg)
20739 +{
20740 +       long err;
20741 +
20742 +       switch (cmd) {
20743 +       case AUFS_CTL_RDU:
20744 +       case AUFS_CTL_RDU_INO:
20745 +               err = au_rdu_compat_ioctl(file, cmd, arg);
20746 +               break;
20747 +
20748 +       case AUFS_CTL_IBUSY:
20749 +               err = au_ibusy_compat_ioctl(file, arg);
20750 +               break;
20751 +
20752 +       case AUFS_CTL_BRINFO:
20753 +               err = au_brinfo_compat_ioctl(file, arg);
20754 +               break;
20755 +
20756 +       default:
20757 +               err = aufs_ioctl_dir(file, cmd, arg);
20758 +       }
20759 +
20760 +       AuTraceErr(err);
20761 +       return err;
20762 +}
20763 +
20764 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
20765 +                             unsigned long arg)
20766 +{
20767 +       return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
20768 +}
20769 +#endif
20770 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
20771 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
20772 +++ linux/fs/aufs/i_op_add.c    2023-02-20 21:05:51.959693785 +0100
20773 @@ -0,0 +1,972 @@
20774 +// SPDX-License-Identifier: GPL-2.0
20775 +/*
20776 + * Copyright (C) 2005-2022 Junjiro R. Okajima
20777 + *
20778 + * This program is free software; you can redistribute it and/or modify
20779 + * it under the terms of the GNU General Public License as published by
20780 + * the Free Software Foundation; either version 2 of the License, or
20781 + * (at your option) any later version.
20782 + *
20783 + * This program is distributed in the hope that it will be useful,
20784 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20785 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20786 + * GNU General Public License for more details.
20787 + *
20788 + * You should have received a copy of the GNU General Public License
20789 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
20790 + */
20791 +
20792 +/*
20793 + * inode operations (add entry)
20794 + */
20795 +
20796 +#include <linux/iversion.h>
20797 +#include "aufs.h"
20798 +
20799 +/*
20800 + * final procedure of adding a new entry, except link(2).
20801 + * remove whiteout, instantiate, copyup the parent dir's times and size
20802 + * and update version.
20803 + * if it failed, re-create the removed whiteout.
20804 + */
20805 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
20806 +                 struct dentry *wh_dentry, struct dentry *dentry)
20807 +{
20808 +       int err, rerr;
20809 +       aufs_bindex_t bwh;
20810 +       struct path h_path;
20811 +       struct super_block *sb;
20812 +       struct inode *inode, *h_dir;
20813 +       struct dentry *wh;
20814 +
20815 +       bwh = -1;
20816 +       sb = dir->i_sb;
20817 +       if (wh_dentry) {
20818 +               h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
20819 +               IMustLock(h_dir);
20820 +               AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
20821 +               bwh = au_dbwh(dentry);
20822 +               h_path.dentry = wh_dentry;
20823 +               h_path.mnt = au_sbr_mnt(sb, bindex);
20824 +               err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
20825 +                                         dentry);
20826 +               if (unlikely(err))
20827 +                       goto out;
20828 +       }
20829 +
20830 +       inode = au_new_inode(dentry, /*must_new*/1);
20831 +       if (!IS_ERR(inode)) {
20832 +               d_instantiate(dentry, inode);
20833 +               dir = d_inode(dentry->d_parent); /* dir inode is locked */
20834 +               IMustLock(dir);
20835 +               au_dir_ts(dir, bindex);
20836 +               inode_inc_iversion(dir);
20837 +               au_fhsm_wrote(sb, bindex, /*force*/0);
20838 +               return 0; /* success */
20839 +       }
20840 +
20841 +       err = PTR_ERR(inode);
20842 +       if (!wh_dentry)
20843 +               goto out;
20844 +
20845 +       /* revert */
20846 +       /* dir inode is locked */
20847 +       wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
20848 +       rerr = PTR_ERR(wh);
20849 +       if (IS_ERR(wh)) {
20850 +               AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
20851 +                       dentry, err, rerr);
20852 +               err = -EIO;
20853 +       } else
20854 +               dput(wh);
20855 +
20856 +out:
20857 +       return err;
20858 +}
20859 +
20860 +static int au_d_may_add(struct dentry *dentry)
20861 +{
20862 +       int err;
20863 +
20864 +       err = 0;
20865 +       if (unlikely(d_unhashed(dentry)))
20866 +               err = -ENOENT;
20867 +       if (unlikely(d_really_is_positive(dentry)))
20868 +               err = -EEXIST;
20869 +       return err;
20870 +}
20871 +
20872 +/*
20873 + * simple tests for the adding inode operations.
20874 + * following the checks in vfs, plus the parent-child relationship.
20875 + */
20876 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20877 +              struct dentry *h_parent, int isdir)
20878 +{
20879 +       int err;
20880 +       umode_t h_mode;
20881 +       struct dentry *h_dentry;
20882 +       struct inode *h_inode;
20883 +
20884 +       err = -ENAMETOOLONG;
20885 +       if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20886 +               goto out;
20887 +
20888 +       h_dentry = au_h_dptr(dentry, bindex);
20889 +       if (d_really_is_negative(dentry)) {
20890 +               err = -EEXIST;
20891 +               if (unlikely(d_is_positive(h_dentry)))
20892 +                       goto out;
20893 +       } else {
20894 +               /* rename(2) case */
20895 +               err = -EIO;
20896 +               if (unlikely(d_is_negative(h_dentry)))
20897 +                       goto out;
20898 +               h_inode = d_inode(h_dentry);
20899 +               if (unlikely(!h_inode->i_nlink))
20900 +                       goto out;
20901 +
20902 +               h_mode = h_inode->i_mode;
20903 +               if (!isdir) {
20904 +                       err = -EISDIR;
20905 +                       if (unlikely(S_ISDIR(h_mode)))
20906 +                               goto out;
20907 +               } else if (unlikely(!S_ISDIR(h_mode))) {
20908 +                       err = -ENOTDIR;
20909 +                       goto out;
20910 +               }
20911 +       }
20912 +
20913 +       err = 0;
20914 +       /* expected parent dir is locked */
20915 +       if (unlikely(h_parent != h_dentry->d_parent))
20916 +               err = -EIO;
20917 +
20918 +out:
20919 +       AuTraceErr(err);
20920 +       return err;
20921 +}
20922 +
20923 +/*
20924 + * initial procedure of adding a new entry.
20925 + * prepare writable branch and the parent dir, lock it,
20926 + * and lookup whiteout for the new entry.
20927 + */
20928 +static struct dentry*
20929 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
20930 +                 struct dentry *src_dentry, struct au_pin *pin,
20931 +                 struct au_wr_dir_args *wr_dir_args)
20932 +{
20933 +       struct dentry *wh_dentry, *h_parent;
20934 +       struct super_block *sb;
20935 +       struct au_branch *br;
20936 +       int err;
20937 +       unsigned int udba;
20938 +       aufs_bindex_t bcpup;
20939 +
20940 +       AuDbg("%pd\n", dentry);
20941 +
20942 +       err = au_wr_dir(dentry, src_dentry, wr_dir_args);
20943 +       bcpup = err;
20944 +       wh_dentry = ERR_PTR(err);
20945 +       if (unlikely(err < 0))
20946 +               goto out;
20947 +
20948 +       sb = dentry->d_sb;
20949 +       udba = au_opt_udba(sb);
20950 +       err = au_pin(pin, dentry, bcpup, udba,
20951 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20952 +       wh_dentry = ERR_PTR(err);
20953 +       if (unlikely(err))
20954 +               goto out;
20955 +
20956 +       h_parent = au_pinned_h_parent(pin);
20957 +       if (udba != AuOpt_UDBA_NONE
20958 +           && au_dbtop(dentry) == bcpup)
20959 +               err = au_may_add(dentry, bcpup, h_parent,
20960 +                                au_ftest_wrdir(wr_dir_args->flags, ISDIR));
20961 +       else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20962 +               err = -ENAMETOOLONG;
20963 +       wh_dentry = ERR_PTR(err);
20964 +       if (unlikely(err))
20965 +               goto out_unpin;
20966 +
20967 +       br = au_sbr(sb, bcpup);
20968 +       if (dt) {
20969 +               struct path tmp = {
20970 +                       .dentry = h_parent,
20971 +                       .mnt    = au_br_mnt(br)
20972 +               };
20973 +               au_dtime_store(dt, au_pinned_parent(pin), &tmp);
20974 +       }
20975 +
20976 +       wh_dentry = NULL;
20977 +       if (bcpup != au_dbwh(dentry))
20978 +               goto out; /* success */
20979 +
20980 +       /*
20981 +        * ENAMETOOLONG here means that if we allowed create such name, then it
20982 +        * would not be able to removed in the future. So we don't allow such
20983 +        * name here and we don't handle ENAMETOOLONG differently here.
20984 +        */
20985 +       wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
20986 +
20987 +out_unpin:
20988 +       if (IS_ERR(wh_dentry))
20989 +               au_unpin(pin);
20990 +out:
20991 +       return wh_dentry;
20992 +}
20993 +
20994 +/* ---------------------------------------------------------------------- */
20995 +
20996 +enum { Mknod, Symlink, Creat };
20997 +struct simple_arg {
20998 +       int type;
20999 +       union {
21000 +               struct {
21001 +                       umode_t                 mode;
21002 +                       bool                    want_excl;
21003 +                       bool                    try_aopen;
21004 +                       struct vfsub_aopen_args *aopen;
21005 +               } c;
21006 +               struct {
21007 +                       const char *symname;
21008 +               } s;
21009 +               struct {
21010 +                       umode_t mode;
21011 +                       dev_t dev;
21012 +               } m;
21013 +       } u;
21014 +};
21015 +
21016 +static int add_simple(struct inode *dir, struct dentry *dentry,
21017 +                     struct simple_arg *arg)
21018 +{
21019 +       int err, rerr;
21020 +       aufs_bindex_t btop;
21021 +       unsigned char created;
21022 +       const unsigned char try_aopen
21023 +               = (arg->type == Creat && arg->u.c.try_aopen);
21024 +       struct vfsub_aopen_args *aopen = arg->u.c.aopen;
21025 +       struct dentry *wh_dentry, *parent;
21026 +       struct inode *h_dir;
21027 +       struct super_block *sb;
21028 +       struct au_branch *br;
21029 +       /* to reduce stack size */
21030 +       struct {
21031 +               struct au_dtime dt;
21032 +               struct au_pin pin;
21033 +               struct path h_path;
21034 +               struct au_wr_dir_args wr_dir_args;
21035 +       } *a;
21036 +
21037 +       AuDbg("%pd\n", dentry);
21038 +       IMustLock(dir);
21039 +
21040 +       err = -ENOMEM;
21041 +       a = kmalloc(sizeof(*a), GFP_NOFS);
21042 +       if (unlikely(!a))
21043 +               goto out;
21044 +       a->wr_dir_args.force_btgt = -1;
21045 +       a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
21046 +
21047 +       parent = dentry->d_parent; /* dir inode is locked */
21048 +       if (!try_aopen) {
21049 +               err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21050 +               if (unlikely(err))
21051 +                       goto out_free;
21052 +       }
21053 +       err = au_d_may_add(dentry);
21054 +       if (unlikely(err))
21055 +               goto out_unlock;
21056 +       if (!try_aopen)
21057 +               di_write_lock_parent(parent);
21058 +       wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21059 +                                     &a->pin, &a->wr_dir_args);
21060 +       err = PTR_ERR(wh_dentry);
21061 +       if (IS_ERR(wh_dentry))
21062 +               goto out_parent;
21063 +
21064 +       btop = au_dbtop(dentry);
21065 +       sb = dentry->d_sb;
21066 +       br = au_sbr(sb, btop);
21067 +       a->h_path.dentry = au_h_dptr(dentry, btop);
21068 +       a->h_path.mnt = au_br_mnt(br);
21069 +       h_dir = au_pinned_h_dir(&a->pin);
21070 +       switch (arg->type) {
21071 +       case Creat:
21072 +               if (!try_aopen || !h_dir->i_op->atomic_open) {
21073 +                       err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
21074 +                                          arg->u.c.want_excl);
21075 +                       created = !err;
21076 +                       if (!err && try_aopen)
21077 +                               aopen->file->f_mode |= FMODE_CREATED;
21078 +               } else {
21079 +                       aopen->br = br;
21080 +                       err = vfsub_atomic_open(h_dir, a->h_path.dentry, aopen);
21081 +                       AuDbg("err %d\n", err);
21082 +                       AuDbgFile(aopen->file);
21083 +                       created = err >= 0
21084 +                               && !!(aopen->file->f_mode & FMODE_CREATED);
21085 +               }
21086 +               break;
21087 +       case Symlink:
21088 +               err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
21089 +               created = !err;
21090 +               break;
21091 +       case Mknod:
21092 +               err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
21093 +                                 arg->u.m.dev);
21094 +               created = !err;
21095 +               break;
21096 +       default:
21097 +               BUG();
21098 +       }
21099 +       if (unlikely(err < 0))
21100 +               goto out_unpin;
21101 +
21102 +       err = epilog(dir, btop, wh_dentry, dentry);
21103 +       if (!err)
21104 +               goto out_unpin; /* success */
21105 +
21106 +       /* revert */
21107 +       if (created /* && d_is_positive(a->h_path.dentry) */) {
21108 +               /* no delegation since it is just created */
21109 +               rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
21110 +                                   /*force*/0);
21111 +               if (rerr) {
21112 +                       AuIOErr("%pd revert failure(%d, %d)\n",
21113 +                               dentry, err, rerr);
21114 +                       err = -EIO;
21115 +               }
21116 +               au_dtime_revert(&a->dt);
21117 +       }
21118 +       if (try_aopen && h_dir->i_op->atomic_open
21119 +           && (aopen->file->f_mode & FMODE_OPENED))
21120 +               /* aopen->file is still opened */
21121 +               au_lcnt_dec(&aopen->br->br_nfiles);
21122 +
21123 +out_unpin:
21124 +       au_unpin(&a->pin);
21125 +       dput(wh_dentry);
21126 +out_parent:
21127 +       if (!try_aopen)
21128 +               di_write_unlock(parent);
21129 +out_unlock:
21130 +       if (unlikely(err)) {
21131 +               au_update_dbtop(dentry);
21132 +               d_drop(dentry);
21133 +       }
21134 +       if (!try_aopen)
21135 +               aufs_read_unlock(dentry, AuLock_DW);
21136 +out_free:
21137 +       au_kfree_rcu(a);
21138 +out:
21139 +       return err;
21140 +}
21141 +
21142 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
21143 +              struct dentry *dentry, umode_t mode, dev_t dev)
21144 +{
21145 +       struct simple_arg arg = {
21146 +               .type = Mknod,
21147 +               .u.m = {
21148 +                       .mode   = mode,
21149 +                       .dev    = dev
21150 +               }
21151 +       };
21152 +       return add_simple(dir, dentry, &arg);
21153 +}
21154 +
21155 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
21156 +                struct dentry *dentry, const char *symname)
21157 +{
21158 +       struct simple_arg arg = {
21159 +               .type = Symlink,
21160 +               .u.s.symname = symname
21161 +       };
21162 +       return add_simple(dir, dentry, &arg);
21163 +}
21164 +
21165 +int aufs_create(struct user_namespace *userns, struct inode *dir,
21166 +               struct dentry *dentry, umode_t mode, bool want_excl)
21167 +{
21168 +       struct simple_arg arg = {
21169 +               .type = Creat,
21170 +               .u.c = {
21171 +                       .mode           = mode,
21172 +                       .want_excl      = want_excl
21173 +               }
21174 +       };
21175 +       return add_simple(dir, dentry, &arg);
21176 +}
21177 +
21178 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
21179 +                      struct vfsub_aopen_args *aopen_args)
21180 +{
21181 +       struct simple_arg arg = {
21182 +               .type = Creat,
21183 +               .u.c = {
21184 +                       .mode           = aopen_args->create_mode,
21185 +                       .want_excl      = aopen_args->open_flag & O_EXCL,
21186 +                       .try_aopen      = true,
21187 +                       .aopen          = aopen_args
21188 +               }
21189 +       };
21190 +       return add_simple(dir, dentry, &arg);
21191 +}
21192 +
21193 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
21194 +                struct file *file, umode_t mode)
21195 +{
21196 +       int err;
21197 +       aufs_bindex_t bindex;
21198 +       struct path h_ppath;
21199 +       struct super_block *sb;
21200 +       struct au_branch *br;
21201 +       struct dentry *dentry, *parent, *h_parent, *h_dentry;
21202 +       struct inode *h_dir, *inode;
21203 +       struct vfsmount *h_mnt;
21204 +       struct user_namespace *h_userns;
21205 +       struct file *h_file;
21206 +       struct au_wr_dir_args wr_dir_args = {
21207 +               .force_btgt     = -1,
21208 +               .flags          = AuWrDir_TMPFILE
21209 +       };
21210 +
21211 +       /* copy-up may happen */
21212 +       inode_lock(dir);
21213 +
21214 +       h_file = NULL;
21215 +       sb = dir->i_sb;
21216 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21217 +       if (unlikely(err))
21218 +               goto out;
21219 +
21220 +       dentry = file->f_path.dentry;
21221 +       err = au_di_init(dentry);
21222 +       if (unlikely(err))
21223 +               goto out_si;
21224 +
21225 +       err = -EBUSY;
21226 +       parent = d_find_any_alias(dir);
21227 +       AuDebugOn(!parent);
21228 +       di_write_lock_parent(parent);
21229 +       if (unlikely(d_inode(parent) != dir))
21230 +               goto out_parent;
21231 +
21232 +       err = au_digen_test(parent, au_sigen(sb));
21233 +       if (unlikely(err))
21234 +               goto out_parent;
21235 +
21236 +       bindex = au_dbtop(parent);
21237 +       au_set_dbtop(dentry, bindex);
21238 +       au_set_dbbot(dentry, bindex);
21239 +       err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
21240 +       bindex = err;
21241 +       if (unlikely(err < 0))
21242 +               goto out_parent;
21243 +
21244 +       err = -EOPNOTSUPP;
21245 +       h_dir = au_h_iptr(dir, bindex);
21246 +       if (unlikely(!h_dir->i_op->tmpfile))
21247 +               goto out_parent;
21248 +
21249 +       br = au_sbr(sb, bindex);
21250 +       h_mnt = au_br_mnt(br);
21251 +       err = vfsub_mnt_want_write(h_mnt);
21252 +       if (unlikely(err))
21253 +               goto out_parent;
21254 +
21255 +       h_userns = mnt_user_ns(h_mnt);
21256 +       h_parent = au_h_dptr(parent, bindex);
21257 +       h_ppath.mnt = h_mnt;
21258 +       h_ppath.dentry = h_parent;
21259 +       h_file = vfs_tmpfile_open(h_userns, &h_ppath, mode, /*open_flag*/0,
21260 +                                 current_cred());
21261 +       if (IS_ERR(h_file)) {
21262 +               err = PTR_ERR(h_file);
21263 +               h_file = NULL;
21264 +               goto out_mnt;
21265 +       }
21266 +
21267 +       h_dentry = h_file->f_path.dentry;
21268 +       au_set_dbtop(dentry, bindex);
21269 +       au_set_dbbot(dentry, bindex);
21270 +       au_set_h_dptr(dentry, bindex, dget(h_dentry));
21271 +       inode = au_new_inode(dentry, /*must_new*/1);
21272 +       if (IS_ERR(inode)) {
21273 +               err = PTR_ERR(inode);
21274 +               au_set_h_dptr(dentry, bindex, NULL);
21275 +               au_set_dbtop(dentry, -1);
21276 +               au_set_dbbot(dentry, -1);
21277 +               goto out_h_file;
21278 +       }
21279 +
21280 +       if (!inode->i_nlink)
21281 +               set_nlink(inode, 1);
21282 +       d_tmpfile(file, inode);
21283 +       au_di(dentry)->di_tmpfile = 1;
21284 +       get_file(h_file);
21285 +       au_di(dentry)->di_htmpfile = h_file;
21286 +
21287 +       /* update without i_mutex */
21288 +       if (au_ibtop(dir) == au_dbtop(dentry))
21289 +               au_cpup_attr_timesizes(dir);
21290 +
21291 +out_h_file:
21292 +       fput(h_file);
21293 +out_mnt:
21294 +       vfsub_mnt_drop_write(h_mnt);
21295 +out_parent:
21296 +       di_write_unlock(parent);
21297 +       dput(parent);
21298 +       di_write_unlock(dentry);
21299 +       if (!err)
21300 +               goto out_si;
21301 +       if (h_file)
21302 +               fput(h_file);
21303 +       au_di(dentry)->di_htmpfile = NULL;
21304 +       au_di_fin(dentry);
21305 +       dentry->d_fsdata = NULL;
21306 +out_si:
21307 +       si_read_unlock(sb);
21308 +       if (!err && h_file) {
21309 +               /* finally... */
21310 +               err = finish_open_simple(file, err);
21311 +               if (!err)
21312 +                       au_lcnt_inc(&br->br_nfiles);
21313 +               else {
21314 +                       fput(h_file);
21315 +                       au_di(dentry)->di_htmpfile = NULL;
21316 +                       au_di_fin(dentry);
21317 +                       dentry->d_fsdata = NULL;
21318 +               }
21319 +       }
21320 +out:
21321 +       inode_unlock(dir);
21322 +       AuTraceErr(err);
21323 +       return err;
21324 +}
21325 +
21326 +/* ---------------------------------------------------------------------- */
21327 +
21328 +struct au_link_args {
21329 +       aufs_bindex_t bdst, bsrc;
21330 +       struct au_pin pin;
21331 +       struct path h_path;
21332 +       struct dentry *src_parent, *parent;
21333 +};
21334 +
21335 +static int au_cpup_before_link(struct dentry *src_dentry,
21336 +                              struct au_link_args *a)
21337 +{
21338 +       int err;
21339 +       struct dentry *h_src_dentry;
21340 +       struct au_cp_generic cpg = {
21341 +               .dentry = src_dentry,
21342 +               .bdst   = a->bdst,
21343 +               .bsrc   = a->bsrc,
21344 +               .len    = -1,
21345 +               .pin    = &a->pin,
21346 +               .flags  = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
21347 +       };
21348 +
21349 +       di_read_lock_parent(a->src_parent, AuLock_IR);
21350 +       err = au_test_and_cpup_dirs(src_dentry, a->bdst);
21351 +       if (unlikely(err))
21352 +               goto out;
21353 +
21354 +       h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
21355 +       err = au_pin(&a->pin, src_dentry, a->bdst,
21356 +                    au_opt_udba(src_dentry->d_sb),
21357 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21358 +       if (unlikely(err))
21359 +               goto out;
21360 +
21361 +       err = au_sio_cpup_simple(&cpg);
21362 +       au_unpin(&a->pin);
21363 +
21364 +out:
21365 +       di_read_unlock(a->src_parent, AuLock_IR);
21366 +       return err;
21367 +}
21368 +
21369 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
21370 +                          struct au_link_args *a)
21371 +{
21372 +       int err;
21373 +       unsigned char plink;
21374 +       aufs_bindex_t bbot;
21375 +       struct dentry *h_src_dentry;
21376 +       struct inode *h_inode, *inode, *delegated;
21377 +       struct super_block *sb;
21378 +       struct file *h_file;
21379 +
21380 +       plink = 0;
21381 +       h_inode = NULL;
21382 +       sb = src_dentry->d_sb;
21383 +       inode = d_inode(src_dentry);
21384 +       if (au_ibtop(inode) <= a->bdst)
21385 +               h_inode = au_h_iptr(inode, a->bdst);
21386 +       if (!h_inode || !h_inode->i_nlink) {
21387 +               /* copyup src_dentry as the name of dentry. */
21388 +               bbot = au_dbbot(dentry);
21389 +               if (bbot < a->bsrc)
21390 +                       au_set_dbbot(dentry, a->bsrc);
21391 +               au_set_h_dptr(dentry, a->bsrc,
21392 +                             dget(au_h_dptr(src_dentry, a->bsrc)));
21393 +               dget(a->h_path.dentry);
21394 +               au_set_h_dptr(dentry, a->bdst, NULL);
21395 +               AuDbg("temporary d_inode...\n");
21396 +               spin_lock(&dentry->d_lock);
21397 +               dentry->d_inode = d_inode(src_dentry); /* tmp */
21398 +               spin_unlock(&dentry->d_lock);
21399 +               h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
21400 +               if (IS_ERR(h_file))
21401 +                       err = PTR_ERR(h_file);
21402 +               else {
21403 +                       struct au_cp_generic cpg = {
21404 +                               .dentry = dentry,
21405 +                               .bdst   = a->bdst,
21406 +                               .bsrc   = -1,
21407 +                               .len    = -1,
21408 +                               .pin    = &a->pin,
21409 +                               .flags  = AuCpup_KEEPLINO
21410 +                       };
21411 +                       err = au_sio_cpup_simple(&cpg);
21412 +                       au_h_open_post(dentry, a->bsrc, h_file);
21413 +                       if (!err) {
21414 +                               dput(a->h_path.dentry);
21415 +                               a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21416 +                       } else
21417 +                               au_set_h_dptr(dentry, a->bdst,
21418 +                                             a->h_path.dentry);
21419 +               }
21420 +               spin_lock(&dentry->d_lock);
21421 +               dentry->d_inode = NULL; /* restore */
21422 +               spin_unlock(&dentry->d_lock);
21423 +               AuDbg("temporary d_inode...done\n");
21424 +               au_set_h_dptr(dentry, a->bsrc, NULL);
21425 +               au_set_dbbot(dentry, bbot);
21426 +       } else {
21427 +               /* the inode of src_dentry already exists on a.bdst branch */
21428 +               h_src_dentry = d_find_alias(h_inode);
21429 +               if (!h_src_dentry && au_plink_test(inode)) {
21430 +                       plink = 1;
21431 +                       h_src_dentry = au_plink_lkup(inode, a->bdst);
21432 +                       err = PTR_ERR(h_src_dentry);
21433 +                       if (IS_ERR(h_src_dentry))
21434 +                               goto out;
21435 +
21436 +                       if (unlikely(d_is_negative(h_src_dentry))) {
21437 +                               dput(h_src_dentry);
21438 +                               h_src_dentry = NULL;
21439 +                       }
21440 +
21441 +               }
21442 +               if (h_src_dentry) {
21443 +                       delegated = NULL;
21444 +                       err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21445 +                                        &a->h_path, &delegated);
21446 +                       if (unlikely(err == -EWOULDBLOCK)) {
21447 +                               pr_warn("cannot retry for NFSv4 delegation"
21448 +                                       " for an internal link\n");
21449 +                               iput(delegated);
21450 +                       }
21451 +                       dput(h_src_dentry);
21452 +               } else {
21453 +                       AuIOErr("no dentry found for hi%lu on b%d\n",
21454 +                               h_inode->i_ino, a->bdst);
21455 +                       err = -EIO;
21456 +               }
21457 +       }
21458 +
21459 +       if (!err && !plink)
21460 +               au_plink_append(inode, a->bdst, a->h_path.dentry);
21461 +
21462 +out:
21463 +       AuTraceErr(err);
21464 +       return err;
21465 +}
21466 +
21467 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
21468 +             struct dentry *dentry)
21469 +{
21470 +       int err, rerr;
21471 +       struct au_dtime dt;
21472 +       struct au_link_args *a;
21473 +       struct dentry *wh_dentry, *h_src_dentry;
21474 +       struct inode *inode, *delegated;
21475 +       struct super_block *sb;
21476 +       struct au_wr_dir_args wr_dir_args = {
21477 +               /* .force_btgt  = -1, */
21478 +               .flags          = AuWrDir_ADD_ENTRY
21479 +       };
21480 +
21481 +       IMustLock(dir);
21482 +       inode = d_inode(src_dentry);
21483 +       IMustLock(inode);
21484 +
21485 +       err = -ENOMEM;
21486 +       a = kzalloc(sizeof(*a), GFP_NOFS);
21487 +       if (unlikely(!a))
21488 +               goto out;
21489 +
21490 +       a->parent = dentry->d_parent; /* dir inode is locked */
21491 +       err = aufs_read_and_write_lock2(dentry, src_dentry,
21492 +                                       AuLock_NOPLM | AuLock_GEN);
21493 +       if (unlikely(err))
21494 +               goto out_kfree;
21495 +       err = au_d_linkable(src_dentry);
21496 +       if (unlikely(err))
21497 +               goto out_unlock;
21498 +       err = au_d_may_add(dentry);
21499 +       if (unlikely(err))
21500 +               goto out_unlock;
21501 +
21502 +       a->src_parent = dget_parent(src_dentry);
21503 +       wr_dir_args.force_btgt = au_ibtop(inode);
21504 +
21505 +       di_write_lock_parent(a->parent);
21506 +       wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
21507 +       wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
21508 +                                     &wr_dir_args);
21509 +       err = PTR_ERR(wh_dentry);
21510 +       if (IS_ERR(wh_dentry))
21511 +               goto out_parent;
21512 +
21513 +       err = 0;
21514 +       sb = dentry->d_sb;
21515 +       a->bdst = au_dbtop(dentry);
21516 +       a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21517 +       a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
21518 +       a->bsrc = au_ibtop(inode);
21519 +       h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21520 +       if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
21521 +               h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
21522 +       if (!h_src_dentry) {
21523 +               a->bsrc = au_dbtop(src_dentry);
21524 +               h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21525 +               AuDebugOn(!h_src_dentry);
21526 +       } else if (IS_ERR(h_src_dentry)) {
21527 +               err = PTR_ERR(h_src_dentry);
21528 +               goto out_parent;
21529 +       }
21530 +
21531 +       /*
21532 +        * aufs doesn't touch the credential so
21533 +        * security_dentry_create_files_as() is unnecessary.
21534 +        */
21535 +       if (au_opt_test(au_mntflags(sb), PLINK)) {
21536 +               if (a->bdst < a->bsrc
21537 +                   /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
21538 +                       err = au_cpup_or_link(src_dentry, dentry, a);
21539 +               else {
21540 +                       delegated = NULL;
21541 +                       err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21542 +                                        &a->h_path, &delegated);
21543 +                       if (unlikely(err == -EWOULDBLOCK)) {
21544 +                               pr_warn("cannot retry for NFSv4 delegation"
21545 +                                       " for an internal link\n");
21546 +                               iput(delegated);
21547 +                       }
21548 +               }
21549 +               dput(h_src_dentry);
21550 +       } else {
21551 +               /*
21552 +                * copyup src_dentry to the branch we process,
21553 +                * and then link(2) to it.
21554 +                */
21555 +               dput(h_src_dentry);
21556 +               if (a->bdst < a->bsrc
21557 +                   /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
21558 +                       au_unpin(&a->pin);
21559 +                       di_write_unlock(a->parent);
21560 +                       err = au_cpup_before_link(src_dentry, a);
21561 +                       di_write_lock_parent(a->parent);
21562 +                       if (!err)
21563 +                               err = au_pin(&a->pin, dentry, a->bdst,
21564 +                                            au_opt_udba(sb),
21565 +                                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21566 +                       if (unlikely(err))
21567 +                               goto out_wh;
21568 +               }
21569 +               if (!err) {
21570 +                       h_src_dentry = au_h_dptr(src_dentry, a->bdst);
21571 +                       err = -ENOENT;
21572 +                       if (h_src_dentry && d_is_positive(h_src_dentry)) {
21573 +                               delegated = NULL;
21574 +                               err = vfsub_link(h_src_dentry,
21575 +                                                au_pinned_h_dir(&a->pin),
21576 +                                                &a->h_path, &delegated);
21577 +                               if (unlikely(err == -EWOULDBLOCK)) {
21578 +                                       pr_warn("cannot retry"
21579 +                                               " for NFSv4 delegation"
21580 +                                               " for an internal link\n");
21581 +                                       iput(delegated);
21582 +                               }
21583 +                       }
21584 +               }
21585 +       }
21586 +       if (unlikely(err))
21587 +               goto out_unpin;
21588 +
21589 +       if (wh_dentry) {
21590 +               a->h_path.dentry = wh_dentry;
21591 +               err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
21592 +                                         dentry);
21593 +               if (unlikely(err))
21594 +                       goto out_revert;
21595 +       }
21596 +
21597 +       au_dir_ts(dir, a->bdst);
21598 +       inode_inc_iversion(dir);
21599 +       inc_nlink(inode);
21600 +       inode->i_ctime = dir->i_ctime;
21601 +       d_instantiate(dentry, au_igrab(inode));
21602 +       if (d_unhashed(a->h_path.dentry))
21603 +               /* some filesystem calls d_drop() */
21604 +               d_drop(dentry);
21605 +       /* some filesystems consume an inode even hardlink */
21606 +       au_fhsm_wrote(sb, a->bdst, /*force*/0);
21607 +       goto out_unpin; /* success */
21608 +
21609 +out_revert:
21610 +       /* no delegation since it is just created */
21611 +       rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
21612 +                           /*delegated*/NULL, /*force*/0);
21613 +       if (unlikely(rerr)) {
21614 +               AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
21615 +               err = -EIO;
21616 +       }
21617 +       au_dtime_revert(&dt);
21618 +out_unpin:
21619 +       au_unpin(&a->pin);
21620 +out_wh:
21621 +       dput(wh_dentry);
21622 +out_parent:
21623 +       di_write_unlock(a->parent);
21624 +       dput(a->src_parent);
21625 +out_unlock:
21626 +       if (unlikely(err)) {
21627 +               au_update_dbtop(dentry);
21628 +               d_drop(dentry);
21629 +       }
21630 +       aufs_read_and_write_unlock2(dentry, src_dentry);
21631 +out_kfree:
21632 +       au_kfree_rcu(a);
21633 +out:
21634 +       AuTraceErr(err);
21635 +       return err;
21636 +}
21637 +
21638 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
21639 +              struct dentry *dentry, umode_t mode)
21640 +{
21641 +       int err, rerr;
21642 +       aufs_bindex_t bindex;
21643 +       unsigned char diropq;
21644 +       struct path h_path;
21645 +       struct dentry *wh_dentry, *parent, *opq_dentry;
21646 +       struct inode *h_inode;
21647 +       struct super_block *sb;
21648 +       struct {
21649 +               struct au_pin pin;
21650 +               struct au_dtime dt;
21651 +       } *a; /* reduce the stack usage */
21652 +       struct au_wr_dir_args wr_dir_args = {
21653 +               .force_btgt     = -1,
21654 +               .flags          = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
21655 +       };
21656 +
21657 +       IMustLock(dir);
21658 +
21659 +       err = -ENOMEM;
21660 +       a = kmalloc(sizeof(*a), GFP_NOFS);
21661 +       if (unlikely(!a))
21662 +               goto out;
21663 +
21664 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21665 +       if (unlikely(err))
21666 +               goto out_free;
21667 +       err = au_d_may_add(dentry);
21668 +       if (unlikely(err))
21669 +               goto out_unlock;
21670 +
21671 +       parent = dentry->d_parent; /* dir inode is locked */
21672 +       di_write_lock_parent(parent);
21673 +       wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21674 +                                     &a->pin, &wr_dir_args);
21675 +       err = PTR_ERR(wh_dentry);
21676 +       if (IS_ERR(wh_dentry))
21677 +               goto out_parent;
21678 +
21679 +       sb = dentry->d_sb;
21680 +       bindex = au_dbtop(dentry);
21681 +       h_path.dentry = au_h_dptr(dentry, bindex);
21682 +       h_path.mnt = au_sbr_mnt(sb, bindex);
21683 +       err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
21684 +       if (unlikely(err))
21685 +               goto out_unpin;
21686 +
21687 +       /* make the dir opaque */
21688 +       diropq = 0;
21689 +       h_inode = d_inode(h_path.dentry);
21690 +       if (wh_dentry
21691 +           || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
21692 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
21693 +               opq_dentry = au_diropq_create(dentry, bindex);
21694 +               inode_unlock(h_inode);
21695 +               err = PTR_ERR(opq_dentry);
21696 +               if (IS_ERR(opq_dentry))
21697 +                       goto out_dir;
21698 +               dput(opq_dentry);
21699 +               diropq = 1;
21700 +       }
21701 +
21702 +       err = epilog(dir, bindex, wh_dentry, dentry);
21703 +       if (!err) {
21704 +               inc_nlink(dir);
21705 +               goto out_unpin; /* success */
21706 +       }
21707 +
21708 +       /* revert */
21709 +       if (diropq) {
21710 +               AuLabel(revert opq);
21711 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
21712 +               rerr = au_diropq_remove(dentry, bindex);
21713 +               inode_unlock(h_inode);
21714 +               if (rerr) {
21715 +                       AuIOErr("%pd reverting diropq failed(%d, %d)\n",
21716 +                               dentry, err, rerr);
21717 +                       err = -EIO;
21718 +               }
21719 +       }
21720 +
21721 +out_dir:
21722 +       AuLabel(revert dir);
21723 +       rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
21724 +       if (rerr) {
21725 +               AuIOErr("%pd reverting dir failed(%d, %d)\n",
21726 +                       dentry, err, rerr);
21727 +               err = -EIO;
21728 +       }
21729 +       au_dtime_revert(&a->dt);
21730 +out_unpin:
21731 +       au_unpin(&a->pin);
21732 +       dput(wh_dentry);
21733 +out_parent:
21734 +       di_write_unlock(parent);
21735 +out_unlock:
21736 +       if (unlikely(err)) {
21737 +               au_update_dbtop(dentry);
21738 +               d_drop(dentry);
21739 +       }
21740 +       aufs_read_unlock(dentry, AuLock_DW);
21741 +out_free:
21742 +       au_kfree_rcu(a);
21743 +out:
21744 +       return err;
21745 +}
21746 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
21747 --- /usr/share/empty/fs/aufs/i_op.c     1970-01-01 01:00:00.000000000 +0100
21748 +++ linux/fs/aufs/i_op.c        2023-02-20 21:05:51.959693785 +0100
21749 @@ -0,0 +1,1516 @@
21750 +// SPDX-License-Identifier: GPL-2.0
21751 +/*
21752 + * Copyright (C) 2005-2022 Junjiro R. Okajima
21753 + *
21754 + * This program is free software; you can redistribute it and/or modify
21755 + * it under the terms of the GNU General Public License as published by
21756 + * the Free Software Foundation; either version 2 of the License, or
21757 + * (at your option) any later version.
21758 + *
21759 + * This program is distributed in the hope that it will be useful,
21760 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21761 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
21762 + * GNU General Public License for more details.
21763 + *
21764 + * You should have received a copy of the GNU General Public License
21765 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
21766 + */
21767 +
21768 +/*
21769 + * inode operations (except add/del/rename)
21770 + */
21771 +
21772 +#include <linux/device_cgroup.h>
21773 +#include <linux/fs_stack.h>
21774 +#include <linux/iversion.h>
21775 +#include <linux/security.h>
21776 +#include "aufs.h"
21777 +
21778 +static int h_permission(struct inode *h_inode, int mask,
21779 +                       struct path *h_path, int brperm)
21780 +{
21781 +       int err;
21782 +       const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21783 +       struct user_namespace *h_userns;
21784 +
21785 +       err = -EPERM;
21786 +       if (write_mask && IS_IMMUTABLE(h_inode))
21787 +               goto out;
21788 +
21789 +       err = -EACCES;
21790 +       if (((mask & MAY_EXEC)
21791 +            && S_ISREG(h_inode->i_mode)
21792 +            && (path_noexec(h_path)
21793 +                || !(h_inode->i_mode & 0111))))
21794 +               goto out;
21795 +
21796 +       /*
21797 +        * - skip the lower fs test in the case of write to ro branch.
21798 +        * - nfs dir permission write check is optimized, but a policy for
21799 +        *   link/rename requires a real check.
21800 +        * - nfs always sets SB_POSIXACL regardless its mount option 'noacl.'
21801 +        *   in this case, generic_permission() returns -EOPNOTSUPP.
21802 +        */
21803 +       h_userns = mnt_user_ns(h_path->mnt);
21804 +       if ((write_mask && !au_br_writable(brperm))
21805 +           || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
21806 +               && write_mask && !(mask & MAY_READ))
21807 +           || !h_inode->i_op->permission) {
21808 +               /* AuLabel(generic_permission); */
21809 +               /* AuDbg("get_inode_acl %ps\n",
21810 +                  h_inode->i_op->get_inode_acl); */
21811 +               err = generic_permission(h_userns, h_inode, mask);
21812 +               if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
21813 +                       err = h_inode->i_op->permission(h_userns, h_inode,
21814 +                                                       mask);
21815 +               AuTraceErr(err);
21816 +       } else {
21817 +               /* AuLabel(h_inode->permission); */
21818 +               err = h_inode->i_op->permission(h_userns, h_inode, mask);
21819 +               AuTraceErr(err);
21820 +       }
21821 +
21822 +       if (!err)
21823 +               err = devcgroup_inode_permission(h_inode, mask);
21824 +       if (!err)
21825 +               err = security_inode_permission(h_inode, mask);
21826 +
21827 +out:
21828 +       return err;
21829 +}
21830 +
21831 +static int aufs_permission(struct user_namespace *userns, struct inode *inode,
21832 +                          int mask)
21833 +{
21834 +       int err;
21835 +       aufs_bindex_t bindex, bbot;
21836 +       const unsigned char isdir = !!S_ISDIR(inode->i_mode),
21837 +               write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21838 +       struct inode *h_inode;
21839 +       struct super_block *sb;
21840 +       struct au_branch *br;
21841 +
21842 +       /* todo: support rcu-walk? */
21843 +       if (mask & MAY_NOT_BLOCK)
21844 +               return -ECHILD;
21845 +
21846 +       sb = inode->i_sb;
21847 +       si_read_lock(sb, AuLock_FLUSH);
21848 +       ii_read_lock_child(inode);
21849 +#if 0 /* reserved for future use */
21850 +       /*
21851 +        * This test may be rather 'too much' since the test is essentially done
21852 +        * in the aufs_lookup().  Theoretically it is possible that the inode
21853 +        * generation doesn't match to the superblock's here.  But it isn't a
21854 +        * big deal I suppose.
21855 +        */
21856 +       err = au_iigen_test(inode, au_sigen(sb));
21857 +       if (unlikely(err))
21858 +               goto out;
21859 +#endif
21860 +
21861 +       if (!isdir
21862 +           || write_mask
21863 +           || au_opt_test(au_mntflags(sb), DIRPERM1)) {
21864 +               err = au_busy_or_stale();
21865 +               h_inode = au_h_iptr(inode, au_ibtop(inode));
21866 +               if (unlikely(!h_inode
21867 +                            || (h_inode->i_mode & S_IFMT)
21868 +                            != (inode->i_mode & S_IFMT)))
21869 +                       goto out;
21870 +
21871 +               err = 0;
21872 +               bindex = au_ibtop(inode);
21873 +               br = au_sbr(sb, bindex);
21874 +               err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
21875 +               if (write_mask
21876 +                   && !err
21877 +                   && !special_file(h_inode->i_mode)) {
21878 +                       /* test whether the upper writable branch exists */
21879 +                       err = -EROFS;
21880 +                       for (; bindex >= 0; bindex--)
21881 +                               if (!au_br_rdonly(au_sbr(sb, bindex))) {
21882 +                                       err = 0;
21883 +                                       break;
21884 +                               }
21885 +               }
21886 +               goto out;
21887 +       }
21888 +
21889 +       /* non-write to dir */
21890 +       err = 0;
21891 +       bbot = au_ibbot(inode);
21892 +       for (bindex = au_ibtop(inode); !err && bindex <= bbot; bindex++) {
21893 +               h_inode = au_h_iptr(inode, bindex);
21894 +               if (h_inode) {
21895 +                       err = au_busy_or_stale();
21896 +                       if (unlikely(!S_ISDIR(h_inode->i_mode)))
21897 +                               break;
21898 +
21899 +                       br = au_sbr(sb, bindex);
21900 +                       err = h_permission(h_inode, mask, &br->br_path,
21901 +                                          br->br_perm);
21902 +               }
21903 +       }
21904 +
21905 +out:
21906 +       ii_read_unlock(inode);
21907 +       si_read_unlock(sb);
21908 +       return err;
21909 +}
21910 +
21911 +/* ---------------------------------------------------------------------- */
21912 +
21913 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
21914 +                                 unsigned int flags)
21915 +{
21916 +       struct dentry *ret, *parent;
21917 +       struct inode *inode;
21918 +       struct super_block *sb;
21919 +       int err, npositive;
21920 +
21921 +       IMustLock(dir);
21922 +
21923 +       /* todo: support rcu-walk? */
21924 +       ret = ERR_PTR(-ECHILD);
21925 +       if (flags & LOOKUP_RCU)
21926 +               goto out;
21927 +
21928 +       ret = ERR_PTR(-ENAMETOOLONG);
21929 +       if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
21930 +               goto out;
21931 +
21932 +       sb = dir->i_sb;
21933 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21934 +       ret = ERR_PTR(err);
21935 +       if (unlikely(err))
21936 +               goto out;
21937 +
21938 +       err = au_di_init(dentry);
21939 +       ret = ERR_PTR(err);
21940 +       if (unlikely(err))
21941 +               goto out_si;
21942 +
21943 +       inode = NULL;
21944 +       npositive = 0; /* suppress a warning */
21945 +       parent = dentry->d_parent; /* dir inode is locked */
21946 +       di_read_lock_parent(parent, AuLock_IR);
21947 +       err = au_alive_dir(parent);
21948 +       if (!err)
21949 +               err = au_digen_test(parent, au_sigen(sb));
21950 +       if (!err) {
21951 +               /* regardless LOOKUP_CREATE, always ALLOW_NEG */
21952 +               npositive = au_lkup_dentry(dentry, au_dbtop(parent),
21953 +                                          AuLkup_ALLOW_NEG);
21954 +               err = npositive;
21955 +       }
21956 +       di_read_unlock(parent, AuLock_IR);
21957 +       ret = ERR_PTR(err);
21958 +       if (unlikely(err < 0))
21959 +               goto out_unlock;
21960 +
21961 +       if (npositive) {
21962 +               inode = au_new_inode(dentry, /*must_new*/0);
21963 +               if (IS_ERR(inode)) {
21964 +                       ret = (void *)inode;
21965 +                       inode = NULL;
21966 +                       goto out_unlock;
21967 +               }
21968 +       }
21969 +
21970 +       if (inode)
21971 +               atomic_inc(&inode->i_count);
21972 +       ret = d_splice_alias(inode, dentry);
21973 +#if 0 /* reserved for future use */
21974 +       if (unlikely(d_need_lookup(dentry))) {
21975 +               spin_lock(&dentry->d_lock);
21976 +               dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
21977 +               spin_unlock(&dentry->d_lock);
21978 +       } else
21979 +#endif
21980 +       if (inode) {
21981 +               if (!IS_ERR(ret)) {
21982 +                       iput(inode);
21983 +                       if (ret && ret != dentry)
21984 +                               ii_write_unlock(inode);
21985 +               } else {
21986 +                       ii_write_unlock(inode);
21987 +                       iput(inode);
21988 +                       inode = NULL;
21989 +               }
21990 +       }
21991 +
21992 +out_unlock:
21993 +       di_write_unlock(dentry);
21994 +out_si:
21995 +       si_read_unlock(sb);
21996 +out:
21997 +       return ret;
21998 +}
21999 +
22000 +/* ---------------------------------------------------------------------- */
22001 +
22002 +/*
22003 + * very dirty and complicated aufs ->atomic_open().
22004 + * aufs_atomic_open()
22005 + * + au_aopen_or_create()
22006 + *   + add_simple()
22007 + *     + vfsub_atomic_open()
22008 + *       + branch fs ->atomic_open()
22009 + *        may call the actual 'open' for h_file
22010 + *       + inc br_nfiles only if opened
22011 + * + au_aopen_no_open() or au_aopen_do_open()
22012 + *
22013 + * au_aopen_do_open()
22014 + * + finish_open()
22015 + *   + au_do_aopen()
22016 + *     + au_do_open() the body of all 'open'
22017 + *       + au_do_open_nondir()
22018 + *        set the passed h_file
22019 + *
22020 + * au_aopen_no_open()
22021 + * + finish_no_open()
22022 + */
22023 +
22024 +struct aopen_node {
22025 +       struct hlist_bl_node hblist;
22026 +       struct file *file, *h_file;
22027 +};
22028 +
22029 +static int au_do_aopen(struct inode *inode, struct file *file)
22030 +{
22031 +       struct hlist_bl_head *aopen;
22032 +       struct hlist_bl_node *pos;
22033 +       struct aopen_node *node;
22034 +       struct au_do_open_args args = {
22035 +               .aopen  = 1,
22036 +               .open   = au_do_open_nondir
22037 +       };
22038 +
22039 +       aopen = &au_sbi(inode->i_sb)->si_aopen;
22040 +       hlist_bl_lock(aopen);
22041 +       hlist_bl_for_each_entry(node, pos, aopen, hblist)
22042 +               if (node->file == file) {
22043 +                       args.h_file = node->h_file;
22044 +                       break;
22045 +               }
22046 +       hlist_bl_unlock(aopen);
22047 +       /* AuDebugOn(!args.h_file); */
22048 +
22049 +       return au_do_open(file, &args);
22050 +}
22051 +
22052 +static int au_aopen_do_open(struct file *file, struct dentry *dentry,
22053 +                           struct aopen_node *aopen_node)
22054 +{
22055 +       int err;
22056 +       struct hlist_bl_head *aopen;
22057 +
22058 +       AuLabel(here);
22059 +       aopen = &au_sbi(dentry->d_sb)->si_aopen;
22060 +       au_hbl_add(&aopen_node->hblist, aopen);
22061 +       err = finish_open(file, dentry, au_do_aopen);
22062 +       au_hbl_del(&aopen_node->hblist, aopen);
22063 +       /* AuDbgFile(file); */
22064 +       AuDbg("%pd%s%s\n", dentry,
22065 +             (file->f_mode & FMODE_CREATED) ? " created" : "",
22066 +             (file->f_mode & FMODE_OPENED) ? " opened" : "");
22067 +
22068 +       AuTraceErr(err);
22069 +       return err;
22070 +}
22071 +
22072 +static int au_aopen_no_open(struct file *file, struct dentry *dentry)
22073 +{
22074 +       int err;
22075 +
22076 +       AuLabel(here);
22077 +       dget(dentry);
22078 +       err = finish_no_open(file, dentry);
22079 +
22080 +       AuTraceErr(err);
22081 +       return err;
22082 +}
22083 +
22084 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
22085 +                           struct file *file, unsigned int open_flag,
22086 +                           umode_t create_mode)
22087 +{
22088 +       int err, did_open;
22089 +       unsigned int lkup_flags;
22090 +       aufs_bindex_t bindex;
22091 +       struct super_block *sb;
22092 +       struct dentry *parent, *d;
22093 +       struct vfsub_aopen_args args = {
22094 +               .open_flag      = open_flag,
22095 +               .create_mode    = create_mode
22096 +       };
22097 +       struct aopen_node aopen_node = {
22098 +               .file   = file
22099 +       };
22100 +
22101 +       IMustLock(dir);
22102 +       AuDbg("open_flag 0%o\n", open_flag);
22103 +       AuDbgDentry(dentry);
22104 +
22105 +       err = 0;
22106 +       if (!au_di(dentry)) {
22107 +               lkup_flags = LOOKUP_OPEN;
22108 +               if (open_flag & O_CREAT)
22109 +                       lkup_flags |= LOOKUP_CREATE;
22110 +               d = aufs_lookup(dir, dentry, lkup_flags);
22111 +               if (IS_ERR(d)) {
22112 +                       err = PTR_ERR(d);
22113 +                       AuTraceErr(err);
22114 +                       goto out;
22115 +               } else if (d) {
22116 +                       /*
22117 +                        * obsoleted dentry found.
22118 +                        * another error will be returned later.
22119 +                        */
22120 +                       d_drop(d);
22121 +                       AuDbgDentry(d);
22122 +                       dput(d);
22123 +               }
22124 +               AuDbgDentry(dentry);
22125 +       }
22126 +
22127 +       if (d_is_positive(dentry)
22128 +           || d_unhashed(dentry)
22129 +           || d_unlinked(dentry)
22130 +           || !(open_flag & O_CREAT)) {
22131 +               err = au_aopen_no_open(file, dentry);
22132 +               goto out; /* success */
22133 +       }
22134 +
22135 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
22136 +       if (unlikely(err))
22137 +               goto out;
22138 +
22139 +       sb = dentry->d_sb;
22140 +       parent = dentry->d_parent;      /* dir is locked */
22141 +       di_write_lock_parent(parent);
22142 +       err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
22143 +       if (unlikely(err < 0))
22144 +               goto out_parent;
22145 +
22146 +       AuDbgDentry(dentry);
22147 +       if (d_is_positive(dentry)) {
22148 +               err = au_aopen_no_open(file, dentry);
22149 +               goto out_parent; /* success */
22150 +       }
22151 +
22152 +       args.file = alloc_empty_file(file->f_flags, current_cred());
22153 +       err = PTR_ERR(args.file);
22154 +       if (IS_ERR(args.file))
22155 +               goto out_parent;
22156 +
22157 +       bindex = au_dbtop(dentry);
22158 +       err = au_aopen_or_create(dir, dentry, &args);
22159 +       AuTraceErr(err);
22160 +       AuDbgFile(args.file);
22161 +       file->f_mode = args.file->f_mode & ~FMODE_OPENED;
22162 +       did_open = !!(args.file->f_mode & FMODE_OPENED);
22163 +       if (!did_open) {
22164 +               fput(args.file);
22165 +               args.file = NULL;
22166 +       }
22167 +       di_write_unlock(parent);
22168 +       di_write_unlock(dentry);
22169 +       if (unlikely(err < 0)) {
22170 +               if (args.file)
22171 +                       fput(args.file);
22172 +               goto out_sb;
22173 +       }
22174 +
22175 +       if (!did_open)
22176 +               err = au_aopen_no_open(file, dentry);
22177 +       else {
22178 +               aopen_node.h_file = args.file;
22179 +               err = au_aopen_do_open(file, dentry, &aopen_node);
22180 +       }
22181 +       if (unlikely(err < 0)) {
22182 +               if (args.file)
22183 +                       fput(args.file);
22184 +               if (did_open)
22185 +                       au_lcnt_dec(&args.br->br_nfiles);
22186 +       }
22187 +       goto out_sb; /* success */
22188 +
22189 +out_parent:
22190 +       di_write_unlock(parent);
22191 +       di_write_unlock(dentry);
22192 +out_sb:
22193 +       si_read_unlock(sb);
22194 +out:
22195 +       AuTraceErr(err);
22196 +       AuDbgFile(file);
22197 +       return err;
22198 +}
22199 +
22200 +
22201 +/* ---------------------------------------------------------------------- */
22202 +
22203 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
22204 +                         const unsigned char add_entry, aufs_bindex_t bcpup,
22205 +                         aufs_bindex_t btop)
22206 +{
22207 +       int err;
22208 +       struct dentry *h_parent;
22209 +       struct inode *h_dir;
22210 +
22211 +       if (add_entry)
22212 +               IMustLock(d_inode(parent));
22213 +       else
22214 +               di_write_lock_parent(parent);
22215 +
22216 +       err = 0;
22217 +       if (!au_h_dptr(parent, bcpup)) {
22218 +               if (btop > bcpup)
22219 +                       err = au_cpup_dirs(dentry, bcpup);
22220 +               else if (btop < bcpup)
22221 +                       err = au_cpdown_dirs(dentry, bcpup);
22222 +               else
22223 +                       BUG();
22224 +       }
22225 +       if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
22226 +               h_parent = au_h_dptr(parent, bcpup);
22227 +               h_dir = d_inode(h_parent);
22228 +               inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
22229 +               err = au_lkup_neg(dentry, bcpup, /*wh*/0);
22230 +               /* todo: no unlock here */
22231 +               inode_unlock_shared(h_dir);
22232 +
22233 +               AuDbg("bcpup %d\n", bcpup);
22234 +               if (!err) {
22235 +                       if (d_really_is_negative(dentry))
22236 +                               au_set_h_dptr(dentry, btop, NULL);
22237 +                       au_update_dbrange(dentry, /*do_put_zero*/0);
22238 +               }
22239 +       }
22240 +
22241 +       if (!add_entry)
22242 +               di_write_unlock(parent);
22243 +       if (!err)
22244 +               err = bcpup; /* success */
22245 +
22246 +       AuTraceErr(err);
22247 +       return err;
22248 +}
22249 +
22250 +/*
22251 + * decide the branch and the parent dir where we will create a new entry.
22252 + * returns new bindex or an error.
22253 + * copyup the parent dir if needed.
22254 + */
22255 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
22256 +             struct au_wr_dir_args *args)
22257 +{
22258 +       int err;
22259 +       unsigned int flags;
22260 +       aufs_bindex_t bcpup, btop, src_btop;
22261 +       const unsigned char add_entry
22262 +               = au_ftest_wrdir(args->flags, ADD_ENTRY)
22263 +               | au_ftest_wrdir(args->flags, TMPFILE);
22264 +       struct super_block *sb;
22265 +       struct dentry *parent;
22266 +       struct au_sbinfo *sbinfo;
22267 +
22268 +       sb = dentry->d_sb;
22269 +       sbinfo = au_sbi(sb);
22270 +       parent = dget_parent(dentry);
22271 +       btop = au_dbtop(dentry);
22272 +       bcpup = btop;
22273 +       if (args->force_btgt < 0) {
22274 +               if (src_dentry) {
22275 +                       src_btop = au_dbtop(src_dentry);
22276 +                       if (src_btop < btop)
22277 +                               bcpup = src_btop;
22278 +               } else if (add_entry) {
22279 +                       flags = 0;
22280 +                       if (au_ftest_wrdir(args->flags, ISDIR))
22281 +                               au_fset_wbr(flags, DIR);
22282 +                       err = AuWbrCreate(sbinfo, dentry, flags);
22283 +                       bcpup = err;
22284 +               }
22285 +
22286 +               if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
22287 +                       if (add_entry)
22288 +                               err = AuWbrCopyup(sbinfo, dentry);
22289 +                       else {
22290 +                               if (!IS_ROOT(dentry)) {
22291 +                                       di_read_lock_parent(parent, !AuLock_IR);
22292 +                                       err = AuWbrCopyup(sbinfo, dentry);
22293 +                                       di_read_unlock(parent, !AuLock_IR);
22294 +                               } else
22295 +                                       err = AuWbrCopyup(sbinfo, dentry);
22296 +                       }
22297 +                       bcpup = err;
22298 +                       if (unlikely(err < 0))
22299 +                               goto out;
22300 +               }
22301 +       } else {
22302 +               bcpup = args->force_btgt;
22303 +               AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
22304 +       }
22305 +
22306 +       AuDbg("btop %d, bcpup %d\n", btop, bcpup);
22307 +       err = bcpup;
22308 +       if (bcpup == btop)
22309 +               goto out; /* success */
22310 +
22311 +       /* copyup the new parent into the branch we process */
22312 +       err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, btop);
22313 +       if (err >= 0) {
22314 +               if (d_really_is_negative(dentry)) {
22315 +                       au_set_h_dptr(dentry, btop, NULL);
22316 +                       au_set_dbtop(dentry, bcpup);
22317 +                       au_set_dbbot(dentry, bcpup);
22318 +               }
22319 +               AuDebugOn(add_entry
22320 +                         && !au_ftest_wrdir(args->flags, TMPFILE)
22321 +                         && !au_h_dptr(dentry, bcpup));
22322 +       }
22323 +
22324 +out:
22325 +       dput(parent);
22326 +       return err;
22327 +}
22328 +
22329 +/* ---------------------------------------------------------------------- */
22330 +
22331 +void au_pin_hdir_unlock(struct au_pin *p)
22332 +{
22333 +       if (p->hdir)
22334 +               au_hn_inode_unlock(p->hdir);
22335 +}
22336 +
22337 +int au_pin_hdir_lock(struct au_pin *p)
22338 +{
22339 +       int err;
22340 +
22341 +       err = 0;
22342 +       if (!p->hdir)
22343 +               goto out;
22344 +
22345 +       /* even if an error happens later, keep this lock */
22346 +       au_hn_inode_lock_nested(p->hdir, p->lsc_hi);
22347 +
22348 +       err = -EBUSY;
22349 +       if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
22350 +               goto out;
22351 +
22352 +       err = 0;
22353 +       if (p->h_dentry)
22354 +               err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
22355 +                                 p->h_parent, p->br);
22356 +
22357 +out:
22358 +       return err;
22359 +}
22360 +
22361 +int au_pin_hdir_relock(struct au_pin *p)
22362 +{
22363 +       int err, i;
22364 +       struct inode *h_i;
22365 +       struct dentry *h_d[] = {
22366 +               p->h_dentry,
22367 +               p->h_parent
22368 +       };
22369 +
22370 +       err = au_pin_hdir_lock(p);
22371 +       if (unlikely(err))
22372 +               goto out;
22373 +
22374 +       for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
22375 +               if (!h_d[i])
22376 +                       continue;
22377 +               if (d_is_positive(h_d[i])) {
22378 +                       h_i = d_inode(h_d[i]);
22379 +                       err = !h_i->i_nlink;
22380 +               }
22381 +       }
22382 +
22383 +out:
22384 +       return err;
22385 +}
22386 +
22387 +static void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
22388 +{
22389 +       atomic_long_set(&p->hdir->hi_inode->i_rwsem.owner, (long)task);
22390 +}
22391 +
22392 +void au_pin_hdir_acquire_nest(struct au_pin *p)
22393 +{
22394 +       if (p->hdir) {
22395 +               rwsem_acquire_nest(&p->hdir->hi_inode->i_rwsem.dep_map,
22396 +                                  p->lsc_hi, 0, NULL, _RET_IP_);
22397 +               au_pin_hdir_set_owner(p, current);
22398 +       }
22399 +}
22400 +
22401 +void au_pin_hdir_release(struct au_pin *p)
22402 +{
22403 +       if (p->hdir) {
22404 +               au_pin_hdir_set_owner(p, p->task);
22405 +               rwsem_release(&p->hdir->hi_inode->i_rwsem.dep_map, _RET_IP_);
22406 +       }
22407 +}
22408 +
22409 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
22410 +{
22411 +       if (pin && pin->parent)
22412 +               return au_h_dptr(pin->parent, pin->bindex);
22413 +       return NULL;
22414 +}
22415 +
22416 +void au_unpin(struct au_pin *p)
22417 +{
22418 +       if (p->hdir)
22419 +               au_pin_hdir_unlock(p);
22420 +       if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
22421 +               vfsub_mnt_drop_write(p->h_mnt);
22422 +       if (!p->hdir)
22423 +               return;
22424 +
22425 +       if (!au_ftest_pin(p->flags, DI_LOCKED))
22426 +               di_read_unlock(p->parent, AuLock_IR);
22427 +       iput(p->hdir->hi_inode);
22428 +       dput(p->parent);
22429 +       p->parent = NULL;
22430 +       p->hdir = NULL;
22431 +       p->h_mnt = NULL;
22432 +       /* do not clear p->task */
22433 +}
22434 +
22435 +int au_do_pin(struct au_pin *p)
22436 +{
22437 +       int err;
22438 +       struct super_block *sb;
22439 +       struct inode *h_dir;
22440 +
22441 +       err = 0;
22442 +       sb = p->dentry->d_sb;
22443 +       p->br = au_sbr(sb, p->bindex);
22444 +       if (IS_ROOT(p->dentry)) {
22445 +               if (au_ftest_pin(p->flags, MNT_WRITE)) {
22446 +                       p->h_mnt = au_br_mnt(p->br);
22447 +                       err = vfsub_mnt_want_write(p->h_mnt);
22448 +                       if (unlikely(err)) {
22449 +                               au_fclr_pin(p->flags, MNT_WRITE);
22450 +                               goto out_err;
22451 +                       }
22452 +               }
22453 +               goto out;
22454 +       }
22455 +
22456 +       p->h_dentry = NULL;
22457 +       if (p->bindex <= au_dbbot(p->dentry))
22458 +               p->h_dentry = au_h_dptr(p->dentry, p->bindex);
22459 +
22460 +       p->parent = dget_parent(p->dentry);
22461 +       if (!au_ftest_pin(p->flags, DI_LOCKED))
22462 +               di_read_lock(p->parent, AuLock_IR, p->lsc_di);
22463 +
22464 +       h_dir = NULL;
22465 +       p->h_parent = au_h_dptr(p->parent, p->bindex);
22466 +       p->hdir = au_hi(d_inode(p->parent), p->bindex);
22467 +       if (p->hdir)
22468 +               h_dir = p->hdir->hi_inode;
22469 +
22470 +       /*
22471 +        * udba case, or
22472 +        * if DI_LOCKED is not set, then p->parent may be different
22473 +        * and h_parent can be NULL.
22474 +        */
22475 +       if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
22476 +               err = -EBUSY;
22477 +               if (!au_ftest_pin(p->flags, DI_LOCKED))
22478 +                       di_read_unlock(p->parent, AuLock_IR);
22479 +               dput(p->parent);
22480 +               p->parent = NULL;
22481 +               goto out_err;
22482 +       }
22483 +
22484 +       if (au_ftest_pin(p->flags, MNT_WRITE)) {
22485 +               p->h_mnt = au_br_mnt(p->br);
22486 +               err = vfsub_mnt_want_write(p->h_mnt);
22487 +               if (unlikely(err)) {
22488 +                       au_fclr_pin(p->flags, MNT_WRITE);
22489 +                       if (!au_ftest_pin(p->flags, DI_LOCKED))
22490 +                               di_read_unlock(p->parent, AuLock_IR);
22491 +                       dput(p->parent);
22492 +                       p->parent = NULL;
22493 +                       goto out_err;
22494 +               }
22495 +       }
22496 +
22497 +       au_igrab(h_dir);
22498 +       err = au_pin_hdir_lock(p);
22499 +       if (!err)
22500 +               goto out; /* success */
22501 +
22502 +       au_unpin(p);
22503 +
22504 +out_err:
22505 +       pr_err("err %d\n", err);
22506 +       err = au_busy_or_stale();
22507 +out:
22508 +       return err;
22509 +}
22510 +
22511 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
22512 +                aufs_bindex_t bindex, int lsc_di, int lsc_hi,
22513 +                unsigned int udba, unsigned char flags)
22514 +{
22515 +       p->dentry = dentry;
22516 +       p->udba = udba;
22517 +       p->lsc_di = lsc_di;
22518 +       p->lsc_hi = lsc_hi;
22519 +       p->flags = flags;
22520 +       p->bindex = bindex;
22521 +
22522 +       p->parent = NULL;
22523 +       p->hdir = NULL;
22524 +       p->h_mnt = NULL;
22525 +
22526 +       p->h_dentry = NULL;
22527 +       p->h_parent = NULL;
22528 +       p->br = NULL;
22529 +       p->task = current;
22530 +}
22531 +
22532 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
22533 +          unsigned int udba, unsigned char flags)
22534 +{
22535 +       au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
22536 +                   udba, flags);
22537 +       return au_do_pin(pin);
22538 +}
22539 +
22540 +/* ---------------------------------------------------------------------- */
22541 +
22542 +/*
22543 + * ->setattr() and ->getattr() are called in various cases.
22544 + * chmod, stat: dentry is revalidated.
22545 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
22546 + *               unhashed.
22547 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
22548 + */
22549 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
22550 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
22551 +{
22552 +       int err;
22553 +       struct dentry *parent;
22554 +
22555 +       err = 0;
22556 +       if (au_digen_test(dentry, sigen)) {
22557 +               parent = dget_parent(dentry);
22558 +               di_read_lock_parent(parent, AuLock_IR);
22559 +               err = au_refresh_dentry(dentry, parent);
22560 +               di_read_unlock(parent, AuLock_IR);
22561 +               dput(parent);
22562 +       }
22563 +
22564 +       AuTraceErr(err);
22565 +       return err;
22566 +}
22567 +
22568 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
22569 +                    struct au_icpup_args *a)
22570 +{
22571 +       int err;
22572 +       loff_t sz;
22573 +       aufs_bindex_t btop, ibtop;
22574 +       struct dentry *hi_wh, *parent;
22575 +       struct inode *inode;
22576 +       struct au_wr_dir_args wr_dir_args = {
22577 +               .force_btgt     = -1,
22578 +               .flags          = 0
22579 +       };
22580 +
22581 +       if (d_is_dir(dentry))
22582 +               au_fset_wrdir(wr_dir_args.flags, ISDIR);
22583 +       /* plink or hi_wh() case */
22584 +       btop = au_dbtop(dentry);
22585 +       inode = d_inode(dentry);
22586 +       ibtop = au_ibtop(inode);
22587 +       if (btop != ibtop && !au_test_ro(inode->i_sb, ibtop, inode))
22588 +               wr_dir_args.force_btgt = ibtop;
22589 +       err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
22590 +       if (unlikely(err < 0))
22591 +               goto out;
22592 +       a->btgt = err;
22593 +       if (err != btop)
22594 +               au_fset_icpup(a->flags, DID_CPUP);
22595 +
22596 +       err = 0;
22597 +       a->pin_flags = AuPin_MNT_WRITE;
22598 +       parent = NULL;
22599 +       if (!IS_ROOT(dentry)) {
22600 +               au_fset_pin(a->pin_flags, DI_LOCKED);
22601 +               parent = dget_parent(dentry);
22602 +               di_write_lock_parent(parent);
22603 +       }
22604 +
22605 +       err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
22606 +       if (unlikely(err))
22607 +               goto out_parent;
22608 +
22609 +       sz = -1;
22610 +       a->h_path.dentry = au_h_dptr(dentry, btop);
22611 +       a->h_inode = d_inode(a->h_path.dentry);
22612 +       if (ia && (ia->ia_valid & ATTR_SIZE)) {
22613 +               inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
22614 +               if (ia->ia_size < i_size_read(a->h_inode))
22615 +                       sz = ia->ia_size;
22616 +               inode_unlock_shared(a->h_inode);
22617 +       }
22618 +
22619 +       hi_wh = NULL;
22620 +       if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
22621 +               hi_wh = au_hi_wh(inode, a->btgt);
22622 +               if (!hi_wh) {
22623 +                       struct au_cp_generic cpg = {
22624 +                               .dentry = dentry,
22625 +                               .bdst   = a->btgt,
22626 +                               .bsrc   = -1,
22627 +                               .len    = sz,
22628 +                               .pin    = &a->pin
22629 +                       };
22630 +                       err = au_sio_cpup_wh(&cpg, /*file*/NULL);
22631 +                       if (unlikely(err))
22632 +                               goto out_unlock;
22633 +                       hi_wh = au_hi_wh(inode, a->btgt);
22634 +                       /* todo: revalidate hi_wh? */
22635 +               }
22636 +       }
22637 +
22638 +       if (parent) {
22639 +               au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
22640 +               di_downgrade_lock(parent, AuLock_IR);
22641 +               dput(parent);
22642 +               parent = NULL;
22643 +       }
22644 +       if (!au_ftest_icpup(a->flags, DID_CPUP))
22645 +               goto out; /* success */
22646 +
22647 +       if (!d_unhashed(dentry)) {
22648 +               struct au_cp_generic cpg = {
22649 +                       .dentry = dentry,
22650 +                       .bdst   = a->btgt,
22651 +                       .bsrc   = btop,
22652 +                       .len    = sz,
22653 +                       .pin    = &a->pin,
22654 +                       .flags  = AuCpup_DTIME | AuCpup_HOPEN
22655 +               };
22656 +               err = au_sio_cpup_simple(&cpg);
22657 +               if (!err)
22658 +                       a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22659 +       } else if (!hi_wh)
22660 +               a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22661 +       else
22662 +               a->h_path.dentry = hi_wh; /* do not dget here */
22663 +
22664 +out_unlock:
22665 +       a->h_inode = d_inode(a->h_path.dentry);
22666 +       if (!err)
22667 +               goto out; /* success */
22668 +       au_unpin(&a->pin);
22669 +out_parent:
22670 +       if (parent) {
22671 +               di_write_unlock(parent);
22672 +               dput(parent);
22673 +       }
22674 +out:
22675 +       if (!err)
22676 +               inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22677 +       return err;
22678 +}
22679 +
22680 +static int aufs_setattr(struct user_namespace *userns, struct dentry *dentry,
22681 +                       struct iattr *ia)
22682 +{
22683 +       int err;
22684 +       struct inode *inode, *delegated;
22685 +       struct super_block *sb;
22686 +       struct file *file;
22687 +       struct au_icpup_args *a;
22688 +       struct user_namespace *h_userns;
22689 +
22690 +       inode = d_inode(dentry);
22691 +       IMustLock(inode);
22692 +
22693 +       err = setattr_prepare(userns, dentry, ia);
22694 +       if (unlikely(err))
22695 +               goto out;
22696 +
22697 +       err = -ENOMEM;
22698 +       a = kzalloc(sizeof(*a), GFP_NOFS);
22699 +       if (unlikely(!a))
22700 +               goto out;
22701 +
22702 +       if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
22703 +               ia->ia_valid &= ~ATTR_MODE;
22704 +
22705 +       file = NULL;
22706 +       sb = dentry->d_sb;
22707 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22708 +       if (unlikely(err))
22709 +               goto out_kfree;
22710 +
22711 +       if (ia->ia_valid & ATTR_FILE) {
22712 +               /* currently ftruncate(2) only */
22713 +               AuDebugOn(!d_is_reg(dentry));
22714 +               file = ia->ia_file;
22715 +               err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1,
22716 +                                           /*fi_lsc*/0);
22717 +               if (unlikely(err))
22718 +                       goto out_si;
22719 +               ia->ia_file = au_hf_top(file);
22720 +               a->udba = AuOpt_UDBA_NONE;
22721 +       } else {
22722 +               /* fchmod() doesn't pass ia_file */
22723 +               a->udba = au_opt_udba(sb);
22724 +               di_write_lock_child(dentry);
22725 +               /* no d_unlinked(), to set UDBA_NONE for root */
22726 +               if (d_unhashed(dentry))
22727 +                       a->udba = AuOpt_UDBA_NONE;
22728 +               if (a->udba != AuOpt_UDBA_NONE) {
22729 +                       AuDebugOn(IS_ROOT(dentry));
22730 +                       err = au_reval_for_attr(dentry, au_sigen(sb));
22731 +                       if (unlikely(err))
22732 +                               goto out_dentry;
22733 +               }
22734 +       }
22735 +
22736 +       err = au_pin_and_icpup(dentry, ia, a);
22737 +       if (unlikely(err < 0))
22738 +               goto out_dentry;
22739 +       if (au_ftest_icpup(a->flags, DID_CPUP)) {
22740 +               ia->ia_file = NULL;
22741 +               ia->ia_valid &= ~ATTR_FILE;
22742 +       }
22743 +
22744 +       a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
22745 +       if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
22746 +           == (ATTR_MODE | ATTR_CTIME)) {
22747 +               err = security_path_chmod(&a->h_path, ia->ia_mode);
22748 +               if (unlikely(err))
22749 +                       goto out_unlock;
22750 +       } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
22751 +                  && (ia->ia_valid & ATTR_CTIME)) {
22752 +               err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
22753 +               if (unlikely(err))
22754 +                       goto out_unlock;
22755 +       }
22756 +
22757 +       if (ia->ia_valid & ATTR_SIZE) {
22758 +               struct file *f;
22759 +
22760 +               if (ia->ia_size < i_size_read(inode))
22761 +                       /* unmap only */
22762 +                       truncate_setsize(inode, ia->ia_size);
22763 +
22764 +               f = NULL;
22765 +               if (ia->ia_valid & ATTR_FILE)
22766 +                       f = ia->ia_file;
22767 +               inode_unlock(a->h_inode);
22768 +               err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
22769 +               inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22770 +       } else {
22771 +               delegated = NULL;
22772 +               while (1) {
22773 +                       err = vfsub_notify_change(&a->h_path, ia, &delegated);
22774 +                       if (delegated) {
22775 +                               err = break_deleg_wait(&delegated);
22776 +                               if (!err)
22777 +                                       continue;
22778 +                       }
22779 +                       break;
22780 +               }
22781 +       }
22782 +       /*
22783 +        * regardless aufs 'acl' option setting.
22784 +        * why don't all acl-aware fs call this func from their ->setattr()?
22785 +        */
22786 +       if (!err && (ia->ia_valid & ATTR_MODE)) {
22787 +               h_userns = mnt_user_ns(a->h_path.mnt);
22788 +               err = vfsub_acl_chmod(h_userns, a->h_path.dentry, ia->ia_mode);
22789 +       }
22790 +       if (!err)
22791 +               au_cpup_attr_changeable(inode);
22792 +
22793 +out_unlock:
22794 +       inode_unlock(a->h_inode);
22795 +       au_unpin(&a->pin);
22796 +       if (unlikely(err))
22797 +               au_update_dbtop(dentry);
22798 +out_dentry:
22799 +       di_write_unlock(dentry);
22800 +       if (file) {
22801 +               fi_write_unlock(file);
22802 +               ia->ia_file = file;
22803 +               ia->ia_valid |= ATTR_FILE;
22804 +       }
22805 +out_si:
22806 +       si_read_unlock(sb);
22807 +out_kfree:
22808 +       au_kfree_rcu(a);
22809 +out:
22810 +       AuTraceErr(err);
22811 +       return err;
22812 +}
22813 +
22814 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
22815 +static int au_h_path_to_set_attr(struct dentry *dentry,
22816 +                                struct au_icpup_args *a, struct path *h_path)
22817 +{
22818 +       int err;
22819 +       struct super_block *sb;
22820 +
22821 +       sb = dentry->d_sb;
22822 +       a->udba = au_opt_udba(sb);
22823 +       /* no d_unlinked(), to set UDBA_NONE for root */
22824 +       if (d_unhashed(dentry))
22825 +               a->udba = AuOpt_UDBA_NONE;
22826 +       if (a->udba != AuOpt_UDBA_NONE) {
22827 +               AuDebugOn(IS_ROOT(dentry));
22828 +               err = au_reval_for_attr(dentry, au_sigen(sb));
22829 +               if (unlikely(err))
22830 +                       goto out;
22831 +       }
22832 +       err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
22833 +       if (unlikely(err < 0))
22834 +               goto out;
22835 +
22836 +       h_path->dentry = a->h_path.dentry;
22837 +       h_path->mnt = au_sbr_mnt(sb, a->btgt);
22838 +
22839 +out:
22840 +       return err;
22841 +}
22842 +
22843 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
22844 +                 struct au_sxattr *arg)
22845 +{
22846 +       int err;
22847 +       struct path h_path;
22848 +       struct super_block *sb;
22849 +       struct au_icpup_args *a;
22850 +       struct inode *h_inode;
22851 +       struct user_namespace *h_userns;
22852 +
22853 +       IMustLock(inode);
22854 +
22855 +       err = -ENOMEM;
22856 +       a = kzalloc(sizeof(*a), GFP_NOFS);
22857 +       if (unlikely(!a))
22858 +               goto out;
22859 +
22860 +       sb = dentry->d_sb;
22861 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22862 +       if (unlikely(err))
22863 +               goto out_kfree;
22864 +
22865 +       h_path.dentry = NULL;   /* silence gcc */
22866 +       di_write_lock_child(dentry);
22867 +       err = au_h_path_to_set_attr(dentry, a, &h_path);
22868 +       if (unlikely(err))
22869 +               goto out_di;
22870 +       h_userns = mnt_user_ns(h_path.mnt);
22871 +
22872 +       inode_unlock(a->h_inode);
22873 +       switch (arg->type) {
22874 +       case AU_XATTR_SET:
22875 +               AuDebugOn(d_is_negative(h_path.dentry));
22876 +               err = vfsub_setxattr(h_userns, h_path.dentry,
22877 +                                    arg->u.set.name, arg->u.set.value,
22878 +                                    arg->u.set.size, arg->u.set.flags);
22879 +               break;
22880 +       case AU_ACL_SET:
22881 +               err = -EOPNOTSUPP;
22882 +               h_inode = d_inode(h_path.dentry);
22883 +               if (h_inode->i_op->set_acl) {
22884 +                       /* this will call posix_acl_update_mode */
22885 +                       err = h_inode->i_op->set_acl(h_userns, h_path.dentry,
22886 +                                                    arg->u.acl_set.acl,
22887 +                                                    arg->u.acl_set.type);
22888 +               }
22889 +               break;
22890 +       }
22891 +       if (!err)
22892 +               au_cpup_attr_timesizes(inode);
22893 +
22894 +       au_unpin(&a->pin);
22895 +       if (unlikely(err))
22896 +               au_update_dbtop(dentry);
22897 +
22898 +out_di:
22899 +       di_write_unlock(dentry);
22900 +       si_read_unlock(sb);
22901 +out_kfree:
22902 +       au_kfree_rcu(a);
22903 +out:
22904 +       AuTraceErr(err);
22905 +       return err;
22906 +}
22907 +#endif
22908 +
22909 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
22910 +                            unsigned int nlink)
22911 +{
22912 +       unsigned int n;
22913 +
22914 +       inode->i_mode = st->mode;
22915 +       /* don't i_[ug]id_write() here */
22916 +       inode->i_uid = st->uid;
22917 +       inode->i_gid = st->gid;
22918 +       inode->i_atime = st->atime;
22919 +       inode->i_mtime = st->mtime;
22920 +       inode->i_ctime = st->ctime;
22921 +
22922 +       au_cpup_attr_nlink(inode, /*force*/0);
22923 +       if (S_ISDIR(inode->i_mode)) {
22924 +               n = inode->i_nlink;
22925 +               n -= nlink;
22926 +               n += st->nlink;
22927 +               smp_mb(); /* for i_nlink */
22928 +               /* 0 can happen */
22929 +               set_nlink(inode, n);
22930 +       }
22931 +
22932 +       spin_lock(&inode->i_lock);
22933 +       inode->i_blocks = st->blocks;
22934 +       i_size_write(inode, st->size);
22935 +       spin_unlock(&inode->i_lock);
22936 +}
22937 +
22938 +/*
22939 + * common routine for aufs_getattr() and au_getxattr().
22940 + * returns zero or negative (an error).
22941 + * @dentry will be read-locked in success.
22942 + */
22943 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
22944 +                     struct path *h_path, int locked)
22945 +{
22946 +       int err;
22947 +       unsigned int mnt_flags, sigen;
22948 +       unsigned char udba_none;
22949 +       aufs_bindex_t bindex;
22950 +       struct super_block *sb, *h_sb;
22951 +
22952 +       h_path->mnt = NULL;
22953 +       h_path->dentry = NULL;
22954 +
22955 +       err = 0;
22956 +       sb = dentry->d_sb;
22957 +       mnt_flags = au_mntflags(sb);
22958 +       udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
22959 +
22960 +       if (unlikely(locked))
22961 +               goto body; /* skip locking dinfo */
22962 +
22963 +       /* support fstat(2) */
22964 +       if (!d_unlinked(dentry) && !udba_none) {
22965 +               sigen = au_sigen(sb);
22966 +               err = au_digen_test(dentry, sigen);
22967 +               if (!err) {
22968 +                       di_read_lock_child(dentry, AuLock_IR);
22969 +                       err = au_dbrange_test(dentry);
22970 +                       if (unlikely(err)) {
22971 +                               di_read_unlock(dentry, AuLock_IR);
22972 +                               goto out;
22973 +                       }
22974 +               } else {
22975 +                       AuDebugOn(IS_ROOT(dentry));
22976 +                       di_write_lock_child(dentry);
22977 +                       err = au_dbrange_test(dentry);
22978 +                       if (!err)
22979 +                               err = au_reval_for_attr(dentry, sigen);
22980 +                       if (!err)
22981 +                               di_downgrade_lock(dentry, AuLock_IR);
22982 +                       else {
22983 +                               di_write_unlock(dentry);
22984 +                               goto out;
22985 +                       }
22986 +               }
22987 +       } else
22988 +               di_read_lock_child(dentry, AuLock_IR);
22989 +
22990 +body:
22991 +       if (!inode) {
22992 +               inode = d_inode(dentry);
22993 +               if (unlikely(!inode))
22994 +                       goto out;
22995 +       }
22996 +       bindex = au_ibtop(inode);
22997 +       h_path->mnt = au_sbr_mnt(sb, bindex);
22998 +       h_sb = h_path->mnt->mnt_sb;
22999 +       if (!force
23000 +           && !au_test_fs_bad_iattr(h_sb)
23001 +           && udba_none)
23002 +               goto out; /* success */
23003 +
23004 +       if (au_dbtop(dentry) == bindex)
23005 +               h_path->dentry = au_h_dptr(dentry, bindex);
23006 +       else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
23007 +               h_path->dentry = au_plink_lkup(inode, bindex);
23008 +               if (IS_ERR(h_path->dentry))
23009 +                       /* pretending success */
23010 +                       h_path->dentry = NULL;
23011 +               else
23012 +                       dput(h_path->dentry);
23013 +       }
23014 +
23015 +out:
23016 +       return err;
23017 +}
23018 +
23019 +static int aufs_getattr(struct user_namespace *userns, const struct path *path,
23020 +                       struct kstat *st, u32 request, unsigned int query)
23021 +{
23022 +       int err;
23023 +       unsigned char positive;
23024 +       struct path h_path;
23025 +       struct dentry *dentry;
23026 +       struct inode *inode;
23027 +       struct super_block *sb;
23028 +
23029 +       dentry = path->dentry;
23030 +       inode = d_inode(dentry);
23031 +       sb = dentry->d_sb;
23032 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
23033 +       if (unlikely(err))
23034 +               goto out;
23035 +       err = au_h_path_getattr(dentry, /*inode*/NULL, /*force*/0, &h_path,
23036 +                               /*locked*/0);
23037 +       if (unlikely(err))
23038 +               goto out_si;
23039 +       if (unlikely(!h_path.dentry))
23040 +               /* illegally overlapped or something */
23041 +               goto out_fill; /* pretending success */
23042 +
23043 +       positive = d_is_positive(h_path.dentry);
23044 +       if (positive)
23045 +               /* no vfsub version */
23046 +               err = vfs_getattr(&h_path, st, request, query);
23047 +       if (!err) {
23048 +               if (positive)
23049 +                       au_refresh_iattr(inode, st,
23050 +                                        d_inode(h_path.dentry)->i_nlink);
23051 +               goto out_fill; /* success */
23052 +       }
23053 +       AuTraceErr(err);
23054 +       goto out_di;
23055 +
23056 +out_fill:
23057 +       generic_fillattr(userns, inode, st);
23058 +out_di:
23059 +       di_read_unlock(dentry, AuLock_IR);
23060 +out_si:
23061 +       si_read_unlock(sb);
23062 +out:
23063 +       AuTraceErr(err);
23064 +       return err;
23065 +}
23066 +
23067 +/* ---------------------------------------------------------------------- */
23068 +
23069 +static const char *aufs_get_link(struct dentry *dentry, struct inode *inode,
23070 +                                struct delayed_call *done)
23071 +{
23072 +       const char *ret;
23073 +       struct dentry *h_dentry;
23074 +       struct inode *h_inode;
23075 +       int err;
23076 +       aufs_bindex_t bindex;
23077 +
23078 +       ret = NULL; /* suppress a warning */
23079 +       err = -ECHILD;
23080 +       if (!dentry)
23081 +               goto out;
23082 +
23083 +       err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
23084 +       if (unlikely(err))
23085 +               goto out;
23086 +
23087 +       err = au_d_hashed_positive(dentry);
23088 +       if (unlikely(err))
23089 +               goto out_unlock;
23090 +
23091 +       err = -EINVAL;
23092 +       inode = d_inode(dentry);
23093 +       bindex = au_ibtop(inode);
23094 +       h_inode = au_h_iptr(inode, bindex);
23095 +       if (unlikely(!h_inode->i_op->get_link))
23096 +               goto out_unlock;
23097 +
23098 +       err = -EBUSY;
23099 +       h_dentry = NULL;
23100 +       if (au_dbtop(dentry) <= bindex) {
23101 +               h_dentry = au_h_dptr(dentry, bindex);
23102 +               if (h_dentry)
23103 +                       dget(h_dentry);
23104 +       }
23105 +       if (!h_dentry) {
23106 +               h_dentry = d_find_any_alias(h_inode);
23107 +               if (IS_ERR(h_dentry)) {
23108 +                       err = PTR_ERR(h_dentry);
23109 +                       goto out_unlock;
23110 +               }
23111 +       }
23112 +       if (unlikely(!h_dentry))
23113 +               goto out_unlock;
23114 +
23115 +       err = 0;
23116 +       AuDbg("%ps\n", h_inode->i_op->get_link);
23117 +       AuDbgDentry(h_dentry);
23118 +       ret = vfs_get_link(h_dentry, done);
23119 +       dput(h_dentry);
23120 +       if (IS_ERR(ret))
23121 +               err = PTR_ERR(ret);
23122 +
23123 +out_unlock:
23124 +       aufs_read_unlock(dentry, AuLock_IR);
23125 +out:
23126 +       if (unlikely(err))
23127 +               ret = ERR_PTR(err);
23128 +       AuTraceErrPtr(ret);
23129 +       return ret;
23130 +}
23131 +
23132 +/* ---------------------------------------------------------------------- */
23133 +
23134 +static int au_is_special(struct inode *inode)
23135 +{
23136 +       return (inode->i_mode & (S_IFBLK | S_IFCHR | S_IFIFO | S_IFSOCK));
23137 +}
23138 +
23139 +static int aufs_update_time(struct inode *inode, struct timespec64 *ts,
23140 +                           int flags)
23141 +{
23142 +       int err;
23143 +       aufs_bindex_t bindex;
23144 +       struct super_block *sb;
23145 +       struct inode *h_inode;
23146 +       struct vfsmount *h_mnt;
23147 +
23148 +       sb = inode->i_sb;
23149 +       WARN_ONCE((flags & S_ATIME) && !IS_NOATIME(inode),
23150 +                 "unexpected s_flags 0x%lx", sb->s_flags);
23151 +
23152 +       /* mmap_sem might be acquired already, cf. aufs_mmap() */
23153 +       lockdep_off();
23154 +       si_read_lock(sb, AuLock_FLUSH);
23155 +       ii_write_lock_child(inode);
23156 +
23157 +       err = 0;
23158 +       bindex = au_ibtop(inode);
23159 +       h_inode = au_h_iptr(inode, bindex);
23160 +       if (!au_test_ro(sb, bindex, inode)) {
23161 +               h_mnt = au_sbr_mnt(sb, bindex);
23162 +               err = vfsub_mnt_want_write(h_mnt);
23163 +               if (!err) {
23164 +                       err = vfsub_update_time(h_inode, ts, flags);
23165 +                       vfsub_mnt_drop_write(h_mnt);
23166 +               }
23167 +       } else if (au_is_special(h_inode)) {
23168 +               /*
23169 +                * Never copy-up here.
23170 +                * These special files may already be opened and used for
23171 +                * communicating. If we copied it up, then the communication
23172 +                * would be corrupted.
23173 +                */
23174 +               AuWarn1("timestamps for i%lu are ignored "
23175 +                       "since it is on readonly branch (hi%lu).\n",
23176 +                       inode->i_ino, h_inode->i_ino);
23177 +       } else if (flags & ~S_ATIME) {
23178 +               err = -EIO;
23179 +               AuIOErr1("unexpected flags 0x%x\n", flags);
23180 +               AuDebugOn(1);
23181 +       }
23182 +
23183 +       if (!err)
23184 +               au_cpup_attr_timesizes(inode);
23185 +       ii_write_unlock(inode);
23186 +       si_read_unlock(sb);
23187 +       lockdep_on();
23188 +
23189 +       if (!err && (flags & S_VERSION))
23190 +               inode_inc_iversion(inode);
23191 +
23192 +       return err;
23193 +}
23194 +
23195 +/* ---------------------------------------------------------------------- */
23196 +
23197 +/* no getattr version will be set by module.c:aufs_init() */
23198 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
23199 +       aufs_iop[] = {
23200 +       [AuIop_SYMLINK] = {
23201 +               .permission     = aufs_permission,
23202 +#ifdef CONFIG_FS_POSIX_ACL
23203 +               .get_inode_acl  = aufs_get_inode_acl,
23204 +               .get_acl        = aufs_get_acl,
23205 +               .set_acl        = aufs_set_acl, /* unsupport for symlink? */
23206 +#endif
23207 +
23208 +               .setattr        = aufs_setattr,
23209 +               .getattr        = aufs_getattr,
23210 +
23211 +#ifdef CONFIG_AUFS_XATTR
23212 +               .listxattr      = aufs_listxattr,
23213 +#endif
23214 +
23215 +               .get_link       = aufs_get_link
23216 +
23217 +               /* .update_time = aufs_update_time */
23218 +       },
23219 +       [AuIop_DIR] = {
23220 +               .create         = aufs_create,
23221 +               .lookup         = aufs_lookup,
23222 +               .link           = aufs_link,
23223 +               .unlink         = aufs_unlink,
23224 +               .symlink        = aufs_symlink,
23225 +               .mkdir          = aufs_mkdir,
23226 +               .rmdir          = aufs_rmdir,
23227 +               .mknod          = aufs_mknod,
23228 +               .rename         = aufs_rename,
23229 +
23230 +               .permission     = aufs_permission,
23231 +#ifdef CONFIG_FS_POSIX_ACL
23232 +               .get_inode_acl  = aufs_get_inode_acl,
23233 +               .get_acl        = aufs_get_acl,
23234 +               .set_acl        = aufs_set_acl,
23235 +#endif
23236 +
23237 +               .setattr        = aufs_setattr,
23238 +               .getattr        = aufs_getattr,
23239 +
23240 +#ifdef CONFIG_AUFS_XATTR
23241 +               .listxattr      = aufs_listxattr,
23242 +#endif
23243 +
23244 +               .update_time    = aufs_update_time,
23245 +               .atomic_open    = aufs_atomic_open,
23246 +               .tmpfile        = aufs_tmpfile
23247 +       },
23248 +       [AuIop_OTHER] = {
23249 +               .permission     = aufs_permission,
23250 +#ifdef CONFIG_FS_POSIX_ACL
23251 +               .get_inode_acl  = aufs_get_inode_acl,
23252 +               .get_acl        = aufs_get_acl,
23253 +               .set_acl        = aufs_set_acl,
23254 +#endif
23255 +
23256 +               .setattr        = aufs_setattr,
23257 +               .getattr        = aufs_getattr,
23258 +
23259 +#ifdef CONFIG_AUFS_XATTR
23260 +               .listxattr      = aufs_listxattr,
23261 +#endif
23262 +
23263 +               .update_time    = aufs_update_time
23264 +       }
23265 +};
23266 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
23267 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
23268 +++ linux/fs/aufs/i_op_del.c    2022-11-05 23:02:18.965889284 +0100
23269 @@ -0,0 +1,522 @@
23270 +// SPDX-License-Identifier: GPL-2.0
23271 +/*
23272 + * Copyright (C) 2005-2022 Junjiro R. Okajima
23273 + *
23274 + * This program is free software; you can redistribute it and/or modify
23275 + * it under the terms of the GNU General Public License as published by
23276 + * the Free Software Foundation; either version 2 of the License, or
23277 + * (at your option) any later version.
23278 + *
23279 + * This program is distributed in the hope that it will be useful,
23280 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23281 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
23282 + * GNU General Public License for more details.
23283 + *
23284 + * You should have received a copy of the GNU General Public License
23285 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
23286 + */
23287 +
23288 +/*
23289 + * inode operations (del entry)
23290 + */
23291 +
23292 +#include <linux/iversion.h>
23293 +#include "aufs.h"
23294 +
23295 +/*
23296 + * decide if a new whiteout for @dentry is necessary or not.
23297 + * when it is necessary, prepare the parent dir for the upper branch whose
23298 + * branch index is @bcpup for creation. the actual creation of the whiteout will
23299 + * be done by caller.
23300 + * return value:
23301 + * 0: wh is unnecessary
23302 + * plus: wh is necessary
23303 + * minus: error
23304 + */
23305 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
23306 +{
23307 +       int need_wh, err;
23308 +       aufs_bindex_t btop;
23309 +       struct super_block *sb;
23310 +
23311 +       sb = dentry->d_sb;
23312 +       btop = au_dbtop(dentry);
23313 +       if (*bcpup < 0) {
23314 +               *bcpup = btop;
23315 +               if (au_test_ro(sb, btop, d_inode(dentry))) {
23316 +                       err = AuWbrCopyup(au_sbi(sb), dentry);
23317 +                       *bcpup = err;
23318 +                       if (unlikely(err < 0))
23319 +                               goto out;
23320 +               }
23321 +       } else
23322 +               AuDebugOn(btop < *bcpup
23323 +                         || au_test_ro(sb, *bcpup, d_inode(dentry)));
23324 +       AuDbg("bcpup %d, btop %d\n", *bcpup, btop);
23325 +
23326 +       if (*bcpup != btop) {
23327 +               err = au_cpup_dirs(dentry, *bcpup);
23328 +               if (unlikely(err))
23329 +                       goto out;
23330 +               need_wh = 1;
23331 +       } else {
23332 +               struct au_dinfo *dinfo, *tmp;
23333 +
23334 +               need_wh = -ENOMEM;
23335 +               dinfo = au_di(dentry);
23336 +               tmp = au_di_alloc(sb, AuLsc_DI_TMP);
23337 +               if (tmp) {
23338 +                       au_di_cp(tmp, dinfo);
23339 +                       au_di_swap(tmp, dinfo);
23340 +                       /* returns the number of positive dentries */
23341 +                       need_wh = au_lkup_dentry(dentry, btop + 1,
23342 +                                                /* AuLkup_IGNORE_PERM */ 0);
23343 +                       au_di_swap(tmp, dinfo);
23344 +                       au_rw_write_unlock(&tmp->di_rwsem);
23345 +                       au_di_free(tmp);
23346 +               }
23347 +       }
23348 +       AuDbg("need_wh %d\n", need_wh);
23349 +       err = need_wh;
23350 +
23351 +out:
23352 +       return err;
23353 +}
23354 +
23355 +/*
23356 + * simple tests for the del-entry operations.
23357 + * following the checks in vfs, plus the parent-child relationship.
23358 + */
23359 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
23360 +              struct dentry *h_parent, int isdir)
23361 +{
23362 +       int err;
23363 +       umode_t h_mode;
23364 +       struct dentry *h_dentry, *h_latest;
23365 +       struct inode *h_inode;
23366 +       struct path h_ppath;
23367 +       struct super_block *sb;
23368 +       struct au_branch *br;
23369 +       struct user_namespace *h_userns;
23370 +
23371 +       h_dentry = au_h_dptr(dentry, bindex);
23372 +       if (d_really_is_positive(dentry)) {
23373 +               err = -ENOENT;
23374 +               if (unlikely(d_is_negative(h_dentry)))
23375 +                       goto out;
23376 +               h_inode = d_inode(h_dentry);
23377 +               if (unlikely(!h_inode->i_nlink))
23378 +                       goto out;
23379 +
23380 +               h_mode = h_inode->i_mode;
23381 +               if (!isdir) {
23382 +                       err = -EISDIR;
23383 +                       if (unlikely(S_ISDIR(h_mode)))
23384 +                               goto out;
23385 +               } else if (unlikely(!S_ISDIR(h_mode))) {
23386 +                       err = -ENOTDIR;
23387 +                       goto out;
23388 +               }
23389 +       } else {
23390 +               /* rename(2) case */
23391 +               err = -EIO;
23392 +               if (unlikely(d_is_positive(h_dentry)))
23393 +                       goto out;
23394 +       }
23395 +
23396 +       err = -ENOENT;
23397 +       /* expected parent dir is locked */
23398 +       if (unlikely(h_parent != h_dentry->d_parent))
23399 +               goto out;
23400 +       err = 0;
23401 +
23402 +       /*
23403 +        * rmdir a dir may break the consistency on some filesystem.
23404 +        * let's try heavy test.
23405 +        */
23406 +       err = -EACCES;
23407 +       sb = dentry->d_sb;
23408 +       br = au_sbr(sb, bindex);
23409 +       h_userns = au_br_userns(br);
23410 +       if (unlikely(!au_opt_test(au_mntflags(sb), DIRPERM1)
23411 +                    && au_test_h_perm(h_userns, d_inode(h_parent),
23412 +                                      MAY_EXEC | MAY_WRITE)))
23413 +               goto out;
23414 +
23415 +       h_ppath.dentry = h_parent;
23416 +       h_ppath.mnt = au_br_mnt(br);
23417 +       h_latest = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
23418 +       err = -EIO;
23419 +       if (IS_ERR(h_latest))
23420 +               goto out;
23421 +       if (h_latest == h_dentry)
23422 +               err = 0;
23423 +       dput(h_latest);
23424 +
23425 +out:
23426 +       return err;
23427 +}
23428 +
23429 +/*
23430 + * decide the branch where we operate for @dentry. the branch index will be set
23431 + * @rbcpup. after deciding it, 'pin' it and store the timestamps of the parent
23432 + * dir for reverting.
23433 + * when a new whiteout is necessary, create it.
23434 + */
23435 +static struct dentry*
23436 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
23437 +                   struct au_dtime *dt, struct au_pin *pin)
23438 +{
23439 +       struct dentry *wh_dentry;
23440 +       struct super_block *sb;
23441 +       struct path h_path;
23442 +       int err, need_wh;
23443 +       unsigned int udba;
23444 +       aufs_bindex_t bcpup;
23445 +
23446 +       need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
23447 +       wh_dentry = ERR_PTR(need_wh);
23448 +       if (unlikely(need_wh < 0))
23449 +               goto out;
23450 +
23451 +       sb = dentry->d_sb;
23452 +       udba = au_opt_udba(sb);
23453 +       bcpup = *rbcpup;
23454 +       err = au_pin(pin, dentry, bcpup, udba,
23455 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23456 +       wh_dentry = ERR_PTR(err);
23457 +       if (unlikely(err))
23458 +               goto out;
23459 +
23460 +       h_path.dentry = au_pinned_h_parent(pin);
23461 +       if (udba != AuOpt_UDBA_NONE
23462 +           && au_dbtop(dentry) == bcpup) {
23463 +               err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
23464 +               wh_dentry = ERR_PTR(err);
23465 +               if (unlikely(err))
23466 +                       goto out_unpin;
23467 +       }
23468 +
23469 +       h_path.mnt = au_sbr_mnt(sb, bcpup);
23470 +       au_dtime_store(dt, au_pinned_parent(pin), &h_path);
23471 +       wh_dentry = NULL;
23472 +       if (!need_wh)
23473 +               goto out; /* success, no need to create whiteout */
23474 +
23475 +       wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
23476 +       if (IS_ERR(wh_dentry))
23477 +               goto out_unpin;
23478 +
23479 +       /* returns with the parent is locked and wh_dentry is dget-ed */
23480 +       goto out; /* success */
23481 +
23482 +out_unpin:
23483 +       au_unpin(pin);
23484 +out:
23485 +       return wh_dentry;
23486 +}
23487 +
23488 +/*
23489 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
23490 + * in order to be revertible and save time for removing many child whiteouts
23491 + * under the dir.
23492 + * returns 1 when there are too many child whiteout and caller should remove
23493 + * them asynchronously. returns 0 when the number of children is enough small to
23494 + * remove now or the branch fs is a remote fs.
23495 + * otherwise return an error.
23496 + */
23497 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
23498 +                          struct au_nhash *whlist, struct inode *dir)
23499 +{
23500 +       int rmdir_later, err, dirwh;
23501 +       struct dentry *h_dentry;
23502 +       struct super_block *sb;
23503 +       struct inode *inode;
23504 +
23505 +       sb = dentry->d_sb;
23506 +       SiMustAnyLock(sb);
23507 +       h_dentry = au_h_dptr(dentry, bindex);
23508 +       err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
23509 +       if (unlikely(err))
23510 +               goto out;
23511 +
23512 +       /* stop monitoring */
23513 +       inode = d_inode(dentry);
23514 +       au_hn_free(au_hi(inode, bindex));
23515 +
23516 +       if (!au_test_fs_remote(h_dentry->d_sb)) {
23517 +               dirwh = au_sbi(sb)->si_dirwh;
23518 +               rmdir_later = (dirwh <= 1);
23519 +               if (!rmdir_later)
23520 +                       rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
23521 +                                                             dirwh);
23522 +               if (rmdir_later)
23523 +                       return rmdir_later;
23524 +       }
23525 +
23526 +       err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
23527 +       if (unlikely(err)) {
23528 +               AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
23529 +                       h_dentry, bindex, err);
23530 +               err = 0;
23531 +       }
23532 +
23533 +out:
23534 +       AuTraceErr(err);
23535 +       return err;
23536 +}
23537 +
23538 +/*
23539 + * final procedure for deleting a entry.
23540 + * maintain dentry and iattr.
23541 + */
23542 +static void epilog(struct inode *dir, struct dentry *dentry,
23543 +                  aufs_bindex_t bindex)
23544 +{
23545 +       struct inode *inode;
23546 +
23547 +       inode = d_inode(dentry);
23548 +       d_drop(dentry);
23549 +       inode->i_ctime = dir->i_ctime;
23550 +
23551 +       au_dir_ts(dir, bindex);
23552 +       inode_inc_iversion(dir);
23553 +}
23554 +
23555 +/*
23556 + * when an error happened, remove the created whiteout and revert everything.
23557 + */
23558 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
23559 +                    aufs_bindex_t bwh, struct dentry *wh_dentry,
23560 +                    struct dentry *dentry, struct au_dtime *dt)
23561 +{
23562 +       int rerr;
23563 +       struct path h_path = {
23564 +               .dentry = wh_dentry,
23565 +               .mnt    = au_sbr_mnt(dir->i_sb, bindex)
23566 +       };
23567 +
23568 +       rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
23569 +       if (!rerr) {
23570 +               au_set_dbwh(dentry, bwh);
23571 +               au_dtime_revert(dt);
23572 +               return 0;
23573 +       }
23574 +
23575 +       AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
23576 +       return -EIO;
23577 +}
23578 +
23579 +/* ---------------------------------------------------------------------- */
23580 +
23581 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
23582 +{
23583 +       int err;
23584 +       aufs_bindex_t bwh, bindex, btop;
23585 +       struct inode *inode, *h_dir, *delegated;
23586 +       struct dentry *parent, *wh_dentry;
23587 +       /* to reduce stack size */
23588 +       struct {
23589 +               struct au_dtime dt;
23590 +               struct au_pin pin;
23591 +               struct path h_path;
23592 +       } *a;
23593 +
23594 +       IMustLock(dir);
23595 +
23596 +       err = -ENOMEM;
23597 +       a = kmalloc(sizeof(*a), GFP_NOFS);
23598 +       if (unlikely(!a))
23599 +               goto out;
23600 +
23601 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
23602 +       if (unlikely(err))
23603 +               goto out_free;
23604 +       err = au_d_hashed_positive(dentry);
23605 +       if (unlikely(err))
23606 +               goto out_unlock;
23607 +       inode = d_inode(dentry);
23608 +       IMustLock(inode);
23609 +       err = -EISDIR;
23610 +       if (unlikely(d_is_dir(dentry)))
23611 +               goto out_unlock; /* possible? */
23612 +
23613 +       btop = au_dbtop(dentry);
23614 +       bwh = au_dbwh(dentry);
23615 +       bindex = -1;
23616 +       parent = dentry->d_parent; /* dir inode is locked */
23617 +       di_write_lock_parent(parent);
23618 +       wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
23619 +                                       &a->pin);
23620 +       err = PTR_ERR(wh_dentry);
23621 +       if (IS_ERR(wh_dentry))
23622 +               goto out_parent;
23623 +
23624 +       a->h_path.mnt = au_sbr_mnt(dentry->d_sb, btop);
23625 +       a->h_path.dentry = au_h_dptr(dentry, btop);
23626 +       dget(a->h_path.dentry);
23627 +       if (bindex == btop) {
23628 +               h_dir = au_pinned_h_dir(&a->pin);
23629 +               delegated = NULL;
23630 +               err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
23631 +               if (unlikely(err == -EWOULDBLOCK)) {
23632 +                       pr_warn("cannot retry for NFSv4 delegation"
23633 +                               " for an internal unlink\n");
23634 +                       iput(delegated);
23635 +               }
23636 +       } else {
23637 +               /* dir inode is locked */
23638 +               h_dir = d_inode(wh_dentry->d_parent);
23639 +               IMustLock(h_dir);
23640 +               err = 0;
23641 +       }
23642 +
23643 +       if (!err) {
23644 +               vfsub_drop_nlink(inode);
23645 +               epilog(dir, dentry, bindex);
23646 +
23647 +               /* update target timestamps */
23648 +               if (bindex == btop) {
23649 +                       vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
23650 +                       /*ignore*/
23651 +                       inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
23652 +               } else
23653 +                       /* todo: this timestamp may be reverted later */
23654 +                       inode->i_ctime = h_dir->i_ctime;
23655 +               goto out_unpin; /* success */
23656 +       }
23657 +
23658 +       /* revert */
23659 +       if (wh_dentry) {
23660 +               int rerr;
23661 +
23662 +               rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23663 +                                &a->dt);
23664 +               if (rerr)
23665 +                       err = rerr;
23666 +       }
23667 +
23668 +out_unpin:
23669 +       au_unpin(&a->pin);
23670 +       dput(wh_dentry);
23671 +       dput(a->h_path.dentry);
23672 +out_parent:
23673 +       di_write_unlock(parent);
23674 +out_unlock:
23675 +       aufs_read_unlock(dentry, AuLock_DW);
23676 +out_free:
23677 +       au_kfree_rcu(a);
23678 +out:
23679 +       return err;
23680 +}
23681 +
23682 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
23683 +{
23684 +       int err, rmdir_later;
23685 +       aufs_bindex_t bwh, bindex, btop;
23686 +       struct inode *inode;
23687 +       struct dentry *parent, *wh_dentry, *h_dentry;
23688 +       struct au_whtmp_rmdir *args;
23689 +       /* to reduce stack size */
23690 +       struct {
23691 +               struct au_dtime dt;
23692 +               struct au_pin pin;
23693 +       } *a;
23694 +
23695 +       IMustLock(dir);
23696 +
23697 +       err = -ENOMEM;
23698 +       a = kmalloc(sizeof(*a), GFP_NOFS);
23699 +       if (unlikely(!a))
23700 +               goto out;
23701 +
23702 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
23703 +       if (unlikely(err))
23704 +               goto out_free;
23705 +       err = au_alive_dir(dentry);
23706 +       if (unlikely(err))
23707 +               goto out_unlock;
23708 +       inode = d_inode(dentry);
23709 +       IMustLock(inode);
23710 +       err = -ENOTDIR;
23711 +       if (unlikely(!d_is_dir(dentry)))
23712 +               goto out_unlock; /* possible? */
23713 +
23714 +       err = -ENOMEM;
23715 +       args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
23716 +       if (unlikely(!args))
23717 +               goto out_unlock;
23718 +
23719 +       parent = dentry->d_parent; /* dir inode is locked */
23720 +       di_write_lock_parent(parent);
23721 +       err = au_test_empty(dentry, &args->whlist);
23722 +       if (unlikely(err))
23723 +               goto out_parent;
23724 +
23725 +       btop = au_dbtop(dentry);
23726 +       bwh = au_dbwh(dentry);
23727 +       bindex = -1;
23728 +       wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
23729 +                                       &a->pin);
23730 +       err = PTR_ERR(wh_dentry);
23731 +       if (IS_ERR(wh_dentry))
23732 +               goto out_parent;
23733 +
23734 +       h_dentry = au_h_dptr(dentry, btop);
23735 +       dget(h_dentry);
23736 +       rmdir_later = 0;
23737 +       if (bindex == btop) {
23738 +               err = renwh_and_rmdir(dentry, btop, &args->whlist, dir);
23739 +               if (err > 0) {
23740 +                       rmdir_later = err;
23741 +                       err = 0;
23742 +               }
23743 +       } else {
23744 +               /* stop monitoring */
23745 +               au_hn_free(au_hi(inode, btop));
23746 +
23747 +               /* dir inode is locked */
23748 +               IMustLock(d_inode(wh_dentry->d_parent));
23749 +               err = 0;
23750 +       }
23751 +
23752 +       if (!err) {
23753 +               vfsub_dead_dir(inode);
23754 +               au_set_dbdiropq(dentry, -1);
23755 +               epilog(dir, dentry, bindex);
23756 +
23757 +               if (rmdir_later) {
23758 +                       au_whtmp_kick_rmdir(dir, btop, h_dentry, args);
23759 +                       args = NULL;
23760 +               }
23761 +
23762 +               goto out_unpin; /* success */
23763 +       }
23764 +
23765 +       /* revert */
23766 +       AuLabel(revert);
23767 +       if (wh_dentry) {
23768 +               int rerr;
23769 +
23770 +               rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23771 +                                &a->dt);
23772 +               if (rerr)
23773 +                       err = rerr;
23774 +       }
23775 +
23776 +out_unpin:
23777 +       au_unpin(&a->pin);
23778 +       dput(wh_dentry);
23779 +       dput(h_dentry);
23780 +out_parent:
23781 +       di_write_unlock(parent);
23782 +       if (args)
23783 +               au_whtmp_rmdir_free(args);
23784 +out_unlock:
23785 +       aufs_read_unlock(dentry, AuLock_DW);
23786 +out_free:
23787 +       au_kfree_rcu(a);
23788 +out:
23789 +       AuTraceErr(err);
23790 +       return err;
23791 +}
23792 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
23793 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
23794 +++ linux/fs/aufs/i_op_ren.c    2022-11-05 23:02:18.965889284 +0100
23795 @@ -0,0 +1,1257 @@
23796 +// SPDX-License-Identifier: GPL-2.0
23797 +/*
23798 + * Copyright (C) 2005-2022 Junjiro R. Okajima
23799 + *
23800 + * This program is free software; you can redistribute it and/or modify
23801 + * it under the terms of the GNU General Public License as published by
23802 + * the Free Software Foundation; either version 2 of the License, or
23803 + * (at your option) any later version.
23804 + *
23805 + * This program is distributed in the hope that it will be useful,
23806 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23807 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
23808 + * GNU General Public License for more details.
23809 + *
23810 + * You should have received a copy of the GNU General Public License
23811 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
23812 + */
23813 +
23814 +/*
23815 + * inode operation (rename entry)
23816 + * todo: this is crazy monster
23817 + */
23818 +
23819 +#include <linux/iversion.h>
23820 +#include "aufs.h"
23821 +
23822 +enum { AuSRC, AuDST, AuSrcDst };
23823 +enum { AuPARENT, AuCHILD, AuParentChild };
23824 +
23825 +#define AuRen_ISDIR_SRC                1
23826 +#define AuRen_ISDIR_DST                (1 << 1)
23827 +#define AuRen_ISSAMEDIR                (1 << 2)
23828 +#define AuRen_WHSRC            (1 << 3)
23829 +#define AuRen_WHDST            (1 << 4)
23830 +#define AuRen_MNT_WRITE                (1 << 5)
23831 +#define AuRen_DT_DSTDIR                (1 << 6)
23832 +#define AuRen_DIROPQ_SRC       (1 << 7)
23833 +#define AuRen_DIROPQ_DST       (1 << 8)
23834 +#define AuRen_DIRREN           (1 << 9)
23835 +#define AuRen_DROPPED_SRC      (1 << 10)
23836 +#define AuRen_DROPPED_DST      (1 << 11)
23837 +#define au_ftest_ren(flags, name)      ((flags) & AuRen_##name)
23838 +#define au_fset_ren(flags, name) \
23839 +       do { (flags) |= AuRen_##name; } while (0)
23840 +#define au_fclr_ren(flags, name) \
23841 +       do { (flags) &= ~AuRen_##name; } while (0)
23842 +
23843 +#ifndef CONFIG_AUFS_DIRREN
23844 +#undef AuRen_DIRREN
23845 +#define AuRen_DIRREN           0
23846 +#endif
23847 +
23848 +struct au_ren_args {
23849 +       struct {
23850 +               struct dentry *dentry, *h_dentry, *parent, *h_parent,
23851 +                       *wh_dentry;
23852 +               struct inode *dir, *inode;
23853 +               struct au_hinode *hdir, *hinode;
23854 +               struct au_dtime dt[AuParentChild];
23855 +               aufs_bindex_t btop, bdiropq;
23856 +       } sd[AuSrcDst];
23857 +
23858 +#define src_dentry     sd[AuSRC].dentry
23859 +#define src_dir                sd[AuSRC].dir
23860 +#define src_inode      sd[AuSRC].inode
23861 +#define src_h_dentry   sd[AuSRC].h_dentry
23862 +#define src_parent     sd[AuSRC].parent
23863 +#define src_h_parent   sd[AuSRC].h_parent
23864 +#define src_wh_dentry  sd[AuSRC].wh_dentry
23865 +#define src_hdir       sd[AuSRC].hdir
23866 +#define src_hinode     sd[AuSRC].hinode
23867 +#define src_h_dir      sd[AuSRC].hdir->hi_inode
23868 +#define src_dt         sd[AuSRC].dt
23869 +#define src_btop       sd[AuSRC].btop
23870 +#define src_bdiropq    sd[AuSRC].bdiropq
23871 +
23872 +#define dst_dentry     sd[AuDST].dentry
23873 +#define dst_dir                sd[AuDST].dir
23874 +#define dst_inode      sd[AuDST].inode
23875 +#define dst_h_dentry   sd[AuDST].h_dentry
23876 +#define dst_parent     sd[AuDST].parent
23877 +#define dst_h_parent   sd[AuDST].h_parent
23878 +#define dst_wh_dentry  sd[AuDST].wh_dentry
23879 +#define dst_hdir       sd[AuDST].hdir
23880 +#define dst_hinode     sd[AuDST].hinode
23881 +#define dst_h_dir      sd[AuDST].hdir->hi_inode
23882 +#define dst_dt         sd[AuDST].dt
23883 +#define dst_btop       sd[AuDST].btop
23884 +#define dst_bdiropq    sd[AuDST].bdiropq
23885 +
23886 +       struct dentry *h_trap;
23887 +       struct au_branch *br;
23888 +       struct path h_path;
23889 +       struct au_nhash whlist;
23890 +       aufs_bindex_t btgt, src_bwh;
23891 +
23892 +       struct {
23893 +               unsigned short auren_flags;
23894 +               unsigned char flags;    /* syscall parameter */
23895 +               unsigned char exchange;
23896 +       } __packed;
23897 +
23898 +       struct au_whtmp_rmdir *thargs;
23899 +       struct dentry *h_dst;
23900 +       struct au_hinode *h_root;
23901 +};
23902 +
23903 +/* ---------------------------------------------------------------------- */
23904 +
23905 +/*
23906 + * functions for reverting.
23907 + * when an error happened in a single rename systemcall, we should revert
23908 + * everything as if nothing happened.
23909 + * we don't need to revert the copied-up/down the parent dir since they are
23910 + * harmless.
23911 + */
23912 +
23913 +#define RevertFailure(fmt, ...) do { \
23914 +       AuIOErr("revert failure: " fmt " (%d, %d)\n", \
23915 +               ##__VA_ARGS__, err, rerr); \
23916 +       err = -EIO; \
23917 +} while (0)
23918 +
23919 +static void au_ren_do_rev_diropq(int err, struct au_ren_args *a, int idx)
23920 +{
23921 +       int rerr;
23922 +       struct dentry *d;
23923 +#define src_or_dst(member) a->sd[idx].member
23924 +
23925 +       d = src_or_dst(dentry); /* {src,dst}_dentry */
23926 +       au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
23927 +       rerr = au_diropq_remove(d, a->btgt);
23928 +       au_hn_inode_unlock(src_or_dst(hinode));
23929 +       au_set_dbdiropq(d, src_or_dst(bdiropq));
23930 +       if (rerr)
23931 +               RevertFailure("remove diropq %pd", d);
23932 +
23933 +#undef src_or_dst_
23934 +}
23935 +
23936 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
23937 +{
23938 +       if (au_ftest_ren(a->auren_flags, DIROPQ_SRC))
23939 +               au_ren_do_rev_diropq(err, a, AuSRC);
23940 +       if (au_ftest_ren(a->auren_flags, DIROPQ_DST))
23941 +               au_ren_do_rev_diropq(err, a, AuDST);
23942 +}
23943 +
23944 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
23945 +{
23946 +       int rerr;
23947 +       struct inode *delegated;
23948 +       struct path h_ppath = {
23949 +               .dentry = a->src_h_parent,
23950 +               .mnt    = a->h_path.mnt
23951 +       };
23952 +
23953 +       a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name, &h_ppath);
23954 +       rerr = PTR_ERR(a->h_path.dentry);
23955 +       if (IS_ERR(a->h_path.dentry)) {
23956 +               RevertFailure("lkup one %pd", a->src_dentry);
23957 +               return;
23958 +       }
23959 +
23960 +       delegated = NULL;
23961 +       rerr = vfsub_rename(a->dst_h_dir,
23962 +                           au_h_dptr(a->src_dentry, a->btgt),
23963 +                           a->src_h_dir, &a->h_path, &delegated, a->flags);
23964 +       if (unlikely(rerr == -EWOULDBLOCK)) {
23965 +               pr_warn("cannot retry for NFSv4 delegation"
23966 +                       " for an internal rename\n");
23967 +               iput(delegated);
23968 +       }
23969 +       d_drop(a->h_path.dentry);
23970 +       dput(a->h_path.dentry);
23971 +       /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
23972 +       if (rerr)
23973 +               RevertFailure("rename %pd", a->src_dentry);
23974 +}
23975 +
23976 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
23977 +{
23978 +       int rerr;
23979 +       struct inode *delegated;
23980 +       struct path h_ppath = {
23981 +               .dentry = a->dst_h_parent,
23982 +               .mnt    = a->h_path.mnt
23983 +       };
23984 +
23985 +       a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name, &h_ppath);
23986 +       rerr = PTR_ERR(a->h_path.dentry);
23987 +       if (IS_ERR(a->h_path.dentry)) {
23988 +               RevertFailure("lkup one %pd", a->dst_dentry);
23989 +               return;
23990 +       }
23991 +       if (d_is_positive(a->h_path.dentry)) {
23992 +               d_drop(a->h_path.dentry);
23993 +               dput(a->h_path.dentry);
23994 +               return;
23995 +       }
23996 +
23997 +       delegated = NULL;
23998 +       rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
23999 +                           &delegated, a->flags);
24000 +       if (unlikely(rerr == -EWOULDBLOCK)) {
24001 +               pr_warn("cannot retry for NFSv4 delegation"
24002 +                       " for an internal rename\n");
24003 +               iput(delegated);
24004 +       }
24005 +       d_drop(a->h_path.dentry);
24006 +       dput(a->h_path.dentry);
24007 +       if (!rerr)
24008 +               au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
24009 +       else
24010 +               RevertFailure("rename %pd", a->h_dst);
24011 +}
24012 +
24013 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
24014 +{
24015 +       int rerr;
24016 +
24017 +       a->h_path.dentry = a->src_wh_dentry;
24018 +       rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
24019 +       au_set_dbwh(a->src_dentry, a->src_bwh);
24020 +       if (rerr)
24021 +               RevertFailure("unlink %pd", a->src_wh_dentry);
24022 +}
24023 +#undef RevertFailure
24024 +
24025 +/* ---------------------------------------------------------------------- */
24026 +
24027 +/*
24028 + * when we have to copyup the renaming entry, do it with the rename-target name
24029 + * in order to minimize the cost (the later actual rename is unnecessary).
24030 + * otherwise rename it on the target branch.
24031 + */
24032 +static int au_ren_or_cpup(struct au_ren_args *a)
24033 +{
24034 +       int err;
24035 +       struct dentry *d;
24036 +       struct inode *delegated;
24037 +
24038 +       d = a->src_dentry;
24039 +       if (au_dbtop(d) == a->btgt) {
24040 +               a->h_path.dentry = a->dst_h_dentry;
24041 +               AuDebugOn(au_dbtop(d) != a->btgt);
24042 +               delegated = NULL;
24043 +               err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
24044 +                                  a->dst_h_dir, &a->h_path, &delegated,
24045 +                                  a->flags);
24046 +               if (unlikely(err == -EWOULDBLOCK)) {
24047 +                       pr_warn("cannot retry for NFSv4 delegation"
24048 +                               " for an internal rename\n");
24049 +                       iput(delegated);
24050 +               }
24051 +       } else
24052 +               BUG();
24053 +
24054 +       if (!err && a->h_dst)
24055 +               /* it will be set to dinfo later */
24056 +               dget(a->h_dst);
24057 +
24058 +       return err;
24059 +}
24060 +
24061 +/* cf. aufs_rmdir() */
24062 +static int au_ren_del_whtmp(struct au_ren_args *a)
24063 +{
24064 +       int err;
24065 +       struct inode *dir;
24066 +
24067 +       dir = a->dst_dir;
24068 +       SiMustAnyLock(dir->i_sb);
24069 +       if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
24070 +                                    au_sbi(dir->i_sb)->si_dirwh)
24071 +           || au_test_fs_remote(a->h_dst->d_sb)) {
24072 +               err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
24073 +               if (unlikely(err))
24074 +                       pr_warn("failed removing whtmp dir %pd (%d), "
24075 +                               "ignored.\n", a->h_dst, err);
24076 +       } else {
24077 +               au_nhash_wh_free(&a->thargs->whlist);
24078 +               a->thargs->whlist = a->whlist;
24079 +               a->whlist.nh_num = 0;
24080 +               au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
24081 +               dput(a->h_dst);
24082 +               a->thargs = NULL;
24083 +       }
24084 +
24085 +       return 0;
24086 +}
24087 +
24088 +/* make it 'opaque' dir. */
24089 +static int au_ren_do_diropq(struct au_ren_args *a, int idx)
24090 +{
24091 +       int err;
24092 +       struct dentry *d, *diropq;
24093 +#define src_or_dst(member) a->sd[idx].member
24094 +
24095 +       err = 0;
24096 +       d = src_or_dst(dentry); /* {src,dst}_dentry */
24097 +       src_or_dst(bdiropq) = au_dbdiropq(d);
24098 +       src_or_dst(hinode) = au_hi(src_or_dst(inode), a->btgt);
24099 +       au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
24100 +       diropq = au_diropq_create(d, a->btgt);
24101 +       au_hn_inode_unlock(src_or_dst(hinode));
24102 +       if (IS_ERR(diropq))
24103 +               err = PTR_ERR(diropq);
24104 +       else
24105 +               dput(diropq);
24106 +
24107 +#undef src_or_dst_
24108 +       return err;
24109 +}
24110 +
24111 +static int au_ren_diropq(struct au_ren_args *a)
24112 +{
24113 +       int err;
24114 +       unsigned char always;
24115 +       struct dentry *d;
24116 +
24117 +       err = 0;
24118 +       d = a->dst_dentry; /* already renamed on the branch */
24119 +       always = !!au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ);
24120 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24121 +           && !au_ftest_ren(a->auren_flags, DIRREN)
24122 +           && a->btgt != au_dbdiropq(a->src_dentry)
24123 +           && (a->dst_wh_dentry
24124 +               || a->btgt <= au_dbdiropq(d)
24125 +               /* hide the lower to keep xino */
24126 +               /* the lowers may not be a dir, but we hide them anyway */
24127 +               || a->btgt < au_dbbot(d)
24128 +               || always)) {
24129 +               AuDbg("here\n");
24130 +               err = au_ren_do_diropq(a, AuSRC);
24131 +               if (unlikely(err))
24132 +                       goto out;
24133 +               au_fset_ren(a->auren_flags, DIROPQ_SRC);
24134 +       }
24135 +       if (!a->exchange)
24136 +               goto out; /* success */
24137 +
24138 +       d = a->src_dentry; /* already renamed on the branch */
24139 +       if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24140 +           && a->btgt != au_dbdiropq(a->dst_dentry)
24141 +           && (a->btgt < au_dbdiropq(d)
24142 +               || a->btgt < au_dbbot(d)
24143 +               || always)) {
24144 +               AuDbgDentry(a->src_dentry);
24145 +               AuDbgDentry(a->dst_dentry);
24146 +               err = au_ren_do_diropq(a, AuDST);
24147 +               if (unlikely(err))
24148 +                       goto out_rev_src;
24149 +               au_fset_ren(a->auren_flags, DIROPQ_DST);
24150 +       }
24151 +       goto out; /* success */
24152 +
24153 +out_rev_src:
24154 +       AuDbg("err %d, reverting src\n", err);
24155 +       au_ren_rev_diropq(err, a);
24156 +out:
24157 +       return err;
24158 +}
24159 +
24160 +static int do_rename(struct au_ren_args *a)
24161 +{
24162 +       int err;
24163 +       struct dentry *d, *h_d;
24164 +
24165 +       if (!a->exchange) {
24166 +               /* prepare workqueue args for asynchronous rmdir */
24167 +               h_d = a->dst_h_dentry;
24168 +               if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24169 +                   /* && !au_ftest_ren(a->auren_flags, DIRREN) */
24170 +                   && d_is_positive(h_d)) {
24171 +                       err = -ENOMEM;
24172 +                       a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb,
24173 +                                                        GFP_NOFS);
24174 +                       if (unlikely(!a->thargs))
24175 +                               goto out;
24176 +                       a->h_dst = dget(h_d);
24177 +               }
24178 +
24179 +               /* create whiteout for src_dentry */
24180 +               if (au_ftest_ren(a->auren_flags, WHSRC)) {
24181 +                       a->src_bwh = au_dbwh(a->src_dentry);
24182 +                       AuDebugOn(a->src_bwh >= 0);
24183 +                       a->src_wh_dentry = au_wh_create(a->src_dentry, a->btgt,
24184 +                                                       a->src_h_parent);
24185 +                       err = PTR_ERR(a->src_wh_dentry);
24186 +                       if (IS_ERR(a->src_wh_dentry))
24187 +                               goto out_thargs;
24188 +               }
24189 +
24190 +               /* lookup whiteout for dentry */
24191 +               if (au_ftest_ren(a->auren_flags, WHDST)) {
24192 +                       h_d = au_wh_lkup(a->dst_h_parent,
24193 +                                        &a->dst_dentry->d_name, a->br);
24194 +                       err = PTR_ERR(h_d);
24195 +                       if (IS_ERR(h_d))
24196 +                               goto out_whsrc;
24197 +                       if (d_is_negative(h_d))
24198 +                               dput(h_d);
24199 +                       else
24200 +                               a->dst_wh_dentry = h_d;
24201 +               }
24202 +
24203 +               /* rename dentry to tmpwh */
24204 +               if (a->thargs) {
24205 +                       err = au_whtmp_ren(a->dst_h_dentry, a->br);
24206 +                       if (unlikely(err))
24207 +                               goto out_whdst;
24208 +
24209 +                       d = a->dst_dentry;
24210 +                       au_set_h_dptr(d, a->btgt, NULL);
24211 +                       err = au_lkup_neg(d, a->btgt, /*wh*/0);
24212 +                       if (unlikely(err))
24213 +                               goto out_whtmp;
24214 +                       a->dst_h_dentry = au_h_dptr(d, a->btgt);
24215 +               }
24216 +       }
24217 +
24218 +       BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_btop != a->btgt);
24219 +#if 0 /* debugging */
24220 +       BUG_ON(!au_ftest_ren(a->auren_flags, DIRREN)
24221 +              && d_is_positive(a->dst_h_dentry)
24222 +              && a->src_btop != a->btgt);
24223 +#endif
24224 +
24225 +       /* rename by vfs_rename or cpup */
24226 +       err = au_ren_or_cpup(a);
24227 +       if (unlikely(err))
24228 +               /* leave the copied-up one */
24229 +               goto out_whtmp;
24230 +
24231 +       /* make dir opaque */
24232 +       err = au_ren_diropq(a);
24233 +       if (unlikely(err))
24234 +               goto out_rename;
24235 +
24236 +       /* update target timestamps */
24237 +       if (a->exchange) {
24238 +               AuDebugOn(au_dbtop(a->dst_dentry) != a->btgt);
24239 +               a->h_path.dentry = au_h_dptr(a->dst_dentry, a->btgt);
24240 +               vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24241 +               a->dst_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24242 +       }
24243 +       AuDebugOn(au_dbtop(a->src_dentry) != a->btgt);
24244 +       a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
24245 +       vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24246 +       a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24247 +
24248 +       if (!a->exchange) {
24249 +               /* remove whiteout for dentry */
24250 +               if (a->dst_wh_dentry) {
24251 +                       a->h_path.dentry = a->dst_wh_dentry;
24252 +                       err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
24253 +                                                 a->dst_dentry);
24254 +                       if (unlikely(err))
24255 +                               goto out_diropq;
24256 +               }
24257 +
24258 +               /* remove whtmp */
24259 +               if (a->thargs)
24260 +                       au_ren_del_whtmp(a); /* ignore this error */
24261 +
24262 +               au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
24263 +       }
24264 +       err = 0;
24265 +       goto out_success;
24266 +
24267 +out_diropq:
24268 +       au_ren_rev_diropq(err, a);
24269 +out_rename:
24270 +       au_ren_rev_rename(err, a);
24271 +       dput(a->h_dst);
24272 +out_whtmp:
24273 +       if (a->thargs)
24274 +               au_ren_rev_whtmp(err, a);
24275 +out_whdst:
24276 +       dput(a->dst_wh_dentry);
24277 +       a->dst_wh_dentry = NULL;
24278 +out_whsrc:
24279 +       if (a->src_wh_dentry)
24280 +               au_ren_rev_whsrc(err, a);
24281 +out_success:
24282 +       dput(a->src_wh_dentry);
24283 +       dput(a->dst_wh_dentry);
24284 +out_thargs:
24285 +       if (a->thargs) {
24286 +               dput(a->h_dst);
24287 +               au_whtmp_rmdir_free(a->thargs);
24288 +               a->thargs = NULL;
24289 +       }
24290 +out:
24291 +       return err;
24292 +}
24293 +
24294 +/* ---------------------------------------------------------------------- */
24295 +
24296 +/*
24297 + * test if @dentry dir can be rename destination or not.
24298 + * success means, it is a logically empty dir.
24299 + */
24300 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
24301 +{
24302 +       return au_test_empty(dentry, whlist);
24303 +}
24304 +
24305 +/*
24306 + * test if @a->src_dentry dir can be rename source or not.
24307 + * if it can, return 0.
24308 + * success means,
24309 + * - it is a logically empty dir.
24310 + * - or, it exists on writable branch and has no children including whiteouts
24311 + *   on the lower branch unless DIRREN is on.
24312 + */
24313 +static int may_rename_srcdir(struct au_ren_args *a)
24314 +{
24315 +       int err;
24316 +       unsigned int rdhash;
24317 +       aufs_bindex_t btop, btgt;
24318 +       struct dentry *dentry;
24319 +       struct super_block *sb;
24320 +       struct au_sbinfo *sbinfo;
24321 +
24322 +       dentry = a->src_dentry;
24323 +       sb = dentry->d_sb;
24324 +       sbinfo = au_sbi(sb);
24325 +       if (au_opt_test(sbinfo->si_mntflags, DIRREN))
24326 +               au_fset_ren(a->auren_flags, DIRREN);
24327 +
24328 +       btgt = a->btgt;
24329 +       btop = au_dbtop(dentry);
24330 +       if (btop != btgt) {
24331 +               struct au_nhash whlist;
24332 +
24333 +               SiMustAnyLock(sb);
24334 +               rdhash = sbinfo->si_rdhash;
24335 +               if (!rdhash)
24336 +                       rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
24337 +                                                          dentry));
24338 +               err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
24339 +               if (unlikely(err))
24340 +                       goto out;
24341 +               err = au_test_empty(dentry, &whlist);
24342 +               au_nhash_wh_free(&whlist);
24343 +               goto out;
24344 +       }
24345 +
24346 +       if (btop == au_dbtaildir(dentry))
24347 +               return 0; /* success */
24348 +
24349 +       err = au_test_empty_lower(dentry);
24350 +
24351 +out:
24352 +       if (err == -ENOTEMPTY) {
24353 +               if (au_ftest_ren(a->auren_flags, DIRREN)) {
24354 +                       err = 0;
24355 +               } else {
24356 +                       AuWarn1("renaming dir who has child(ren) on multiple "
24357 +                               "branches, is not supported\n");
24358 +                       err = -EXDEV;
24359 +               }
24360 +       }
24361 +       return err;
24362 +}
24363 +
24364 +/* side effect: sets whlist and h_dentry */
24365 +static int au_ren_may_dir(struct au_ren_args *a)
24366 +{
24367 +       int err;
24368 +       unsigned int rdhash;
24369 +       struct dentry *d;
24370 +
24371 +       d = a->dst_dentry;
24372 +       SiMustAnyLock(d->d_sb);
24373 +
24374 +       err = 0;
24375 +       if (au_ftest_ren(a->auren_flags, ISDIR_DST) && a->dst_inode) {
24376 +               rdhash = au_sbi(d->d_sb)->si_rdhash;
24377 +               if (!rdhash)
24378 +                       rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
24379 +               err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
24380 +               if (unlikely(err))
24381 +                       goto out;
24382 +
24383 +               if (!a->exchange) {
24384 +                       au_set_dbtop(d, a->dst_btop);
24385 +                       err = may_rename_dstdir(d, &a->whlist);
24386 +                       au_set_dbtop(d, a->btgt);
24387 +               } else
24388 +                       err = may_rename_srcdir(a);
24389 +       }
24390 +       a->dst_h_dentry = au_h_dptr(d, au_dbtop(d));
24391 +       if (unlikely(err))
24392 +               goto out;
24393 +
24394 +       d = a->src_dentry;
24395 +       a->src_h_dentry = au_h_dptr(d, au_dbtop(d));
24396 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24397 +               err = may_rename_srcdir(a);
24398 +               if (unlikely(err)) {
24399 +                       au_nhash_wh_free(&a->whlist);
24400 +                       a->whlist.nh_num = 0;
24401 +               }
24402 +       }
24403 +out:
24404 +       return err;
24405 +}
24406 +
24407 +/* ---------------------------------------------------------------------- */
24408 +
24409 +/*
24410 + * simple tests for rename.
24411 + * following the checks in vfs, plus the parent-child relationship.
24412 + */
24413 +static int au_may_ren(struct au_ren_args *a)
24414 +{
24415 +       int err, isdir;
24416 +       struct inode *h_inode;
24417 +
24418 +       if (a->src_btop == a->btgt) {
24419 +               err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
24420 +                                au_ftest_ren(a->auren_flags, ISDIR_SRC));
24421 +               if (unlikely(err))
24422 +                       goto out;
24423 +               err = -EINVAL;
24424 +               if (unlikely(a->src_h_dentry == a->h_trap))
24425 +                       goto out;
24426 +       }
24427 +
24428 +       err = 0;
24429 +       if (a->dst_btop != a->btgt)
24430 +               goto out;
24431 +
24432 +       err = -ENOTEMPTY;
24433 +       if (unlikely(a->dst_h_dentry == a->h_trap))
24434 +               goto out;
24435 +
24436 +       err = -EIO;
24437 +       isdir = !!au_ftest_ren(a->auren_flags, ISDIR_DST);
24438 +       if (d_really_is_negative(a->dst_dentry)) {
24439 +               if (d_is_negative(a->dst_h_dentry))
24440 +                       err = au_may_add(a->dst_dentry, a->btgt,
24441 +                                        a->dst_h_parent, isdir);
24442 +       } else {
24443 +               if (unlikely(d_is_negative(a->dst_h_dentry)))
24444 +                       goto out;
24445 +               h_inode = d_inode(a->dst_h_dentry);
24446 +               if (h_inode->i_nlink)
24447 +                       err = au_may_del(a->dst_dentry, a->btgt,
24448 +                                        a->dst_h_parent, isdir);
24449 +       }
24450 +
24451 +out:
24452 +       if (unlikely(err == -ENOENT || err == -EEXIST))
24453 +               err = -EIO;
24454 +       AuTraceErr(err);
24455 +       return err;
24456 +}
24457 +
24458 +/* ---------------------------------------------------------------------- */
24459 +
24460 +/*
24461 + * locking order
24462 + * (VFS)
24463 + * - src_dir and dir by lock_rename()
24464 + * - inode if exists
24465 + * (aufs)
24466 + * - lock all
24467 + *   + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
24468 + *     + si_read_lock
24469 + *     + di_write_lock2_child()
24470 + *       + di_write_lock_child()
24471 + *        + ii_write_lock_child()
24472 + *       + di_write_lock_child2()
24473 + *        + ii_write_lock_child2()
24474 + *     + src_parent and parent
24475 + *       + di_write_lock_parent()
24476 + *        + ii_write_lock_parent()
24477 + *       + di_write_lock_parent2()
24478 + *        + ii_write_lock_parent2()
24479 + *   + lower src_dir and dir by vfsub_lock_rename()
24480 + *   + verify the every relationships between child and parent. if any
24481 + *     of them failed, unlock all and return -EBUSY.
24482 + */
24483 +static void au_ren_unlock(struct au_ren_args *a)
24484 +{
24485 +       vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
24486 +                           a->dst_h_parent, a->dst_hdir);
24487 +       if (au_ftest_ren(a->auren_flags, DIRREN)
24488 +           && a->h_root)
24489 +               au_hn_inode_unlock(a->h_root);
24490 +       if (au_ftest_ren(a->auren_flags, MNT_WRITE))
24491 +               vfsub_mnt_drop_write(au_br_mnt(a->br));
24492 +}
24493 +
24494 +static int au_ren_lock(struct au_ren_args *a)
24495 +{
24496 +       int err;
24497 +       unsigned int udba;
24498 +
24499 +       err = 0;
24500 +       a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
24501 +       a->src_hdir = au_hi(a->src_dir, a->btgt);
24502 +       a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
24503 +       a->dst_hdir = au_hi(a->dst_dir, a->btgt);
24504 +
24505 +       err = vfsub_mnt_want_write(au_br_mnt(a->br));
24506 +       if (unlikely(err))
24507 +               goto out;
24508 +       au_fset_ren(a->auren_flags, MNT_WRITE);
24509 +       if (au_ftest_ren(a->auren_flags, DIRREN)) {
24510 +               struct dentry *root;
24511 +               struct inode *dir;
24512 +
24513 +               /*
24514 +                * sbinfo is already locked, so this ii_read_lock is
24515 +                * unnecessary. but our debugging feature checks it.
24516 +                */
24517 +               root = a->src_inode->i_sb->s_root;
24518 +               if (root != a->src_parent && root != a->dst_parent) {
24519 +                       dir = d_inode(root);
24520 +                       ii_read_lock_parent3(dir);
24521 +                       a->h_root = au_hi(dir, a->btgt);
24522 +                       ii_read_unlock(dir);
24523 +                       au_hn_inode_lock_nested(a->h_root, AuLsc_I_PARENT3);
24524 +               }
24525 +       }
24526 +       a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
24527 +                                     a->dst_h_parent, a->dst_hdir);
24528 +       udba = au_opt_udba(a->src_dentry->d_sb);
24529 +       if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
24530 +                    || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
24531 +               err = au_busy_or_stale();
24532 +       if (!err && au_dbtop(a->src_dentry) == a->btgt)
24533 +               err = au_h_verify(a->src_h_dentry, udba,
24534 +                                 d_inode(a->src_h_parent), a->src_h_parent,
24535 +                                 a->br);
24536 +       if (!err && au_dbtop(a->dst_dentry) == a->btgt)
24537 +               err = au_h_verify(a->dst_h_dentry, udba,
24538 +                                 d_inode(a->dst_h_parent), a->dst_h_parent,
24539 +                                 a->br);
24540 +       if (!err)
24541 +               goto out; /* success */
24542 +
24543 +       err = au_busy_or_stale();
24544 +       au_ren_unlock(a);
24545 +
24546 +out:
24547 +       return err;
24548 +}
24549 +
24550 +/* ---------------------------------------------------------------------- */
24551 +
24552 +static void au_ren_refresh_dir(struct au_ren_args *a)
24553 +{
24554 +       struct inode *dir;
24555 +
24556 +       dir = a->dst_dir;
24557 +       inode_inc_iversion(dir);
24558 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24559 +               /* is this updating defined in POSIX? */
24560 +               au_cpup_attr_timesizes(a->src_inode);
24561 +               au_cpup_attr_nlink(dir, /*force*/1);
24562 +       }
24563 +       au_dir_ts(dir, a->btgt);
24564 +
24565 +       if (a->exchange) {
24566 +               dir = a->src_dir;
24567 +               inode_inc_iversion(dir);
24568 +               if (au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24569 +                       /* is this updating defined in POSIX? */
24570 +                       au_cpup_attr_timesizes(a->dst_inode);
24571 +                       au_cpup_attr_nlink(dir, /*force*/1);
24572 +               }
24573 +               au_dir_ts(dir, a->btgt);
24574 +       }
24575 +
24576 +       if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24577 +               return;
24578 +
24579 +       dir = a->src_dir;
24580 +       inode_inc_iversion(dir);
24581 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC))
24582 +               au_cpup_attr_nlink(dir, /*force*/1);
24583 +       au_dir_ts(dir, a->btgt);
24584 +}
24585 +
24586 +static void au_ren_refresh(struct au_ren_args *a)
24587 +{
24588 +       aufs_bindex_t bbot, bindex;
24589 +       struct dentry *d, *h_d;
24590 +       struct inode *i, *h_i;
24591 +       struct super_block *sb;
24592 +
24593 +       d = a->dst_dentry;
24594 +       d_drop(d);
24595 +       if (a->h_dst)
24596 +               /* already dget-ed by au_ren_or_cpup() */
24597 +               au_set_h_dptr(d, a->btgt, a->h_dst);
24598 +
24599 +       i = a->dst_inode;
24600 +       if (i) {
24601 +               if (!a->exchange) {
24602 +                       if (!au_ftest_ren(a->auren_flags, ISDIR_DST))
24603 +                               vfsub_drop_nlink(i);
24604 +                       else {
24605 +                               vfsub_dead_dir(i);
24606 +                               au_cpup_attr_timesizes(i);
24607 +                       }
24608 +                       au_update_dbrange(d, /*do_put_zero*/1);
24609 +               } else
24610 +                       au_cpup_attr_nlink(i, /*force*/1);
24611 +       } else {
24612 +               bbot = a->btgt;
24613 +               for (bindex = au_dbtop(d); bindex < bbot; bindex++)
24614 +                       au_set_h_dptr(d, bindex, NULL);
24615 +               bbot = au_dbbot(d);
24616 +               for (bindex = a->btgt + 1; bindex <= bbot; bindex++)
24617 +                       au_set_h_dptr(d, bindex, NULL);
24618 +               au_update_dbrange(d, /*do_put_zero*/0);
24619 +       }
24620 +
24621 +       if (a->exchange
24622 +           || au_ftest_ren(a->auren_flags, DIRREN)) {
24623 +               d_drop(a->src_dentry);
24624 +               if (au_ftest_ren(a->auren_flags, DIRREN))
24625 +                       au_set_dbwh(a->src_dentry, -1);
24626 +               return;
24627 +       }
24628 +
24629 +       d = a->src_dentry;
24630 +       au_set_dbwh(d, -1);
24631 +       bbot = au_dbbot(d);
24632 +       for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24633 +               h_d = au_h_dptr(d, bindex);
24634 +               if (h_d)
24635 +                       au_set_h_dptr(d, bindex, NULL);
24636 +       }
24637 +       au_set_dbbot(d, a->btgt);
24638 +
24639 +       sb = d->d_sb;
24640 +       i = a->src_inode;
24641 +       if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
24642 +               return; /* success */
24643 +
24644 +       bbot = au_ibbot(i);
24645 +       for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24646 +               h_i = au_h_iptr(i, bindex);
24647 +               if (h_i) {
24648 +                       au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
24649 +                       /* ignore this error */
24650 +                       au_set_h_iptr(i, bindex, NULL, 0);
24651 +               }
24652 +       }
24653 +       au_set_ibbot(i, a->btgt);
24654 +}
24655 +
24656 +/* ---------------------------------------------------------------------- */
24657 +
24658 +/* mainly for link(2) and rename(2) */
24659 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
24660 +{
24661 +       aufs_bindex_t bdiropq, bwh;
24662 +       struct dentry *parent;
24663 +       struct au_branch *br;
24664 +
24665 +       parent = dentry->d_parent;
24666 +       IMustLock(d_inode(parent)); /* dir is locked */
24667 +
24668 +       bdiropq = au_dbdiropq(parent);
24669 +       bwh = au_dbwh(dentry);
24670 +       br = au_sbr(dentry->d_sb, btgt);
24671 +       if (au_br_rdonly(br)
24672 +           || (0 <= bdiropq && bdiropq < btgt)
24673 +           || (0 <= bwh && bwh < btgt))
24674 +               btgt = -1;
24675 +
24676 +       AuDbg("btgt %d\n", btgt);
24677 +       return btgt;
24678 +}
24679 +
24680 +/* sets src_btop, dst_btop and btgt */
24681 +static int au_ren_wbr(struct au_ren_args *a)
24682 +{
24683 +       int err;
24684 +       struct au_wr_dir_args wr_dir_args = {
24685 +               /* .force_btgt  = -1, */
24686 +               .flags          = AuWrDir_ADD_ENTRY
24687 +       };
24688 +
24689 +       a->src_btop = au_dbtop(a->src_dentry);
24690 +       a->dst_btop = au_dbtop(a->dst_dentry);
24691 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24692 +           || au_ftest_ren(a->auren_flags, ISDIR_DST))
24693 +               au_fset_wrdir(wr_dir_args.flags, ISDIR);
24694 +       wr_dir_args.force_btgt = a->src_btop;
24695 +       if (a->dst_inode && a->dst_btop < a->src_btop)
24696 +               wr_dir_args.force_btgt = a->dst_btop;
24697 +       wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
24698 +       err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
24699 +       a->btgt = err;
24700 +       if (a->exchange)
24701 +               au_update_dbtop(a->dst_dentry);
24702 +
24703 +       return err;
24704 +}
24705 +
24706 +static void au_ren_dt(struct au_ren_args *a)
24707 +{
24708 +       a->h_path.dentry = a->src_h_parent;
24709 +       au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
24710 +       if (!au_ftest_ren(a->auren_flags, ISSAMEDIR)) {
24711 +               a->h_path.dentry = a->dst_h_parent;
24712 +               au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
24713 +       }
24714 +
24715 +       au_fclr_ren(a->auren_flags, DT_DSTDIR);
24716 +       if (!au_ftest_ren(a->auren_flags, ISDIR_SRC)
24717 +           && !a->exchange)
24718 +               return;
24719 +
24720 +       a->h_path.dentry = a->src_h_dentry;
24721 +       au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
24722 +       if (d_is_positive(a->dst_h_dentry)) {
24723 +               au_fset_ren(a->auren_flags, DT_DSTDIR);
24724 +               a->h_path.dentry = a->dst_h_dentry;
24725 +               au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
24726 +       }
24727 +}
24728 +
24729 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
24730 +{
24731 +       struct dentry *h_d;
24732 +       struct inode *h_inode;
24733 +
24734 +       au_dtime_revert(a->src_dt + AuPARENT);
24735 +       if (!au_ftest_ren(a->auren_flags, ISSAMEDIR))
24736 +               au_dtime_revert(a->dst_dt + AuPARENT);
24737 +
24738 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC) && err != -EIO) {
24739 +               h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
24740 +               h_inode = d_inode(h_d);
24741 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
24742 +               au_dtime_revert(a->src_dt + AuCHILD);
24743 +               inode_unlock(h_inode);
24744 +
24745 +               if (au_ftest_ren(a->auren_flags, DT_DSTDIR)) {
24746 +                       h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
24747 +                       h_inode = d_inode(h_d);
24748 +                       inode_lock_nested(h_inode, AuLsc_I_CHILD);
24749 +                       au_dtime_revert(a->dst_dt + AuCHILD);
24750 +                       inode_unlock(h_inode);
24751 +               }
24752 +       }
24753 +}
24754 +
24755 +/* ---------------------------------------------------------------------- */
24756 +
24757 +int aufs_rename(struct user_namespace *userns,
24758 +               struct inode *_src_dir, struct dentry *_src_dentry,
24759 +               struct inode *_dst_dir, struct dentry *_dst_dentry,
24760 +               unsigned int _flags)
24761 +{
24762 +       int err, lock_flags;
24763 +       void *rev;
24764 +       /* reduce stack space */
24765 +       struct au_ren_args *a;
24766 +       struct au_pin pin;
24767 +
24768 +       AuDbg("%pd, %pd, 0x%x\n", _src_dentry, _dst_dentry, _flags);
24769 +       IMustLock(_src_dir);
24770 +       IMustLock(_dst_dir);
24771 +
24772 +       err = -EINVAL;
24773 +       if (unlikely(_flags & RENAME_WHITEOUT))
24774 +               goto out;
24775 +
24776 +       err = -ENOMEM;
24777 +       BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
24778 +       a = kzalloc(sizeof(*a), GFP_NOFS);
24779 +       if (unlikely(!a))
24780 +               goto out;
24781 +
24782 +       a->flags = _flags;
24783 +       BUILD_BUG_ON(sizeof(a->exchange) == sizeof(u8)
24784 +                    && RENAME_EXCHANGE > U8_MAX);
24785 +       a->exchange = _flags & RENAME_EXCHANGE;
24786 +       a->src_dir = _src_dir;
24787 +       a->src_dentry = _src_dentry;
24788 +       a->src_inode = NULL;
24789 +       if (d_really_is_positive(a->src_dentry))
24790 +               a->src_inode = d_inode(a->src_dentry);
24791 +       a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
24792 +       a->dst_dir = _dst_dir;
24793 +       a->dst_dentry = _dst_dentry;
24794 +       a->dst_inode = NULL;
24795 +       if (d_really_is_positive(a->dst_dentry))
24796 +               a->dst_inode = d_inode(a->dst_dentry);
24797 +       a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
24798 +       if (a->dst_inode) {
24799 +               /*
24800 +                * if EXCHANGE && src is non-dir && dst is dir,
24801 +                * dst is not locked.
24802 +                */
24803 +               /* IMustLock(a->dst_inode); */
24804 +               au_igrab(a->dst_inode);
24805 +       }
24806 +
24807 +       err = -ENOTDIR;
24808 +       lock_flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
24809 +       if (d_is_dir(a->src_dentry)) {
24810 +               au_fset_ren(a->auren_flags, ISDIR_SRC);
24811 +               if (unlikely(!a->exchange
24812 +                            && d_really_is_positive(a->dst_dentry)
24813 +                            && !d_is_dir(a->dst_dentry)))
24814 +                       goto out_free;
24815 +               lock_flags |= AuLock_DIRS;
24816 +       }
24817 +       if (a->dst_inode && d_is_dir(a->dst_dentry)) {
24818 +               au_fset_ren(a->auren_flags, ISDIR_DST);
24819 +               if (unlikely(!a->exchange
24820 +                            && d_really_is_positive(a->src_dentry)
24821 +                            && !d_is_dir(a->src_dentry)))
24822 +                       goto out_free;
24823 +               lock_flags |= AuLock_DIRS;
24824 +       }
24825 +       err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
24826 +                                       lock_flags);
24827 +       if (unlikely(err))
24828 +               goto out_free;
24829 +
24830 +       err = au_d_hashed_positive(a->src_dentry);
24831 +       if (unlikely(err))
24832 +               goto out_unlock;
24833 +       err = -ENOENT;
24834 +       if (a->dst_inode) {
24835 +               /*
24836 +                * If it is a dir, VFS unhash it before this
24837 +                * function. It means we cannot rely upon d_unhashed().
24838 +                */
24839 +               if (unlikely(!a->dst_inode->i_nlink))
24840 +                       goto out_unlock;
24841 +               if (!au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24842 +                       err = au_d_hashed_positive(a->dst_dentry);
24843 +                       if (unlikely(err && !a->exchange))
24844 +                               goto out_unlock;
24845 +               } else if (unlikely(IS_DEADDIR(a->dst_inode)))
24846 +                       goto out_unlock;
24847 +       } else if (unlikely(d_unhashed(a->dst_dentry)))
24848 +               goto out_unlock;
24849 +
24850 +       /*
24851 +        * is it possible?
24852 +        * yes, it happened (in linux-3.3-rcN) but I don't know why.
24853 +        * there may exist a problem somewhere else.
24854 +        */
24855 +       err = -EINVAL;
24856 +       if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
24857 +               goto out_unlock;
24858 +
24859 +       au_fset_ren(a->auren_flags, ISSAMEDIR); /* temporary */
24860 +       di_write_lock_parent(a->dst_parent);
24861 +
24862 +       /* which branch we process */
24863 +       err = au_ren_wbr(a);
24864 +       if (unlikely(err < 0))
24865 +               goto out_parent;
24866 +       a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
24867 +       a->h_path.mnt = au_br_mnt(a->br);
24868 +
24869 +       /* are they available to be renamed */
24870 +       err = au_ren_may_dir(a);
24871 +       if (unlikely(err))
24872 +               goto out_children;
24873 +
24874 +       /* prepare the writable parent dir on the same branch */
24875 +       if (a->dst_btop == a->btgt) {
24876 +               au_fset_ren(a->auren_flags, WHDST);
24877 +       } else {
24878 +               err = au_cpup_dirs(a->dst_dentry, a->btgt);
24879 +               if (unlikely(err))
24880 +                       goto out_children;
24881 +       }
24882 +
24883 +       err = 0;
24884 +       if (!a->exchange) {
24885 +               if (a->src_dir != a->dst_dir) {
24886 +                       /*
24887 +                        * this temporary unlock is safe,
24888 +                        * because both dir->i_mutex are locked.
24889 +                        */
24890 +                       di_write_unlock(a->dst_parent);
24891 +                       di_write_lock_parent(a->src_parent);
24892 +                       err = au_wr_dir_need_wh(a->src_dentry,
24893 +                                               au_ftest_ren(a->auren_flags,
24894 +                                                            ISDIR_SRC),
24895 +                                               &a->btgt);
24896 +                       di_write_unlock(a->src_parent);
24897 +                       di_write_lock2_parent(a->src_parent, a->dst_parent,
24898 +                                             /*isdir*/1);
24899 +                       au_fclr_ren(a->auren_flags, ISSAMEDIR);
24900 +               } else
24901 +                       err = au_wr_dir_need_wh(a->src_dentry,
24902 +                                               au_ftest_ren(a->auren_flags,
24903 +                                                            ISDIR_SRC),
24904 +                                               &a->btgt);
24905 +       }
24906 +       if (unlikely(err < 0))
24907 +               goto out_children;
24908 +       if (err)
24909 +               au_fset_ren(a->auren_flags, WHSRC);
24910 +
24911 +       /* cpup src */
24912 +       if (a->src_btop != a->btgt) {
24913 +               err = au_pin(&pin, a->src_dentry, a->btgt,
24914 +                            au_opt_udba(a->src_dentry->d_sb),
24915 +                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24916 +               if (!err) {
24917 +                       struct au_cp_generic cpg = {
24918 +                               .dentry = a->src_dentry,
24919 +                               .bdst   = a->btgt,
24920 +                               .bsrc   = a->src_btop,
24921 +                               .len    = -1,
24922 +                               .pin    = &pin,
24923 +                               .flags  = AuCpup_DTIME | AuCpup_HOPEN
24924 +                       };
24925 +                       AuDebugOn(au_dbtop(a->src_dentry) != a->src_btop);
24926 +                       err = au_sio_cpup_simple(&cpg);
24927 +                       au_unpin(&pin);
24928 +               }
24929 +               if (unlikely(err))
24930 +                       goto out_children;
24931 +               a->src_btop = a->btgt;
24932 +               a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
24933 +               if (!a->exchange)
24934 +                       au_fset_ren(a->auren_flags, WHSRC);
24935 +       }
24936 +
24937 +       /* cpup dst */
24938 +       if (a->exchange && a->dst_inode
24939 +           && a->dst_btop != a->btgt) {
24940 +               err = au_pin(&pin, a->dst_dentry, a->btgt,
24941 +                            au_opt_udba(a->dst_dentry->d_sb),
24942 +                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24943 +               if (!err) {
24944 +                       struct au_cp_generic cpg = {
24945 +                               .dentry = a->dst_dentry,
24946 +                               .bdst   = a->btgt,
24947 +                               .bsrc   = a->dst_btop,
24948 +                               .len    = -1,
24949 +                               .pin    = &pin,
24950 +                               .flags  = AuCpup_DTIME | AuCpup_HOPEN
24951 +                       };
24952 +                       err = au_sio_cpup_simple(&cpg);
24953 +                       au_unpin(&pin);
24954 +               }
24955 +               if (unlikely(err))
24956 +                       goto out_children;
24957 +               a->dst_btop = a->btgt;
24958 +               a->dst_h_dentry = au_h_dptr(a->dst_dentry, a->btgt);
24959 +       }
24960 +
24961 +       /* lock them all */
24962 +       err = au_ren_lock(a);
24963 +       if (unlikely(err))
24964 +               /* leave the copied-up one */
24965 +               goto out_children;
24966 +
24967 +       if (!a->exchange) {
24968 +               if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
24969 +                       err = au_may_ren(a);
24970 +               else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
24971 +                       err = -ENAMETOOLONG;
24972 +               if (unlikely(err))
24973 +                       goto out_hdir;
24974 +       }
24975 +
24976 +       /* store timestamps to be revertible */
24977 +       au_ren_dt(a);
24978 +
24979 +       /* store dirren info */
24980 +       if (au_ftest_ren(a->auren_flags, DIRREN)) {
24981 +               err = au_dr_rename(a->src_dentry, a->btgt,
24982 +                                  &a->dst_dentry->d_name, &rev);
24983 +               AuTraceErr(err);
24984 +               if (unlikely(err))
24985 +                       goto out_dt;
24986 +       }
24987 +
24988 +       /* here we go */
24989 +       err = do_rename(a);
24990 +       if (unlikely(err))
24991 +               goto out_dirren;
24992 +
24993 +       if (au_ftest_ren(a->auren_flags, DIRREN))
24994 +               au_dr_rename_fin(a->src_dentry, a->btgt, rev);
24995 +
24996 +       /* update dir attributes */
24997 +       au_ren_refresh_dir(a);
24998 +
24999 +       /* dput/iput all lower dentries */
25000 +       au_ren_refresh(a);
25001 +
25002 +       goto out_hdir; /* success */
25003 +
25004 +out_dirren:
25005 +       if (au_ftest_ren(a->auren_flags, DIRREN))
25006 +               au_dr_rename_rev(a->src_dentry, a->btgt, rev);
25007 +out_dt:
25008 +       au_ren_rev_dt(err, a);
25009 +out_hdir:
25010 +       au_ren_unlock(a);
25011 +out_children:
25012 +       au_nhash_wh_free(&a->whlist);
25013 +       if (err && a->dst_inode && a->dst_btop != a->btgt) {
25014 +               AuDbg("btop %d, btgt %d\n", a->dst_btop, a->btgt);
25015 +               au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
25016 +               au_set_dbtop(a->dst_dentry, a->dst_btop);
25017 +       }
25018 +out_parent:
25019 +       if (!err) {
25020 +               if (d_unhashed(a->src_dentry))
25021 +                       au_fset_ren(a->auren_flags, DROPPED_SRC);
25022 +               if (d_unhashed(a->dst_dentry))
25023 +                       au_fset_ren(a->auren_flags, DROPPED_DST);
25024 +               if (!a->exchange)
25025 +                       d_move(a->src_dentry, a->dst_dentry);
25026 +               else {
25027 +                       d_exchange(a->src_dentry, a->dst_dentry);
25028 +                       if (au_ftest_ren(a->auren_flags, DROPPED_DST))
25029 +                               d_drop(a->dst_dentry);
25030 +               }
25031 +               if (au_ftest_ren(a->auren_flags, DROPPED_SRC))
25032 +                       d_drop(a->src_dentry);
25033 +       } else {
25034 +               au_update_dbtop(a->dst_dentry);
25035 +               if (!a->dst_inode)
25036 +                       d_drop(a->dst_dentry);
25037 +       }
25038 +       if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
25039 +               di_write_unlock(a->dst_parent);
25040 +       else
25041 +               di_write_unlock2(a->src_parent, a->dst_parent);
25042 +out_unlock:
25043 +       aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
25044 +out_free:
25045 +       iput(a->dst_inode);
25046 +       if (a->thargs)
25047 +               au_whtmp_rmdir_free(a->thargs);
25048 +       au_kfree_rcu(a);
25049 +out:
25050 +       AuTraceErr(err);
25051 +       return err;
25052 +}
25053 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
25054 --- /usr/share/empty/fs/aufs/Kconfig    1970-01-01 01:00:00.000000000 +0100
25055 +++ linux/fs/aufs/Kconfig       2022-11-05 23:02:18.959222617 +0100
25056 @@ -0,0 +1,199 @@
25057 +# SPDX-License-Identifier: GPL-2.0
25058 +config AUFS_FS
25059 +       tristate "Aufs (Advanced multi layered unification filesystem) support"
25060 +       help
25061 +       Aufs is a stackable unification filesystem such as Unionfs,
25062 +       which unifies several directories and provides a merged single
25063 +       directory.
25064 +       In the early days, aufs was entirely re-designed and
25065 +       re-implemented Unionfs Version 1.x series. Introducing many
25066 +       original ideas, approaches and improvements, it becomes totally
25067 +       different from Unionfs while keeping the basic features.
25068 +
25069 +if AUFS_FS
25070 +choice
25071 +       prompt "Maximum number of branches"
25072 +       default AUFS_BRANCH_MAX_127
25073 +       help
25074 +       Specifies the maximum number of branches (or member directories)
25075 +       in a single aufs. The larger value consumes more system
25076 +       resources and has a minor impact to performance.
25077 +config AUFS_BRANCH_MAX_127
25078 +       bool "127"
25079 +       help
25080 +       Specifies the maximum number of branches (or member directories)
25081 +       in a single aufs. The larger value consumes more system
25082 +       resources and has a minor impact to performance.
25083 +config AUFS_BRANCH_MAX_511
25084 +       bool "511"
25085 +       help
25086 +       Specifies the maximum number of branches (or member directories)
25087 +       in a single aufs. The larger value consumes more system
25088 +       resources and has a minor impact to performance.
25089 +config AUFS_BRANCH_MAX_1023
25090 +       bool "1023"
25091 +       help
25092 +       Specifies the maximum number of branches (or member directories)
25093 +       in a single aufs. The larger value consumes more system
25094 +       resources and has a minor impact to performance.
25095 +config AUFS_BRANCH_MAX_32767
25096 +       bool "32767"
25097 +       help
25098 +       Specifies the maximum number of branches (or member directories)
25099 +       in a single aufs. The larger value consumes more system
25100 +       resources and has a minor impact to performance.
25101 +endchoice
25102 +
25103 +config AUFS_SBILIST
25104 +       bool
25105 +       depends on AUFS_MAGIC_SYSRQ || PROC_FS
25106 +       default y
25107 +       help
25108 +       Automatic configuration for internal use.
25109 +       When aufs supports Magic SysRq or /proc, enabled automatically.
25110 +
25111 +config AUFS_HNOTIFY
25112 +       bool "Detect direct branch access (bypassing aufs)"
25113 +       help
25114 +       If you want to modify files on branches directly, eg. bypassing aufs,
25115 +       and want aufs to detect the changes of them fully, then enable this
25116 +       option and use 'udba=notify' mount option.
25117 +       Currently there is only one available configuration, "fsnotify".
25118 +       It will have a negative impact to the performance.
25119 +       See detail in aufs.5.
25120 +
25121 +choice
25122 +       prompt "method" if AUFS_HNOTIFY
25123 +       default AUFS_HFSNOTIFY
25124 +config AUFS_HFSNOTIFY
25125 +       bool "fsnotify"
25126 +       select FSNOTIFY
25127 +endchoice
25128 +
25129 +config AUFS_EXPORT
25130 +       bool "NFS-exportable aufs"
25131 +       depends on EXPORTFS
25132 +       help
25133 +       If you want to export your mounted aufs via NFS, then enable this
25134 +       option. There are several requirements for this configuration.
25135 +       See detail in aufs.5.
25136 +
25137 +config AUFS_INO_T_64
25138 +       bool
25139 +       depends on AUFS_EXPORT
25140 +       depends on 64BIT && !(ALPHA || S390)
25141 +       default y
25142 +       help
25143 +       Automatic configuration for internal use.
25144 +       /* typedef unsigned long/int __kernel_ino_t */
25145 +       /* alpha and s390x are int */
25146 +
25147 +config AUFS_XATTR
25148 +       bool "support for XATTR/EA (including Security Labels)"
25149 +       help
25150 +       If your branch fs supports XATTR/EA and you want to make them
25151 +       available in aufs too, then enable this opsion and specify the
25152 +       branch attributes for EA.
25153 +       See detail in aufs.5.
25154 +
25155 +config AUFS_FHSM
25156 +       bool "File-based Hierarchical Storage Management"
25157 +       help
25158 +       Hierarchical Storage Management (or HSM) is a well-known feature
25159 +       in the storage world. Aufs provides this feature as file-based.
25160 +       with multiple branches.
25161 +       These multiple branches are prioritized, ie. the topmost one
25162 +       should be the fastest drive and be used heavily.
25163 +
25164 +config AUFS_RDU
25165 +       bool "Readdir in userspace"
25166 +       help
25167 +       Aufs has two methods to provide a merged view for a directory,
25168 +       by a user-space library and by kernel-space natively. The latter
25169 +       is always enabled but sometimes large and slow.
25170 +       If you enable this option, install the library in aufs2-util
25171 +       package, and set some environment variables for your readdir(3),
25172 +       then the work will be handled in user-space which generally
25173 +       shows better performance in most cases.
25174 +       See detail in aufs.5.
25175 +
25176 +config AUFS_DIRREN
25177 +       bool "Workaround for rename(2)-ing a directory"
25178 +       help
25179 +       By default, aufs returns EXDEV error in renameing a dir who has
25180 +       his child on the lower branch, since it is a bad idea to issue
25181 +       rename(2) internally for every lower branch. But user may not
25182 +       accept this behaviour. So here is a workaround to allow such
25183 +       rename(2) and store some extra information on the writable
25184 +       branch. Obviously this costs high (and I don't like it).
25185 +       To use this feature, you need to enable this configuration AND
25186 +       to specify the mount option `dirren.'
25187 +       See details in aufs.5 and the design documents.
25188 +
25189 +config AUFS_SHWH
25190 +       bool "Show whiteouts"
25191 +       help
25192 +       If you want to make the whiteouts in aufs visible, then enable
25193 +       this option and specify 'shwh' mount option. Although it may
25194 +       sounds like philosophy or something, but in technically it
25195 +       simply shows the name of whiteout with keeping its behaviour.
25196 +
25197 +config AUFS_BR_RAMFS
25198 +       bool "Ramfs (initramfs/rootfs) as an aufs branch"
25199 +       help
25200 +       If you want to use ramfs as an aufs branch fs, then enable this
25201 +       option. Generally tmpfs is recommended.
25202 +       Aufs prohibited them to be a branch fs by default, because
25203 +       initramfs becomes unusable after switch_root or something
25204 +       generally. If you sets initramfs as an aufs branch and boot your
25205 +       system by switch_root, you will meet a problem easily since the
25206 +       files in initramfs may be inaccessible.
25207 +       Unless you are going to use ramfs as an aufs branch fs without
25208 +       switch_root or something, leave it N.
25209 +
25210 +config AUFS_BR_FUSE
25211 +       bool "Fuse fs as an aufs branch"
25212 +       depends on FUSE_FS
25213 +       select AUFS_POLL
25214 +       help
25215 +       If you want to use fuse-based userspace filesystem as an aufs
25216 +       branch fs, then enable this option.
25217 +       It implements the internal poll(2) operation which is
25218 +       implemented by fuse only (curretnly).
25219 +
25220 +config AUFS_POLL
25221 +       bool
25222 +       help
25223 +       Automatic configuration for internal use.
25224 +
25225 +config AUFS_BR_HFSPLUS
25226 +       bool "Hfsplus as an aufs branch"
25227 +       depends on HFSPLUS_FS
25228 +       default y
25229 +       help
25230 +       If you want to use hfsplus fs as an aufs branch fs, then enable
25231 +       this option. This option introduces a small overhead at
25232 +       copying-up a file on hfsplus.
25233 +
25234 +config AUFS_BDEV_LOOP
25235 +       bool
25236 +       depends on BLK_DEV_LOOP
25237 +       default y
25238 +       help
25239 +       Automatic configuration for internal use.
25240 +       Convert =[ym] into =y.
25241 +
25242 +config AUFS_DEBUG
25243 +       bool "Debug aufs"
25244 +       help
25245 +       Enable this to compile aufs internal debug code.
25246 +       It will have a negative impact to the performance.
25247 +
25248 +config AUFS_MAGIC_SYSRQ
25249 +       bool
25250 +       depends on AUFS_DEBUG && MAGIC_SYSRQ
25251 +       default y
25252 +       help
25253 +       Automatic configuration for internal use.
25254 +       When aufs supports Magic SysRq, enabled automatically.
25255 +endif
25256 diff -urN /usr/share/empty/fs/aufs/lcnt.h linux/fs/aufs/lcnt.h
25257 --- /usr/share/empty/fs/aufs/lcnt.h     1970-01-01 01:00:00.000000000 +0100
25258 +++ linux/fs/aufs/lcnt.h        2022-11-05 23:02:18.965889284 +0100
25259 @@ -0,0 +1,186 @@
25260 +/* SPDX-License-Identifier: GPL-2.0 */
25261 +/*
25262 + * Copyright (C) 2018-2022 Junjiro R. Okajima
25263 + *
25264 + * This program is free software; you can redistribute it and/or modify
25265 + * it under the terms of the GNU General Public License as published by
25266 + * the Free Software Foundation; either version 2 of the License, or
25267 + * (at your option) any later version.
25268 + *
25269 + * This program is distributed in the hope that it will be useful,
25270 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25271 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25272 + * GNU General Public License for more details.
25273 + *
25274 + * You should have received a copy of the GNU General Public License
25275 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25276 + */
25277 +
25278 +/*
25279 + * simple long counter wrapper
25280 + */
25281 +
25282 +#ifndef __AUFS_LCNT_H__
25283 +#define __AUFS_LCNT_H__
25284 +
25285 +#ifdef __KERNEL__
25286 +
25287 +#include "debug.h"
25288 +
25289 +#define AuLCntATOMIC   1
25290 +#define AuLCntPCPUCNT  2
25291 +/*
25292 + * why does percpu_refcount require extra synchronize_rcu()s in
25293 + * au_br_do_free()
25294 + */
25295 +#define AuLCntPCPUREF  3
25296 +
25297 +/* #define AuLCntChosen        AuLCntATOMIC */
25298 +#define AuLCntChosen   AuLCntPCPUCNT
25299 +/* #define AuLCntChosen        AuLCntPCPUREF */
25300 +
25301 +#if AuLCntChosen == AuLCntATOMIC
25302 +#include <linux/atomic.h>
25303 +
25304 +typedef atomic_long_t au_lcnt_t;
25305 +
25306 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25307 +{
25308 +       atomic_long_set(cnt, 0);
25309 +       return 0;
25310 +}
25311 +
25312 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25313 +{
25314 +       /* empty */
25315 +}
25316 +
25317 +static inline void au_lcnt_fin(au_lcnt_t *cnt __maybe_unused,
25318 +                              int do_sync __maybe_unused)
25319 +{
25320 +       /* empty */
25321 +}
25322 +
25323 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25324 +{
25325 +       atomic_long_inc(cnt);
25326 +}
25327 +
25328 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25329 +{
25330 +       atomic_long_dec(cnt);
25331 +}
25332 +
25333 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25334 +{
25335 +       return atomic_long_read(cnt);
25336 +}
25337 +#endif
25338 +
25339 +#if AuLCntChosen == AuLCntPCPUCNT
25340 +#include <linux/percpu_counter.h>
25341 +
25342 +typedef struct percpu_counter au_lcnt_t;
25343 +
25344 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25345 +{
25346 +       return percpu_counter_init(cnt, 0, GFP_NOFS);
25347 +}
25348 +
25349 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25350 +{
25351 +       /* empty */
25352 +}
25353 +
25354 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync __maybe_unused)
25355 +{
25356 +       percpu_counter_destroy(cnt);
25357 +}
25358 +
25359 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25360 +{
25361 +       percpu_counter_inc(cnt);
25362 +}
25363 +
25364 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25365 +{
25366 +       percpu_counter_dec(cnt);
25367 +}
25368 +
25369 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25370 +{
25371 +       s64 n;
25372 +
25373 +       n = percpu_counter_sum(cnt);
25374 +       BUG_ON(n < 0);
25375 +       if (LONG_MAX != LLONG_MAX
25376 +           && n > LONG_MAX)
25377 +               AuWarn1("%s\n", "wrap-around");
25378 +
25379 +       return n;
25380 +}
25381 +#endif
25382 +
25383 +#if AuLCntChosen == AuLCntPCPUREF
25384 +#include <linux/percpu-refcount.h>
25385 +
25386 +typedef struct percpu_ref au_lcnt_t;
25387 +
25388 +static inline int au_lcnt_init(au_lcnt_t *cnt, percpu_ref_func_t *release)
25389 +{
25390 +       if (!release)
25391 +               release = percpu_ref_exit;
25392 +       return percpu_ref_init(cnt, release, /*percpu mode*/0, GFP_NOFS);
25393 +}
25394 +
25395 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25396 +{
25397 +       synchronize_rcu();
25398 +}
25399 +
25400 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync)
25401 +{
25402 +       percpu_ref_kill(cnt);
25403 +       if (do_sync)
25404 +               au_lcnt_wait_for_fin(cnt);
25405 +}
25406 +
25407 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25408 +{
25409 +       percpu_ref_get(cnt);
25410 +}
25411 +
25412 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25413 +{
25414 +       percpu_ref_put(cnt);
25415 +}
25416 +
25417 +/*
25418 + * avoid calling this func as possible.
25419 + */
25420 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev)
25421 +{
25422 +       long l;
25423 +
25424 +       percpu_ref_switch_to_atomic_sync(cnt);
25425 +       l = atomic_long_read(&cnt->count);
25426 +       if (do_rev)
25427 +               percpu_ref_switch_to_percpu(cnt);
25428 +
25429 +       /* percpu_ref is initialized by 1 instead of 0 */
25430 +       return l - 1;
25431 +}
25432 +#endif
25433 +
25434 +#ifdef CONFIG_AUFS_DEBUG
25435 +#define AuLCntZero(val) do {                   \
25436 +       long l = val;                           \
25437 +       if (l)                                  \
25438 +               AuDbg("%s = %ld\n", #val, l);   \
25439 +} while (0)
25440 +#else
25441 +#define AuLCntZero(val)                do {} while (0)
25442 +#endif
25443 +
25444 +#endif /* __KERNEL__ */
25445 +#endif /* __AUFS_LCNT_H__ */
25446 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
25447 --- /usr/share/empty/fs/aufs/loop.c     1970-01-01 01:00:00.000000000 +0100
25448 +++ linux/fs/aufs/loop.c        2022-11-05 23:02:18.965889284 +0100
25449 @@ -0,0 +1,148 @@
25450 +// SPDX-License-Identifier: GPL-2.0
25451 +/*
25452 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25453 + *
25454 + * This program is free software; you can redistribute it and/or modify
25455 + * it under the terms of the GNU General Public License as published by
25456 + * the Free Software Foundation; either version 2 of the License, or
25457 + * (at your option) any later version.
25458 + *
25459 + * This program is distributed in the hope that it will be useful,
25460 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25461 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25462 + * GNU General Public License for more details.
25463 + *
25464 + * You should have received a copy of the GNU General Public License
25465 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25466 + */
25467 +
25468 +/*
25469 + * support for loopback block device as a branch
25470 + */
25471 +
25472 +#include "aufs.h"
25473 +
25474 +/* added into drivers/block/loop.c */
25475 +static struct file *(*backing_file_func)(struct super_block *sb);
25476 +
25477 +/*
25478 + * test if two lower dentries have overlapping branches.
25479 + */
25480 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
25481 +{
25482 +       struct super_block *h_sb;
25483 +       struct file *backing_file;
25484 +
25485 +       if (unlikely(!backing_file_func)) {
25486 +               /* don't load "loop" module here */
25487 +               backing_file_func = symbol_get(loop_backing_file);
25488 +               if (unlikely(!backing_file_func))
25489 +                       /* "loop" module is not loaded */
25490 +                       return 0;
25491 +       }
25492 +
25493 +       h_sb = h_adding->d_sb;
25494 +       backing_file = backing_file_func(h_sb);
25495 +       if (!backing_file)
25496 +               return 0;
25497 +
25498 +       h_adding = backing_file->f_path.dentry;
25499 +       /*
25500 +        * h_adding can be local NFS.
25501 +        * in this case aufs cannot detect the loop.
25502 +        */
25503 +       if (unlikely(h_adding->d_sb == sb))
25504 +               return 1;
25505 +       return !!au_test_subdir(h_adding, sb->s_root);
25506 +}
25507 +
25508 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
25509 +int au_test_loopback_kthread(void)
25510 +{
25511 +       int ret;
25512 +       struct task_struct *tsk = current;
25513 +       char c, comm[sizeof(tsk->comm)];
25514 +
25515 +       ret = 0;
25516 +       if (tsk->flags & PF_KTHREAD) {
25517 +               get_task_comm(comm, tsk);
25518 +               c = comm[4];
25519 +               ret = ('0' <= c && c <= '9'
25520 +                      && !strncmp(comm, "loop", 4));
25521 +       }
25522 +
25523 +       return ret;
25524 +}
25525 +
25526 +/* ---------------------------------------------------------------------- */
25527 +
25528 +#define au_warn_loopback_step  16
25529 +static int au_warn_loopback_nelem = au_warn_loopback_step;
25530 +static unsigned long *au_warn_loopback_array;
25531 +
25532 +void au_warn_loopback(struct super_block *h_sb)
25533 +{
25534 +       int i, new_nelem;
25535 +       unsigned long *a, magic;
25536 +       static DEFINE_SPINLOCK(spin);
25537 +
25538 +       magic = h_sb->s_magic;
25539 +       spin_lock(&spin);
25540 +       a = au_warn_loopback_array;
25541 +       for (i = 0; i < au_warn_loopback_nelem && *a; i++)
25542 +               if (a[i] == magic) {
25543 +                       spin_unlock(&spin);
25544 +                       return;
25545 +               }
25546 +
25547 +       /* h_sb is new to us, print it */
25548 +       if (i < au_warn_loopback_nelem) {
25549 +               a[i] = magic;
25550 +               goto pr;
25551 +       }
25552 +
25553 +       /* expand the array */
25554 +       new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
25555 +       a = au_kzrealloc(au_warn_loopback_array,
25556 +                        au_warn_loopback_nelem * sizeof(unsigned long),
25557 +                        new_nelem * sizeof(unsigned long), GFP_ATOMIC,
25558 +                        /*may_shrink*/0);
25559 +       if (a) {
25560 +               au_warn_loopback_nelem = new_nelem;
25561 +               au_warn_loopback_array = a;
25562 +               a[i] = magic;
25563 +               goto pr;
25564 +       }
25565 +
25566 +       spin_unlock(&spin);
25567 +       AuWarn1("realloc failed, ignored\n");
25568 +       return;
25569 +
25570 +pr:
25571 +       spin_unlock(&spin);
25572 +       pr_warn("you may want to try another patch for loopback file "
25573 +               "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
25574 +}
25575 +
25576 +int au_loopback_init(void)
25577 +{
25578 +       int err;
25579 +       struct super_block *sb __maybe_unused;
25580 +
25581 +       BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(*au_warn_loopback_array));
25582 +
25583 +       err = 0;
25584 +       au_warn_loopback_array = kcalloc(au_warn_loopback_step,
25585 +                                        sizeof(unsigned long), GFP_NOFS);
25586 +       if (unlikely(!au_warn_loopback_array))
25587 +               err = -ENOMEM;
25588 +
25589 +       return err;
25590 +}
25591 +
25592 +void au_loopback_fin(void)
25593 +{
25594 +       if (backing_file_func)
25595 +               symbol_put(loop_backing_file);
25596 +       au_kfree_try_rcu(au_warn_loopback_array);
25597 +}
25598 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
25599 --- /usr/share/empty/fs/aufs/loop.h     1970-01-01 01:00:00.000000000 +0100
25600 +++ linux/fs/aufs/loop.h        2022-11-05 23:02:18.965889284 +0100
25601 @@ -0,0 +1,55 @@
25602 +/* SPDX-License-Identifier: GPL-2.0 */
25603 +/*
25604 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25605 + *
25606 + * This program is free software; you can redistribute it and/or modify
25607 + * it under the terms of the GNU General Public License as published by
25608 + * the Free Software Foundation; either version 2 of the License, or
25609 + * (at your option) any later version.
25610 + *
25611 + * This program is distributed in the hope that it will be useful,
25612 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25613 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25614 + * GNU General Public License for more details.
25615 + *
25616 + * You should have received a copy of the GNU General Public License
25617 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25618 + */
25619 +
25620 +/*
25621 + * support for loopback mount as a branch
25622 + */
25623 +
25624 +#ifndef __AUFS_LOOP_H__
25625 +#define __AUFS_LOOP_H__
25626 +
25627 +#ifdef __KERNEL__
25628 +
25629 +struct dentry;
25630 +struct super_block;
25631 +
25632 +#ifdef CONFIG_AUFS_BDEV_LOOP
25633 +/* drivers/block/loop.c */
25634 +struct file *loop_backing_file(struct super_block *sb);
25635 +
25636 +/* loop.c */
25637 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
25638 +int au_test_loopback_kthread(void);
25639 +void au_warn_loopback(struct super_block *h_sb);
25640 +
25641 +int au_loopback_init(void);
25642 +void au_loopback_fin(void);
25643 +#else
25644 +AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
25645 +
25646 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
25647 +          struct dentry *h_adding)
25648 +AuStubInt0(au_test_loopback_kthread, void)
25649 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
25650 +
25651 +AuStubInt0(au_loopback_init, void)
25652 +AuStubVoid(au_loopback_fin, void)
25653 +#endif /* BLK_DEV_LOOP */
25654 +
25655 +#endif /* __KERNEL__ */
25656 +#endif /* __AUFS_LOOP_H__ */
25657 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
25658 --- /usr/share/empty/fs/aufs/magic.mk   1970-01-01 01:00:00.000000000 +0100
25659 +++ linux/fs/aufs/magic.mk      2022-11-05 23:02:18.965889284 +0100
25660 @@ -0,0 +1,31 @@
25661 +# SPDX-License-Identifier: GPL-2.0
25662 +
25663 +# defined in ${srctree}/fs/fuse/inode.c
25664 +# tristate
25665 +ifdef CONFIG_FUSE_FS
25666 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
25667 +endif
25668 +
25669 +# defined in ${srctree}/fs/xfs/xfs_sb.h
25670 +# tristate
25671 +ifdef CONFIG_XFS_FS
25672 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
25673 +endif
25674 +
25675 +# defined in ${srctree}/fs/configfs/mount.c
25676 +# tristate
25677 +ifdef CONFIG_CONFIGFS_FS
25678 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
25679 +endif
25680 +
25681 +# defined in ${srctree}/fs/ubifs/ubifs.h
25682 +# tristate
25683 +ifdef CONFIG_UBIFS_FS
25684 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
25685 +endif
25686 +
25687 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
25688 +# tristate
25689 +ifdef CONFIG_HFSPLUS_FS
25690 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
25691 +endif
25692 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
25693 --- /usr/share/empty/fs/aufs/Makefile   1970-01-01 01:00:00.000000000 +0100
25694 +++ linux/fs/aufs/Makefile      2022-11-05 23:02:18.959222617 +0100
25695 @@ -0,0 +1,46 @@
25696 +# SPDX-License-Identifier: GPL-2.0
25697 +
25698 +include ${src}/magic.mk
25699 +ifeq (${CONFIG_AUFS_FS},m)
25700 +include ${src}/conf.mk
25701 +endif
25702 +-include ${src}/priv_def.mk
25703 +
25704 +# cf. include/linux/kernel.h
25705 +# enable pr_debug
25706 +ccflags-y += -DDEBUG
25707 +# sparse requires the full pathname
25708 +ifdef M
25709 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
25710 +else
25711 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
25712 +endif
25713 +
25714 +obj-$(CONFIG_AUFS_FS) += aufs.o
25715 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o fsctx.o \
25716 +       wkq.o vfsub.o dcsub.o \
25717 +       cpup.o whout.o wbr_policy.o \
25718 +       dinfo.o dentry.o \
25719 +       dynop.o \
25720 +       finfo.o file.o f_op.o \
25721 +       dir.o vdir.o \
25722 +       iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
25723 +       mvdown.o ioctl.o
25724 +
25725 +# all are boolean
25726 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
25727 +aufs-$(CONFIG_SYSFS) += sysfs.o
25728 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
25729 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
25730 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
25731 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
25732 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
25733 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
25734 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
25735 +aufs-$(CONFIG_AUFS_DIRREN) += dirren.o
25736 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
25737 +aufs-$(CONFIG_AUFS_POLL) += poll.o
25738 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
25739 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
25740 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
25741 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
25742 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
25743 --- /usr/share/empty/fs/aufs/module.c   1970-01-01 01:00:00.000000000 +0100
25744 +++ linux/fs/aufs/module.c      2022-11-05 23:02:18.965889284 +0100
25745 @@ -0,0 +1,273 @@
25746 +// SPDX-License-Identifier: GPL-2.0
25747 +/*
25748 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25749 + *
25750 + * This program is free software; you can redistribute it and/or modify
25751 + * it under the terms of the GNU General Public License as published by
25752 + * the Free Software Foundation; either version 2 of the License, or
25753 + * (at your option) any later version.
25754 + *
25755 + * This program is distributed in the hope that it will be useful,
25756 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25757 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25758 + * GNU General Public License for more details.
25759 + *
25760 + * You should have received a copy of the GNU General Public License
25761 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25762 + */
25763 +
25764 +/*
25765 + * module global variables and operations
25766 + */
25767 +
25768 +#include <linux/module.h>
25769 +#include <linux/seq_file.h>
25770 +#include "aufs.h"
25771 +
25772 +/* shrinkable realloc */
25773 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink)
25774 +{
25775 +       size_t sz;
25776 +       int diff;
25777 +
25778 +       sz = 0;
25779 +       diff = -1;
25780 +       if (p) {
25781 +#if 0 /* unused */
25782 +               if (!new_sz) {
25783 +                       au_kfree_rcu(p);
25784 +                       p = NULL;
25785 +                       goto out;
25786 +               }
25787 +#else
25788 +               AuDebugOn(!new_sz);
25789 +#endif
25790 +               sz = ksize(p);
25791 +               diff = au_kmidx_sub(sz, new_sz);
25792 +       }
25793 +       if (sz && !diff)
25794 +               goto out;
25795 +
25796 +       if (sz < new_sz)
25797 +               /* expand or SLOB */
25798 +               p = krealloc(p, new_sz, gfp);
25799 +       else if (new_sz < sz && may_shrink) {
25800 +               /* shrink */
25801 +               void *q;
25802 +
25803 +               q = kmalloc(new_sz, gfp);
25804 +               if (q) {
25805 +                       if (p) {
25806 +                               memcpy(q, p, new_sz);
25807 +                               au_kfree_try_rcu(p);
25808 +                       }
25809 +                       p = q;
25810 +               } else
25811 +                       p = NULL;
25812 +       }
25813 +
25814 +out:
25815 +       return p;
25816 +}
25817 +
25818 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
25819 +                  int may_shrink)
25820 +{
25821 +       p = au_krealloc(p, new_sz, gfp, may_shrink);
25822 +       if (p && new_sz > nused)
25823 +               memset(p + nused, 0, new_sz - nused);
25824 +       return p;
25825 +}
25826 +
25827 +/* ---------------------------------------------------------------------- */
25828 +/*
25829 + * aufs caches
25830 + */
25831 +struct kmem_cache *au_cache[AuCache_Last];
25832 +
25833 +static void au_cache_fin(void)
25834 +{
25835 +       int i;
25836 +
25837 +       /*
25838 +        * Make sure all delayed rcu free inodes are flushed before we
25839 +        * destroy cache.
25840 +        */
25841 +       rcu_barrier();
25842 +
25843 +       /* excluding AuCache_HNOTIFY */
25844 +       BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
25845 +       for (i = 0; i < AuCache_HNOTIFY; i++) {
25846 +               kmem_cache_destroy(au_cache[i]);
25847 +               au_cache[i] = NULL;
25848 +       }
25849 +}
25850 +
25851 +static int __init au_cache_init(void)
25852 +{
25853 +       au_cache[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
25854 +       if (au_cache[AuCache_DINFO])
25855 +               /* SLAB_DESTROY_BY_RCU */
25856 +               au_cache[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
25857 +                                                      au_icntnr_init_once);
25858 +       if (au_cache[AuCache_ICNTNR])
25859 +               au_cache[AuCache_FINFO] = AuCacheCtor(au_finfo,
25860 +                                                     au_fi_init_once);
25861 +       if (au_cache[AuCache_FINFO])
25862 +               au_cache[AuCache_VDIR] = AuCache(au_vdir);
25863 +       if (au_cache[AuCache_VDIR])
25864 +               au_cache[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
25865 +       if (au_cache[AuCache_DEHSTR])
25866 +               return 0;
25867 +
25868 +       au_cache_fin();
25869 +       return -ENOMEM;
25870 +}
25871 +
25872 +/* ---------------------------------------------------------------------- */
25873 +
25874 +int au_dir_roflags;
25875 +
25876 +#ifdef CONFIG_AUFS_SBILIST
25877 +/*
25878 + * iterate_supers_type() doesn't protect us from
25879 + * remounting (branch management)
25880 + */
25881 +struct hlist_bl_head au_sbilist;
25882 +#endif
25883 +
25884 +/*
25885 + * functions for module interface.
25886 + */
25887 +MODULE_LICENSE("GPL");
25888 +/* MODULE_LICENSE("GPL v2"); */
25889 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
25890 +MODULE_DESCRIPTION(AUFS_NAME
25891 +       " -- Advanced multi layered unification filesystem");
25892 +MODULE_VERSION(AUFS_VERSION);
25893 +MODULE_ALIAS_FS(AUFS_NAME);
25894 +
25895 +/* this module parameter has no meaning when SYSFS is disabled */
25896 +int sysaufs_brs = 1;
25897 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
25898 +module_param_named(brs, sysaufs_brs, int, 0444);
25899 +
25900 +/* this module parameter has no meaning when USER_NS is disabled */
25901 +bool au_userns;
25902 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
25903 +module_param_named(allow_userns, au_userns, bool, 0444);
25904 +
25905 +/* ---------------------------------------------------------------------- */
25906 +
25907 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
25908 +
25909 +int au_seq_path(struct seq_file *seq, struct path *path)
25910 +{
25911 +       int err;
25912 +
25913 +       err = seq_path(seq, path, au_esc_chars);
25914 +       if (err >= 0)
25915 +               err = 0;
25916 +       else
25917 +               err = -ENOMEM;
25918 +
25919 +       return err;
25920 +}
25921 +
25922 +/* ---------------------------------------------------------------------- */
25923 +
25924 +static int __init aufs_init(void)
25925 +{
25926 +       int err, i;
25927 +       char *p;
25928 +
25929 +       p = au_esc_chars;
25930 +       for (i = 1; i <= ' '; i++)
25931 +               *p++ = i;
25932 +       *p++ = '\\';
25933 +       *p++ = '\x7f';
25934 +       *p = 0;
25935 +
25936 +       au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
25937 +
25938 +       memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
25939 +       for (i = 0; i < AuIop_Last; i++)
25940 +               aufs_iop_nogetattr[i].getattr = NULL;
25941 +
25942 +       memset(au_cache, 0, sizeof(au_cache));  /* including hnotify */
25943 +
25944 +       au_sbilist_init();
25945 +       sysaufs_brs_init();
25946 +       au_debug_init();
25947 +       au_dy_init();
25948 +       err = sysaufs_init();
25949 +       if (unlikely(err))
25950 +               goto out;
25951 +       err = dbgaufs_init();
25952 +       if (unlikely(err))
25953 +               goto out_sysaufs;
25954 +       err = au_procfs_init();
25955 +       if (unlikely(err))
25956 +               goto out_dbgaufs;
25957 +       err = au_wkq_init();
25958 +       if (unlikely(err))
25959 +               goto out_procfs;
25960 +       err = au_loopback_init();
25961 +       if (unlikely(err))
25962 +               goto out_wkq;
25963 +       err = au_hnotify_init();
25964 +       if (unlikely(err))
25965 +               goto out_loopback;
25966 +       err = au_sysrq_init();
25967 +       if (unlikely(err))
25968 +               goto out_hin;
25969 +       err = au_cache_init();
25970 +       if (unlikely(err))
25971 +               goto out_sysrq;
25972 +
25973 +       aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
25974 +       err = register_filesystem(&aufs_fs_type);
25975 +       if (unlikely(err))
25976 +               goto out_cache;
25977 +
25978 +       /* since we define pr_fmt, call printk directly */
25979 +       printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
25980 +       goto out; /* success */
25981 +
25982 +out_cache:
25983 +       au_cache_fin();
25984 +out_sysrq:
25985 +       au_sysrq_fin();
25986 +out_hin:
25987 +       au_hnotify_fin();
25988 +out_loopback:
25989 +       au_loopback_fin();
25990 +out_wkq:
25991 +       au_wkq_fin();
25992 +out_procfs:
25993 +       au_procfs_fin();
25994 +out_dbgaufs:
25995 +       dbgaufs_fin();
25996 +out_sysaufs:
25997 +       sysaufs_fin();
25998 +       au_dy_fin();
25999 +out:
26000 +       return err;
26001 +}
26002 +
26003 +static void __exit aufs_exit(void)
26004 +{
26005 +       unregister_filesystem(&aufs_fs_type);
26006 +       au_cache_fin();
26007 +       au_sysrq_fin();
26008 +       au_hnotify_fin();
26009 +       au_loopback_fin();
26010 +       au_wkq_fin();
26011 +       au_procfs_fin();
26012 +       dbgaufs_fin();
26013 +       sysaufs_fin();
26014 +       au_dy_fin();
26015 +}
26016 +
26017 +module_init(aufs_init);
26018 +module_exit(aufs_exit);
26019 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
26020 --- /usr/share/empty/fs/aufs/module.h   1970-01-01 01:00:00.000000000 +0100
26021 +++ linux/fs/aufs/module.h      2022-11-05 23:02:18.969222617 +0100
26022 @@ -0,0 +1,180 @@
26023 +/* SPDX-License-Identifier: GPL-2.0 */
26024 +/*
26025 + * Copyright (C) 2005-2022 Junjiro R. Okajima
26026 + *
26027 + * This program is free software; you can redistribute it and/or modify
26028 + * it under the terms of the GNU General Public License as published by
26029 + * the Free Software Foundation; either version 2 of the License, or
26030 + * (at your option) any later version.
26031 + *
26032 + * This program is distributed in the hope that it will be useful,
26033 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26034 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
26035 + * GNU General Public License for more details.
26036 + *
26037 + * You should have received a copy of the GNU General Public License
26038 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
26039 + */
26040 +
26041 +/*
26042 + * module initialization and module-global
26043 + */
26044 +
26045 +#ifndef __AUFS_MODULE_H__
26046 +#define __AUFS_MODULE_H__
26047 +
26048 +#ifdef __KERNEL__
26049 +
26050 +#include <linux/slab.h>
26051 +#include "debug.h"
26052 +#include "dentry.h"
26053 +#include "dir.h"
26054 +#include "file.h"
26055 +#include "inode.h"
26056 +
26057 +struct path;
26058 +struct seq_file;
26059 +
26060 +/* module parameters */
26061 +extern int sysaufs_brs;
26062 +extern bool au_userns;
26063 +
26064 +/* ---------------------------------------------------------------------- */
26065 +
26066 +extern int au_dir_roflags;
26067 +
26068 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink);
26069 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
26070 +                  int may_shrink);
26071 +
26072 +/*
26073 + * Comparing the size of the object with sizeof(struct rcu_head)
26074 + * case 1: object is always larger
26075 + *     --> au_kfree_rcu() or au_kfree_do_rcu()
26076 + * case 2: object is always smaller
26077 + *     --> au_kfree_small()
26078 + * case 3: object can be any size
26079 + *     --> au_kfree_try_rcu()
26080 + */
26081 +
26082 +static inline void au_kfree_do_rcu(const void *p)
26083 +{
26084 +       struct {
26085 +               struct rcu_head rcu;
26086 +       } *a = (void *)p;
26087 +
26088 +       kfree_rcu(a, rcu);
26089 +}
26090 +
26091 +#define au_kfree_rcu(_p) do {                                          \
26092 +               typeof(_p) p = (_p);                                    \
26093 +               BUILD_BUG_ON(sizeof(*p) < sizeof(struct rcu_head));     \
26094 +               if (p)                                                  \
26095 +                       au_kfree_do_rcu(p);                             \
26096 +       } while (0)
26097 +
26098 +#define au_kfree_do_sz_test(sz)        (sz >= sizeof(struct rcu_head))
26099 +#define au_kfree_sz_test(p)    (p && au_kfree_do_sz_test(ksize(p)))
26100 +
26101 +static inline void au_kfree_try_rcu(const void *p)
26102 +{
26103 +       if (!p)
26104 +               return;
26105 +       if (au_kfree_sz_test(p))
26106 +               au_kfree_do_rcu(p);
26107 +       else
26108 +               kfree(p);
26109 +}
26110 +
26111 +static inline void au_kfree_small(const void *p)
26112 +{
26113 +       if (!p)
26114 +               return;
26115 +       AuDebugOn(au_kfree_sz_test(p));
26116 +       kfree(p);
26117 +}
26118 +
26119 +static inline int au_kmidx_sub(size_t sz, size_t new_sz)
26120 +{
26121 +#ifndef CONFIG_SLOB
26122 +       return __kmalloc_index(sz, false) - __kmalloc_index(new_sz, false);
26123 +#else
26124 +       return -1; /* SLOB is untested */
26125 +#endif
26126 +}
26127 +
26128 +int au_seq_path(struct seq_file *seq, struct path *path);
26129 +
26130 +#ifdef CONFIG_PROC_FS
26131 +/* procfs.c */
26132 +int __init au_procfs_init(void);
26133 +void au_procfs_fin(void);
26134 +#else
26135 +AuStubInt0(au_procfs_init, void);
26136 +AuStubVoid(au_procfs_fin, void);
26137 +#endif
26138 +
26139 +/* ---------------------------------------------------------------------- */
26140 +
26141 +/* kmem cache */
26142 +enum {
26143 +       AuCache_DINFO,
26144 +       AuCache_ICNTNR,
26145 +       AuCache_FINFO,
26146 +       AuCache_VDIR,
26147 +       AuCache_DEHSTR,
26148 +       AuCache_HNOTIFY, /* must be last */
26149 +       AuCache_Last
26150 +};
26151 +
26152 +extern struct kmem_cache *au_cache[AuCache_Last];
26153 +
26154 +#define AuCacheFlags           (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
26155 +#define AuCache(type)          KMEM_CACHE(type, AuCacheFlags)
26156 +#define AuCacheCtor(type, ctor)        \
26157 +       kmem_cache_create(#type, sizeof(struct type), \
26158 +                         __alignof__(struct type), AuCacheFlags, ctor)
26159 +
26160 +#define AuCacheFuncAlloc(name, index)                                  \
26161 +       static inline struct au_##name *au_cache_alloc_##name(void)     \
26162 +       { return kmem_cache_alloc(au_cache[AuCache_##index], GFP_NOFS); }
26163 +
26164 +#define AuCacheFuncs(name, index)                                      \
26165 +       static inline void au_cache_free_##name##_norcu(struct au_##name *p) \
26166 +       { kmem_cache_free(au_cache[AuCache_##index], p); }              \
26167 +                                                                       \
26168 +       static inline void au_cache_free_##name##_rcu_cb(struct rcu_head *rcu) \
26169 +       { void *p = rcu;                                                \
26170 +               p -= offsetof(struct au_##name, rcu);                   \
26171 +               kmem_cache_free(au_cache[AuCache_##index], p); }        \
26172 +       static inline void au_cache_free_##name##_rcu(struct au_##name *p) \
26173 +       { BUILD_BUG_ON(sizeof(struct au_##name) < sizeof(struct rcu_head)); \
26174 +               call_rcu(&p->rcu, au_cache_free_##name##_rcu_cb); }     \
26175 +                                                                       \
26176 +       static inline void au_cache_free_##name(struct au_##name *p)    \
26177 +       { /* au_cache_free_##name##_norcu(p); */                        \
26178 +               au_cache_free_##name##_rcu(p); }
26179 +
26180 +AuCacheFuncs(dinfo, DINFO);
26181 +AuCacheFuncAlloc(dinfo, DINFO);
26182 +
26183 +AuCacheFuncs(icntnr, ICNTNR);
26184 +static inline struct au_icntnr *au_cache_alloc_icntnr(struct super_block *sb)
26185 +{ return alloc_inode_sb(sb, au_cache[AuCache_ICNTNR], GFP_NOFS); }
26186 +
26187 +AuCacheFuncs(finfo, FINFO);
26188 +AuCacheFuncAlloc(finfo, FINFO);
26189 +
26190 +AuCacheFuncs(vdir, VDIR);
26191 +AuCacheFuncAlloc(vdir, VDIR);
26192 +
26193 +AuCacheFuncs(vdir_dehstr, DEHSTR);
26194 +AuCacheFuncAlloc(vdir_dehstr, DEHSTR);
26195 +
26196 +#ifdef CONFIG_AUFS_HNOTIFY
26197 +AuCacheFuncs(hnotify, HNOTIFY);
26198 +AuCacheFuncAlloc(hnotify, HNOTIFY);
26199 +#endif
26200 +
26201 +#endif /* __KERNEL__ */
26202 +#endif /* __AUFS_MODULE_H__ */
26203 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
26204 --- /usr/share/empty/fs/aufs/mvdown.c   1970-01-01 01:00:00.000000000 +0100
26205 +++ linux/fs/aufs/mvdown.c      2022-11-05 23:02:18.969222617 +0100
26206 @@ -0,0 +1,706 @@
26207 +// SPDX-License-Identifier: GPL-2.0
26208 +/*
26209 + * Copyright (C) 2011-2022 Junjiro R. Okajima
26210 + *
26211 + * This program is free software; you can redistribute it and/or modify
26212 + * it under the terms of the GNU General Public License as published by
26213 + * the Free Software Foundation; either version 2 of the License, or
26214 + * (at your option) any later version.
26215 + *
26216 + * This program is distributed in the hope that it will be useful,
26217 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26218 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
26219 + * GNU General Public License for more details.
26220 + *
26221 + * You should have received a copy of the GNU General Public License
26222 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
26223 + */
26224 +
26225 +/*
26226 + * move-down, opposite of copy-up
26227 + */
26228 +
26229 +#include "aufs.h"
26230 +
26231 +struct au_mvd_args {
26232 +       struct {
26233 +               struct super_block *h_sb;
26234 +               struct dentry *h_parent;
26235 +               struct au_hinode *hdir;
26236 +               struct inode *h_dir, *h_inode;
26237 +               struct au_pin pin;
26238 +       } info[AUFS_MVDOWN_NARRAY];
26239 +
26240 +       struct aufs_mvdown mvdown;
26241 +       struct dentry *dentry, *parent;
26242 +       struct inode *inode, *dir;
26243 +       struct super_block *sb;
26244 +       aufs_bindex_t bopq, bwh, bfound;
26245 +       unsigned char rename_lock;
26246 +};
26247 +
26248 +#define mvd_errno              mvdown.au_errno
26249 +#define mvd_bsrc               mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
26250 +#define mvd_src_brid           mvdown.stbr[AUFS_MVDOWN_UPPER].brid
26251 +#define mvd_bdst               mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
26252 +#define mvd_dst_brid           mvdown.stbr[AUFS_MVDOWN_LOWER].brid
26253 +
26254 +#define mvd_h_src_sb           info[AUFS_MVDOWN_UPPER].h_sb
26255 +#define mvd_h_src_parent       info[AUFS_MVDOWN_UPPER].h_parent
26256 +#define mvd_hdir_src           info[AUFS_MVDOWN_UPPER].hdir
26257 +#define mvd_h_src_dir          info[AUFS_MVDOWN_UPPER].h_dir
26258 +#define mvd_h_src_inode                info[AUFS_MVDOWN_UPPER].h_inode
26259 +#define mvd_pin_src            info[AUFS_MVDOWN_UPPER].pin
26260 +
26261 +#define mvd_h_dst_sb           info[AUFS_MVDOWN_LOWER].h_sb
26262 +#define mvd_h_dst_parent       info[AUFS_MVDOWN_LOWER].h_parent
26263 +#define mvd_hdir_dst           info[AUFS_MVDOWN_LOWER].hdir
26264 +#define mvd_h_dst_dir          info[AUFS_MVDOWN_LOWER].h_dir
26265 +#define mvd_h_dst_inode                info[AUFS_MVDOWN_LOWER].h_inode
26266 +#define mvd_pin_dst            info[AUFS_MVDOWN_LOWER].pin
26267 +
26268 +#define AU_MVD_PR(flag, ...) do {                      \
26269 +               if (flag)                               \
26270 +                       pr_err(__VA_ARGS__);            \
26271 +       } while (0)
26272 +
26273 +static int find_lower_writable(struct au_mvd_args *a)
26274 +{
26275 +       struct super_block *sb;
26276 +       aufs_bindex_t bindex, bbot;
26277 +       struct au_branch *br;
26278 +
26279 +       sb = a->sb;
26280 +       bindex = a->mvd_bsrc;
26281 +       bbot = au_sbbot(sb);
26282 +       if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
26283 +               for (bindex++; bindex <= bbot; bindex++) {
26284 +                       br = au_sbr(sb, bindex);
26285 +                       if (au_br_fhsm(br->br_perm)
26286 +                           && !sb_rdonly(au_br_sb(br)))
26287 +                               return bindex;
26288 +               }
26289 +       else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
26290 +               for (bindex++; bindex <= bbot; bindex++) {
26291 +                       br = au_sbr(sb, bindex);
26292 +                       if (!au_br_rdonly(br))
26293 +                               return bindex;
26294 +               }
26295 +       else
26296 +               for (bindex++; bindex <= bbot; bindex++) {
26297 +                       br = au_sbr(sb, bindex);
26298 +                       if (!sb_rdonly(au_br_sb(br))) {
26299 +                               if (au_br_rdonly(br))
26300 +                                       a->mvdown.flags
26301 +                                               |= AUFS_MVDOWN_ROLOWER_R;
26302 +                               return bindex;
26303 +                       }
26304 +               }
26305 +
26306 +       return -1;
26307 +}
26308 +
26309 +/* make the parent dir on bdst */
26310 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
26311 +{
26312 +       int err;
26313 +
26314 +       err = 0;
26315 +       a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
26316 +       a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
26317 +       a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
26318 +       a->mvd_h_dst_parent = NULL;
26319 +       if (au_dbbot(a->parent) >= a->mvd_bdst)
26320 +               a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26321 +       if (!a->mvd_h_dst_parent) {
26322 +               err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
26323 +               if (unlikely(err)) {
26324 +                       AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
26325 +                       goto out;
26326 +               }
26327 +               a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26328 +       }
26329 +
26330 +out:
26331 +       AuTraceErr(err);
26332 +       return err;
26333 +}
26334 +
26335 +/* lock them all */
26336 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
26337 +{
26338 +       int err;
26339 +       struct dentry *h_trap;
26340 +
26341 +       a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
26342 +       a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
26343 +       err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
26344 +                    au_opt_udba(a->sb),
26345 +                    AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26346 +       AuTraceErr(err);
26347 +       if (unlikely(err)) {
26348 +               AU_MVD_PR(dmsg, "pin_dst failed\n");
26349 +               goto out;
26350 +       }
26351 +
26352 +       if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
26353 +               a->rename_lock = 0;
26354 +               au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26355 +                           AuLsc_DI_PARENT, AuLsc_I_PARENT3,
26356 +                           au_opt_udba(a->sb),
26357 +                           AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26358 +               err = au_do_pin(&a->mvd_pin_src);
26359 +               AuTraceErr(err);
26360 +               a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26361 +               if (unlikely(err)) {
26362 +                       AU_MVD_PR(dmsg, "pin_src failed\n");
26363 +                       goto out_dst;
26364 +               }
26365 +               goto out; /* success */
26366 +       }
26367 +
26368 +       a->rename_lock = 1;
26369 +       au_pin_hdir_unlock(&a->mvd_pin_dst);
26370 +       err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26371 +                    au_opt_udba(a->sb),
26372 +                    AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26373 +       AuTraceErr(err);
26374 +       a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26375 +       if (unlikely(err)) {
26376 +               AU_MVD_PR(dmsg, "pin_src failed\n");
26377 +               au_pin_hdir_lock(&a->mvd_pin_dst);
26378 +               goto out_dst;
26379 +       }
26380 +       au_pin_hdir_unlock(&a->mvd_pin_src);
26381 +       h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26382 +                                  a->mvd_h_dst_parent, a->mvd_hdir_dst);
26383 +       if (h_trap) {
26384 +               err = (h_trap != a->mvd_h_src_parent);
26385 +               if (err)
26386 +                       err = (h_trap != a->mvd_h_dst_parent);
26387 +       }
26388 +       BUG_ON(err); /* it should never happen */
26389 +       if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
26390 +               err = -EBUSY;
26391 +               AuTraceErr(err);
26392 +               vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26393 +                                   a->mvd_h_dst_parent, a->mvd_hdir_dst);
26394 +               au_pin_hdir_lock(&a->mvd_pin_src);
26395 +               au_unpin(&a->mvd_pin_src);
26396 +               au_pin_hdir_lock(&a->mvd_pin_dst);
26397 +               goto out_dst;
26398 +       }
26399 +       goto out; /* success */
26400 +
26401 +out_dst:
26402 +       au_unpin(&a->mvd_pin_dst);
26403 +out:
26404 +       AuTraceErr(err);
26405 +       return err;
26406 +}
26407 +
26408 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
26409 +{
26410 +       if (!a->rename_lock)
26411 +               au_unpin(&a->mvd_pin_src);
26412 +       else {
26413 +               vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26414 +                                   a->mvd_h_dst_parent, a->mvd_hdir_dst);
26415 +               au_pin_hdir_lock(&a->mvd_pin_src);
26416 +               au_unpin(&a->mvd_pin_src);
26417 +               au_pin_hdir_lock(&a->mvd_pin_dst);
26418 +       }
26419 +       au_unpin(&a->mvd_pin_dst);
26420 +}
26421 +
26422 +/* copy-down the file */
26423 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
26424 +{
26425 +       int err;
26426 +       struct au_cp_generic cpg = {
26427 +               .dentry = a->dentry,
26428 +               .bdst   = a->mvd_bdst,
26429 +               .bsrc   = a->mvd_bsrc,
26430 +               .len    = -1,
26431 +               .pin    = &a->mvd_pin_dst,
26432 +               .flags  = AuCpup_DTIME | AuCpup_HOPEN
26433 +       };
26434 +
26435 +       AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
26436 +       if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26437 +               au_fset_cpup(cpg.flags, OVERWRITE);
26438 +       if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
26439 +               au_fset_cpup(cpg.flags, RWDST);
26440 +       err = au_sio_cpdown_simple(&cpg);
26441 +       if (unlikely(err))
26442 +               AU_MVD_PR(dmsg, "cpdown failed\n");
26443 +
26444 +       AuTraceErr(err);
26445 +       return err;
26446 +}
26447 +
26448 +/*
26449 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
26450 + * were sleeping
26451 + */
26452 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
26453 +{
26454 +       int err;
26455 +       struct path h_path;
26456 +       struct au_branch *br;
26457 +       struct inode *delegated;
26458 +
26459 +       br = au_sbr(a->sb, a->mvd_bdst);
26460 +       h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
26461 +       err = PTR_ERR(h_path.dentry);
26462 +       if (IS_ERR(h_path.dentry)) {
26463 +               AU_MVD_PR(dmsg, "wh_lkup failed\n");
26464 +               goto out;
26465 +       }
26466 +
26467 +       err = 0;
26468 +       if (d_is_positive(h_path.dentry)) {
26469 +               h_path.mnt = au_br_mnt(br);
26470 +               delegated = NULL;
26471 +               err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
26472 +                                  &delegated, /*force*/0);
26473 +               if (unlikely(err == -EWOULDBLOCK)) {
26474 +                       pr_warn("cannot retry for NFSv4 delegation"
26475 +                               " for an internal unlink\n");
26476 +                       iput(delegated);
26477 +               }
26478 +               if (unlikely(err))
26479 +                       AU_MVD_PR(dmsg, "wh_unlink failed\n");
26480 +       }
26481 +       dput(h_path.dentry);
26482 +
26483 +out:
26484 +       AuTraceErr(err);
26485 +       return err;
26486 +}
26487 +
26488 +/*
26489 + * unlink the topmost h_dentry
26490 + */
26491 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
26492 +{
26493 +       int err;
26494 +       struct path h_path;
26495 +       struct inode *delegated;
26496 +
26497 +       h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
26498 +       h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
26499 +       delegated = NULL;
26500 +       err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
26501 +       if (unlikely(err == -EWOULDBLOCK)) {
26502 +               pr_warn("cannot retry for NFSv4 delegation"
26503 +                       " for an internal unlink\n");
26504 +               iput(delegated);
26505 +       }
26506 +       if (unlikely(err))
26507 +               AU_MVD_PR(dmsg, "unlink failed\n");
26508 +
26509 +       AuTraceErr(err);
26510 +       return err;
26511 +}
26512 +
26513 +/* Since mvdown succeeded, we ignore an error of this function */
26514 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
26515 +{
26516 +       int err;
26517 +       struct au_branch *br;
26518 +
26519 +       a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
26520 +       br = au_sbr(a->sb, a->mvd_bsrc);
26521 +       err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
26522 +       if (!err) {
26523 +               br = au_sbr(a->sb, a->mvd_bdst);
26524 +               a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
26525 +               err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
26526 +       }
26527 +       if (!err)
26528 +               a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
26529 +       else
26530 +               AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
26531 +}
26532 +
26533 +/*
26534 + * copy-down the file and unlink the bsrc file.
26535 + * - unlink the bdst whout if exist
26536 + * - copy-down the file (with whtmp name and rename)
26537 + * - unlink the bsrc file
26538 + */
26539 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
26540 +{
26541 +       int err;
26542 +
26543 +       err = au_do_mkdir(dmsg, a);
26544 +       if (!err)
26545 +               err = au_do_lock(dmsg, a);
26546 +       if (unlikely(err))
26547 +               goto out;
26548 +
26549 +       /*
26550 +        * do not revert the activities we made on bdst since they should be
26551 +        * harmless in aufs.
26552 +        */
26553 +
26554 +       err = au_do_cpdown(dmsg, a);
26555 +       if (!err)
26556 +               err = au_do_unlink_wh(dmsg, a);
26557 +       if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
26558 +               err = au_do_unlink(dmsg, a);
26559 +       if (unlikely(err))
26560 +               goto out_unlock;
26561 +
26562 +       AuDbg("%pd2, 0x%x, %d --> %d\n",
26563 +             a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
26564 +       if (find_lower_writable(a) < 0)
26565 +               a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
26566 +
26567 +       if (a->mvdown.flags & AUFS_MVDOWN_STFS)
26568 +               au_do_stfs(dmsg, a);
26569 +
26570 +       /* maintain internal array */
26571 +       if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
26572 +               au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
26573 +               au_set_dbtop(a->dentry, a->mvd_bdst);
26574 +               au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
26575 +               au_set_ibtop(a->inode, a->mvd_bdst);
26576 +       } else {
26577 +               /* hide the lower */
26578 +               au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
26579 +               au_set_dbbot(a->dentry, a->mvd_bsrc);
26580 +               au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
26581 +               au_set_ibbot(a->inode, a->mvd_bsrc);
26582 +       }
26583 +       if (au_dbbot(a->dentry) < a->mvd_bdst)
26584 +               au_set_dbbot(a->dentry, a->mvd_bdst);
26585 +       if (au_ibbot(a->inode) < a->mvd_bdst)
26586 +               au_set_ibbot(a->inode, a->mvd_bdst);
26587 +
26588 +out_unlock:
26589 +       au_do_unlock(dmsg, a);
26590 +out:
26591 +       AuTraceErr(err);
26592 +       return err;
26593 +}
26594 +
26595 +/* ---------------------------------------------------------------------- */
26596 +
26597 +/* make sure the file is idle */
26598 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
26599 +{
26600 +       int err, plinked;
26601 +
26602 +       err = 0;
26603 +       plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
26604 +       if (au_dbtop(a->dentry) == a->mvd_bsrc
26605 +           && au_dcount(a->dentry) == 1
26606 +           && atomic_read(&a->inode->i_count) == 1
26607 +           /* && a->mvd_h_src_inode->i_nlink == 1 */
26608 +           && (!plinked || !au_plink_test(a->inode))
26609 +           && a->inode->i_nlink == 1)
26610 +               goto out;
26611 +
26612 +       err = -EBUSY;
26613 +       AU_MVD_PR(dmsg,
26614 +                 "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
26615 +                 a->mvd_bsrc, au_dbtop(a->dentry), au_dcount(a->dentry),
26616 +                 atomic_read(&a->inode->i_count), a->inode->i_nlink,
26617 +                 a->mvd_h_src_inode->i_nlink,
26618 +                 plinked, plinked ? au_plink_test(a->inode) : 0);
26619 +
26620 +out:
26621 +       AuTraceErr(err);
26622 +       return err;
26623 +}
26624 +
26625 +/* make sure the parent dir is fine */
26626 +static int au_mvd_args_parent(const unsigned char dmsg,
26627 +                             struct au_mvd_args *a)
26628 +{
26629 +       int err;
26630 +       aufs_bindex_t bindex;
26631 +
26632 +       err = 0;
26633 +       if (unlikely(au_alive_dir(a->parent))) {
26634 +               err = -ENOENT;
26635 +               AU_MVD_PR(dmsg, "parent dir is dead\n");
26636 +               goto out;
26637 +       }
26638 +
26639 +       a->bopq = au_dbdiropq(a->parent);
26640 +       bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
26641 +       AuDbg("b%d\n", bindex);
26642 +       if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
26643 +                    || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
26644 +               err = -EINVAL;
26645 +               a->mvd_errno = EAU_MVDOWN_OPAQUE;
26646 +               AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
26647 +                         a->bopq, a->mvd_bdst);
26648 +       }
26649 +
26650 +out:
26651 +       AuTraceErr(err);
26652 +       return err;
26653 +}
26654 +
26655 +static int au_mvd_args_intermediate(const unsigned char dmsg,
26656 +                                   struct au_mvd_args *a)
26657 +{
26658 +       int err;
26659 +       struct au_dinfo *dinfo, *tmp;
26660 +
26661 +       /* lookup the next lower positive entry */
26662 +       err = -ENOMEM;
26663 +       tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
26664 +       if (unlikely(!tmp))
26665 +               goto out;
26666 +
26667 +       a->bfound = -1;
26668 +       a->bwh = -1;
26669 +       dinfo = au_di(a->dentry);
26670 +       au_di_cp(tmp, dinfo);
26671 +       au_di_swap(tmp, dinfo);
26672 +
26673 +       /* returns the number of positive dentries */
26674 +       err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1,
26675 +                            /* AuLkup_IGNORE_PERM */ 0);
26676 +       if (!err)
26677 +               a->bwh = au_dbwh(a->dentry);
26678 +       else if (err > 0)
26679 +               a->bfound = au_dbtop(a->dentry);
26680 +
26681 +       au_di_swap(tmp, dinfo);
26682 +       au_rw_write_unlock(&tmp->di_rwsem);
26683 +       au_di_free(tmp);
26684 +       if (unlikely(err < 0))
26685 +               AU_MVD_PR(dmsg, "failed look-up lower\n");
26686 +
26687 +       /*
26688 +        * here, we have these cases.
26689 +        * bfound == -1
26690 +        *      no positive dentry under bsrc. there are more sub-cases.
26691 +        *      bwh < 0
26692 +        *              there no whiteout, we can safely move-down.
26693 +        *      bwh <= bsrc
26694 +        *              impossible
26695 +        *      bsrc < bwh && bwh < bdst
26696 +        *              there is a whiteout on RO branch. cannot proceed.
26697 +        *      bwh == bdst
26698 +        *              there is a whiteout on the RW target branch. it should
26699 +        *              be removed.
26700 +        *      bdst < bwh
26701 +        *              there is a whiteout somewhere unrelated branch.
26702 +        * -1 < bfound && bfound <= bsrc
26703 +        *      impossible.
26704 +        * bfound < bdst
26705 +        *      found, but it is on RO branch between bsrc and bdst. cannot
26706 +        *      proceed.
26707 +        * bfound == bdst
26708 +        *      found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
26709 +        *      error.
26710 +        * bdst < bfound
26711 +        *      found, after we create the file on bdst, it will be hidden.
26712 +        */
26713 +
26714 +       AuDebugOn(a->bfound == -1
26715 +                 && a->bwh != -1
26716 +                 && a->bwh <= a->mvd_bsrc);
26717 +       AuDebugOn(-1 < a->bfound
26718 +                 && a->bfound <= a->mvd_bsrc);
26719 +
26720 +       err = -EINVAL;
26721 +       if (a->bfound == -1
26722 +           && a->mvd_bsrc < a->bwh
26723 +           && a->bwh != -1
26724 +           && a->bwh < a->mvd_bdst) {
26725 +               a->mvd_errno = EAU_MVDOWN_WHITEOUT;
26726 +               AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
26727 +                         a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
26728 +               goto out;
26729 +       } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
26730 +               a->mvd_errno = EAU_MVDOWN_UPPER;
26731 +               AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
26732 +                         a->mvd_bdst, a->bfound);
26733 +               goto out;
26734 +       }
26735 +
26736 +       err = 0; /* success */
26737 +
26738 +out:
26739 +       AuTraceErr(err);
26740 +       return err;
26741 +}
26742 +
26743 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
26744 +{
26745 +       int err;
26746 +
26747 +       err = 0;
26748 +       if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26749 +           && a->bfound == a->mvd_bdst)
26750 +               err = -EEXIST;
26751 +       AuTraceErr(err);
26752 +       return err;
26753 +}
26754 +
26755 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
26756 +{
26757 +       int err;
26758 +       struct au_branch *br;
26759 +
26760 +       err = -EISDIR;
26761 +       if (unlikely(S_ISDIR(a->inode->i_mode)))
26762 +               goto out;
26763 +
26764 +       err = -EINVAL;
26765 +       if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
26766 +               a->mvd_bsrc = au_ibtop(a->inode);
26767 +       else {
26768 +               a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
26769 +               if (unlikely(a->mvd_bsrc < 0
26770 +                            || (a->mvd_bsrc < au_dbtop(a->dentry)
26771 +                                || au_dbbot(a->dentry) < a->mvd_bsrc
26772 +                                || !au_h_dptr(a->dentry, a->mvd_bsrc))
26773 +                            || (a->mvd_bsrc < au_ibtop(a->inode)
26774 +                                || au_ibbot(a->inode) < a->mvd_bsrc
26775 +                                || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
26776 +                       a->mvd_errno = EAU_MVDOWN_NOUPPER;
26777 +                       AU_MVD_PR(dmsg, "no upper\n");
26778 +                       goto out;
26779 +               }
26780 +       }
26781 +       if (unlikely(a->mvd_bsrc == au_sbbot(a->sb))) {
26782 +               a->mvd_errno = EAU_MVDOWN_BOTTOM;
26783 +               AU_MVD_PR(dmsg, "on the bottom\n");
26784 +               goto out;
26785 +       }
26786 +       a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
26787 +       br = au_sbr(a->sb, a->mvd_bsrc);
26788 +       err = au_br_rdonly(br);
26789 +       if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
26790 +               if (unlikely(err))
26791 +                       goto out;
26792 +       } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
26793 +                    || IS_APPEND(a->mvd_h_src_inode))) {
26794 +               if (err)
26795 +                       a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
26796 +               /* go on */
26797 +       } else
26798 +               goto out;
26799 +
26800 +       err = -EINVAL;
26801 +       if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
26802 +               a->mvd_bdst = find_lower_writable(a);
26803 +               if (unlikely(a->mvd_bdst < 0)) {
26804 +                       a->mvd_errno = EAU_MVDOWN_BOTTOM;
26805 +                       AU_MVD_PR(dmsg, "no writable lower branch\n");
26806 +                       goto out;
26807 +               }
26808 +       } else {
26809 +               a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
26810 +               if (unlikely(a->mvd_bdst < 0
26811 +                            || au_sbbot(a->sb) < a->mvd_bdst)) {
26812 +                       a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
26813 +                       AU_MVD_PR(dmsg, "no lower brid\n");
26814 +                       goto out;
26815 +               }
26816 +       }
26817 +
26818 +       err = au_mvd_args_busy(dmsg, a);
26819 +       if (!err)
26820 +               err = au_mvd_args_parent(dmsg, a);
26821 +       if (!err)
26822 +               err = au_mvd_args_intermediate(dmsg, a);
26823 +       if (!err)
26824 +               err = au_mvd_args_exist(dmsg, a);
26825 +       if (!err)
26826 +               AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
26827 +
26828 +out:
26829 +       AuTraceErr(err);
26830 +       return err;
26831 +}
26832 +
26833 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
26834 +{
26835 +       int err, e;
26836 +       unsigned char dmsg;
26837 +       struct au_mvd_args *args;
26838 +       struct inode *inode;
26839 +
26840 +       inode = d_inode(dentry);
26841 +       err = -EPERM;
26842 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
26843 +               goto out;
26844 +
26845 +       err = -ENOMEM;
26846 +       args = kmalloc(sizeof(*args), GFP_NOFS);
26847 +       if (unlikely(!args))
26848 +               goto out;
26849 +
26850 +       err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
26851 +       if (!err)
26852 +               /* VERIFY_WRITE */
26853 +               err = !access_ok(uarg, sizeof(*uarg));
26854 +       if (unlikely(err)) {
26855 +               err = -EFAULT;
26856 +               AuTraceErr(err);
26857 +               goto out_free;
26858 +       }
26859 +       AuDbg("flags 0x%x\n", args->mvdown.flags);
26860 +       args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
26861 +       args->mvdown.au_errno = 0;
26862 +       args->dentry = dentry;
26863 +       args->inode = inode;
26864 +       args->sb = dentry->d_sb;
26865 +
26866 +       err = -ENOENT;
26867 +       dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
26868 +       args->parent = dget_parent(dentry);
26869 +       args->dir = d_inode(args->parent);
26870 +       inode_lock_nested(args->dir, I_MUTEX_PARENT);
26871 +       dput(args->parent);
26872 +       if (unlikely(args->parent != dentry->d_parent)) {
26873 +               AU_MVD_PR(dmsg, "parent dir is moved\n");
26874 +               goto out_dir;
26875 +       }
26876 +
26877 +       inode_lock_nested(inode, I_MUTEX_CHILD);
26878 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
26879 +       if (unlikely(err))
26880 +               goto out_inode;
26881 +
26882 +       di_write_lock_parent(args->parent);
26883 +       err = au_mvd_args(dmsg, args);
26884 +       if (unlikely(err))
26885 +               goto out_parent;
26886 +
26887 +       err = au_do_mvdown(dmsg, args);
26888 +       if (unlikely(err))
26889 +               goto out_parent;
26890 +
26891 +       au_cpup_attr_timesizes(args->dir);
26892 +       au_cpup_attr_timesizes(inode);
26893 +       if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
26894 +               au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
26895 +       /* au_digen_dec(dentry); */
26896 +
26897 +out_parent:
26898 +       di_write_unlock(args->parent);
26899 +       aufs_read_unlock(dentry, AuLock_DW);
26900 +out_inode:
26901 +       inode_unlock(inode);
26902 +out_dir:
26903 +       inode_unlock(args->dir);
26904 +out_free:
26905 +       e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
26906 +       if (unlikely(e))
26907 +               err = -EFAULT;
26908 +       au_kfree_rcu(args);
26909 +out:
26910 +       AuTraceErr(err);
26911 +       return err;
26912 +}
26913 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
26914 --- /usr/share/empty/fs/aufs/opts.c     1970-01-01 01:00:00.000000000 +0100
26915 +++ linux/fs/aufs/opts.c        2022-11-05 23:02:18.969222617 +0100
26916 @@ -0,0 +1,1032 @@
26917 +// SPDX-License-Identifier: GPL-2.0
26918 +/*
26919 + * Copyright (C) 2005-2022 Junjiro R. Okajima
26920 + *
26921 + * This program is free software; you can redistribute it and/or modify
26922 + * it under the terms of the GNU General Public License as published by
26923 + * the Free Software Foundation; either version 2 of the License, or
26924 + * (at your option) any later version.
26925 + *
26926 + * This program is distributed in the hope that it will be useful,
26927 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26928 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
26929 + * GNU General Public License for more details.
26930 + *
26931 + * You should have received a copy of the GNU General Public License
26932 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
26933 + */
26934 +
26935 +/*
26936 + * mount options/flags
26937 + */
26938 +
26939 +#include <linux/types.h> /* a distribution requires */
26940 +#include <linux/parser.h>
26941 +#include "aufs.h"
26942 +
26943 +/* ---------------------------------------------------------------------- */
26944 +
26945 +static const char *au_parser_pattern(int val, match_table_t tbl)
26946 +{
26947 +       struct match_token *p;
26948 +
26949 +       p = tbl;
26950 +       while (p->pattern) {
26951 +               if (p->token == val)
26952 +                       return p->pattern;
26953 +               p++;
26954 +       }
26955 +       BUG();
26956 +       return "??";
26957 +}
26958 +
26959 +static const char *au_optstr(int *val, match_table_t tbl)
26960 +{
26961 +       struct match_token *p;
26962 +       int v;
26963 +
26964 +       v = *val;
26965 +       if (!v)
26966 +               goto out;
26967 +       p = tbl;
26968 +       while (p->pattern) {
26969 +               if (p->token
26970 +                   && (v & p->token) == p->token) {
26971 +                       *val &= ~p->token;
26972 +                       return p->pattern;
26973 +               }
26974 +               p++;
26975 +       }
26976 +
26977 +out:
26978 +       return NULL;
26979 +}
26980 +
26981 +/* ---------------------------------------------------------------------- */
26982 +
26983 +static match_table_t brperm = {
26984 +       {AuBrPerm_RO, AUFS_BRPERM_RO},
26985 +       {AuBrPerm_RR, AUFS_BRPERM_RR},
26986 +       {AuBrPerm_RW, AUFS_BRPERM_RW},
26987 +       {0, NULL}
26988 +};
26989 +
26990 +static match_table_t brattr = {
26991 +       /* general */
26992 +       {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
26993 +       {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
26994 +       /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
26995 +       {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
26996 +#ifdef CONFIG_AUFS_FHSM
26997 +       {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
26998 +#endif
26999 +#ifdef CONFIG_AUFS_XATTR
27000 +       {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
27001 +       {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
27002 +       {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
27003 +       {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
27004 +       {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
27005 +       {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
27006 +#endif
27007 +
27008 +       /* ro/rr branch */
27009 +       {AuBrRAttr_WH, AUFS_BRRATTR_WH},
27010 +
27011 +       /* rw branch */
27012 +       {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
27013 +       {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
27014 +
27015 +       {0, NULL}
27016 +};
27017 +
27018 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
27019 +{
27020 +       int attr, v;
27021 +       char *p;
27022 +
27023 +       attr = 0;
27024 +       do {
27025 +               p = strchr(str, '+');
27026 +               if (p)
27027 +                       *p = 0;
27028 +               v = match_token(str, table, args);
27029 +               if (v) {
27030 +                       if (v & AuBrAttr_CMOO_Mask)
27031 +                               attr &= ~AuBrAttr_CMOO_Mask;
27032 +                       attr |= v;
27033 +               } else {
27034 +                       if (p)
27035 +                               *p = '+';
27036 +                       pr_warn("ignored branch attribute %s\n", str);
27037 +                       break;
27038 +               }
27039 +               if (p)
27040 +                       str = p + 1;
27041 +       } while (p);
27042 +
27043 +       return attr;
27044 +}
27045 +
27046 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
27047 +{
27048 +       int sz;
27049 +       const char *p;
27050 +       char *q;
27051 +
27052 +       q = str->a;
27053 +       *q = 0;
27054 +       p = au_optstr(&perm, brattr);
27055 +       if (p) {
27056 +               sz = strlen(p);
27057 +               memcpy(q, p, sz + 1);
27058 +               q += sz;
27059 +       } else
27060 +               goto out;
27061 +
27062 +       do {
27063 +               p = au_optstr(&perm, brattr);
27064 +               if (p) {
27065 +                       *q++ = '+';
27066 +                       sz = strlen(p);
27067 +                       memcpy(q, p, sz + 1);
27068 +                       q += sz;
27069 +               }
27070 +       } while (p);
27071 +
27072 +out:
27073 +       return q - str->a;
27074 +}
27075 +
27076 +int au_br_perm_val(char *perm)
27077 +{
27078 +       int val, bad, sz;
27079 +       char *p;
27080 +       substring_t args[MAX_OPT_ARGS];
27081 +       au_br_perm_str_t attr;
27082 +
27083 +       p = strchr(perm, '+');
27084 +       if (p)
27085 +               *p = 0;
27086 +       val = match_token(perm, brperm, args);
27087 +       if (!val) {
27088 +               if (p)
27089 +                       *p = '+';
27090 +               pr_warn("ignored branch permission %s\n", perm);
27091 +               val = AuBrPerm_RO;
27092 +               goto out;
27093 +       }
27094 +       if (!p)
27095 +               goto out;
27096 +
27097 +       val |= br_attr_val(p + 1, brattr, args);
27098 +
27099 +       bad = 0;
27100 +       switch (val & AuBrPerm_Mask) {
27101 +       case AuBrPerm_RO:
27102 +       case AuBrPerm_RR:
27103 +               bad = val & AuBrWAttr_Mask;
27104 +               val &= ~AuBrWAttr_Mask;
27105 +               break;
27106 +       case AuBrPerm_RW:
27107 +               bad = val & AuBrRAttr_Mask;
27108 +               val &= ~AuBrRAttr_Mask;
27109 +               break;
27110 +       }
27111 +
27112 +       /*
27113 +        * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
27114 +        * does not treat it as an error, just warning.
27115 +        * this is a tiny guard for the user operation.
27116 +        */
27117 +       if (val & AuBrAttr_UNPIN) {
27118 +               bad |= AuBrAttr_UNPIN;
27119 +               val &= ~AuBrAttr_UNPIN;
27120 +       }
27121 +
27122 +       if (unlikely(bad)) {
27123 +               sz = au_do_optstr_br_attr(&attr, bad);
27124 +               AuDebugOn(!sz);
27125 +               pr_warn("ignored branch attribute %s\n", attr.a);
27126 +       }
27127 +
27128 +out:
27129 +       return val;
27130 +}
27131 +
27132 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
27133 +{
27134 +       au_br_perm_str_t attr;
27135 +       const char *p;
27136 +       char *q;
27137 +       int sz;
27138 +
27139 +       q = str->a;
27140 +       p = au_optstr(&perm, brperm);
27141 +       AuDebugOn(!p || !*p);
27142 +       sz = strlen(p);
27143 +       memcpy(q, p, sz + 1);
27144 +       q += sz;
27145 +
27146 +       sz = au_do_optstr_br_attr(&attr, perm);
27147 +       if (sz) {
27148 +               *q++ = '+';
27149 +               memcpy(q, attr.a, sz + 1);
27150 +       }
27151 +
27152 +       AuDebugOn(strlen(str->a) >= sizeof(str->a));
27153 +}
27154 +
27155 +/* ---------------------------------------------------------------------- */
27156 +
27157 +static match_table_t udbalevel = {
27158 +       {AuOpt_UDBA_REVAL, "reval"},
27159 +       {AuOpt_UDBA_NONE, "none"},
27160 +#ifdef CONFIG_AUFS_HNOTIFY
27161 +       {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
27162 +#ifdef CONFIG_AUFS_HFSNOTIFY
27163 +       {AuOpt_UDBA_HNOTIFY, "fsnotify"},
27164 +#endif
27165 +#endif
27166 +       {-1, NULL}
27167 +};
27168 +
27169 +int au_udba_val(char *str)
27170 +{
27171 +       substring_t args[MAX_OPT_ARGS];
27172 +
27173 +       return match_token(str, udbalevel, args);
27174 +}
27175 +
27176 +const char *au_optstr_udba(int udba)
27177 +{
27178 +       return au_parser_pattern(udba, udbalevel);
27179 +}
27180 +
27181 +/* ---------------------------------------------------------------------- */
27182 +
27183 +static match_table_t au_wbr_create_policy = {
27184 +       {AuWbrCreate_TDP, "tdp"},
27185 +       {AuWbrCreate_TDP, "top-down-parent"},
27186 +       {AuWbrCreate_RR, "rr"},
27187 +       {AuWbrCreate_RR, "round-robin"},
27188 +       {AuWbrCreate_MFS, "mfs"},
27189 +       {AuWbrCreate_MFS, "most-free-space"},
27190 +       {AuWbrCreate_MFSV, "mfs:%d"},
27191 +       {AuWbrCreate_MFSV, "most-free-space:%d"},
27192 +
27193 +       /* top-down regardless the parent, and then mfs */
27194 +       {AuWbrCreate_TDMFS, "tdmfs:%d"},
27195 +       {AuWbrCreate_TDMFSV, "tdmfs:%d:%d"},
27196 +
27197 +       {AuWbrCreate_MFSRR, "mfsrr:%d"},
27198 +       {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
27199 +       {AuWbrCreate_PMFS, "pmfs"},
27200 +       {AuWbrCreate_PMFSV, "pmfs:%d"},
27201 +       {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
27202 +       {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
27203 +
27204 +       {-1, NULL}
27205 +};
27206 +
27207 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
27208 +                           struct au_opt_wbr_create *create)
27209 +{
27210 +       int err;
27211 +       unsigned long long ull;
27212 +
27213 +       err = 0;
27214 +       if (!match_u64(arg, &ull))
27215 +               create->mfsrr_watermark = ull;
27216 +       else {
27217 +               pr_err("bad integer in %s\n", str);
27218 +               err = -EINVAL;
27219 +       }
27220 +
27221 +       return err;
27222 +}
27223 +
27224 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
27225 +                         struct au_opt_wbr_create *create)
27226 +{
27227 +       int n, err;
27228 +
27229 +       err = 0;
27230 +       if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
27231 +               create->mfs_second = n;
27232 +       else {
27233 +               pr_err("bad integer in %s\n", str);
27234 +               err = -EINVAL;
27235 +       }
27236 +
27237 +       return err;
27238 +}
27239 +
27240 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
27241 +{
27242 +       int err, e;
27243 +       substring_t args[MAX_OPT_ARGS];
27244 +
27245 +       err = match_token(str, au_wbr_create_policy, args);
27246 +       create->wbr_create = err;
27247 +       switch (err) {
27248 +       case AuWbrCreate_MFSRRV:
27249 +       case AuWbrCreate_TDMFSV:
27250 +       case AuWbrCreate_PMFSRRV:
27251 +               e = au_wbr_mfs_wmark(&args[0], str, create);
27252 +               if (!e)
27253 +                       e = au_wbr_mfs_sec(&args[1], str, create);
27254 +               if (unlikely(e))
27255 +                       err = e;
27256 +               break;
27257 +       case AuWbrCreate_MFSRR:
27258 +       case AuWbrCreate_TDMFS:
27259 +       case AuWbrCreate_PMFSRR:
27260 +               e = au_wbr_mfs_wmark(&args[0], str, create);
27261 +               if (unlikely(e)) {
27262 +                       err = e;
27263 +                       break;
27264 +               }
27265 +               fallthrough;
27266 +       case AuWbrCreate_MFS:
27267 +       case AuWbrCreate_PMFS:
27268 +               create->mfs_second = AUFS_MFS_DEF_SEC;
27269 +               break;
27270 +       case AuWbrCreate_MFSV:
27271 +       case AuWbrCreate_PMFSV:
27272 +               e = au_wbr_mfs_sec(&args[0], str, create);
27273 +               if (unlikely(e))
27274 +                       err = e;
27275 +               break;
27276 +       }
27277 +
27278 +       return err;
27279 +}
27280 +
27281 +const char *au_optstr_wbr_create(int wbr_create)
27282 +{
27283 +       return au_parser_pattern(wbr_create, au_wbr_create_policy);
27284 +}
27285 +
27286 +static match_table_t au_wbr_copyup_policy = {
27287 +       {AuWbrCopyup_TDP, "tdp"},
27288 +       {AuWbrCopyup_TDP, "top-down-parent"},
27289 +       {AuWbrCopyup_BUP, "bup"},
27290 +       {AuWbrCopyup_BUP, "bottom-up-parent"},
27291 +       {AuWbrCopyup_BU, "bu"},
27292 +       {AuWbrCopyup_BU, "bottom-up"},
27293 +       {-1, NULL}
27294 +};
27295 +
27296 +int au_wbr_copyup_val(char *str)
27297 +{
27298 +       substring_t args[MAX_OPT_ARGS];
27299 +
27300 +       return match_token(str, au_wbr_copyup_policy, args);
27301 +}
27302 +
27303 +const char *au_optstr_wbr_copyup(int wbr_copyup)
27304 +{
27305 +       return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
27306 +}
27307 +
27308 +/* ---------------------------------------------------------------------- */
27309 +
27310 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
27311 +              aufs_bindex_t bindex)
27312 +{
27313 +       int err;
27314 +       struct au_opt_add *add = &opt->add;
27315 +       char *p;
27316 +
27317 +       add->bindex = bindex;
27318 +       add->perm = AuBrPerm_RO;
27319 +       add->pathname = opt_str;
27320 +       p = strchr(opt_str, '=');
27321 +       if (p) {
27322 +               *p++ = 0;
27323 +               if (*p)
27324 +                       add->perm = au_br_perm_val(p);
27325 +       }
27326 +
27327 +       err = vfsub_kern_path(add->pathname, AuOpt_LkupDirFlags, &add->path);
27328 +       if (!err) {
27329 +               if (!p) {
27330 +                       add->perm = AuBrPerm_RO;
27331 +                       if (au_test_fs_rr(add->path.dentry->d_sb))
27332 +                               add->perm = AuBrPerm_RR;
27333 +                       else if (!bindex && !(sb_flags & SB_RDONLY))
27334 +                               add->perm = AuBrPerm_RW;
27335 +               }
27336 +               opt->type = Opt_add;
27337 +               goto out;
27338 +       }
27339 +       pr_err("lookup failed %s (%d)\n", add->pathname, err);
27340 +       err = -EINVAL;
27341 +
27342 +out:
27343 +       return err;
27344 +}
27345 +
27346 +static int au_opt_wbr_create(struct super_block *sb,
27347 +                            struct au_opt_wbr_create *create)
27348 +{
27349 +       int err;
27350 +       struct au_sbinfo *sbinfo;
27351 +
27352 +       SiMustWriteLock(sb);
27353 +
27354 +       err = 1; /* handled */
27355 +       sbinfo = au_sbi(sb);
27356 +       if (sbinfo->si_wbr_create_ops->fin) {
27357 +               err = sbinfo->si_wbr_create_ops->fin(sb);
27358 +               if (!err)
27359 +                       err = 1;
27360 +       }
27361 +
27362 +       sbinfo->si_wbr_create = create->wbr_create;
27363 +       sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
27364 +       switch (create->wbr_create) {
27365 +       case AuWbrCreate_MFSRRV:
27366 +       case AuWbrCreate_MFSRR:
27367 +       case AuWbrCreate_TDMFS:
27368 +       case AuWbrCreate_TDMFSV:
27369 +       case AuWbrCreate_PMFSRR:
27370 +       case AuWbrCreate_PMFSRRV:
27371 +               sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
27372 +               fallthrough;
27373 +       case AuWbrCreate_MFS:
27374 +       case AuWbrCreate_MFSV:
27375 +       case AuWbrCreate_PMFS:
27376 +       case AuWbrCreate_PMFSV:
27377 +               sbinfo->si_wbr_mfs.mfs_expire
27378 +                       = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
27379 +               break;
27380 +       }
27381 +
27382 +       if (sbinfo->si_wbr_create_ops->init)
27383 +               sbinfo->si_wbr_create_ops->init(sb); /* ignore */
27384 +
27385 +       return err;
27386 +}
27387 +
27388 +/*
27389 + * returns,
27390 + * plus: processed without an error
27391 + * zero: unprocessed
27392 + */
27393 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
27394 +                        struct au_opts *opts)
27395 +{
27396 +       int err;
27397 +       struct au_sbinfo *sbinfo;
27398 +
27399 +       SiMustWriteLock(sb);
27400 +
27401 +       err = 1; /* handled */
27402 +       sbinfo = au_sbi(sb);
27403 +       switch (opt->type) {
27404 +       case Opt_udba:
27405 +               sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27406 +               sbinfo->si_mntflags |= opt->udba;
27407 +               opts->given_udba |= opt->udba;
27408 +               break;
27409 +
27410 +       case Opt_plink:
27411 +               if (opt->tf)
27412 +                       au_opt_set(sbinfo->si_mntflags, PLINK);
27413 +               else {
27414 +                       if (au_opt_test(sbinfo->si_mntflags, PLINK))
27415 +                               au_plink_put(sb, /*verbose*/1);
27416 +                       au_opt_clr(sbinfo->si_mntflags, PLINK);
27417 +               }
27418 +               break;
27419 +       case Opt_list_plink:
27420 +               if (au_opt_test(sbinfo->si_mntflags, PLINK))
27421 +                       au_plink_list(sb);
27422 +               break;
27423 +
27424 +       case Opt_dio:
27425 +               if (opt->tf) {
27426 +                       au_opt_set(sbinfo->si_mntflags, DIO);
27427 +                       au_fset_opts(opts->flags, REFRESH_DYAOP);
27428 +               } else {
27429 +                       au_opt_clr(sbinfo->si_mntflags, DIO);
27430 +                       au_fset_opts(opts->flags, REFRESH_DYAOP);
27431 +               }
27432 +               break;
27433 +
27434 +       case Opt_fhsm_sec:
27435 +               au_fhsm_set(sbinfo, opt->fhsm_second);
27436 +               break;
27437 +
27438 +       case Opt_diropq_a:
27439 +               au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27440 +               break;
27441 +       case Opt_diropq_w:
27442 +               au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27443 +               break;
27444 +
27445 +       case Opt_warn_perm:
27446 +               if (opt->tf)
27447 +                       au_opt_set(sbinfo->si_mntflags, WARN_PERM);
27448 +               else
27449 +                       au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
27450 +               break;
27451 +
27452 +       case Opt_verbose:
27453 +               if (opt->tf)
27454 +                       au_opt_set(sbinfo->si_mntflags, VERBOSE);
27455 +               else
27456 +                       au_opt_clr(sbinfo->si_mntflags, VERBOSE);
27457 +               break;
27458 +
27459 +       case Opt_sum:
27460 +               if (opt->tf)
27461 +                       au_opt_set(sbinfo->si_mntflags, SUM);
27462 +               else {
27463 +                       au_opt_clr(sbinfo->si_mntflags, SUM);
27464 +                       au_opt_clr(sbinfo->si_mntflags, SUM_W);
27465 +               }
27466 +               break;
27467 +       case Opt_wsum:
27468 +               au_opt_clr(sbinfo->si_mntflags, SUM);
27469 +               au_opt_set(sbinfo->si_mntflags, SUM_W);
27470 +               break;
27471 +
27472 +       case Opt_wbr_create:
27473 +               err = au_opt_wbr_create(sb, &opt->wbr_create);
27474 +               break;
27475 +       case Opt_wbr_copyup:
27476 +               sbinfo->si_wbr_copyup = opt->wbr_copyup;
27477 +               sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
27478 +               break;
27479 +
27480 +       case Opt_dirwh:
27481 +               sbinfo->si_dirwh = opt->dirwh;
27482 +               break;
27483 +
27484 +       case Opt_rdcache:
27485 +               sbinfo->si_rdcache
27486 +                       = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
27487 +               break;
27488 +       case Opt_rdblk:
27489 +               sbinfo->si_rdblk = opt->rdblk;
27490 +               break;
27491 +       case Opt_rdhash:
27492 +               sbinfo->si_rdhash = opt->rdhash;
27493 +               break;
27494 +
27495 +       case Opt_shwh:
27496 +               if (opt->tf)
27497 +                       au_opt_set(sbinfo->si_mntflags, SHWH);
27498 +               else
27499 +                       au_opt_clr(sbinfo->si_mntflags, SHWH);
27500 +               break;
27501 +
27502 +       case Opt_dirperm1:
27503 +               if (opt->tf)
27504 +                       au_opt_set(sbinfo->si_mntflags, DIRPERM1);
27505 +               else
27506 +                       au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
27507 +               break;
27508 +
27509 +       case Opt_trunc_xino:
27510 +               if (opt->tf)
27511 +                       au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
27512 +               else
27513 +                       au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
27514 +               break;
27515 +
27516 +       case Opt_trunc_xino_path:
27517 +       case Opt_itrunc_xino:
27518 +               err = au_xino_trunc(sb, opt->xino_itrunc.bindex,
27519 +                                   /*idx_begin*/0);
27520 +               if (!err)
27521 +                       err = 1;
27522 +               break;
27523 +
27524 +       case Opt_trunc_xib:
27525 +               if (opt->tf)
27526 +                       au_fset_opts(opts->flags, TRUNC_XIB);
27527 +               else
27528 +                       au_fclr_opts(opts->flags, TRUNC_XIB);
27529 +               break;
27530 +
27531 +       case Opt_dirren:
27532 +               err = 1;
27533 +               if (opt->tf) {
27534 +                       if (!au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27535 +                               err = au_dr_opt_set(sb);
27536 +                               if (!err)
27537 +                                       err = 1;
27538 +                       }
27539 +                       if (err == 1)
27540 +                               au_opt_set(sbinfo->si_mntflags, DIRREN);
27541 +               } else {
27542 +                       if (au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27543 +                               err = au_dr_opt_clr(sb, au_ftest_opts(opts->flags,
27544 +                                                                     DR_FLUSHED));
27545 +                               if (!err)
27546 +                                       err = 1;
27547 +                       }
27548 +                       if (err == 1)
27549 +                               au_opt_clr(sbinfo->si_mntflags, DIRREN);
27550 +               }
27551 +               break;
27552 +
27553 +       case Opt_acl:
27554 +               if (opt->tf)
27555 +                       sb->s_flags |= SB_POSIXACL;
27556 +               else
27557 +                       sb->s_flags &= ~SB_POSIXACL;
27558 +               break;
27559 +
27560 +       default:
27561 +               err = 0;
27562 +               break;
27563 +       }
27564 +
27565 +       return err;
27566 +}
27567 +
27568 +/*
27569 + * returns tri-state.
27570 + * plus: processed without an error
27571 + * zero: unprocessed
27572 + * minus: error
27573 + */
27574 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
27575 +                    struct au_opts *opts)
27576 +{
27577 +       int err, do_refresh;
27578 +
27579 +       err = 0;
27580 +       switch (opt->type) {
27581 +       case Opt_append:
27582 +               opt->add.bindex = au_sbbot(sb) + 1;
27583 +               if (opt->add.bindex < 0)
27584 +                       opt->add.bindex = 0;
27585 +               goto add;
27586 +               /* Always goto add, not fallthrough */
27587 +       case Opt_prepend:
27588 +               opt->add.bindex = 0;
27589 +               fallthrough;
27590 +       add: /* indented label */
27591 +       case Opt_add:
27592 +               err = au_br_add(sb, &opt->add,
27593 +                               au_ftest_opts(opts->flags, REMOUNT));
27594 +               if (!err) {
27595 +                       err = 1;
27596 +                       au_fset_opts(opts->flags, REFRESH);
27597 +               }
27598 +               break;
27599 +
27600 +       case Opt_del:
27601 +       case Opt_idel:
27602 +               err = au_br_del(sb, &opt->del,
27603 +                               au_ftest_opts(opts->flags, REMOUNT));
27604 +               if (!err) {
27605 +                       err = 1;
27606 +                       au_fset_opts(opts->flags, TRUNC_XIB);
27607 +                       au_fset_opts(opts->flags, REFRESH);
27608 +               }
27609 +               break;
27610 +
27611 +       case Opt_mod:
27612 +       case Opt_imod:
27613 +               err = au_br_mod(sb, &opt->mod,
27614 +                               au_ftest_opts(opts->flags, REMOUNT),
27615 +                               &do_refresh);
27616 +               if (!err) {
27617 +                       err = 1;
27618 +                       if (do_refresh)
27619 +                               au_fset_opts(opts->flags, REFRESH);
27620 +               }
27621 +               break;
27622 +       }
27623 +       return err;
27624 +}
27625 +
27626 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
27627 +                      struct au_opt_xino **opt_xino,
27628 +                      struct au_opts *opts)
27629 +{
27630 +       int err;
27631 +
27632 +       err = 0;
27633 +       switch (opt->type) {
27634 +       case Opt_xino:
27635 +               err = au_xino_set(sb, &opt->xino,
27636 +                                 !!au_ftest_opts(opts->flags, REMOUNT));
27637 +               if (!err)
27638 +                       *opt_xino = &opt->xino;
27639 +               break;
27640 +       case Opt_noxino:
27641 +               au_xino_clr(sb);
27642 +               *opt_xino = (void *)-1;
27643 +               break;
27644 +       }
27645 +
27646 +       return err;
27647 +}
27648 +
27649 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
27650 +                  unsigned int pending)
27651 +{
27652 +       int err, fhsm;
27653 +       aufs_bindex_t bindex, bbot;
27654 +       unsigned char do_plink, skip, do_free, can_no_dreval;
27655 +       struct au_branch *br;
27656 +       struct au_wbr *wbr;
27657 +       struct dentry *root, *dentry;
27658 +       struct inode *dir, *h_dir;
27659 +       struct au_sbinfo *sbinfo;
27660 +       struct au_hinode *hdir;
27661 +
27662 +       SiMustAnyLock(sb);
27663 +
27664 +       sbinfo = au_sbi(sb);
27665 +       AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
27666 +
27667 +       if (!(sb_flags & SB_RDONLY)) {
27668 +               if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
27669 +                       pr_warn("first branch should be rw\n");
27670 +               if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
27671 +                       pr_warn_once("shwh should be used with ro\n");
27672 +       }
27673 +
27674 +       if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
27675 +           && !au_opt_test(sbinfo->si_mntflags, XINO))
27676 +               pr_warn_once("udba=*notify requires xino\n");
27677 +
27678 +       if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
27679 +               pr_warn_once("dirperm1 breaks the protection"
27680 +                            " by the permission bits on the lower branch\n");
27681 +
27682 +       err = 0;
27683 +       fhsm = 0;
27684 +       root = sb->s_root;
27685 +       dir = d_inode(root);
27686 +       do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
27687 +       can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
27688 +                                     UDBA_NONE);
27689 +       bbot = au_sbbot(sb);
27690 +       for (bindex = 0; !err && bindex <= bbot; bindex++) {
27691 +               skip = 0;
27692 +               h_dir = au_h_iptr(dir, bindex);
27693 +               br = au_sbr(sb, bindex);
27694 +
27695 +               if ((br->br_perm & AuBrAttr_ICEX)
27696 +                   && !h_dir->i_op->listxattr)
27697 +                       br->br_perm &= ~AuBrAttr_ICEX;
27698 +#if 0 /* untested */
27699 +               if ((br->br_perm & AuBrAttr_ICEX_SEC)
27700 +                   && (au_br_sb(br)->s_flags & SB_NOSEC))
27701 +                       br->br_perm &= ~AuBrAttr_ICEX_SEC;
27702 +#endif
27703 +
27704 +               do_free = 0;
27705 +               wbr = br->br_wbr;
27706 +               if (wbr)
27707 +                       wbr_wh_read_lock(wbr);
27708 +
27709 +               if (!au_br_writable(br->br_perm)) {
27710 +                       do_free = !!wbr;
27711 +                       skip = (!wbr
27712 +                               || (!wbr->wbr_whbase
27713 +                                   && !wbr->wbr_plink
27714 +                                   && !wbr->wbr_orph));
27715 +               } else if (!au_br_wh_linkable(br->br_perm)) {
27716 +                       /* skip = (!br->br_whbase && !br->br_orph); */
27717 +                       skip = (!wbr || !wbr->wbr_whbase);
27718 +                       if (skip && wbr) {
27719 +                               if (do_plink)
27720 +                                       skip = !!wbr->wbr_plink;
27721 +                               else
27722 +                                       skip = !wbr->wbr_plink;
27723 +                       }
27724 +               } else {
27725 +                       /* skip = (br->br_whbase && br->br_ohph); */
27726 +                       skip = (wbr && wbr->wbr_whbase);
27727 +                       if (skip) {
27728 +                               if (do_plink)
27729 +                                       skip = !!wbr->wbr_plink;
27730 +                               else
27731 +                                       skip = !wbr->wbr_plink;
27732 +                       }
27733 +               }
27734 +               if (wbr)
27735 +                       wbr_wh_read_unlock(wbr);
27736 +
27737 +               if (can_no_dreval) {
27738 +                       dentry = br->br_path.dentry;
27739 +                       spin_lock(&dentry->d_lock);
27740 +                       if (dentry->d_flags &
27741 +                           (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
27742 +                               can_no_dreval = 0;
27743 +                       spin_unlock(&dentry->d_lock);
27744 +               }
27745 +
27746 +               if (au_br_fhsm(br->br_perm)) {
27747 +                       fhsm++;
27748 +                       AuDebugOn(!br->br_fhsm);
27749 +               }
27750 +
27751 +               if (skip)
27752 +                       continue;
27753 +
27754 +               hdir = au_hi(dir, bindex);
27755 +               au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
27756 +               if (wbr)
27757 +                       wbr_wh_write_lock(wbr);
27758 +               err = au_wh_init(br, sb);
27759 +               if (wbr)
27760 +                       wbr_wh_write_unlock(wbr);
27761 +               au_hn_inode_unlock(hdir);
27762 +
27763 +               if (!err && do_free) {
27764 +                       au_kfree_rcu(wbr);
27765 +                       br->br_wbr = NULL;
27766 +               }
27767 +       }
27768 +
27769 +       if (can_no_dreval)
27770 +               au_fset_si(sbinfo, NO_DREVAL);
27771 +       else
27772 +               au_fclr_si(sbinfo, NO_DREVAL);
27773 +
27774 +       if (fhsm >= 2) {
27775 +               au_fset_si(sbinfo, FHSM);
27776 +               for (bindex = bbot; bindex >= 0; bindex--) {
27777 +                       br = au_sbr(sb, bindex);
27778 +                       if (au_br_fhsm(br->br_perm)) {
27779 +                               au_fhsm_set_bottom(sb, bindex);
27780 +                               break;
27781 +                       }
27782 +               }
27783 +       } else {
27784 +               au_fclr_si(sbinfo, FHSM);
27785 +               au_fhsm_set_bottom(sb, -1);
27786 +       }
27787 +
27788 +       return err;
27789 +}
27790 +
27791 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
27792 +{
27793 +       int err;
27794 +       unsigned int tmp;
27795 +       aufs_bindex_t bindex, bbot;
27796 +       struct au_opt *opt;
27797 +       struct au_opt_xino *opt_xino, xino;
27798 +       struct au_sbinfo *sbinfo;
27799 +       struct au_branch *br;
27800 +       struct inode *dir;
27801 +
27802 +       SiMustWriteLock(sb);
27803 +
27804 +       err = 0;
27805 +       opt_xino = NULL;
27806 +       opt = opts->opt;
27807 +       while (err >= 0 && opt->type != Opt_tail)
27808 +               err = au_opt_simple(sb, opt++, opts);
27809 +       if (err > 0)
27810 +               err = 0;
27811 +       else if (unlikely(err < 0))
27812 +               goto out;
27813 +
27814 +       /* disable xino and udba temporary */
27815 +       sbinfo = au_sbi(sb);
27816 +       tmp = sbinfo->si_mntflags;
27817 +       au_opt_clr(sbinfo->si_mntflags, XINO);
27818 +       au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
27819 +
27820 +       opt = opts->opt;
27821 +       while (err >= 0 && opt->type != Opt_tail)
27822 +               err = au_opt_br(sb, opt++, opts);
27823 +       if (err > 0)
27824 +               err = 0;
27825 +       else if (unlikely(err < 0))
27826 +               goto out;
27827 +
27828 +       bbot = au_sbbot(sb);
27829 +       if (unlikely(bbot < 0)) {
27830 +               err = -EINVAL;
27831 +               pr_err("no branches\n");
27832 +               goto out;
27833 +       }
27834 +
27835 +       if (au_opt_test(tmp, XINO))
27836 +               au_opt_set(sbinfo->si_mntflags, XINO);
27837 +       opt = opts->opt;
27838 +       while (!err && opt->type != Opt_tail)
27839 +               err = au_opt_xino(sb, opt++, &opt_xino, opts);
27840 +       if (unlikely(err))
27841 +               goto out;
27842 +
27843 +       err = au_opts_verify(sb, sb->s_flags, tmp);
27844 +       if (unlikely(err))
27845 +               goto out;
27846 +
27847 +       /* restore xino */
27848 +       if (au_opt_test(tmp, XINO) && !opt_xino) {
27849 +               xino.file = au_xino_def(sb);
27850 +               err = PTR_ERR(xino.file);
27851 +               if (IS_ERR(xino.file))
27852 +                       goto out;
27853 +
27854 +               err = au_xino_set(sb, &xino, /*remount*/0);
27855 +               fput(xino.file);
27856 +               if (unlikely(err))
27857 +                       goto out;
27858 +       }
27859 +
27860 +       /* restore udba */
27861 +       tmp &= AuOptMask_UDBA;
27862 +       sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27863 +       sbinfo->si_mntflags |= tmp;
27864 +       bbot = au_sbbot(sb);
27865 +       for (bindex = 0; bindex <= bbot; bindex++) {
27866 +               br = au_sbr(sb, bindex);
27867 +               err = au_hnotify_reset_br(tmp, br, br->br_perm);
27868 +               if (unlikely(err))
27869 +                       AuIOErr("hnotify failed on br %d, %d, ignored\n",
27870 +                               bindex, err);
27871 +               /* go on even if err */
27872 +       }
27873 +       if (au_opt_test(tmp, UDBA_HNOTIFY)) {
27874 +               dir = d_inode(sb->s_root);
27875 +               au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
27876 +       }
27877 +
27878 +out:
27879 +       return err;
27880 +}
27881 +
27882 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
27883 +{
27884 +       int err, rerr;
27885 +       unsigned char no_dreval;
27886 +       struct inode *dir;
27887 +       struct au_opt_xino *opt_xino;
27888 +       struct au_opt *opt;
27889 +       struct au_sbinfo *sbinfo;
27890 +
27891 +       SiMustWriteLock(sb);
27892 +
27893 +       err = au_dr_opt_flush(sb);
27894 +       if (unlikely(err))
27895 +               goto out;
27896 +       au_fset_opts(opts->flags, DR_FLUSHED);
27897 +
27898 +       dir = d_inode(sb->s_root);
27899 +       sbinfo = au_sbi(sb);
27900 +       opt_xino = NULL;
27901 +       opt = opts->opt;
27902 +       while (err >= 0 && opt->type != Opt_tail) {
27903 +               err = au_opt_simple(sb, opt, opts);
27904 +               if (!err)
27905 +                       err = au_opt_br(sb, opt, opts);
27906 +               if (!err)
27907 +                       err = au_opt_xino(sb, opt, &opt_xino, opts);
27908 +               opt++;
27909 +       }
27910 +       if (err > 0)
27911 +               err = 0;
27912 +       AuTraceErr(err);
27913 +       /* go on even err */
27914 +
27915 +       no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
27916 +       rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
27917 +       if (unlikely(rerr && !err))
27918 +               err = rerr;
27919 +
27920 +       if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
27921 +               au_fset_opts(opts->flags, REFRESH_IDOP);
27922 +
27923 +       if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
27924 +               rerr = au_xib_trunc(sb);
27925 +               if (unlikely(rerr && !err))
27926 +                       err = rerr;
27927 +       }
27928 +
27929 +       /* will be handled by the caller */
27930 +       if (!au_ftest_opts(opts->flags, REFRESH)
27931 +           && (opts->given_udba
27932 +               || au_opt_test(sbinfo->si_mntflags, XINO)
27933 +               || au_ftest_opts(opts->flags, REFRESH_IDOP)
27934 +                   ))
27935 +               au_fset_opts(opts->flags, REFRESH);
27936 +
27937 +       AuDbg("status 0x%x\n", opts->flags);
27938 +
27939 +out:
27940 +       return err;
27941 +}
27942 +
27943 +/* ---------------------------------------------------------------------- */
27944 +
27945 +unsigned int au_opt_udba(struct super_block *sb)
27946 +{
27947 +       return au_mntflags(sb) & AuOptMask_UDBA;
27948 +}
27949 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
27950 --- /usr/share/empty/fs/aufs/opts.h     1970-01-01 01:00:00.000000000 +0100
27951 +++ linux/fs/aufs/opts.h        2022-11-05 23:02:18.969222617 +0100
27952 @@ -0,0 +1,263 @@
27953 +/* SPDX-License-Identifier: GPL-2.0 */
27954 +/*
27955 + * Copyright (C) 2005-2022 Junjiro R. Okajima
27956 + *
27957 + * This program is free software; you can redistribute it and/or modify
27958 + * it under the terms of the GNU General Public License as published by
27959 + * the Free Software Foundation; either version 2 of the License, or
27960 + * (at your option) any later version.
27961 + *
27962 + * This program is distributed in the hope that it will be useful,
27963 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27964 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
27965 + * GNU General Public License for more details.
27966 + *
27967 + * You should have received a copy of the GNU General Public License
27968 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
27969 + */
27970 +
27971 +/*
27972 + * mount options/flags
27973 + */
27974 +
27975 +#ifndef __AUFS_OPTS_H__
27976 +#define __AUFS_OPTS_H__
27977 +
27978 +#ifdef __KERNEL__
27979 +
27980 +#include <linux/fs_parser.h>
27981 +#include <linux/namei.h>
27982 +#include <linux/path.h>
27983 +
27984 +enum {
27985 +       Opt_br,
27986 +       Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
27987 +       Opt_idel, Opt_imod,
27988 +       Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
27989 +       Opt_xino, Opt_noxino,
27990 +       Opt_trunc_xino, Opt_trunc_xino_v,
27991 +       Opt_trunc_xino_path, Opt_itrunc_xino,
27992 +       Opt_trunc_xib,
27993 +       Opt_shwh,
27994 +       Opt_plink, Opt_list_plink,
27995 +       Opt_udba,
27996 +       Opt_dio,
27997 +       Opt_diropq, Opt_diropq_a, Opt_diropq_w,
27998 +       Opt_warn_perm,
27999 +       Opt_wbr_copyup, Opt_wbr_create,
28000 +       Opt_fhsm_sec,
28001 +       Opt_verbose, Opt_noverbose,
28002 +       Opt_sum, Opt_wsum,
28003 +       Opt_dirperm1,
28004 +       Opt_dirren,
28005 +       Opt_acl,
28006 +       Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
28007 +};
28008 +
28009 +/* ---------------------------------------------------------------------- */
28010 +
28011 +/* mount flags */
28012 +#define AuOpt_XINO             1               /* external inode number bitmap
28013 +                                                  and translation table */
28014 +#define AuOpt_TRUNC_XINO       (1 << 1)        /* truncate xino files */
28015 +#define AuOpt_UDBA_NONE                (1 << 2)        /* users direct branch access */
28016 +#define AuOpt_UDBA_REVAL       (1 << 3)
28017 +#define AuOpt_UDBA_HNOTIFY     (1 << 4)
28018 +#define AuOpt_SHWH             (1 << 5)        /* show whiteout */
28019 +#define AuOpt_PLINK            (1 << 6)        /* pseudo-link */
28020 +#define AuOpt_DIRPERM1         (1 << 7)        /* ignore the lower dir's perm
28021 +                                                  bits */
28022 +#define AuOpt_ALWAYS_DIROPQ    (1 << 9)        /* policy to creating diropq */
28023 +#define AuOpt_SUM              (1 << 10)       /* summation for statfs(2) */
28024 +#define AuOpt_SUM_W            (1 << 11)       /* unimplemented */
28025 +#define AuOpt_WARN_PERM                (1 << 12)       /* warn when add-branch */
28026 +#define AuOpt_VERBOSE          (1 << 13)       /* print the cause of error */
28027 +#define AuOpt_DIO              (1 << 14)       /* direct io */
28028 +#define AuOpt_DIRREN           (1 << 15)       /* directory rename */
28029 +
28030 +#ifndef CONFIG_AUFS_HNOTIFY
28031 +#undef AuOpt_UDBA_HNOTIFY
28032 +#define AuOpt_UDBA_HNOTIFY     0
28033 +#endif
28034 +#ifndef CONFIG_AUFS_DIRREN
28035 +#undef AuOpt_DIRREN
28036 +#define AuOpt_DIRREN           0
28037 +#endif
28038 +#ifndef CONFIG_AUFS_SHWH
28039 +#undef AuOpt_SHWH
28040 +#define AuOpt_SHWH             0
28041 +#endif
28042 +
28043 +#define AuOpt_Def      (AuOpt_XINO \
28044 +                        | AuOpt_UDBA_REVAL \
28045 +                        | AuOpt_PLINK \
28046 +                        /* | AuOpt_DIRPERM1 */ \
28047 +                        | AuOpt_WARN_PERM)
28048 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
28049 +                        | AuOpt_UDBA_REVAL \
28050 +                        | AuOpt_UDBA_HNOTIFY)
28051 +
28052 +#define AuOpt_LkupDirFlags     (LOOKUP_FOLLOW | LOOKUP_DIRECTORY)
28053 +
28054 +#define au_opt_test(flags, name)       (flags & AuOpt_##name)
28055 +#define au_opt_set(flags, name) do { \
28056 +       BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
28057 +       ((flags) |= AuOpt_##name); \
28058 +} while (0)
28059 +#define au_opt_set_udba(flags, name) do { \
28060 +       (flags) &= ~AuOptMask_UDBA; \
28061 +       ((flags) |= AuOpt_##name); \
28062 +} while (0)
28063 +#define au_opt_clr(flags, name) do { \
28064 +       ((flags) &= ~AuOpt_##name); \
28065 +} while (0)
28066 +
28067 +static inline unsigned int au_opts_plink(unsigned int mntflags)
28068 +{
28069 +#ifdef CONFIG_PROC_FS
28070 +       return mntflags;
28071 +#else
28072 +       return mntflags & ~AuOpt_PLINK;
28073 +#endif
28074 +}
28075 +
28076 +/* ---------------------------------------------------------------------- */
28077 +
28078 +/* policies to select one among multiple writable branches */
28079 +enum {
28080 +       AuWbrCreate_TDP,        /* top down parent */
28081 +       AuWbrCreate_RR,         /* round robin */
28082 +       AuWbrCreate_MFS,        /* most free space */
28083 +       AuWbrCreate_MFSV,       /* mfs with seconds */
28084 +       AuWbrCreate_MFSRR,      /* mfs then rr */
28085 +       AuWbrCreate_MFSRRV,     /* mfs then rr with seconds */
28086 +       AuWbrCreate_TDMFS,      /* top down regardless parent and mfs */
28087 +       AuWbrCreate_TDMFSV,     /* top down regardless parent and mfs */
28088 +       AuWbrCreate_PMFS,       /* parent and mfs */
28089 +       AuWbrCreate_PMFSV,      /* parent and mfs with seconds */
28090 +       AuWbrCreate_PMFSRR,     /* parent, mfs and round-robin */
28091 +       AuWbrCreate_PMFSRRV,    /* plus seconds */
28092 +
28093 +       AuWbrCreate_Def = AuWbrCreate_TDP
28094 +};
28095 +
28096 +enum {
28097 +       AuWbrCopyup_TDP,        /* top down parent */
28098 +       AuWbrCopyup_BUP,        /* bottom up parent */
28099 +       AuWbrCopyup_BU,         /* bottom up */
28100 +
28101 +       AuWbrCopyup_Def = AuWbrCopyup_TDP
28102 +};
28103 +
28104 +/* ---------------------------------------------------------------------- */
28105 +
28106 +struct file;
28107 +
28108 +struct au_opt_add {
28109 +       aufs_bindex_t   bindex;
28110 +       char            *pathname;
28111 +       int             perm;
28112 +       struct path     path;
28113 +};
28114 +
28115 +struct au_opt_del {
28116 +       char            *pathname;
28117 +       struct path     h_path;
28118 +};
28119 +
28120 +struct au_opt_mod {
28121 +       char            *path;
28122 +       int             perm;
28123 +       struct dentry   *h_root;
28124 +};
28125 +
28126 +struct au_opt_xino {
28127 +       char            *path;
28128 +       struct file     *file;
28129 +};
28130 +
28131 +struct au_opt_xino_itrunc {
28132 +       aufs_bindex_t   bindex;
28133 +};
28134 +
28135 +struct au_opt_wbr_create {
28136 +       int                     wbr_create;
28137 +       int                     mfs_second;
28138 +       unsigned long long      mfsrr_watermark;
28139 +};
28140 +
28141 +struct au_opt {
28142 +       int type;
28143 +       union {
28144 +               struct au_opt_xino      xino;
28145 +               struct au_opt_xino_itrunc xino_itrunc;
28146 +               struct au_opt_add       add;
28147 +               struct au_opt_del       del;
28148 +               struct au_opt_mod       mod;
28149 +               int                     dirwh;
28150 +               int                     rdcache;
28151 +               unsigned int            rdblk;
28152 +               unsigned int            rdhash;
28153 +               int                     udba;
28154 +               struct au_opt_wbr_create wbr_create;
28155 +               int                     wbr_copyup;
28156 +               unsigned int            fhsm_second;
28157 +               bool                    tf; /* generic flag, true or false */
28158 +       };
28159 +};
28160 +
28161 +/* opts flags */
28162 +#define AuOpts_REMOUNT         1
28163 +#define AuOpts_REFRESH         (1 << 1)
28164 +#define AuOpts_TRUNC_XIB       (1 << 2)
28165 +#define AuOpts_REFRESH_DYAOP   (1 << 3)
28166 +#define AuOpts_REFRESH_IDOP    (1 << 4)
28167 +#define AuOpts_DR_FLUSHED      (1 << 5)
28168 +#define au_ftest_opts(flags, name)     ((flags) & AuOpts_##name)
28169 +#define au_fset_opts(flags, name) \
28170 +       do { (flags) |= AuOpts_##name; } while (0)
28171 +#define au_fclr_opts(flags, name) \
28172 +       do { (flags) &= ~AuOpts_##name; } while (0)
28173 +
28174 +#ifndef CONFIG_AUFS_DIRREN
28175 +#undef AuOpts_DR_FLUSHED
28176 +#define AuOpts_DR_FLUSHED      0
28177 +#endif
28178 +
28179 +struct au_opts {
28180 +       struct au_opt   *opt;
28181 +       int             max_opt;
28182 +
28183 +       unsigned int    given_udba;
28184 +       unsigned int    flags;
28185 +       unsigned long   sb_flags;
28186 +};
28187 +
28188 +/* ---------------------------------------------------------------------- */
28189 +
28190 +/* opts.c */
28191 +int au_br_perm_val(char *perm);
28192 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
28193 +int au_udba_val(char *str);
28194 +const char *au_optstr_udba(int udba);
28195 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create);
28196 +const char *au_optstr_wbr_create(int wbr_create);
28197 +int au_wbr_copyup_val(char *str);
28198 +const char *au_optstr_wbr_copyup(int wbr_copyup);
28199 +
28200 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
28201 +              aufs_bindex_t bindex);
28202 +struct super_block;
28203 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
28204 +                  unsigned int pending);
28205 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
28206 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
28207 +
28208 +unsigned int au_opt_udba(struct super_block *sb);
28209 +
28210 +/* fsctx.c */
28211 +int aufs_fsctx_init(struct fs_context *fc);
28212 +extern const struct fs_parameter_spec aufs_fsctx_paramspec[];
28213 +
28214 +#endif /* __KERNEL__ */
28215 +#endif /* __AUFS_OPTS_H__ */
28216 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
28217 --- /usr/share/empty/fs/aufs/plink.c    1970-01-01 01:00:00.000000000 +0100
28218 +++ linux/fs/aufs/plink.c       2022-11-05 23:02:18.969222617 +0100
28219 @@ -0,0 +1,516 @@
28220 +// SPDX-License-Identifier: GPL-2.0
28221 +/*
28222 + * Copyright (C) 2005-2022 Junjiro R. Okajima
28223 + *
28224 + * This program is free software; you can redistribute it and/or modify
28225 + * it under the terms of the GNU General Public License as published by
28226 + * the Free Software Foundation; either version 2 of the License, or
28227 + * (at your option) any later version.
28228 + *
28229 + * This program is distributed in the hope that it will be useful,
28230 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28231 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28232 + * GNU General Public License for more details.
28233 + *
28234 + * You should have received a copy of the GNU General Public License
28235 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28236 + */
28237 +
28238 +/*
28239 + * pseudo-link
28240 + */
28241 +
28242 +#include "aufs.h"
28243 +
28244 +/*
28245 + * the pseudo-link maintenance mode.
28246 + * during a user process maintains the pseudo-links,
28247 + * prohibit adding a new plink and branch manipulation.
28248 + *
28249 + * Flags
28250 + * NOPLM:
28251 + *     For entry functions which will handle plink, and i_mutex is already held
28252 + *     in VFS.
28253 + *     They cannot wait and should return an error at once.
28254 + *     Callers has to check the error.
28255 + * NOPLMW:
28256 + *     For entry functions which will handle plink, but i_mutex is not held
28257 + *     in VFS.
28258 + *     They can wait the plink maintenance mode to finish.
28259 + *
28260 + * They behave like F_SETLK and F_SETLKW.
28261 + * If the caller never handle plink, then both flags are unnecessary.
28262 + */
28263 +
28264 +int au_plink_maint(struct super_block *sb, int flags)
28265 +{
28266 +       int err;
28267 +       pid_t pid, ppid;
28268 +       struct task_struct *parent, *prev;
28269 +       struct au_sbinfo *sbi;
28270 +
28271 +       SiMustAnyLock(sb);
28272 +
28273 +       err = 0;
28274 +       if (!au_opt_test(au_mntflags(sb), PLINK))
28275 +               goto out;
28276 +
28277 +       sbi = au_sbi(sb);
28278 +       pid = sbi->si_plink_maint_pid;
28279 +       if (!pid || pid == current->pid)
28280 +               goto out;
28281 +
28282 +       /* todo: it highly depends upon /sbin/mount.aufs */
28283 +       prev = NULL;
28284 +       parent = current;
28285 +       ppid = 0;
28286 +       rcu_read_lock();
28287 +       while (1) {
28288 +               parent = rcu_dereference(parent->real_parent);
28289 +               if (parent == prev)
28290 +                       break;
28291 +               ppid = task_pid_vnr(parent);
28292 +               if (pid == ppid) {
28293 +                       rcu_read_unlock();
28294 +                       goto out;
28295 +               }
28296 +               prev = parent;
28297 +       }
28298 +       rcu_read_unlock();
28299 +
28300 +       if (au_ftest_lock(flags, NOPLMW)) {
28301 +               /* if there is no i_mutex lock in VFS, we don't need to wait */
28302 +               /* AuDebugOn(!lockdep_depth(current)); */
28303 +               while (sbi->si_plink_maint_pid) {
28304 +                       si_read_unlock(sb);
28305 +                       /* gave up wake_up_bit() */
28306 +                       wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
28307 +
28308 +                       if (au_ftest_lock(flags, FLUSH))
28309 +                               au_nwt_flush(&sbi->si_nowait);
28310 +                       si_noflush_read_lock(sb);
28311 +               }
28312 +       } else if (au_ftest_lock(flags, NOPLM)) {
28313 +               AuDbg("ppid %d, pid %d\n", ppid, pid);
28314 +               err = -EAGAIN;
28315 +       }
28316 +
28317 +out:
28318 +       return err;
28319 +}
28320 +
28321 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
28322 +{
28323 +       spin_lock(&sbinfo->si_plink_maint_lock);
28324 +       sbinfo->si_plink_maint_pid = 0;
28325 +       spin_unlock(&sbinfo->si_plink_maint_lock);
28326 +       wake_up_all(&sbinfo->si_plink_wq);
28327 +}
28328 +
28329 +int au_plink_maint_enter(struct super_block *sb)
28330 +{
28331 +       int err;
28332 +       struct au_sbinfo *sbinfo;
28333 +
28334 +       err = 0;
28335 +       sbinfo = au_sbi(sb);
28336 +       /* make sure i am the only one in this fs */
28337 +       si_write_lock(sb, AuLock_FLUSH);
28338 +       if (au_opt_test(au_mntflags(sb), PLINK)) {
28339 +               spin_lock(&sbinfo->si_plink_maint_lock);
28340 +               if (!sbinfo->si_plink_maint_pid)
28341 +                       sbinfo->si_plink_maint_pid = current->pid;
28342 +               else
28343 +                       err = -EBUSY;
28344 +               spin_unlock(&sbinfo->si_plink_maint_lock);
28345 +       }
28346 +       si_write_unlock(sb);
28347 +
28348 +       return err;
28349 +}
28350 +
28351 +/* ---------------------------------------------------------------------- */
28352 +
28353 +#ifdef CONFIG_AUFS_DEBUG
28354 +void au_plink_list(struct super_block *sb)
28355 +{
28356 +       int i;
28357 +       struct au_sbinfo *sbinfo;
28358 +       struct hlist_bl_head *hbl;
28359 +       struct hlist_bl_node *pos;
28360 +       struct au_icntnr *icntnr;
28361 +
28362 +       SiMustAnyLock(sb);
28363 +
28364 +       sbinfo = au_sbi(sb);
28365 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28366 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28367 +
28368 +       for (i = 0; i < AuPlink_NHASH; i++) {
28369 +               hbl = sbinfo->si_plink + i;
28370 +               hlist_bl_lock(hbl);
28371 +               hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28372 +                       AuDbg("%lu\n", icntnr->vfs_inode.i_ino);
28373 +               hlist_bl_unlock(hbl);
28374 +       }
28375 +}
28376 +#endif
28377 +
28378 +/* is the inode pseudo-linked? */
28379 +int au_plink_test(struct inode *inode)
28380 +{
28381 +       int found, i;
28382 +       struct au_sbinfo *sbinfo;
28383 +       struct hlist_bl_head *hbl;
28384 +       struct hlist_bl_node *pos;
28385 +       struct au_icntnr *icntnr;
28386 +
28387 +       sbinfo = au_sbi(inode->i_sb);
28388 +       AuRwMustAnyLock(&sbinfo->si_rwsem);
28389 +       AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
28390 +       AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28391 +
28392 +       found = 0;
28393 +       i = au_plink_hash(inode->i_ino);
28394 +       hbl =  sbinfo->si_plink + i;
28395 +       hlist_bl_lock(hbl);
28396 +       hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28397 +               if (&icntnr->vfs_inode == inode) {
28398 +                       found = 1;
28399 +                       break;
28400 +               }
28401 +       hlist_bl_unlock(hbl);
28402 +       return found;
28403 +}
28404 +
28405 +/* ---------------------------------------------------------------------- */
28406 +
28407 +/*
28408 + * generate a name for plink.
28409 + * the file will be stored under AUFS_WH_PLINKDIR.
28410 + */
28411 +/* 20 is max digits length of ulong 64 */
28412 +#define PLINK_NAME_LEN ((20 + 1) * 2)
28413 +
28414 +static int plink_name(char *name, int len, struct inode *inode,
28415 +                     aufs_bindex_t bindex)
28416 +{
28417 +       int rlen;
28418 +       struct inode *h_inode;
28419 +
28420 +       h_inode = au_h_iptr(inode, bindex);
28421 +       rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
28422 +       return rlen;
28423 +}
28424 +
28425 +struct au_do_plink_lkup_args {
28426 +       struct dentry **errp;
28427 +       struct qstr *tgtname;
28428 +       struct path *h_ppath;
28429 +};
28430 +
28431 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
28432 +                                      struct path *h_ppath)
28433 +{
28434 +       struct dentry *h_dentry;
28435 +       struct inode *h_inode;
28436 +
28437 +       h_inode = d_inode(h_ppath->dentry);
28438 +       inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2);
28439 +       h_dentry = vfsub_lkup_one(tgtname, h_ppath);
28440 +       inode_unlock_shared(h_inode);
28441 +
28442 +       return h_dentry;
28443 +}
28444 +
28445 +static void au_call_do_plink_lkup(void *args)
28446 +{
28447 +       struct au_do_plink_lkup_args *a = args;
28448 +       *a->errp = au_do_plink_lkup(a->tgtname, a->h_ppath);
28449 +}
28450 +
28451 +/* lookup the plink-ed @inode under the branch at @bindex */
28452 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
28453 +{
28454 +       struct dentry *h_dentry;
28455 +       struct au_branch *br;
28456 +       struct path h_ppath;
28457 +       int wkq_err;
28458 +       char a[PLINK_NAME_LEN];
28459 +       struct qstr tgtname = QSTR_INIT(a, 0);
28460 +
28461 +       AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28462 +
28463 +       br = au_sbr(inode->i_sb, bindex);
28464 +       h_ppath.dentry = br->br_wbr->wbr_plink;
28465 +       h_ppath.mnt = au_br_mnt(br);
28466 +       tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28467 +
28468 +       if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28469 +               struct au_do_plink_lkup_args args = {
28470 +                       .errp           = &h_dentry,
28471 +                       .tgtname        = &tgtname,
28472 +                       .h_ppath        = &h_ppath
28473 +               };
28474 +
28475 +               wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
28476 +               if (unlikely(wkq_err))
28477 +                       h_dentry = ERR_PTR(wkq_err);
28478 +       } else
28479 +               h_dentry = au_do_plink_lkup(&tgtname, &h_ppath);
28480 +
28481 +       return h_dentry;
28482 +}
28483 +
28484 +/* create a pseudo-link */
28485 +static int do_whplink(struct qstr *tgt, struct path *h_ppath,
28486 +                     struct dentry *h_dentry)
28487 +{
28488 +       int err;
28489 +       struct path h_path;
28490 +       struct inode *h_dir, *delegated;
28491 +
28492 +       h_dir = d_inode(h_ppath->dentry);
28493 +       inode_lock_nested(h_dir, AuLsc_I_CHILD2);
28494 +       h_path.mnt = h_ppath->mnt;
28495 +again:
28496 +       h_path.dentry = vfsub_lkup_one(tgt, h_ppath);
28497 +       err = PTR_ERR(h_path.dentry);
28498 +       if (IS_ERR(h_path.dentry))
28499 +               goto out;
28500 +
28501 +       err = 0;
28502 +       /* wh.plink dir is not monitored */
28503 +       /* todo: is it really safe? */
28504 +       if (d_is_positive(h_path.dentry)
28505 +           && d_inode(h_path.dentry) != d_inode(h_dentry)) {
28506 +               delegated = NULL;
28507 +               err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
28508 +               if (unlikely(err == -EWOULDBLOCK)) {
28509 +                       pr_warn("cannot retry for NFSv4 delegation"
28510 +                               " for an internal unlink\n");
28511 +                       iput(delegated);
28512 +               }
28513 +               dput(h_path.dentry);
28514 +               h_path.dentry = NULL;
28515 +               if (!err)
28516 +                       goto again;
28517 +       }
28518 +       if (!err && d_is_negative(h_path.dentry)) {
28519 +               delegated = NULL;
28520 +               err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
28521 +               if (unlikely(err == -EWOULDBLOCK)) {
28522 +                       pr_warn("cannot retry for NFSv4 delegation"
28523 +                               " for an internal link\n");
28524 +                       iput(delegated);
28525 +               }
28526 +       }
28527 +       dput(h_path.dentry);
28528 +
28529 +out:
28530 +       inode_unlock(h_dir);
28531 +       return err;
28532 +}
28533 +
28534 +struct do_whplink_args {
28535 +       int *errp;
28536 +       struct qstr *tgt;
28537 +       struct path *h_ppath;
28538 +       struct dentry *h_dentry;
28539 +};
28540 +
28541 +static void call_do_whplink(void *args)
28542 +{
28543 +       struct do_whplink_args *a = args;
28544 +       *a->errp = do_whplink(a->tgt, a->h_ppath, a->h_dentry);
28545 +}
28546 +
28547 +static int whplink(struct dentry *h_dentry, struct inode *inode,
28548 +                  aufs_bindex_t bindex)
28549 +{
28550 +       int err, wkq_err;
28551 +       struct au_branch *br;
28552 +       struct au_wbr *wbr;
28553 +       struct path h_ppath;
28554 +       char a[PLINK_NAME_LEN];
28555 +       struct qstr tgtname = QSTR_INIT(a, 0);
28556 +
28557 +       br = au_sbr(inode->i_sb, bindex);
28558 +       wbr = br->br_wbr;
28559 +       h_ppath.dentry = wbr->wbr_plink;
28560 +       h_ppath.mnt = au_br_mnt(br);
28561 +       tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28562 +
28563 +       /* always superio. */
28564 +       if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28565 +               struct do_whplink_args args = {
28566 +                       .errp           = &err,
28567 +                       .tgt            = &tgtname,
28568 +                       .h_ppath        = &h_ppath,
28569 +                       .h_dentry       = h_dentry
28570 +               };
28571 +               wkq_err = au_wkq_wait(call_do_whplink, &args);
28572 +               if (unlikely(wkq_err))
28573 +                       err = wkq_err;
28574 +       } else
28575 +               err = do_whplink(&tgtname, &h_ppath, h_dentry);
28576 +
28577 +       return err;
28578 +}
28579 +
28580 +/*
28581 + * create a new pseudo-link for @h_dentry on @bindex.
28582 + * the linked inode is held in aufs @inode.
28583 + */
28584 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
28585 +                    struct dentry *h_dentry)
28586 +{
28587 +       struct super_block *sb;
28588 +       struct au_sbinfo *sbinfo;
28589 +       struct hlist_bl_head *hbl;
28590 +       struct hlist_bl_node *pos;
28591 +       struct au_icntnr *icntnr;
28592 +       int found, err, cnt, i;
28593 +
28594 +       sb = inode->i_sb;
28595 +       sbinfo = au_sbi(sb);
28596 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28597 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28598 +
28599 +       found = au_plink_test(inode);
28600 +       if (found)
28601 +               return;
28602 +
28603 +       i = au_plink_hash(inode->i_ino);
28604 +       hbl = sbinfo->si_plink + i;
28605 +       au_igrab(inode);
28606 +
28607 +       hlist_bl_lock(hbl);
28608 +       hlist_bl_for_each_entry(icntnr, pos, hbl, plink) {
28609 +               if (&icntnr->vfs_inode == inode) {
28610 +                       found = 1;
28611 +                       break;
28612 +               }
28613 +       }
28614 +       if (!found) {
28615 +               icntnr = container_of(inode, struct au_icntnr, vfs_inode);
28616 +               hlist_bl_add_head(&icntnr->plink, hbl);
28617 +       }
28618 +       hlist_bl_unlock(hbl);
28619 +       if (!found) {
28620 +               cnt = au_hbl_count(hbl);
28621 +#define msg "unexpectedly unbalanced or too many pseudo-links"
28622 +               if (cnt > AUFS_PLINK_WARN)
28623 +                       AuWarn1(msg ", %d\n", cnt);
28624 +#undef msg
28625 +               err = whplink(h_dentry, inode, bindex);
28626 +               if (unlikely(err)) {
28627 +                       pr_warn("err %d, damaged pseudo link.\n", err);
28628 +                       au_hbl_del(&icntnr->plink, hbl);
28629 +                       iput(&icntnr->vfs_inode);
28630 +               }
28631 +       } else
28632 +               iput(&icntnr->vfs_inode);
28633 +}
28634 +
28635 +/* free all plinks */
28636 +void au_plink_put(struct super_block *sb, int verbose)
28637 +{
28638 +       int i, warned;
28639 +       struct au_sbinfo *sbinfo;
28640 +       struct hlist_bl_head *hbl;
28641 +       struct hlist_bl_node *pos, *tmp;
28642 +       struct au_icntnr *icntnr;
28643 +
28644 +       SiMustWriteLock(sb);
28645 +
28646 +       sbinfo = au_sbi(sb);
28647 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28648 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28649 +
28650 +       /* no spin_lock since sbinfo is write-locked */
28651 +       warned = 0;
28652 +       for (i = 0; i < AuPlink_NHASH; i++) {
28653 +               hbl = sbinfo->si_plink + i;
28654 +               if (!warned && verbose && !hlist_bl_empty(hbl)) {
28655 +                       pr_warn("pseudo-link is not flushed");
28656 +                       warned = 1;
28657 +               }
28658 +               hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink)
28659 +                       iput(&icntnr->vfs_inode);
28660 +               INIT_HLIST_BL_HEAD(hbl);
28661 +       }
28662 +}
28663 +
28664 +void au_plink_clean(struct super_block *sb, int verbose)
28665 +{
28666 +       struct dentry *root;
28667 +
28668 +       root = sb->s_root;
28669 +       aufs_write_lock(root);
28670 +       if (au_opt_test(au_mntflags(sb), PLINK))
28671 +               au_plink_put(sb, verbose);
28672 +       aufs_write_unlock(root);
28673 +}
28674 +
28675 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
28676 +{
28677 +       int do_put;
28678 +       aufs_bindex_t btop, bbot, bindex;
28679 +
28680 +       do_put = 0;
28681 +       btop = au_ibtop(inode);
28682 +       bbot = au_ibbot(inode);
28683 +       if (btop >= 0) {
28684 +               for (bindex = btop; bindex <= bbot; bindex++) {
28685 +                       if (!au_h_iptr(inode, bindex)
28686 +                           || au_ii_br_id(inode, bindex) != br_id)
28687 +                               continue;
28688 +                       au_set_h_iptr(inode, bindex, NULL, 0);
28689 +                       do_put = 1;
28690 +                       break;
28691 +               }
28692 +               if (do_put)
28693 +                       for (bindex = btop; bindex <= bbot; bindex++)
28694 +                               if (au_h_iptr(inode, bindex)) {
28695 +                                       do_put = 0;
28696 +                                       break;
28697 +                               }
28698 +       } else
28699 +               do_put = 1;
28700 +
28701 +       return do_put;
28702 +}
28703 +
28704 +/* free the plinks on a branch specified by @br_id */
28705 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
28706 +{
28707 +       struct au_sbinfo *sbinfo;
28708 +       struct hlist_bl_head *hbl;
28709 +       struct hlist_bl_node *pos, *tmp;
28710 +       struct au_icntnr *icntnr;
28711 +       struct inode *inode;
28712 +       int i, do_put;
28713 +
28714 +       SiMustWriteLock(sb);
28715 +
28716 +       sbinfo = au_sbi(sb);
28717 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28718 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28719 +
28720 +       /* no bit_lock since sbinfo is write-locked */
28721 +       for (i = 0; i < AuPlink_NHASH; i++) {
28722 +               hbl = sbinfo->si_plink + i;
28723 +               hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) {
28724 +                       inode = au_igrab(&icntnr->vfs_inode);
28725 +                       ii_write_lock_child(inode);
28726 +                       do_put = au_plink_do_half_refresh(inode, br_id);
28727 +                       if (do_put) {
28728 +                               hlist_bl_del(&icntnr->plink);
28729 +                               iput(inode);
28730 +                       }
28731 +                       ii_write_unlock(inode);
28732 +                       iput(inode);
28733 +               }
28734 +       }
28735 +}
28736 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
28737 --- /usr/share/empty/fs/aufs/poll.c     1970-01-01 01:00:00.000000000 +0100
28738 +++ linux/fs/aufs/poll.c        2022-11-05 23:02:18.969222617 +0100
28739 @@ -0,0 +1,51 @@
28740 +// SPDX-License-Identifier: GPL-2.0
28741 +/*
28742 + * Copyright (C) 2005-2022 Junjiro R. Okajima
28743 + *
28744 + * This program is free software; you can redistribute it and/or modify
28745 + * it under the terms of the GNU General Public License as published by
28746 + * the Free Software Foundation; either version 2 of the License, or
28747 + * (at your option) any later version.
28748 + *
28749 + * This program is distributed in the hope that it will be useful,
28750 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28751 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28752 + * GNU General Public License for more details.
28753 + *
28754 + * You should have received a copy of the GNU General Public License
28755 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28756 + */
28757 +
28758 +/*
28759 + * poll operation
28760 + * There is only one filesystem which implements ->poll operation, currently.
28761 + */
28762 +
28763 +#include "aufs.h"
28764 +
28765 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt)
28766 +{
28767 +       __poll_t mask;
28768 +       struct file *h_file;
28769 +       struct super_block *sb;
28770 +
28771 +       /* We should pretend an error happened. */
28772 +       mask = EPOLLERR /* | EPOLLIN | EPOLLOUT */;
28773 +       sb = file->f_path.dentry->d_sb;
28774 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
28775 +
28776 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
28777 +       if (IS_ERR(h_file)) {
28778 +               AuDbg("h_file %ld\n", PTR_ERR(h_file));
28779 +               goto out;
28780 +       }
28781 +
28782 +       mask = vfs_poll(h_file, pt);
28783 +       fput(h_file); /* instead of au_read_post() */
28784 +
28785 +out:
28786 +       si_read_unlock(sb);
28787 +       if (mask & EPOLLERR)
28788 +               AuDbg("mask 0x%x\n", mask);
28789 +       return mask;
28790 +}
28791 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
28792 --- /usr/share/empty/fs/aufs/posix_acl.c        1970-01-01 01:00:00.000000000 +0100
28793 +++ linux/fs/aufs/posix_acl.c   2023-02-20 21:05:51.959693785 +0100
28794 @@ -0,0 +1,108 @@
28795 +// SPDX-License-Identifier: GPL-2.0
28796 +/*
28797 + * Copyright (C) 2014-2022 Junjiro R. Okajima
28798 + *
28799 + * This program is free software; you can redistribute it and/or modify
28800 + * it under the terms of the GNU General Public License as published by
28801 + * the Free Software Foundation; either version 2 of the License, or
28802 + * (at your option) any later version.
28803 + *
28804 + * This program is distributed in the hope that it will be useful,
28805 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28806 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28807 + * GNU General Public License for more details.
28808 + *
28809 + * You should have received a copy of the GNU General Public License
28810 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28811 + */
28812 +
28813 +/*
28814 + * posix acl operations
28815 + */
28816 +
28817 +#include <linux/fs.h>
28818 +#include "aufs.h"
28819 +
28820 +struct posix_acl *aufs_get_inode_acl(struct inode *inode, int type, bool rcu)
28821 +{
28822 +       struct posix_acl *acl;
28823 +       int err;
28824 +       aufs_bindex_t bindex;
28825 +       struct inode *h_inode;
28826 +       struct super_block *sb;
28827 +
28828 +       acl = ERR_PTR(-ECHILD);
28829 +       if (rcu)
28830 +               goto out;
28831 +
28832 +       acl = NULL;
28833 +       sb = inode->i_sb;
28834 +       si_read_lock(sb, AuLock_FLUSH);
28835 +       ii_read_lock_child(inode);
28836 +       if (!(sb->s_flags & SB_POSIXACL))
28837 +               goto unlock;
28838 +
28839 +       bindex = au_ibtop(inode);
28840 +       h_inode = au_h_iptr(inode, bindex);
28841 +       if (unlikely(!h_inode
28842 +                    || ((h_inode->i_mode & S_IFMT)
28843 +                        != (inode->i_mode & S_IFMT)))) {
28844 +               err = au_busy_or_stale();
28845 +               acl = ERR_PTR(err);
28846 +               goto unlock;
28847 +       }
28848 +
28849 +       /* always topmost only */
28850 +       acl = get_inode_acl(h_inode, type);
28851 +       if (IS_ERR(acl))
28852 +               forget_cached_acl(inode, type);
28853 +       else
28854 +               set_cached_acl(inode, type, acl);
28855 +
28856 +unlock:
28857 +       ii_read_unlock(inode);
28858 +       si_read_unlock(sb);
28859 +
28860 +out:
28861 +       AuTraceErrPtr(acl);
28862 +       return acl;
28863 +}
28864 +
28865 +struct posix_acl *aufs_get_acl(struct user_namespace *userns,
28866 +                              struct dentry *dentry, int type)
28867 +{
28868 +       struct posix_acl *acl;
28869 +       struct inode *inode;
28870 +
28871 +       inode = d_inode(dentry);
28872 +       acl = aufs_get_inode_acl(inode, type, /*rcu*/false);
28873 +
28874 +       return acl;
28875 +}
28876 +
28877 +int aufs_set_acl(struct user_namespace *userns, struct dentry *dentry,
28878 +                struct posix_acl *acl, int type)
28879 +{
28880 +       int err;
28881 +       ssize_t ssz;
28882 +       struct inode *inode;
28883 +       struct au_sxattr arg = {
28884 +               .type = AU_ACL_SET,
28885 +               .u.acl_set = {
28886 +                       .acl    = acl,
28887 +                       .type   = type
28888 +               },
28889 +       };
28890 +
28891 +       inode = d_inode(dentry);
28892 +       IMustLock(inode);
28893 +
28894 +       ssz = au_sxattr(dentry, inode, &arg);
28895 +       /* forget even it if succeeds since the branch might set differently */
28896 +       forget_cached_acl(inode, type);
28897 +       err = ssz;
28898 +       if (ssz >= 0)
28899 +               err = 0;
28900 +
28901 +       return err;
28902 +}
28903 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
28904 --- /usr/share/empty/fs/aufs/procfs.c   1970-01-01 01:00:00.000000000 +0100
28905 +++ linux/fs/aufs/procfs.c      2022-11-05 23:02:18.969222617 +0100
28906 @@ -0,0 +1,170 @@
28907 +// SPDX-License-Identifier: GPL-2.0
28908 +/*
28909 + * Copyright (C) 2010-2022 Junjiro R. Okajima
28910 + *
28911 + * This program is free software; you can redistribute it and/or modify
28912 + * it under the terms of the GNU General Public License as published by
28913 + * the Free Software Foundation; either version 2 of the License, or
28914 + * (at your option) any later version.
28915 + *
28916 + * This program is distributed in the hope that it will be useful,
28917 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28918 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28919 + * GNU General Public License for more details.
28920 + *
28921 + * You should have received a copy of the GNU General Public License
28922 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28923 + */
28924 +
28925 +/*
28926 + * procfs interfaces
28927 + */
28928 +
28929 +#include <linux/proc_fs.h>
28930 +#include "aufs.h"
28931 +
28932 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
28933 +{
28934 +       struct au_sbinfo *sbinfo;
28935 +
28936 +       sbinfo = file->private_data;
28937 +       if (sbinfo) {
28938 +               au_plink_maint_leave(sbinfo);
28939 +               kobject_put(&sbinfo->si_kobj);
28940 +       }
28941 +
28942 +       return 0;
28943 +}
28944 +
28945 +static void au_procfs_plm_write_clean(struct file *file)
28946 +{
28947 +       struct au_sbinfo *sbinfo;
28948 +
28949 +       sbinfo = file->private_data;
28950 +       if (sbinfo)
28951 +               au_plink_clean(sbinfo->si_sb, /*verbose*/0);
28952 +}
28953 +
28954 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
28955 +{
28956 +       int err;
28957 +       struct super_block *sb;
28958 +       struct au_sbinfo *sbinfo;
28959 +       struct hlist_bl_node *pos;
28960 +
28961 +       err = -EBUSY;
28962 +       if (unlikely(file->private_data))
28963 +               goto out;
28964 +
28965 +       sb = NULL;
28966 +       /* don't use au_sbilist_lock() here */
28967 +       hlist_bl_lock(&au_sbilist);
28968 +       hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
28969 +               if (id == sysaufs_si_id(sbinfo)) {
28970 +                       if (kobject_get_unless_zero(&sbinfo->si_kobj))
28971 +                               sb = sbinfo->si_sb;
28972 +                       break;
28973 +               }
28974 +       hlist_bl_unlock(&au_sbilist);
28975 +
28976 +       err = -EINVAL;
28977 +       if (unlikely(!sb))
28978 +               goto out;
28979 +
28980 +       err = au_plink_maint_enter(sb);
28981 +       if (!err)
28982 +               /* keep kobject_get() */
28983 +               file->private_data = sbinfo;
28984 +       else
28985 +               kobject_put(&sbinfo->si_kobj);
28986 +out:
28987 +       return err;
28988 +}
28989 +
28990 +/*
28991 + * Accept a valid "si=xxxx" only.
28992 + * Once it is accepted successfully, accept "clean" too.
28993 + */
28994 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
28995 +                                  size_t count, loff_t *ppos)
28996 +{
28997 +       ssize_t err;
28998 +       unsigned long id;
28999 +       /* last newline is allowed */
29000 +       char buf[3 + sizeof(unsigned long) * 2 + 1];
29001 +
29002 +       err = -EACCES;
29003 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
29004 +               goto out;
29005 +
29006 +       err = -EINVAL;
29007 +       if (unlikely(count > sizeof(buf)))
29008 +               goto out;
29009 +
29010 +       err = copy_from_user(buf, ubuf, count);
29011 +       if (unlikely(err)) {
29012 +               err = -EFAULT;
29013 +               goto out;
29014 +       }
29015 +       buf[count] = 0;
29016 +
29017 +       err = -EINVAL;
29018 +       if (!strcmp("clean", buf)) {
29019 +               au_procfs_plm_write_clean(file);
29020 +               goto out_success;
29021 +       } else if (unlikely(strncmp("si=", buf, 3)))
29022 +               goto out;
29023 +
29024 +       err = kstrtoul(buf + 3, 16, &id);
29025 +       if (unlikely(err))
29026 +               goto out;
29027 +
29028 +       err = au_procfs_plm_write_si(file, id);
29029 +       if (unlikely(err))
29030 +               goto out;
29031 +
29032 +out_success:
29033 +       err = count; /* success */
29034 +out:
29035 +       return err;
29036 +}
29037 +
29038 +static const struct proc_ops au_procfs_plm_op = {
29039 +       .proc_write     = au_procfs_plm_write,
29040 +       .proc_release   = au_procfs_plm_release
29041 +};
29042 +
29043 +/* ---------------------------------------------------------------------- */
29044 +
29045 +static struct proc_dir_entry *au_procfs_dir;
29046 +
29047 +void au_procfs_fin(void)
29048 +{
29049 +       remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
29050 +       remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29051 +}
29052 +
29053 +int __init au_procfs_init(void)
29054 +{
29055 +       int err;
29056 +       struct proc_dir_entry *entry;
29057 +
29058 +       err = -ENOMEM;
29059 +       au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
29060 +       if (unlikely(!au_procfs_dir))
29061 +               goto out;
29062 +
29063 +       entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | 0200,
29064 +                           au_procfs_dir, &au_procfs_plm_op);
29065 +       if (unlikely(!entry))
29066 +               goto out_dir;
29067 +
29068 +       err = 0;
29069 +       goto out; /* success */
29070 +
29071 +
29072 +out_dir:
29073 +       remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29074 +out:
29075 +       return err;
29076 +}
29077 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
29078 --- /usr/share/empty/fs/aufs/rdu.c      1970-01-01 01:00:00.000000000 +0100
29079 +++ linux/fs/aufs/rdu.c 2022-12-17 09:21:34.799855195 +0100
29080 @@ -0,0 +1,384 @@
29081 +// SPDX-License-Identifier: GPL-2.0
29082 +/*
29083 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29084 + *
29085 + * This program is free software; you can redistribute it and/or modify
29086 + * it under the terms of the GNU General Public License as published by
29087 + * the Free Software Foundation; either version 2 of the License, or
29088 + * (at your option) any later version.
29089 + *
29090 + * This program is distributed in the hope that it will be useful,
29091 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29092 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29093 + * GNU General Public License for more details.
29094 + *
29095 + * You should have received a copy of the GNU General Public License
29096 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29097 + */
29098 +
29099 +/*
29100 + * readdir in userspace.
29101 + */
29102 +
29103 +#include <linux/compat.h>
29104 +#include <linux/fs_stack.h>
29105 +#include <linux/security.h>
29106 +#include "aufs.h"
29107 +
29108 +/* bits for struct aufs_rdu.flags */
29109 +#define        AuRdu_CALLED    1
29110 +#define        AuRdu_CONT      (1 << 1)
29111 +#define        AuRdu_FULL      (1 << 2)
29112 +#define au_ftest_rdu(flags, name)      ((flags) & AuRdu_##name)
29113 +#define au_fset_rdu(flags, name) \
29114 +       do { (flags) |= AuRdu_##name; } while (0)
29115 +#define au_fclr_rdu(flags, name) \
29116 +       do { (flags) &= ~AuRdu_##name; } while (0)
29117 +
29118 +struct au_rdu_arg {
29119 +       struct dir_context              ctx;
29120 +       struct aufs_rdu                 *rdu;
29121 +       union au_rdu_ent_ul             ent;
29122 +       unsigned long                   end;
29123 +
29124 +       struct super_block              *sb;
29125 +       int                             err;
29126 +};
29127 +
29128 +static bool au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
29129 +                       loff_t offset, u64 h_ino, unsigned int d_type)
29130 +{
29131 +       int err, len;
29132 +       struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
29133 +       struct aufs_rdu *rdu = arg->rdu;
29134 +       struct au_rdu_ent ent;
29135 +
29136 +       err = 0;
29137 +       arg->err = 0;
29138 +       au_fset_rdu(rdu->cookie.flags, CALLED);
29139 +       len = au_rdu_len(nlen);
29140 +       if (arg->ent.ul + len  < arg->end) {
29141 +               ent.ino = h_ino;
29142 +               ent.bindex = rdu->cookie.bindex;
29143 +               ent.type = d_type;
29144 +               ent.nlen = nlen;
29145 +               if (unlikely(nlen > AUFS_MAX_NAMELEN))
29146 +                       ent.type = DT_UNKNOWN;
29147 +
29148 +               /* unnecessary to support mmap_sem since this is a dir */
29149 +               err = -EFAULT;
29150 +               if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
29151 +                       goto out;
29152 +               if (copy_to_user(arg->ent.e->name, name, nlen))
29153 +                       goto out;
29154 +               /* the terminating NULL */
29155 +               if (__put_user(0, arg->ent.e->name + nlen))
29156 +                       goto out;
29157 +               err = 0;
29158 +               /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
29159 +               arg->ent.ul += len;
29160 +               rdu->rent++;
29161 +       } else {
29162 +               err = -EFAULT;
29163 +               au_fset_rdu(rdu->cookie.flags, FULL);
29164 +               rdu->full = 1;
29165 +               rdu->tail = arg->ent;
29166 +       }
29167 +
29168 +out:
29169 +       /* AuTraceErr(err); */
29170 +       return !err;
29171 +}
29172 +
29173 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
29174 +{
29175 +       int err;
29176 +       loff_t offset;
29177 +       struct au_rdu_cookie *cookie = &arg->rdu->cookie;
29178 +
29179 +       /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
29180 +       offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
29181 +       err = offset;
29182 +       if (unlikely(offset != cookie->h_pos))
29183 +               goto out;
29184 +
29185 +       err = 0;
29186 +       do {
29187 +               arg->err = 0;
29188 +               au_fclr_rdu(cookie->flags, CALLED);
29189 +               /* smp_mb(); */
29190 +               err = vfsub_iterate_dir(h_file, &arg->ctx);
29191 +               if (err >= 0)
29192 +                       err = arg->err;
29193 +       } while (!err
29194 +                && au_ftest_rdu(cookie->flags, CALLED)
29195 +                && !au_ftest_rdu(cookie->flags, FULL));
29196 +       cookie->h_pos = h_file->f_pos;
29197 +
29198 +out:
29199 +       AuTraceErr(err);
29200 +       return err;
29201 +}
29202 +
29203 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
29204 +{
29205 +       int err;
29206 +       aufs_bindex_t bbot;
29207 +       struct au_rdu_arg arg = {
29208 +               .ctx = {
29209 +                       .actor = au_rdu_fill
29210 +               }
29211 +       };
29212 +       struct dentry *dentry;
29213 +       struct inode *inode;
29214 +       struct file *h_file;
29215 +       struct au_rdu_cookie *cookie = &rdu->cookie;
29216 +
29217 +       /* VERIFY_WRITE */
29218 +       err = !access_ok(rdu->ent.e, rdu->sz);
29219 +       if (unlikely(err)) {
29220 +               err = -EFAULT;
29221 +               AuTraceErr(err);
29222 +               goto out;
29223 +       }
29224 +       rdu->rent = 0;
29225 +       rdu->tail = rdu->ent;
29226 +       rdu->full = 0;
29227 +       arg.rdu = rdu;
29228 +       arg.ent = rdu->ent;
29229 +       arg.end = arg.ent.ul;
29230 +       arg.end += rdu->sz;
29231 +
29232 +       err = -ENOTDIR;
29233 +       if (unlikely(!file->f_op->iterate && !file->f_op->iterate_shared))
29234 +               goto out;
29235 +
29236 +       err = security_file_permission(file, MAY_READ);
29237 +       AuTraceErr(err);
29238 +       if (unlikely(err))
29239 +               goto out;
29240 +
29241 +       dentry = file->f_path.dentry;
29242 +       inode = d_inode(dentry);
29243 +       inode_lock_shared(inode);
29244 +
29245 +       arg.sb = inode->i_sb;
29246 +       err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
29247 +       if (unlikely(err))
29248 +               goto out_mtx;
29249 +       err = au_alive_dir(dentry);
29250 +       if (unlikely(err))
29251 +               goto out_si;
29252 +       /* todo: reval? */
29253 +       fi_read_lock(file);
29254 +
29255 +       err = -EAGAIN;
29256 +       if (unlikely(au_ftest_rdu(cookie->flags, CONT)
29257 +                    && cookie->generation != au_figen(file)))
29258 +               goto out_unlock;
29259 +
29260 +       err = 0;
29261 +       if (!rdu->blk) {
29262 +               rdu->blk = au_sbi(arg.sb)->si_rdblk;
29263 +               if (!rdu->blk)
29264 +                       rdu->blk = au_dir_size(file, /*dentry*/NULL);
29265 +       }
29266 +       bbot = au_fbtop(file);
29267 +       if (cookie->bindex < bbot)
29268 +               cookie->bindex = bbot;
29269 +       bbot = au_fbbot_dir(file);
29270 +       /* AuDbg("b%d, b%d\n", cookie->bindex, bbot); */
29271 +       for (; !err && cookie->bindex <= bbot;
29272 +            cookie->bindex++, cookie->h_pos = 0) {
29273 +               h_file = au_hf_dir(file, cookie->bindex);
29274 +               if (!h_file)
29275 +                       continue;
29276 +
29277 +               au_fclr_rdu(cookie->flags, FULL);
29278 +               err = au_rdu_do(h_file, &arg);
29279 +               AuTraceErr(err);
29280 +               if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
29281 +                       break;
29282 +       }
29283 +       AuDbg("rent %llu\n", rdu->rent);
29284 +
29285 +       if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
29286 +               rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
29287 +               au_fset_rdu(cookie->flags, CONT);
29288 +               cookie->generation = au_figen(file);
29289 +       }
29290 +
29291 +       ii_read_lock_child(inode);
29292 +       fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibtop(inode)));
29293 +       ii_read_unlock(inode);
29294 +
29295 +out_unlock:
29296 +       fi_read_unlock(file);
29297 +out_si:
29298 +       si_read_unlock(arg.sb);
29299 +out_mtx:
29300 +       inode_unlock_shared(inode);
29301 +out:
29302 +       AuTraceErr(err);
29303 +       return err;
29304 +}
29305 +
29306 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
29307 +{
29308 +       int err;
29309 +       ino_t ino;
29310 +       unsigned long long nent;
29311 +       union au_rdu_ent_ul *u;
29312 +       struct au_rdu_ent ent;
29313 +       struct super_block *sb;
29314 +
29315 +       err = 0;
29316 +       nent = rdu->nent;
29317 +       u = &rdu->ent;
29318 +       sb = file->f_path.dentry->d_sb;
29319 +       si_read_lock(sb, AuLock_FLUSH);
29320 +       while (nent-- > 0) {
29321 +               /* unnecessary to support mmap_sem since this is a dir */
29322 +               err = copy_from_user(&ent, u->e, sizeof(ent));
29323 +               if (!err)
29324 +                       /* VERIFY_WRITE */
29325 +                       err = !access_ok(&u->e->ino, sizeof(ino));
29326 +               if (unlikely(err)) {
29327 +                       err = -EFAULT;
29328 +                       AuTraceErr(err);
29329 +                       break;
29330 +               }
29331 +
29332 +               /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
29333 +               if (!ent.wh)
29334 +                       err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
29335 +               else
29336 +                       err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
29337 +                                       &ino);
29338 +               if (unlikely(err)) {
29339 +                       AuTraceErr(err);
29340 +                       break;
29341 +               }
29342 +
29343 +               err = __put_user(ino, &u->e->ino);
29344 +               if (unlikely(err)) {
29345 +                       err = -EFAULT;
29346 +                       AuTraceErr(err);
29347 +                       break;
29348 +               }
29349 +               u->ul += au_rdu_len(ent.nlen);
29350 +       }
29351 +       si_read_unlock(sb);
29352 +
29353 +       return err;
29354 +}
29355 +
29356 +/* ---------------------------------------------------------------------- */
29357 +
29358 +static int au_rdu_verify(struct aufs_rdu *rdu)
29359 +{
29360 +       AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
29361 +             "%llu, b%d, 0x%x, g%u}\n",
29362 +             rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
29363 +             rdu->blk,
29364 +             rdu->rent, rdu->shwh, rdu->full,
29365 +             rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
29366 +             rdu->cookie.generation);
29367 +
29368 +       if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
29369 +               return 0;
29370 +
29371 +       AuDbg("%u:%u\n",
29372 +             rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
29373 +       return -EINVAL;
29374 +}
29375 +
29376 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29377 +{
29378 +       long err, e;
29379 +       struct aufs_rdu rdu;
29380 +       void __user *p = (void __user *)arg;
29381 +
29382 +       err = copy_from_user(&rdu, p, sizeof(rdu));
29383 +       if (unlikely(err)) {
29384 +               err = -EFAULT;
29385 +               AuTraceErr(err);
29386 +               goto out;
29387 +       }
29388 +       err = au_rdu_verify(&rdu);
29389 +       if (unlikely(err))
29390 +               goto out;
29391 +
29392 +       switch (cmd) {
29393 +       case AUFS_CTL_RDU:
29394 +               err = au_rdu(file, &rdu);
29395 +               if (unlikely(err))
29396 +                       break;
29397 +
29398 +               e = copy_to_user(p, &rdu, sizeof(rdu));
29399 +               if (unlikely(e)) {
29400 +                       err = -EFAULT;
29401 +                       AuTraceErr(err);
29402 +               }
29403 +               break;
29404 +       case AUFS_CTL_RDU_INO:
29405 +               err = au_rdu_ino(file, &rdu);
29406 +               break;
29407 +
29408 +       default:
29409 +               /* err = -ENOTTY; */
29410 +               err = -EINVAL;
29411 +       }
29412 +
29413 +out:
29414 +       AuTraceErr(err);
29415 +       return err;
29416 +}
29417 +
29418 +#ifdef CONFIG_COMPAT
29419 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29420 +{
29421 +       long err, e;
29422 +       struct aufs_rdu rdu;
29423 +       void __user *p = compat_ptr(arg);
29424 +
29425 +       /* todo: get_user()? */
29426 +       err = copy_from_user(&rdu, p, sizeof(rdu));
29427 +       if (unlikely(err)) {
29428 +               err = -EFAULT;
29429 +               AuTraceErr(err);
29430 +               goto out;
29431 +       }
29432 +       rdu.ent.e = compat_ptr(rdu.ent.ul);
29433 +       err = au_rdu_verify(&rdu);
29434 +       if (unlikely(err))
29435 +               goto out;
29436 +
29437 +       switch (cmd) {
29438 +       case AUFS_CTL_RDU:
29439 +               err = au_rdu(file, &rdu);
29440 +               if (unlikely(err))
29441 +                       break;
29442 +
29443 +               rdu.ent.ul = ptr_to_compat(rdu.ent.e);
29444 +               rdu.tail.ul = ptr_to_compat(rdu.tail.e);
29445 +               e = copy_to_user(p, &rdu, sizeof(rdu));
29446 +               if (unlikely(e)) {
29447 +                       err = -EFAULT;
29448 +                       AuTraceErr(err);
29449 +               }
29450 +               break;
29451 +       case AUFS_CTL_RDU_INO:
29452 +               err = au_rdu_ino(file, &rdu);
29453 +               break;
29454 +
29455 +       default:
29456 +               /* err = -ENOTTY; */
29457 +               err = -EINVAL;
29458 +       }
29459 +
29460 +out:
29461 +       AuTraceErr(err);
29462 +       return err;
29463 +}
29464 +#endif
29465 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
29466 --- /usr/share/empty/fs/aufs/rwsem.h    1970-01-01 01:00:00.000000000 +0100
29467 +++ linux/fs/aufs/rwsem.h       2022-11-05 23:02:18.969222617 +0100
29468 @@ -0,0 +1,85 @@
29469 +/* SPDX-License-Identifier: GPL-2.0 */
29470 +/*
29471 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29472 + *
29473 + * This program is free software; you can redistribute it and/or modify
29474 + * it under the terms of the GNU General Public License as published by
29475 + * the Free Software Foundation; either version 2 of the License, or
29476 + * (at your option) any later version.
29477 + *
29478 + * This program is distributed in the hope that it will be useful,
29479 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29480 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29481 + * GNU General Public License for more details.
29482 + *
29483 + * You should have received a copy of the GNU General Public License
29484 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29485 + */
29486 +
29487 +/*
29488 + * simple read-write semaphore wrappers
29489 + */
29490 +
29491 +#ifndef __AUFS_RWSEM_H__
29492 +#define __AUFS_RWSEM_H__
29493 +
29494 +#ifdef __KERNEL__
29495 +
29496 +#include "debug.h"
29497 +
29498 +/* in the future, the name 'au_rwsem' will be totally gone */
29499 +#define au_rwsem       rw_semaphore
29500 +
29501 +/* to debug easier, do not make them inlined functions */
29502 +#define AuRwMustNoWaiters(rw)  AuDebugOn(rwsem_is_contended(rw))
29503 +
29504 +#ifdef CONFIG_LOCKDEP
29505 +/* rwsem_is_locked() is unusable */
29506 +#define AuRwMustReadLock(rw)   AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29507 +                                         && !lockdep_recursing(current) \
29508 +                                         && debug_locks                \
29509 +                                         && !lockdep_is_held_type(rw, 1))
29510 +#define AuRwMustWriteLock(rw)  AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29511 +                                         && !lockdep_recursing(current) \
29512 +                                         && debug_locks                \
29513 +                                         && !lockdep_is_held_type(rw, 0))
29514 +#define AuRwMustAnyLock(rw)    AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29515 +                                         && !lockdep_recursing(current) \
29516 +                                         && debug_locks                \
29517 +                                         && !lockdep_is_held(rw))
29518 +#define AuRwDestroy(rw)                AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29519 +                                         && !lockdep_recursing(current) \
29520 +                                         && debug_locks                \
29521 +                                         && lockdep_is_held(rw))
29522 +#else
29523 +#define AuRwMustReadLock(rw)   do {} while (0)
29524 +#define AuRwMustWriteLock(rw)  do {} while (0)
29525 +#define AuRwMustAnyLock(rw)    do {} while (0)
29526 +#define AuRwDestroy(rw)                do {} while (0)
29527 +#endif
29528 +
29529 +#define au_rw_init(rw) init_rwsem(rw)
29530 +
29531 +#define au_rw_init_wlock(rw) do {              \
29532 +               au_rw_init(rw);                 \
29533 +               down_write(rw);                 \
29534 +       } while (0)
29535 +
29536 +#define au_rw_init_wlock_nested(rw, lsc) do {  \
29537 +               au_rw_init(rw);                 \
29538 +               down_write_nested(rw, lsc);     \
29539 +       } while (0)
29540 +
29541 +#define au_rw_read_lock(rw)            down_read(rw)
29542 +#define au_rw_read_lock_nested(rw, lsc)        down_read_nested(rw, lsc)
29543 +#define au_rw_read_unlock(rw)          up_read(rw)
29544 +#define au_rw_dgrade_lock(rw)          downgrade_write(rw)
29545 +#define au_rw_write_lock(rw)           down_write(rw)
29546 +#define au_rw_write_lock_nested(rw, lsc) down_write_nested(rw, lsc)
29547 +#define au_rw_write_unlock(rw)         up_write(rw)
29548 +/* why is not _nested version defined? */
29549 +#define au_rw_read_trylock(rw)         down_read_trylock(rw)
29550 +#define au_rw_write_trylock(rw)                down_write_trylock(rw)
29551 +
29552 +#endif /* __KERNEL__ */
29553 +#endif /* __AUFS_RWSEM_H__ */
29554 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
29555 --- /usr/share/empty/fs/aufs/sbinfo.c   1970-01-01 01:00:00.000000000 +0100
29556 +++ linux/fs/aufs/sbinfo.c      2022-11-05 23:02:18.969222617 +0100
29557 @@ -0,0 +1,316 @@
29558 +// SPDX-License-Identifier: GPL-2.0
29559 +/*
29560 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29561 + *
29562 + * This program is free software; you can redistribute it and/or modify
29563 + * it under the terms of the GNU General Public License as published by
29564 + * the Free Software Foundation; either version 2 of the License, or
29565 + * (at your option) any later version.
29566 + *
29567 + * This program is distributed in the hope that it will be useful,
29568 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29569 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29570 + * GNU General Public License for more details.
29571 + *
29572 + * You should have received a copy of the GNU General Public License
29573 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29574 + */
29575 +
29576 +/*
29577 + * superblock private data
29578 + */
29579 +
29580 +#include <linux/iversion.h>
29581 +#include "aufs.h"
29582 +
29583 +/*
29584 + * they are necessary regardless sysfs is disabled.
29585 + */
29586 +void au_si_free(struct kobject *kobj)
29587 +{
29588 +       int i;
29589 +       struct au_sbinfo *sbinfo;
29590 +       char *locked __maybe_unused; /* debug only */
29591 +
29592 +       sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
29593 +       for (i = 0; i < AuPlink_NHASH; i++)
29594 +               AuDebugOn(!hlist_bl_empty(sbinfo->si_plink + i));
29595 +       AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
29596 +
29597 +       AuLCntZero(au_lcnt_read(&sbinfo->si_ninodes, /*do_rev*/0));
29598 +       au_lcnt_fin(&sbinfo->si_ninodes, /*do_sync*/0);
29599 +       AuLCntZero(au_lcnt_read(&sbinfo->si_nfiles, /*do_rev*/0));
29600 +       au_lcnt_fin(&sbinfo->si_nfiles, /*do_sync*/0);
29601 +
29602 +       dbgaufs_si_fin(sbinfo);
29603 +       au_rw_write_lock(&sbinfo->si_rwsem);
29604 +       au_br_free(sbinfo);
29605 +       au_rw_write_unlock(&sbinfo->si_rwsem);
29606 +
29607 +       au_kfree_try_rcu(sbinfo->si_branch);
29608 +       mutex_destroy(&sbinfo->si_xib_mtx);
29609 +       AuRwDestroy(&sbinfo->si_rwsem);
29610 +
29611 +       au_lcnt_wait_for_fin(&sbinfo->si_ninodes);
29612 +       /* si_nfiles is waited too */
29613 +       au_kfree_rcu(sbinfo);
29614 +}
29615 +
29616 +struct au_sbinfo *au_si_alloc(struct super_block *sb)
29617 +{
29618 +       struct au_sbinfo *sbinfo;
29619 +       int err, i;
29620 +
29621 +       err = -ENOMEM;
29622 +       sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
29623 +       if (unlikely(!sbinfo))
29624 +               goto out;
29625 +
29626 +       /* will be reallocated separately */
29627 +       sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
29628 +       if (unlikely(!sbinfo->si_branch))
29629 +               goto out_sbinfo;
29630 +
29631 +       err = sysaufs_si_init(sbinfo);
29632 +       if (!err) {
29633 +               dbgaufs_si_null(sbinfo);
29634 +               err = dbgaufs_si_init(sbinfo);
29635 +               if (unlikely(err))
29636 +                       kobject_put(&sbinfo->si_kobj);
29637 +       }
29638 +       if (unlikely(err))
29639 +               goto out_br;
29640 +
29641 +       au_nwt_init(&sbinfo->si_nowait);
29642 +       au_rw_init_wlock(&sbinfo->si_rwsem);
29643 +
29644 +       au_lcnt_init(&sbinfo->si_ninodes, /*release*/NULL);
29645 +       au_lcnt_init(&sbinfo->si_nfiles, /*release*/NULL);
29646 +
29647 +       sbinfo->si_bbot = -1;
29648 +       sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
29649 +
29650 +       sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
29651 +       sbinfo->si_wbr_create = AuWbrCreate_Def;
29652 +       sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
29653 +       sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
29654 +
29655 +       au_fhsm_init(sbinfo);
29656 +
29657 +       sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
29658 +
29659 +       sbinfo->si_xino_jiffy = jiffies;
29660 +       sbinfo->si_xino_expire
29661 +               = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
29662 +       mutex_init(&sbinfo->si_xib_mtx);
29663 +       /* leave si_xib_last_pindex and si_xib_next_bit */
29664 +
29665 +       INIT_HLIST_BL_HEAD(&sbinfo->si_aopen);
29666 +
29667 +       sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
29668 +       sbinfo->si_rdblk = AUFS_RDBLK_DEF;
29669 +       sbinfo->si_rdhash = AUFS_RDHASH_DEF;
29670 +       sbinfo->si_dirwh = AUFS_DIRWH_DEF;
29671 +
29672 +       for (i = 0; i < AuPlink_NHASH; i++)
29673 +               INIT_HLIST_BL_HEAD(sbinfo->si_plink + i);
29674 +       init_waitqueue_head(&sbinfo->si_plink_wq);
29675 +       spin_lock_init(&sbinfo->si_plink_maint_lock);
29676 +
29677 +       INIT_HLIST_BL_HEAD(&sbinfo->si_files);
29678 +
29679 +       /* with getattr by default */
29680 +       sbinfo->si_iop_array = aufs_iop;
29681 +
29682 +       /* leave other members for sysaufs and si_mnt. */
29683 +       sbinfo->si_sb = sb;
29684 +       if (sb) {
29685 +               sb->s_fs_info = sbinfo;
29686 +               si_pid_set(sb);
29687 +       }
29688 +       return sbinfo; /* success */
29689 +
29690 +out_br:
29691 +       au_kfree_try_rcu(sbinfo->si_branch);
29692 +out_sbinfo:
29693 +       au_kfree_rcu(sbinfo);
29694 +out:
29695 +       return ERR_PTR(err);
29696 +}
29697 +
29698 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink)
29699 +{
29700 +       int err, sz;
29701 +       struct au_branch **brp;
29702 +
29703 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
29704 +
29705 +       err = -ENOMEM;
29706 +       sz = sizeof(*brp) * (sbinfo->si_bbot + 1);
29707 +       if (unlikely(!sz))
29708 +               sz = sizeof(*brp);
29709 +       brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS,
29710 +                          may_shrink);
29711 +       if (brp) {
29712 +               sbinfo->si_branch = brp;
29713 +               err = 0;
29714 +       }
29715 +
29716 +       return err;
29717 +}
29718 +
29719 +/* ---------------------------------------------------------------------- */
29720 +
29721 +unsigned int au_sigen_inc(struct super_block *sb)
29722 +{
29723 +       unsigned int gen;
29724 +       struct inode *inode;
29725 +
29726 +       SiMustWriteLock(sb);
29727 +
29728 +       gen = ++au_sbi(sb)->si_generation;
29729 +       au_update_digen(sb->s_root);
29730 +       inode = d_inode(sb->s_root);
29731 +       au_update_iigen(inode, /*half*/0);
29732 +       inode_inc_iversion(inode);
29733 +       return gen;
29734 +}
29735 +
29736 +aufs_bindex_t au_new_br_id(struct super_block *sb)
29737 +{
29738 +       aufs_bindex_t br_id;
29739 +       int i;
29740 +       struct au_sbinfo *sbinfo;
29741 +
29742 +       SiMustWriteLock(sb);
29743 +
29744 +       sbinfo = au_sbi(sb);
29745 +       for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
29746 +               br_id = ++sbinfo->si_last_br_id;
29747 +               AuDebugOn(br_id < 0);
29748 +               if (br_id && au_br_index(sb, br_id) < 0)
29749 +                       return br_id;
29750 +       }
29751 +
29752 +       return -1;
29753 +}
29754 +
29755 +/* ---------------------------------------------------------------------- */
29756 +
29757 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
29758 +int si_read_lock(struct super_block *sb, int flags)
29759 +{
29760 +       int err;
29761 +
29762 +       err = 0;
29763 +       if (au_ftest_lock(flags, FLUSH))
29764 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
29765 +
29766 +       si_noflush_read_lock(sb);
29767 +       err = au_plink_maint(sb, flags);
29768 +       if (unlikely(err))
29769 +               si_read_unlock(sb);
29770 +
29771 +       return err;
29772 +}
29773 +
29774 +int si_write_lock(struct super_block *sb, int flags)
29775 +{
29776 +       int err;
29777 +
29778 +       if (au_ftest_lock(flags, FLUSH))
29779 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
29780 +
29781 +       si_noflush_write_lock(sb);
29782 +       err = au_plink_maint(sb, flags);
29783 +       if (unlikely(err))
29784 +               si_write_unlock(sb);
29785 +
29786 +       return err;
29787 +}
29788 +
29789 +/* dentry and super_block lock. call at entry point */
29790 +int aufs_read_lock(struct dentry *dentry, int flags)
29791 +{
29792 +       int err;
29793 +       struct super_block *sb;
29794 +
29795 +       sb = dentry->d_sb;
29796 +       err = si_read_lock(sb, flags);
29797 +       if (unlikely(err))
29798 +               goto out;
29799 +
29800 +       if (au_ftest_lock(flags, DW))
29801 +               di_write_lock_child(dentry);
29802 +       else
29803 +               di_read_lock_child(dentry, flags);
29804 +
29805 +       if (au_ftest_lock(flags, GEN)) {
29806 +               err = au_digen_test(dentry, au_sigen(sb));
29807 +               if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
29808 +                       AuDebugOn(!err && au_dbrange_test(dentry));
29809 +               else if (!err)
29810 +                       err = au_dbrange_test(dentry);
29811 +               if (unlikely(err))
29812 +                       aufs_read_unlock(dentry, flags);
29813 +       }
29814 +
29815 +out:
29816 +       return err;
29817 +}
29818 +
29819 +void aufs_read_unlock(struct dentry *dentry, int flags)
29820 +{
29821 +       if (au_ftest_lock(flags, DW))
29822 +               di_write_unlock(dentry);
29823 +       else
29824 +               di_read_unlock(dentry, flags);
29825 +       si_read_unlock(dentry->d_sb);
29826 +}
29827 +
29828 +void aufs_write_lock(struct dentry *dentry)
29829 +{
29830 +       si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
29831 +       di_write_lock_child(dentry);
29832 +}
29833 +
29834 +void aufs_write_unlock(struct dentry *dentry)
29835 +{
29836 +       di_write_unlock(dentry);
29837 +       si_write_unlock(dentry->d_sb);
29838 +}
29839 +
29840 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
29841 +{
29842 +       int err;
29843 +       unsigned int sigen;
29844 +       struct super_block *sb;
29845 +
29846 +       sb = d1->d_sb;
29847 +       err = si_read_lock(sb, flags);
29848 +       if (unlikely(err))
29849 +               goto out;
29850 +
29851 +       di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
29852 +
29853 +       if (au_ftest_lock(flags, GEN)) {
29854 +               sigen = au_sigen(sb);
29855 +               err = au_digen_test(d1, sigen);
29856 +               AuDebugOn(!err && au_dbrange_test(d1));
29857 +               if (!err) {
29858 +                       err = au_digen_test(d2, sigen);
29859 +                       AuDebugOn(!err && au_dbrange_test(d2));
29860 +               }
29861 +               if (unlikely(err))
29862 +                       aufs_read_and_write_unlock2(d1, d2);
29863 +       }
29864 +
29865 +out:
29866 +       return err;
29867 +}
29868 +
29869 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
29870 +{
29871 +       di_write_unlock2(d1, d2);
29872 +       si_read_unlock(d1->d_sb);
29873 +}
29874 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
29875 --- /usr/share/empty/fs/aufs/super.c    1970-01-01 01:00:00.000000000 +0100
29876 +++ linux/fs/aufs/super.c       2022-11-05 23:02:18.969222617 +0100
29877 @@ -0,0 +1,871 @@
29878 +// SPDX-License-Identifier: GPL-2.0
29879 +/*
29880 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29881 + *
29882 + * This program is free software; you can redistribute it and/or modify
29883 + * it under the terms of the GNU General Public License as published by
29884 + * the Free Software Foundation; either version 2 of the License, or
29885 + * (at your option) any later version.
29886 + *
29887 + * This program is distributed in the hope that it will be useful,
29888 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29889 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29890 + * GNU General Public License for more details.
29891 + *
29892 + * You should have received a copy of the GNU General Public License
29893 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29894 + */
29895 +
29896 +/*
29897 + * mount and super_block operations
29898 + */
29899 +
29900 +#include <linux/iversion.h>
29901 +#include <linux/mm.h>
29902 +#include <linux/seq_file.h>
29903 +#include <linux/statfs.h>
29904 +#include <linux/vmalloc.h>
29905 +#include "aufs.h"
29906 +
29907 +/*
29908 + * super_operations
29909 + */
29910 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
29911 +{
29912 +       struct au_icntnr *c;
29913 +
29914 +       c = au_cache_alloc_icntnr(sb);
29915 +       if (c) {
29916 +               au_icntnr_init(c);
29917 +               inode_set_iversion(&c->vfs_inode, 1); /* sigen(sb); */
29918 +               c->iinfo.ii_hinode = NULL;
29919 +               return &c->vfs_inode;
29920 +       }
29921 +       return NULL;
29922 +}
29923 +
29924 +static void aufs_destroy_inode(struct inode *inode)
29925 +{
29926 +       if (!au_is_bad_inode(inode))
29927 +               au_iinfo_fin(inode);
29928 +}
29929 +
29930 +static void aufs_free_inode(struct inode *inode)
29931 +{
29932 +       au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
29933 +}
29934 +
29935 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
29936 +{
29937 +       struct inode *inode;
29938 +       int err;
29939 +
29940 +       inode = iget_locked(sb, ino);
29941 +       if (unlikely(!inode)) {
29942 +               inode = ERR_PTR(-ENOMEM);
29943 +               goto out;
29944 +       }
29945 +       if (!(inode->i_state & I_NEW))
29946 +               goto out;
29947 +
29948 +       err = au_xigen_new(inode);
29949 +       if (!err)
29950 +               err = au_iinfo_init(inode);
29951 +       if (!err)
29952 +               inode_inc_iversion(inode);
29953 +       else {
29954 +               iget_failed(inode);
29955 +               inode = ERR_PTR(err);
29956 +       }
29957 +
29958 +out:
29959 +       /* never return NULL */
29960 +       AuDebugOn(!inode);
29961 +       AuTraceErrPtr(inode);
29962 +       return inode;
29963 +}
29964 +
29965 +/* lock free root dinfo */
29966 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
29967 +{
29968 +       int err;
29969 +       aufs_bindex_t bindex, bbot;
29970 +       struct path path;
29971 +       struct au_hdentry *hdp;
29972 +       struct au_branch *br;
29973 +       au_br_perm_str_t perm;
29974 +
29975 +       err = 0;
29976 +       bbot = au_sbbot(sb);
29977 +       bindex = 0;
29978 +       hdp = au_hdentry(au_di(sb->s_root), bindex);
29979 +       for (; !err && bindex <= bbot; bindex++, hdp++) {
29980 +               br = au_sbr(sb, bindex);
29981 +               path.mnt = au_br_mnt(br);
29982 +               path.dentry = hdp->hd_dentry;
29983 +               err = au_seq_path(seq, &path);
29984 +               if (!err) {
29985 +                       au_optstr_br_perm(&perm, br->br_perm);
29986 +                       seq_printf(seq, "=%s", perm.a);
29987 +                       if (bindex != bbot)
29988 +                               seq_putc(seq, ':');
29989 +               }
29990 +       }
29991 +       if (unlikely(err || seq_has_overflowed(seq)))
29992 +               err = -E2BIG;
29993 +
29994 +       return err;
29995 +}
29996 +
29997 +static void au_gen_fmt(char *fmt, int len __maybe_unused, const char *pat,
29998 +                      const char *append)
29999 +{
30000 +       char *p;
30001 +
30002 +       p = fmt;
30003 +       while (*pat != ':')
30004 +               *p++ = *pat++;
30005 +       *p++ = *pat++;
30006 +       strcpy(p, append);
30007 +       AuDebugOn(strlen(fmt) >= len);
30008 +}
30009 +
30010 +static void au_show_wbr_create(struct seq_file *m, int v,
30011 +                              struct au_sbinfo *sbinfo)
30012 +{
30013 +       const char *pat;
30014 +       char fmt[32];
30015 +       struct au_wbr_mfs *mfs;
30016 +
30017 +       AuRwMustAnyLock(&sbinfo->si_rwsem);
30018 +
30019 +       seq_puts(m, ",create=");
30020 +       pat = au_optstr_wbr_create(v);
30021 +       mfs = &sbinfo->si_wbr_mfs;
30022 +       switch (v) {
30023 +       case AuWbrCreate_TDP:
30024 +       case AuWbrCreate_RR:
30025 +       case AuWbrCreate_MFS:
30026 +       case AuWbrCreate_PMFS:
30027 +               seq_puts(m, pat);
30028 +               break;
30029 +       case AuWbrCreate_MFSRR:
30030 +       case AuWbrCreate_TDMFS:
30031 +       case AuWbrCreate_PMFSRR:
30032 +               au_gen_fmt(fmt, sizeof(fmt), pat, "%llu");
30033 +               seq_printf(m, fmt, mfs->mfsrr_watermark);
30034 +               break;
30035 +       case AuWbrCreate_MFSV:
30036 +       case AuWbrCreate_PMFSV:
30037 +               au_gen_fmt(fmt, sizeof(fmt), pat, "%lu");
30038 +               seq_printf(m, fmt,
30039 +                          jiffies_to_msecs(mfs->mfs_expire)
30040 +                          / MSEC_PER_SEC);
30041 +               break;
30042 +       case AuWbrCreate_MFSRRV:
30043 +       case AuWbrCreate_TDMFSV:
30044 +       case AuWbrCreate_PMFSRRV:
30045 +               au_gen_fmt(fmt, sizeof(fmt), pat, "%llu:%lu");
30046 +               seq_printf(m, fmt, mfs->mfsrr_watermark,
30047 +                          jiffies_to_msecs(mfs->mfs_expire) / MSEC_PER_SEC);
30048 +               break;
30049 +       default:
30050 +               BUG();
30051 +       }
30052 +}
30053 +
30054 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
30055 +{
30056 +#ifdef CONFIG_SYSFS
30057 +       return 0;
30058 +#else
30059 +       int err;
30060 +       const int len = sizeof(AUFS_XINO_FNAME) - 1;
30061 +       aufs_bindex_t bindex, brid;
30062 +       struct qstr *name;
30063 +       struct file *f;
30064 +       struct dentry *d, *h_root;
30065 +       struct au_branch *br;
30066 +
30067 +       AuRwMustAnyLock(&sbinfo->si_rwsem);
30068 +
30069 +       err = 0;
30070 +       f = au_sbi(sb)->si_xib;
30071 +       if (!f)
30072 +               goto out;
30073 +
30074 +       /* stop printing the default xino path on the first writable branch */
30075 +       h_root = NULL;
30076 +       bindex = au_xi_root(sb, f->f_path.dentry);
30077 +       if (bindex >= 0) {
30078 +               br = au_sbr_sb(sb, bindex);
30079 +               h_root = au_br_dentry(br);
30080 +       }
30081 +
30082 +       d = f->f_path.dentry;
30083 +       name = &d->d_name;
30084 +       /* safe ->d_parent because the file is unlinked */
30085 +       if (d->d_parent == h_root
30086 +           && name->len == len
30087 +           && !memcmp(name->name, AUFS_XINO_FNAME, len))
30088 +               goto out;
30089 +
30090 +       seq_puts(seq, ",xino=");
30091 +       err = au_xino_path(seq, f);
30092 +
30093 +out:
30094 +       return err;
30095 +#endif
30096 +}
30097 +
30098 +/* seq_file will re-call me in case of too long string */
30099 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
30100 +{
30101 +       int err;
30102 +       unsigned int mnt_flags, v;
30103 +       struct super_block *sb;
30104 +       struct au_sbinfo *sbinfo;
30105 +
30106 +#define AuBool(name, str) do { \
30107 +       v = au_opt_test(mnt_flags, name); \
30108 +       if (v != au_opt_test(AuOpt_Def, name)) \
30109 +               seq_printf(m, ",%s" #str, v ? "" : "no"); \
30110 +} while (0)
30111 +
30112 +#define AuStr(name, str) do { \
30113 +       v = mnt_flags & AuOptMask_##name; \
30114 +       if (v != (AuOpt_Def & AuOptMask_##name)) \
30115 +               seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
30116 +} while (0)
30117 +
30118 +#define AuUInt(name, str, val) do { \
30119 +       if (val != AUFS_##name##_DEF) \
30120 +               seq_printf(m, "," #str "=%u", val); \
30121 +} while (0)
30122 +
30123 +       sb = dentry->d_sb;
30124 +       if (sb->s_flags & SB_POSIXACL)
30125 +               seq_puts(m, ",acl");
30126 +#if 0 /* reserved for future use */
30127 +       if (sb->s_flags & SB_I_VERSION)
30128 +               seq_puts(m, ",i_version");
30129 +#endif
30130 +
30131 +       /* lock free root dinfo */
30132 +       si_noflush_read_lock(sb);
30133 +       sbinfo = au_sbi(sb);
30134 +       seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
30135 +
30136 +       mnt_flags = au_mntflags(sb);
30137 +       if (au_opt_test(mnt_flags, XINO)) {
30138 +               err = au_show_xino(m, sb);
30139 +               if (unlikely(err))
30140 +                       goto out;
30141 +       } else
30142 +               seq_puts(m, ",noxino");
30143 +
30144 +       AuBool(TRUNC_XINO, trunc_xino);
30145 +       AuStr(UDBA, udba);
30146 +       AuBool(SHWH, shwh);
30147 +       AuBool(PLINK, plink);
30148 +       AuBool(DIO, dio);
30149 +       AuBool(DIRPERM1, dirperm1);
30150 +
30151 +       v = sbinfo->si_wbr_create;
30152 +       if (v != AuWbrCreate_Def)
30153 +               au_show_wbr_create(m, v, sbinfo);
30154 +
30155 +       v = sbinfo->si_wbr_copyup;
30156 +       if (v != AuWbrCopyup_Def)
30157 +               seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
30158 +
30159 +       v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
30160 +       if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
30161 +               seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
30162 +
30163 +       AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
30164 +
30165 +       v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
30166 +       AuUInt(RDCACHE, rdcache, v);
30167 +
30168 +       AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
30169 +       AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
30170 +
30171 +       au_fhsm_show(m, sbinfo);
30172 +
30173 +       AuBool(DIRREN, dirren);
30174 +       AuBool(SUM, sum);
30175 +       /* AuBool(SUM_W, wsum); */
30176 +       AuBool(WARN_PERM, warn_perm);
30177 +       AuBool(VERBOSE, verbose);
30178 +
30179 +out:
30180 +       /* be sure to print "br:" last */
30181 +       if (!sysaufs_brs) {
30182 +               seq_puts(m, ",br:");
30183 +               au_show_brs(m, sb);
30184 +       }
30185 +       si_read_unlock(sb);
30186 +       return 0;
30187 +
30188 +#undef AuBool
30189 +#undef AuStr
30190 +#undef AuUInt
30191 +}
30192 +
30193 +/* ---------------------------------------------------------------------- */
30194 +
30195 +/* sum mode which returns the summation for statfs(2) */
30196 +
30197 +static u64 au_add_till_max(u64 a, u64 b)
30198 +{
30199 +       u64 old;
30200 +
30201 +       old = a;
30202 +       a += b;
30203 +       if (old <= a)
30204 +               return a;
30205 +       return ULLONG_MAX;
30206 +}
30207 +
30208 +static u64 au_mul_till_max(u64 a, long mul)
30209 +{
30210 +       u64 old;
30211 +
30212 +       old = a;
30213 +       a *= mul;
30214 +       if (old <= a)
30215 +               return a;
30216 +       return ULLONG_MAX;
30217 +}
30218 +
30219 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
30220 +{
30221 +       int err;
30222 +       long bsize, factor;
30223 +       u64 blocks, bfree, bavail, files, ffree;
30224 +       aufs_bindex_t bbot, bindex, i;
30225 +       unsigned char shared;
30226 +       struct path h_path;
30227 +       struct super_block *h_sb;
30228 +
30229 +       err = 0;
30230 +       bsize = LONG_MAX;
30231 +       files = 0;
30232 +       ffree = 0;
30233 +       blocks = 0;
30234 +       bfree = 0;
30235 +       bavail = 0;
30236 +       bbot = au_sbbot(sb);
30237 +       for (bindex = 0; bindex <= bbot; bindex++) {
30238 +               h_path.mnt = au_sbr_mnt(sb, bindex);
30239 +               h_sb = h_path.mnt->mnt_sb;
30240 +               shared = 0;
30241 +               for (i = 0; !shared && i < bindex; i++)
30242 +                       shared = (au_sbr_sb(sb, i) == h_sb);
30243 +               if (shared)
30244 +                       continue;
30245 +
30246 +               /* sb->s_root for NFS is unreliable */
30247 +               h_path.dentry = h_path.mnt->mnt_root;
30248 +               err = vfs_statfs(&h_path, buf);
30249 +               if (unlikely(err))
30250 +                       goto out;
30251 +
30252 +               if (bsize > buf->f_bsize) {
30253 +                       /*
30254 +                        * we will reduce bsize, so we have to expand blocks
30255 +                        * etc. to match them again
30256 +                        */
30257 +                       factor = (bsize / buf->f_bsize);
30258 +                       blocks = au_mul_till_max(blocks, factor);
30259 +                       bfree = au_mul_till_max(bfree, factor);
30260 +                       bavail = au_mul_till_max(bavail, factor);
30261 +                       bsize = buf->f_bsize;
30262 +               }
30263 +
30264 +               factor = (buf->f_bsize / bsize);
30265 +               blocks = au_add_till_max(blocks,
30266 +                               au_mul_till_max(buf->f_blocks, factor));
30267 +               bfree = au_add_till_max(bfree,
30268 +                               au_mul_till_max(buf->f_bfree, factor));
30269 +               bavail = au_add_till_max(bavail,
30270 +                               au_mul_till_max(buf->f_bavail, factor));
30271 +               files = au_add_till_max(files, buf->f_files);
30272 +               ffree = au_add_till_max(ffree, buf->f_ffree);
30273 +       }
30274 +
30275 +       buf->f_bsize = bsize;
30276 +       buf->f_blocks = blocks;
30277 +       buf->f_bfree = bfree;
30278 +       buf->f_bavail = bavail;
30279 +       buf->f_files = files;
30280 +       buf->f_ffree = ffree;
30281 +       buf->f_frsize = 0;
30282 +
30283 +out:
30284 +       return err;
30285 +}
30286 +
30287 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
30288 +{
30289 +       int err;
30290 +       struct path h_path;
30291 +       struct super_block *sb;
30292 +
30293 +       /* lock free root dinfo */
30294 +       sb = dentry->d_sb;
30295 +       si_noflush_read_lock(sb);
30296 +       if (!au_opt_test(au_mntflags(sb), SUM)) {
30297 +               /* sb->s_root for NFS is unreliable */
30298 +               h_path.mnt = au_sbr_mnt(sb, 0);
30299 +               h_path.dentry = h_path.mnt->mnt_root;
30300 +               err = vfs_statfs(&h_path, buf);
30301 +       } else
30302 +               err = au_statfs_sum(sb, buf);
30303 +       si_read_unlock(sb);
30304 +
30305 +       if (!err) {
30306 +               buf->f_type = AUFS_SUPER_MAGIC;
30307 +               buf->f_namelen = AUFS_MAX_NAMELEN;
30308 +               memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
30309 +       }
30310 +       /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
30311 +
30312 +       return err;
30313 +}
30314 +
30315 +/* ---------------------------------------------------------------------- */
30316 +
30317 +static int aufs_sync_fs(struct super_block *sb, int wait)
30318 +{
30319 +       int err, e;
30320 +       aufs_bindex_t bbot, bindex;
30321 +       struct au_branch *br;
30322 +       struct super_block *h_sb;
30323 +
30324 +       err = 0;
30325 +       si_noflush_read_lock(sb);
30326 +       bbot = au_sbbot(sb);
30327 +       for (bindex = 0; bindex <= bbot; bindex++) {
30328 +               br = au_sbr(sb, bindex);
30329 +               if (!au_br_writable(br->br_perm))
30330 +                       continue;
30331 +
30332 +               h_sb = au_sbr_sb(sb, bindex);
30333 +               e = vfsub_sync_filesystem(h_sb);
30334 +               if (unlikely(e && !err))
30335 +                       err = e;
30336 +               /* go on even if an error happens */
30337 +       }
30338 +       si_read_unlock(sb);
30339 +
30340 +       return err;
30341 +}
30342 +
30343 +/* ---------------------------------------------------------------------- */
30344 +
30345 +/* final actions when unmounting a file system */
30346 +static void aufs_put_super(struct super_block *sb)
30347 +{
30348 +       struct au_sbinfo *sbinfo;
30349 +
30350 +       sbinfo = au_sbi(sb);
30351 +       if (sbinfo)
30352 +               kobject_put(&sbinfo->si_kobj);
30353 +}
30354 +
30355 +/* ---------------------------------------------------------------------- */
30356 +
30357 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30358 +                    struct super_block *sb, void *arg)
30359 +{
30360 +       void *array;
30361 +       unsigned long long n, sz;
30362 +
30363 +       array = NULL;
30364 +       n = 0;
30365 +       if (!*hint)
30366 +               goto out;
30367 +
30368 +       if (*hint > ULLONG_MAX / sizeof(array)) {
30369 +               array = ERR_PTR(-EMFILE);
30370 +               pr_err("hint %llu\n", *hint);
30371 +               goto out;
30372 +       }
30373 +
30374 +       sz = sizeof(array) * *hint;
30375 +       array = kzalloc(sz, GFP_NOFS);
30376 +       if (unlikely(!array))
30377 +               array = vzalloc(sz);
30378 +       if (unlikely(!array)) {
30379 +               array = ERR_PTR(-ENOMEM);
30380 +               goto out;
30381 +       }
30382 +
30383 +       n = cb(sb, array, *hint, arg);
30384 +       AuDebugOn(n > *hint);
30385 +
30386 +out:
30387 +       *hint = n;
30388 +       return array;
30389 +}
30390 +
30391 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
30392 +                                      unsigned long long max __maybe_unused,
30393 +                                      void *arg)
30394 +{
30395 +       unsigned long long n;
30396 +       struct inode **p, *inode;
30397 +       struct list_head *head;
30398 +
30399 +       n = 0;
30400 +       p = a;
30401 +       head = arg;
30402 +       spin_lock(&sb->s_inode_list_lock);
30403 +       list_for_each_entry(inode, head, i_sb_list) {
30404 +               if (!au_is_bad_inode(inode)
30405 +                   && au_ii(inode)->ii_btop >= 0) {
30406 +                       spin_lock(&inode->i_lock);
30407 +                       if (atomic_read(&inode->i_count)) {
30408 +                               au_igrab(inode);
30409 +                               *p++ = inode;
30410 +                               n++;
30411 +                               AuDebugOn(n > max);
30412 +                       }
30413 +                       spin_unlock(&inode->i_lock);
30414 +               }
30415 +       }
30416 +       spin_unlock(&sb->s_inode_list_lock);
30417 +
30418 +       return n;
30419 +}
30420 +
30421 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
30422 +{
30423 +       struct au_sbinfo *sbi;
30424 +
30425 +       sbi = au_sbi(sb);
30426 +       *max = au_lcnt_read(&sbi->si_ninodes, /*do_rev*/1);
30427 +       return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
30428 +}
30429 +
30430 +void au_iarray_free(struct inode **a, unsigned long long max)
30431 +{
30432 +       unsigned long long ull;
30433 +
30434 +       for (ull = 0; ull < max; ull++)
30435 +               iput(a[ull]);
30436 +       kvfree(a);
30437 +}
30438 +
30439 +/* ---------------------------------------------------------------------- */
30440 +
30441 +/*
30442 + * refresh dentry and inode at remount time.
30443 + */
30444 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
30445 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
30446 +                     struct dentry *parent)
30447 +{
30448 +       int err;
30449 +
30450 +       di_write_lock_child(dentry);
30451 +       di_read_lock_parent(parent, AuLock_IR);
30452 +       err = au_refresh_dentry(dentry, parent);
30453 +       if (!err && dir_flags)
30454 +               au_hn_reset(d_inode(dentry), dir_flags);
30455 +       di_read_unlock(parent, AuLock_IR);
30456 +       di_write_unlock(dentry);
30457 +
30458 +       return err;
30459 +}
30460 +
30461 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
30462 +                          struct au_sbinfo *sbinfo,
30463 +                          const unsigned int dir_flags, unsigned int do_idop)
30464 +{
30465 +       int err;
30466 +       struct dentry *parent;
30467 +
30468 +       err = 0;
30469 +       parent = dget_parent(dentry);
30470 +       if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
30471 +               if (d_really_is_positive(dentry)) {
30472 +                       if (!d_is_dir(dentry))
30473 +                               err = au_do_refresh(dentry, /*dir_flags*/0,
30474 +                                                parent);
30475 +                       else {
30476 +                               err = au_do_refresh(dentry, dir_flags, parent);
30477 +                               if (unlikely(err))
30478 +                                       au_fset_si(sbinfo, FAILED_REFRESH_DIR);
30479 +                       }
30480 +               } else
30481 +                       err = au_do_refresh(dentry, /*dir_flags*/0, parent);
30482 +               AuDbgDentry(dentry);
30483 +       }
30484 +       dput(parent);
30485 +
30486 +       if (!err) {
30487 +               if (do_idop)
30488 +                       au_refresh_dop(dentry, /*force_reval*/0);
30489 +       } else
30490 +               au_refresh_dop(dentry, /*force_reval*/1);
30491 +
30492 +       AuTraceErr(err);
30493 +       return err;
30494 +}
30495 +
30496 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
30497 +{
30498 +       int err, i, j, ndentry, e;
30499 +       unsigned int sigen;
30500 +       struct au_dcsub_pages dpages;
30501 +       struct au_dpage *dpage;
30502 +       struct dentry **dentries, *d;
30503 +       struct au_sbinfo *sbinfo;
30504 +       struct dentry *root = sb->s_root;
30505 +       const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
30506 +
30507 +       if (do_idop)
30508 +               au_refresh_dop(root, /*force_reval*/0);
30509 +
30510 +       err = au_dpages_init(&dpages, GFP_NOFS);
30511 +       if (unlikely(err))
30512 +               goto out;
30513 +       err = au_dcsub_pages(&dpages, root, NULL, NULL);
30514 +       if (unlikely(err))
30515 +               goto out_dpages;
30516 +
30517 +       sigen = au_sigen(sb);
30518 +       sbinfo = au_sbi(sb);
30519 +       for (i = 0; i < dpages.ndpage; i++) {
30520 +               dpage = dpages.dpages + i;
30521 +               dentries = dpage->dentries;
30522 +               ndentry = dpage->ndentry;
30523 +               for (j = 0; j < ndentry; j++) {
30524 +                       d = dentries[j];
30525 +                       e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
30526 +                                           do_idop);
30527 +                       if (unlikely(e && !err))
30528 +                               err = e;
30529 +                       /* go on even err */
30530 +               }
30531 +       }
30532 +
30533 +out_dpages:
30534 +       au_dpages_free(&dpages);
30535 +out:
30536 +       return err;
30537 +}
30538 +
30539 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
30540 +{
30541 +       int err, e;
30542 +       unsigned int sigen;
30543 +       unsigned long long max, ull;
30544 +       struct inode *inode, **array;
30545 +
30546 +       array = au_iarray_alloc(sb, &max);
30547 +       err = PTR_ERR(array);
30548 +       if (IS_ERR(array))
30549 +               goto out;
30550 +
30551 +       err = 0;
30552 +       sigen = au_sigen(sb);
30553 +       for (ull = 0; ull < max; ull++) {
30554 +               inode = array[ull];
30555 +               if (unlikely(!inode))
30556 +                       break;
30557 +
30558 +               e = 0;
30559 +               ii_write_lock_child(inode);
30560 +               if (au_iigen(inode, NULL) != sigen) {
30561 +                       e = au_refresh_hinode_self(inode);
30562 +                       if (unlikely(e)) {
30563 +                               au_refresh_iop(inode, /*force_getattr*/1);
30564 +                               pr_err("error %d, i%lu\n", e, inode->i_ino);
30565 +                               if (!err)
30566 +                                       err = e;
30567 +                               /* go on even if err */
30568 +                       }
30569 +               }
30570 +               if (!e && do_idop)
30571 +                       au_refresh_iop(inode, /*force_getattr*/0);
30572 +               ii_write_unlock(inode);
30573 +       }
30574 +
30575 +       au_iarray_free(array, max);
30576 +
30577 +out:
30578 +       return err;
30579 +}
30580 +
30581 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
30582 +{
30583 +       int err, e;
30584 +       unsigned int udba;
30585 +       aufs_bindex_t bindex, bbot;
30586 +       struct dentry *root;
30587 +       struct inode *inode;
30588 +       struct au_branch *br;
30589 +       struct au_sbinfo *sbi;
30590 +
30591 +       au_sigen_inc(sb);
30592 +       sbi = au_sbi(sb);
30593 +       au_fclr_si(sbi, FAILED_REFRESH_DIR);
30594 +
30595 +       root = sb->s_root;
30596 +       DiMustNoWaiters(root);
30597 +       inode = d_inode(root);
30598 +       IiMustNoWaiters(inode);
30599 +
30600 +       udba = au_opt_udba(sb);
30601 +       bbot = au_sbbot(sb);
30602 +       for (bindex = 0; bindex <= bbot; bindex++) {
30603 +               br = au_sbr(sb, bindex);
30604 +               err = au_hnotify_reset_br(udba, br, br->br_perm);
30605 +               if (unlikely(err))
30606 +                       AuIOErr("hnotify failed on br %d, %d, ignored\n",
30607 +                               bindex, err);
30608 +               /* go on even if err */
30609 +       }
30610 +       au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
30611 +
30612 +       if (do_idop) {
30613 +               if (au_ftest_si(sbi, NO_DREVAL)) {
30614 +                       AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
30615 +                       sb->s_d_op = &aufs_dop_noreval;
30616 +                       AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
30617 +                       sbi->si_iop_array = aufs_iop_nogetattr;
30618 +               } else {
30619 +                       AuDebugOn(sb->s_d_op == &aufs_dop);
30620 +                       sb->s_d_op = &aufs_dop;
30621 +                       AuDebugOn(sbi->si_iop_array == aufs_iop);
30622 +                       sbi->si_iop_array = aufs_iop;
30623 +               }
30624 +               pr_info("reset to %ps and %ps\n",
30625 +                       sb->s_d_op, sbi->si_iop_array);
30626 +       }
30627 +
30628 +       di_write_unlock(root);
30629 +       err = au_refresh_d(sb, do_idop);
30630 +       e = au_refresh_i(sb, do_idop);
30631 +       if (unlikely(e && !err))
30632 +               err = e;
30633 +       /* aufs_write_lock() calls ..._child() */
30634 +       di_write_lock_child(root);
30635 +
30636 +       au_cpup_attr_all(inode, /*force*/1);
30637 +
30638 +       if (unlikely(err))
30639 +               AuIOErr("refresh failed, ignored, %d\n", err);
30640 +}
30641 +
30642 +const struct super_operations aufs_sop = {
30643 +       .alloc_inode    = aufs_alloc_inode,
30644 +       .destroy_inode  = aufs_destroy_inode,
30645 +       .free_inode     = aufs_free_inode,
30646 +       /* always deleting, no clearing */
30647 +       .drop_inode     = generic_delete_inode,
30648 +       .show_options   = aufs_show_options,
30649 +       .statfs         = aufs_statfs,
30650 +       .put_super      = aufs_put_super,
30651 +       .sync_fs        = aufs_sync_fs
30652 +};
30653 +
30654 +/* ---------------------------------------------------------------------- */
30655 +
30656 +int au_alloc_root(struct super_block *sb)
30657 +{
30658 +       int err;
30659 +       struct inode *inode;
30660 +       struct dentry *root;
30661 +
30662 +       err = -ENOMEM;
30663 +       inode = au_iget_locked(sb, AUFS_ROOT_INO);
30664 +       err = PTR_ERR(inode);
30665 +       if (IS_ERR(inode))
30666 +               goto out;
30667 +
30668 +       inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
30669 +       inode->i_fop = &aufs_dir_fop;
30670 +       inode->i_mode = S_IFDIR;
30671 +       set_nlink(inode, 2);
30672 +       unlock_new_inode(inode);
30673 +
30674 +       root = d_make_root(inode);
30675 +       if (unlikely(!root))
30676 +               goto out;
30677 +       err = PTR_ERR(root);
30678 +       if (IS_ERR(root))
30679 +               goto out;
30680 +
30681 +       err = au_di_init(root);
30682 +       if (!err) {
30683 +               sb->s_root = root;
30684 +               return 0; /* success */
30685 +       }
30686 +       dput(root);
30687 +
30688 +out:
30689 +       return err;
30690 +}
30691 +
30692 +/* ---------------------------------------------------------------------- */
30693 +
30694 +static void aufs_kill_sb(struct super_block *sb)
30695 +{
30696 +       struct au_sbinfo *sbinfo;
30697 +       struct dentry *root;
30698 +
30699 +       sbinfo = au_sbi(sb);
30700 +       if (!sbinfo)
30701 +               goto out;
30702 +
30703 +       au_sbilist_del(sb);
30704 +
30705 +       root = sb->s_root;
30706 +       if (root)
30707 +               aufs_write_lock(root);
30708 +       else
30709 +               __si_write_lock(sb);
30710 +
30711 +       au_fhsm_fin(sb);
30712 +       if (sbinfo->si_wbr_create_ops->fin)
30713 +               sbinfo->si_wbr_create_ops->fin(sb);
30714 +       if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
30715 +               au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
30716 +               au_remount_refresh(sb, /*do_idop*/0);
30717 +       }
30718 +       if (au_opt_test(sbinfo->si_mntflags, PLINK))
30719 +               au_plink_put(sb, /*verbose*/1);
30720 +       au_xino_clr(sb);
30721 +       if (root)
30722 +               au_dr_opt_flush(sb);
30723 +
30724 +       if (root)
30725 +               aufs_write_unlock(root);
30726 +       else
30727 +               __si_write_unlock(sb);
30728 +
30729 +       sbinfo->si_sb = NULL;
30730 +       au_nwt_flush(&sbinfo->si_nowait);
30731 +
30732 +out:
30733 +       kill_anon_super(sb);
30734 +}
30735 +
30736 +struct file_system_type aufs_fs_type = {
30737 +       .name           = AUFS_FSTYPE,
30738 +       /* a race between rename and others */
30739 +       .fs_flags       = FS_RENAME_DOES_D_MOVE
30740 +                               /* untested */
30741 +                               /*| FS_ALLOW_IDMAP*/
30742 +                               ,
30743 +       .init_fs_context = aufs_fsctx_init,
30744 +       .parameters     = aufs_fsctx_paramspec,
30745 +       .kill_sb        = aufs_kill_sb,
30746 +       /* no need to __module_get() and module_put(). */
30747 +       .owner          = THIS_MODULE,
30748 +};
30749 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
30750 --- /usr/share/empty/fs/aufs/super.h    1970-01-01 01:00:00.000000000 +0100
30751 +++ linux/fs/aufs/super.h       2022-11-05 23:02:18.969222617 +0100
30752 @@ -0,0 +1,592 @@
30753 +/* SPDX-License-Identifier: GPL-2.0 */
30754 +/*
30755 + * Copyright (C) 2005-2022 Junjiro R. Okajima
30756 + *
30757 + * This program is free software; you can redistribute it and/or modify
30758 + * it under the terms of the GNU General Public License as published by
30759 + * the Free Software Foundation; either version 2 of the License, or
30760 + * (at your option) any later version.
30761 + *
30762 + * This program is distributed in the hope that it will be useful,
30763 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30764 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
30765 + * GNU General Public License for more details.
30766 + *
30767 + * You should have received a copy of the GNU General Public License
30768 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
30769 + */
30770 +
30771 +/*
30772 + * super_block operations
30773 + */
30774 +
30775 +#ifndef __AUFS_SUPER_H__
30776 +#define __AUFS_SUPER_H__
30777 +
30778 +#ifdef __KERNEL__
30779 +
30780 +#include <linux/fs.h>
30781 +#include <linux/kobject.h>
30782 +#include "hbl.h"
30783 +#include "lcnt.h"
30784 +#include "rwsem.h"
30785 +#include "wkq.h"
30786 +
30787 +/* policies to select one among multiple writable branches */
30788 +struct au_wbr_copyup_operations {
30789 +       int (*copyup)(struct dentry *dentry);
30790 +};
30791 +
30792 +#define AuWbr_DIR      1               /* target is a dir */
30793 +#define AuWbr_PARENT   (1 << 1)        /* always require a parent */
30794 +
30795 +#define au_ftest_wbr(flags, name)      ((flags) & AuWbr_##name)
30796 +#define au_fset_wbr(flags, name)       { (flags) |= AuWbr_##name; }
30797 +#define au_fclr_wbr(flags, name)       { (flags) &= ~AuWbr_##name; }
30798 +
30799 +struct au_wbr_create_operations {
30800 +       int (*create)(struct dentry *dentry, unsigned int flags);
30801 +       int (*init)(struct super_block *sb);
30802 +       int (*fin)(struct super_block *sb);
30803 +};
30804 +
30805 +struct au_wbr_mfs {
30806 +       struct mutex    mfs_lock; /* protect this structure */
30807 +       unsigned long   mfs_jiffy;
30808 +       unsigned long   mfs_expire;
30809 +       aufs_bindex_t   mfs_bindex;
30810 +
30811 +       unsigned long long      mfsrr_bytes;
30812 +       unsigned long long      mfsrr_watermark;
30813 +};
30814 +
30815 +#define AuPlink_NHASH 100
30816 +static inline int au_plink_hash(ino_t ino)
30817 +{
30818 +       return ino % AuPlink_NHASH;
30819 +}
30820 +
30821 +/* File-based Hierarchical Storage Management */
30822 +struct au_fhsm {
30823 +#ifdef CONFIG_AUFS_FHSM
30824 +       /* allow only one process who can receive the notification */
30825 +       spinlock_t              fhsm_spin;
30826 +       pid_t                   fhsm_pid;
30827 +       wait_queue_head_t       fhsm_wqh;
30828 +       atomic_t                fhsm_readable;
30829 +
30830 +       /* these are protected by si_rwsem */
30831 +       unsigned long           fhsm_expire;
30832 +       aufs_bindex_t           fhsm_bottom;
30833 +#endif
30834 +};
30835 +
30836 +struct au_branch;
30837 +struct au_sbinfo {
30838 +       /* nowait tasks in the system-wide workqueue */
30839 +       struct au_nowait_tasks  si_nowait;
30840 +
30841 +       /*
30842 +        * tried sb->s_umount, but failed due to the dependency between i_mutex.
30843 +        * rwsem for au_sbinfo is necessary.
30844 +        */
30845 +       struct au_rwsem         si_rwsem;
30846 +
30847 +       /*
30848 +        * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
30849 +        * remount.
30850 +        */
30851 +       au_lcnt_t               si_ninodes, si_nfiles;
30852 +
30853 +       /* branch management */
30854 +       unsigned int            si_generation;
30855 +
30856 +       /* see AuSi_ flags */
30857 +       unsigned char           au_si_status;
30858 +
30859 +       aufs_bindex_t           si_bbot;
30860 +
30861 +       /* dirty trick to keep br_id plus */
30862 +       unsigned int            si_last_br_id :
30863 +                               sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
30864 +       struct au_branch        **si_branch;
30865 +
30866 +       /* policy to select a writable branch */
30867 +       unsigned char           si_wbr_copyup;
30868 +       unsigned char           si_wbr_create;
30869 +       struct au_wbr_copyup_operations *si_wbr_copyup_ops;
30870 +       struct au_wbr_create_operations *si_wbr_create_ops;
30871 +
30872 +       /* round robin */
30873 +       atomic_t                si_wbr_rr_next;
30874 +
30875 +       /* most free space */
30876 +       struct au_wbr_mfs       si_wbr_mfs;
30877 +
30878 +       /* File-based Hierarchical Storage Management */
30879 +       struct au_fhsm          si_fhsm;
30880 +
30881 +       /* mount flags */
30882 +       /* include/asm-ia64/siginfo.h defines a macro named si_flags */
30883 +       unsigned int            si_mntflags;
30884 +
30885 +       /* external inode number (bitmap and translation table) */
30886 +       loff_t                  si_ximaxent;    /* max entries in a xino */
30887 +
30888 +       struct file             *si_xib;
30889 +       struct mutex            si_xib_mtx; /* protect xib members */
30890 +       unsigned long           *si_xib_buf;
30891 +       unsigned long           si_xib_last_pindex;
30892 +       int                     si_xib_next_bit;
30893 +
30894 +       unsigned long           si_xino_jiffy;
30895 +       unsigned long           si_xino_expire;
30896 +       /* reserved for future use */
30897 +       /* unsigned long long   si_xib_limit; */        /* Max xib file size */
30898 +
30899 +#ifdef CONFIG_AUFS_EXPORT
30900 +       /* i_generation */
30901 +       /* todo: make xigen file an array to support many inode numbers */
30902 +       struct file             *si_xigen;
30903 +       atomic_t                si_xigen_next;
30904 +#endif
30905 +
30906 +       /* dirty trick to support atomic_open */
30907 +       struct hlist_bl_head    si_aopen;
30908 +
30909 +       /* vdir parameters */
30910 +       unsigned long           si_rdcache;     /* max cache time in jiffies */
30911 +       unsigned int            si_rdblk;       /* deblk size */
30912 +       unsigned int            si_rdhash;      /* hash size */
30913 +
30914 +       /*
30915 +        * If the number of whiteouts are larger than si_dirwh, leave all of
30916 +        * them after au_whtmp_ren to reduce the cost of rmdir(2).
30917 +        * future fsck.aufs or kernel thread will remove them later.
30918 +        * Otherwise, remove all whiteouts and the dir in rmdir(2).
30919 +        */
30920 +       unsigned int            si_dirwh;
30921 +
30922 +       /* pseudo_link list */
30923 +       struct hlist_bl_head    si_plink[AuPlink_NHASH];
30924 +       wait_queue_head_t       si_plink_wq;
30925 +       spinlock_t              si_plink_maint_lock;
30926 +       pid_t                   si_plink_maint_pid;
30927 +
30928 +       /* file list */
30929 +       struct hlist_bl_head    si_files;
30930 +
30931 +       /* with/without getattr, brother of sb->s_d_op */
30932 +       const struct inode_operations *si_iop_array;
30933 +
30934 +       /*
30935 +        * sysfs and lifetime management.
30936 +        * this is not a small structure and it may be a waste of memory in case
30937 +        * of sysfs is disabled, particularly when many aufs-es are mounted.
30938 +        * but using sysfs is majority.
30939 +        */
30940 +       struct kobject          si_kobj;
30941 +#ifdef CONFIG_DEBUG_FS
30942 +       struct dentry            *si_dbgaufs;
30943 +       struct dentry            *si_dbgaufs_plink;
30944 +       struct dentry            *si_dbgaufs_xib;
30945 +#ifdef CONFIG_AUFS_EXPORT
30946 +       struct dentry            *si_dbgaufs_xigen;
30947 +#endif
30948 +#endif
30949 +
30950 +#ifdef CONFIG_AUFS_SBILIST
30951 +       struct hlist_bl_node    si_list;
30952 +#endif
30953 +
30954 +       /* dirty, necessary for unmounting, sysfs and sysrq */
30955 +       struct super_block      *si_sb;
30956 +};
30957 +
30958 +/* sbinfo status flags */
30959 +/*
30960 + * set true when refresh_dirs() failed at remount time.
30961 + * then try refreshing dirs at access time again.
30962 + * if it is false, refreshing dirs at access time is unnecessary
30963 + */
30964 +#define AuSi_FAILED_REFRESH_DIR        1
30965 +#define AuSi_FHSM              (1 << 1)        /* fhsm is active now */
30966 +#define AuSi_NO_DREVAL         (1 << 2)        /* disable all d_revalidate */
30967 +
30968 +#ifndef CONFIG_AUFS_FHSM
30969 +#undef AuSi_FHSM
30970 +#define AuSi_FHSM              0
30971 +#endif
30972 +
30973 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
30974 +                                          unsigned int flag)
30975 +{
30976 +       AuRwMustAnyLock(&sbi->si_rwsem);
30977 +       return sbi->au_si_status & flag;
30978 +}
30979 +#define au_ftest_si(sbinfo, name)      au_do_ftest_si(sbinfo, AuSi_##name)
30980 +#define au_fset_si(sbinfo, name) do { \
30981 +       AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30982 +       (sbinfo)->au_si_status |= AuSi_##name; \
30983 +} while (0)
30984 +#define au_fclr_si(sbinfo, name) do { \
30985 +       AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30986 +       (sbinfo)->au_si_status &= ~AuSi_##name; \
30987 +} while (0)
30988 +
30989 +/* ---------------------------------------------------------------------- */
30990 +
30991 +/* policy to select one among writable branches */
30992 +#define AuWbrCopyup(sbinfo, ...) \
30993 +       ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
30994 +#define AuWbrCreate(sbinfo, ...) \
30995 +       ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
30996 +
30997 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
30998 +#define AuLock_DW              1               /* write-lock dentry */
30999 +#define AuLock_IR              (1 << 1)        /* read-lock inode */
31000 +#define AuLock_IW              (1 << 2)        /* write-lock inode */
31001 +#define AuLock_FLUSH           (1 << 3)        /* wait for 'nowait' tasks */
31002 +#define AuLock_DIRS            (1 << 4)        /* target is a pair of dirs */
31003 +                                               /* except RENAME_EXCHANGE */
31004 +#define AuLock_NOPLM           (1 << 5)        /* return err in plm mode */
31005 +#define AuLock_NOPLMW          (1 << 6)        /* wait for plm mode ends */
31006 +#define AuLock_GEN             (1 << 7)        /* test digen/iigen */
31007 +#define au_ftest_lock(flags, name)     ((flags) & AuLock_##name)
31008 +#define au_fset_lock(flags, name) \
31009 +       do { (flags) |= AuLock_##name; } while (0)
31010 +#define au_fclr_lock(flags, name) \
31011 +       do { (flags) &= ~AuLock_##name; } while (0)
31012 +
31013 +/* ---------------------------------------------------------------------- */
31014 +
31015 +/* super.c */
31016 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
31017 +
31018 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
31019 +                                          unsigned long long max, void *arg);
31020 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
31021 +                    struct super_block *sb, void *arg);
31022 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
31023 +void au_iarray_free(struct inode **a, unsigned long long max);
31024 +
31025 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop);
31026 +extern const struct super_operations aufs_sop;
31027 +int au_alloc_root(struct super_block *sb);
31028 +extern struct file_system_type aufs_fs_type;
31029 +
31030 +/* sbinfo.c */
31031 +void au_si_free(struct kobject *kobj);
31032 +struct au_sbinfo *au_si_alloc(struct super_block *sb);
31033 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink);
31034 +
31035 +unsigned int au_sigen_inc(struct super_block *sb);
31036 +aufs_bindex_t au_new_br_id(struct super_block *sb);
31037 +
31038 +int si_read_lock(struct super_block *sb, int flags);
31039 +int si_write_lock(struct super_block *sb, int flags);
31040 +int aufs_read_lock(struct dentry *dentry, int flags);
31041 +void aufs_read_unlock(struct dentry *dentry, int flags);
31042 +void aufs_write_lock(struct dentry *dentry);
31043 +void aufs_write_unlock(struct dentry *dentry);
31044 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
31045 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
31046 +
31047 +/* wbr_policy.c */
31048 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
31049 +extern struct au_wbr_create_operations au_wbr_create_ops[];
31050 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
31051 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
31052 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop);
31053 +
31054 +/* mvdown.c */
31055 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
31056 +
31057 +#ifdef CONFIG_AUFS_FHSM
31058 +/* fhsm.c */
31059 +
31060 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
31061 +{
31062 +       pid_t pid;
31063 +
31064 +       spin_lock(&fhsm->fhsm_spin);
31065 +       pid = fhsm->fhsm_pid;
31066 +       spin_unlock(&fhsm->fhsm_spin);
31067 +
31068 +       return pid;
31069 +}
31070 +
31071 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
31072 +void au_fhsm_wrote_all(struct super_block *sb, int force);
31073 +int au_fhsm_fd(struct super_block *sb, int oflags);
31074 +int au_fhsm_br_alloc(struct au_branch *br);
31075 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
31076 +void au_fhsm_fin(struct super_block *sb);
31077 +void au_fhsm_init(struct au_sbinfo *sbinfo);
31078 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
31079 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
31080 +#else
31081 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
31082 +          int force)
31083 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
31084 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
31085 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
31086 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
31087 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
31088 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
31089 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
31090 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
31091 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
31092 +#endif
31093 +
31094 +/* ---------------------------------------------------------------------- */
31095 +
31096 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
31097 +{
31098 +       return sb->s_fs_info;
31099 +}
31100 +
31101 +/* ---------------------------------------------------------------------- */
31102 +
31103 +#ifdef CONFIG_AUFS_EXPORT
31104 +int au_test_nfsd(void);
31105 +void au_export_init(struct super_block *sb);
31106 +void au_xigen_inc(struct inode *inode);
31107 +int au_xigen_new(struct inode *inode);
31108 +int au_xigen_set(struct super_block *sb, struct path *path);
31109 +void au_xigen_clr(struct super_block *sb);
31110 +
31111 +static inline int au_busy_or_stale(void)
31112 +{
31113 +       if (!au_test_nfsd())
31114 +               return -EBUSY;
31115 +       return -ESTALE;
31116 +}
31117 +#else
31118 +AuStubInt0(au_test_nfsd, void)
31119 +AuStubVoid(au_export_init, struct super_block *sb)
31120 +AuStubVoid(au_xigen_inc, struct inode *inode)
31121 +AuStubInt0(au_xigen_new, struct inode *inode)
31122 +AuStubInt0(au_xigen_set, struct super_block *sb, struct path *path)
31123 +AuStubVoid(au_xigen_clr, struct super_block *sb)
31124 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
31125 +#endif /* CONFIG_AUFS_EXPORT */
31126 +
31127 +/* ---------------------------------------------------------------------- */
31128 +
31129 +#ifdef CONFIG_AUFS_SBILIST
31130 +/* module.c */
31131 +extern struct hlist_bl_head au_sbilist;
31132 +
31133 +static inline void au_sbilist_init(void)
31134 +{
31135 +       INIT_HLIST_BL_HEAD(&au_sbilist);
31136 +}
31137 +
31138 +static inline void au_sbilist_add(struct super_block *sb)
31139 +{
31140 +       au_hbl_add(&au_sbi(sb)->si_list, &au_sbilist);
31141 +}
31142 +
31143 +static inline void au_sbilist_del(struct super_block *sb)
31144 +{
31145 +       au_hbl_del(&au_sbi(sb)->si_list, &au_sbilist);
31146 +}
31147 +
31148 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
31149 +static inline void au_sbilist_lock(void)
31150 +{
31151 +       hlist_bl_lock(&au_sbilist);
31152 +}
31153 +
31154 +static inline void au_sbilist_unlock(void)
31155 +{
31156 +       hlist_bl_unlock(&au_sbilist);
31157 +}
31158 +#define AuGFP_SBILIST  GFP_ATOMIC
31159 +#else
31160 +AuStubVoid(au_sbilist_lock, void)
31161 +AuStubVoid(au_sbilist_unlock, void)
31162 +#define AuGFP_SBILIST  GFP_NOFS
31163 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
31164 +#else
31165 +AuStubVoid(au_sbilist_init, void)
31166 +AuStubVoid(au_sbilist_add, struct super_block *sb)
31167 +AuStubVoid(au_sbilist_del, struct super_block *sb)
31168 +AuStubVoid(au_sbilist_lock, void)
31169 +AuStubVoid(au_sbilist_unlock, void)
31170 +#define AuGFP_SBILIST  GFP_NOFS
31171 +#endif
31172 +
31173 +/* ---------------------------------------------------------------------- */
31174 +
31175 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
31176 +{
31177 +       /*
31178 +        * This function is a dynamic '__init' function actually,
31179 +        * so the tiny check for si_rwsem is unnecessary.
31180 +        */
31181 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
31182 +#ifdef CONFIG_DEBUG_FS
31183 +       sbinfo->si_dbgaufs = NULL;
31184 +       sbinfo->si_dbgaufs_plink = NULL;
31185 +       sbinfo->si_dbgaufs_xib = NULL;
31186 +#ifdef CONFIG_AUFS_EXPORT
31187 +       sbinfo->si_dbgaufs_xigen = NULL;
31188 +#endif
31189 +#endif
31190 +}
31191 +
31192 +/* ---------------------------------------------------------------------- */
31193 +
31194 +/* current->atomic_flags */
31195 +/* this value should never corrupt the ones defined in linux/sched.h */
31196 +#define PFA_AUFS       0x10
31197 +
31198 +TASK_PFA_TEST(AUFS, test_aufs) /* task_test_aufs */
31199 +TASK_PFA_SET(AUFS, aufs)       /* task_set_aufs */
31200 +TASK_PFA_CLEAR(AUFS, aufs)     /* task_clear_aufs */
31201 +
31202 +static inline int si_pid_test(struct super_block *sb)
31203 +{
31204 +       return !!task_test_aufs(current);
31205 +}
31206 +
31207 +static inline void si_pid_clr(struct super_block *sb)
31208 +{
31209 +       AuDebugOn(!task_test_aufs(current));
31210 +       task_clear_aufs(current);
31211 +}
31212 +
31213 +static inline void si_pid_set(struct super_block *sb)
31214 +{
31215 +       AuDebugOn(task_test_aufs(current));
31216 +       task_set_aufs(current);
31217 +}
31218 +
31219 +/* ---------------------------------------------------------------------- */
31220 +
31221 +/* lock superblock. mainly for entry point functions */
31222 +#define __si_read_lock(sb)     au_rw_read_lock(&au_sbi(sb)->si_rwsem)
31223 +#define __si_write_lock(sb)    au_rw_write_lock(&au_sbi(sb)->si_rwsem)
31224 +#define __si_read_trylock(sb)  au_rw_read_trylock(&au_sbi(sb)->si_rwsem)
31225 +#define __si_write_trylock(sb) au_rw_write_trylock(&au_sbi(sb)->si_rwsem)
31226 +/*
31227 +#define __si_read_trylock_nested(sb) \
31228 +       au_rw_read_trylock_nested(&au_sbi(sb)->si_rwsem)
31229 +#define __si_write_trylock_nested(sb) \
31230 +       au_rw_write_trylock_nested(&au_sbi(sb)->si_rwsem)
31231 +*/
31232 +
31233 +#define __si_read_unlock(sb)   au_rw_read_unlock(&au_sbi(sb)->si_rwsem)
31234 +#define __si_write_unlock(sb)  au_rw_write_unlock(&au_sbi(sb)->si_rwsem)
31235 +#define __si_downgrade_lock(sb)        au_rw_dgrade_lock(&au_sbi(sb)->si_rwsem)
31236 +
31237 +#define SiMustNoWaiters(sb)    AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
31238 +#define SiMustAnyLock(sb)      AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
31239 +#define SiMustWriteLock(sb)    AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
31240 +
31241 +static inline void si_noflush_read_lock(struct super_block *sb)
31242 +{
31243 +       __si_read_lock(sb);
31244 +       si_pid_set(sb);
31245 +}
31246 +
31247 +static inline int si_noflush_read_trylock(struct super_block *sb)
31248 +{
31249 +       int locked;
31250 +
31251 +       locked = __si_read_trylock(sb);
31252 +       if (locked)
31253 +               si_pid_set(sb);
31254 +       return locked;
31255 +}
31256 +
31257 +static inline void si_noflush_write_lock(struct super_block *sb)
31258 +{
31259 +       __si_write_lock(sb);
31260 +       si_pid_set(sb);
31261 +}
31262 +
31263 +static inline int si_noflush_write_trylock(struct super_block *sb)
31264 +{
31265 +       int locked;
31266 +
31267 +       locked = __si_write_trylock(sb);
31268 +       if (locked)
31269 +               si_pid_set(sb);
31270 +       return locked;
31271 +}
31272 +
31273 +#if 0 /* reserved */
31274 +static inline int si_read_trylock(struct super_block *sb, int flags)
31275 +{
31276 +       if (au_ftest_lock(flags, FLUSH))
31277 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
31278 +       return si_noflush_read_trylock(sb);
31279 +}
31280 +#endif
31281 +
31282 +static inline void si_read_unlock(struct super_block *sb)
31283 +{
31284 +       si_pid_clr(sb);
31285 +       __si_read_unlock(sb);
31286 +}
31287 +
31288 +#if 0 /* reserved */
31289 +static inline int si_write_trylock(struct super_block *sb, int flags)
31290 +{
31291 +       if (au_ftest_lock(flags, FLUSH))
31292 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
31293 +       return si_noflush_write_trylock(sb);
31294 +}
31295 +#endif
31296 +
31297 +static inline void si_write_unlock(struct super_block *sb)
31298 +{
31299 +       si_pid_clr(sb);
31300 +       __si_write_unlock(sb);
31301 +}
31302 +
31303 +#if 0 /* reserved */
31304 +static inline void si_downgrade_lock(struct super_block *sb)
31305 +{
31306 +       __si_downgrade_lock(sb);
31307 +}
31308 +#endif
31309 +
31310 +/* ---------------------------------------------------------------------- */
31311 +
31312 +static inline aufs_bindex_t au_sbbot(struct super_block *sb)
31313 +{
31314 +       SiMustAnyLock(sb);
31315 +       return au_sbi(sb)->si_bbot;
31316 +}
31317 +
31318 +static inline unsigned int au_mntflags(struct super_block *sb)
31319 +{
31320 +       SiMustAnyLock(sb);
31321 +       return au_sbi(sb)->si_mntflags;
31322 +}
31323 +
31324 +static inline unsigned int au_sigen(struct super_block *sb)
31325 +{
31326 +       SiMustAnyLock(sb);
31327 +       return au_sbi(sb)->si_generation;
31328 +}
31329 +
31330 +static inline struct au_branch *au_sbr(struct super_block *sb,
31331 +                                      aufs_bindex_t bindex)
31332 +{
31333 +       SiMustAnyLock(sb);
31334 +       return au_sbi(sb)->si_branch[0 + bindex];
31335 +}
31336 +
31337 +static inline loff_t au_xi_maxent(struct super_block *sb)
31338 +{
31339 +       SiMustAnyLock(sb);
31340 +       return au_sbi(sb)->si_ximaxent;
31341 +}
31342 +
31343 +#endif /* __KERNEL__ */
31344 +#endif /* __AUFS_SUPER_H__ */
31345 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
31346 --- /usr/share/empty/fs/aufs/sysaufs.c  1970-01-01 01:00:00.000000000 +0100
31347 +++ linux/fs/aufs/sysaufs.c     2022-11-05 23:02:18.969222617 +0100
31348 @@ -0,0 +1,94 @@
31349 +// SPDX-License-Identifier: GPL-2.0
31350 +/*
31351 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31352 + *
31353 + * This program is free software; you can redistribute it and/or modify
31354 + * it under the terms of the GNU General Public License as published by
31355 + * the Free Software Foundation; either version 2 of the License, or
31356 + * (at your option) any later version.
31357 + *
31358 + * This program is distributed in the hope that it will be useful,
31359 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31360 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31361 + * GNU General Public License for more details.
31362 + *
31363 + * You should have received a copy of the GNU General Public License
31364 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31365 + */
31366 +
31367 +/*
31368 + * sysfs interface and lifetime management
31369 + * they are necessary regardless sysfs is disabled.
31370 + */
31371 +
31372 +#include <linux/random.h>
31373 +#include "aufs.h"
31374 +
31375 +unsigned long sysaufs_si_mask;
31376 +struct kset *sysaufs_kset;
31377 +
31378 +#define AuSiAttr(_name) { \
31379 +       .attr   = { .name = __stringify(_name), .mode = 0444 }, \
31380 +       .show   = sysaufs_si_##_name,                           \
31381 +}
31382 +
31383 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
31384 +struct attribute *sysaufs_si_attrs[] = {
31385 +       &sysaufs_si_attr_xi_path.attr,
31386 +       NULL,
31387 +};
31388 +ATTRIBUTE_GROUPS(sysaufs_si);
31389 +
31390 +static const struct sysfs_ops au_sbi_ops = {
31391 +       .show   = sysaufs_si_show
31392 +};
31393 +
31394 +static struct kobj_type au_sbi_ktype = {
31395 +       .release        = au_si_free,
31396 +       .sysfs_ops      = &au_sbi_ops,
31397 +       .default_groups = sysaufs_si_groups
31398 +};
31399 +
31400 +/* ---------------------------------------------------------------------- */
31401 +
31402 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
31403 +{
31404 +       int err;
31405 +
31406 +       sbinfo->si_kobj.kset = sysaufs_kset;
31407 +       /* cf. sysaufs_name() */
31408 +       err = kobject_init_and_add
31409 +               (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
31410 +                SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
31411 +
31412 +       return err;
31413 +}
31414 +
31415 +void sysaufs_fin(void)
31416 +{
31417 +       sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31418 +       kset_unregister(sysaufs_kset);
31419 +}
31420 +
31421 +int __init sysaufs_init(void)
31422 +{
31423 +       int err;
31424 +
31425 +       do {
31426 +               get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
31427 +       } while (!sysaufs_si_mask);
31428 +
31429 +       err = -EINVAL;
31430 +       sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
31431 +       if (unlikely(!sysaufs_kset))
31432 +               goto out;
31433 +       err = PTR_ERR(sysaufs_kset);
31434 +       if (IS_ERR(sysaufs_kset))
31435 +               goto out;
31436 +       err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31437 +       if (unlikely(err))
31438 +               kset_unregister(sysaufs_kset);
31439 +
31440 +out:
31441 +       return err;
31442 +}
31443 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
31444 --- /usr/share/empty/fs/aufs/sysaufs.h  1970-01-01 01:00:00.000000000 +0100
31445 +++ linux/fs/aufs/sysaufs.h     2022-11-05 23:02:18.969222617 +0100
31446 @@ -0,0 +1,102 @@
31447 +/* SPDX-License-Identifier: GPL-2.0 */
31448 +/*
31449 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31450 + *
31451 + * This program is free software; you can redistribute it and/or modify
31452 + * it under the terms of the GNU General Public License as published by
31453 + * the Free Software Foundation; either version 2 of the License, or
31454 + * (at your option) any later version.
31455 + *
31456 + * This program is distributed in the hope that it will be useful,
31457 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31458 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31459 + * GNU General Public License for more details.
31460 + *
31461 + * You should have received a copy of the GNU General Public License
31462 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31463 + */
31464 +
31465 +/*
31466 + * sysfs interface and mount lifetime management
31467 + */
31468 +
31469 +#ifndef __SYSAUFS_H__
31470 +#define __SYSAUFS_H__
31471 +
31472 +#ifdef __KERNEL__
31473 +
31474 +#include <linux/sysfs.h>
31475 +#include "module.h"
31476 +
31477 +struct super_block;
31478 +struct au_sbinfo;
31479 +
31480 +struct sysaufs_si_attr {
31481 +       struct attribute attr;
31482 +       int (*show)(struct seq_file *seq, struct super_block *sb);
31483 +};
31484 +
31485 +/* ---------------------------------------------------------------------- */
31486 +
31487 +/* sysaufs.c */
31488 +extern unsigned long sysaufs_si_mask;
31489 +extern struct kset *sysaufs_kset;
31490 +extern struct attribute *sysaufs_si_attrs[];
31491 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
31492 +int __init sysaufs_init(void);
31493 +void sysaufs_fin(void);
31494 +
31495 +/* ---------------------------------------------------------------------- */
31496 +
31497 +/* some people doesn't like to show a pointer in kernel */
31498 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
31499 +{
31500 +       return sysaufs_si_mask ^ (unsigned long)sbinfo;
31501 +}
31502 +
31503 +#define SysaufsSiNamePrefix    "si_"
31504 +#define SysaufsSiNameLen       (sizeof(SysaufsSiNamePrefix) + 16)
31505 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
31506 +{
31507 +       snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
31508 +                sysaufs_si_id(sbinfo));
31509 +}
31510 +
31511 +struct au_branch;
31512 +#ifdef CONFIG_SYSFS
31513 +/* sysfs.c */
31514 +extern struct attribute_group *sysaufs_attr_group;
31515 +
31516 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
31517 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31518 +                        char *buf);
31519 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
31520 +#ifdef CONFIG_COMPAT
31521 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
31522 +#endif
31523 +
31524 +void sysaufs_br_init(struct au_branch *br);
31525 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
31526 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
31527 +
31528 +#define sysaufs_brs_init()     do {} while (0)
31529 +
31530 +#else
31531 +#define sysaufs_attr_group     NULL
31532 +
31533 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
31534 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
31535 +       struct attribute *attr, char *buf)
31536 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
31537 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
31538 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
31539 +
31540 +static inline void sysaufs_brs_init(void)
31541 +{
31542 +       sysaufs_brs = 0;
31543 +}
31544 +
31545 +#endif /* CONFIG_SYSFS */
31546 +
31547 +#endif /* __KERNEL__ */
31548 +#endif /* __SYSAUFS_H__ */
31549 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
31550 --- /usr/share/empty/fs/aufs/sysfs.c    1970-01-01 01:00:00.000000000 +0100
31551 +++ linux/fs/aufs/sysfs.c       2022-11-05 23:02:18.969222617 +0100
31552 @@ -0,0 +1,374 @@
31553 +// SPDX-License-Identifier: GPL-2.0
31554 +/*
31555 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31556 + *
31557 + * This program is free software; you can redistribute it and/or modify
31558 + * it under the terms of the GNU General Public License as published by
31559 + * the Free Software Foundation; either version 2 of the License, or
31560 + * (at your option) any later version.
31561 + *
31562 + * This program is distributed in the hope that it will be useful,
31563 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31564 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31565 + * GNU General Public License for more details.
31566 + *
31567 + * You should have received a copy of the GNU General Public License
31568 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31569 + */
31570 +
31571 +/*
31572 + * sysfs interface
31573 + */
31574 +
31575 +#include <linux/compat.h>
31576 +#include <linux/seq_file.h>
31577 +#include "aufs.h"
31578 +
31579 +#ifdef CONFIG_AUFS_FS_MODULE
31580 +/* this entry violates the "one line per file" policy of sysfs */
31581 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
31582 +                          char *buf)
31583 +{
31584 +       ssize_t err;
31585 +       static char *conf =
31586 +/* this file is generated at compiling */
31587 +#include "conf.str"
31588 +               ;
31589 +
31590 +       err = snprintf(buf, PAGE_SIZE, conf);
31591 +       if (unlikely(err >= PAGE_SIZE))
31592 +               err = -EFBIG;
31593 +       return err;
31594 +}
31595 +
31596 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
31597 +#endif
31598 +
31599 +static struct attribute *au_attr[] = {
31600 +#ifdef CONFIG_AUFS_FS_MODULE
31601 +       &au_config_attr.attr,
31602 +#endif
31603 +       NULL,   /* need to NULL terminate the list of attributes */
31604 +};
31605 +
31606 +static struct attribute_group sysaufs_attr_group_body = {
31607 +       .attrs = au_attr
31608 +};
31609 +
31610 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
31611 +
31612 +/* ---------------------------------------------------------------------- */
31613 +
31614 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
31615 +{
31616 +       int err;
31617 +
31618 +       SiMustAnyLock(sb);
31619 +
31620 +       err = 0;
31621 +       if (au_opt_test(au_mntflags(sb), XINO)) {
31622 +               err = au_xino_path(seq, au_sbi(sb)->si_xib);
31623 +               seq_putc(seq, '\n');
31624 +       }
31625 +       return err;
31626 +}
31627 +
31628 +/*
31629 + * the lifetime of branch is independent from the entry under sysfs.
31630 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
31631 + * unlinked.
31632 + */
31633 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
31634 +                        aufs_bindex_t bindex, int idx)
31635 +{
31636 +       int err;
31637 +       struct path path;
31638 +       struct dentry *root;
31639 +       struct au_branch *br;
31640 +       au_br_perm_str_t perm;
31641 +
31642 +       AuDbg("b%d\n", bindex);
31643 +
31644 +       err = 0;
31645 +       root = sb->s_root;
31646 +       di_read_lock_parent(root, !AuLock_IR);
31647 +       br = au_sbr(sb, bindex);
31648 +
31649 +       switch (idx) {
31650 +       case AuBrSysfs_BR:
31651 +               path.mnt = au_br_mnt(br);
31652 +               path.dentry = au_h_dptr(root, bindex);
31653 +               err = au_seq_path(seq, &path);
31654 +               if (!err) {
31655 +                       au_optstr_br_perm(&perm, br->br_perm);
31656 +                       seq_printf(seq, "=%s\n", perm.a);
31657 +               }
31658 +               break;
31659 +       case AuBrSysfs_BRID:
31660 +               seq_printf(seq, "%d\n", br->br_id);
31661 +               break;
31662 +       }
31663 +       di_read_unlock(root, !AuLock_IR);
31664 +       if (unlikely(err || seq_has_overflowed(seq)))
31665 +               err = -E2BIG;
31666 +
31667 +       return err;
31668 +}
31669 +
31670 +/* ---------------------------------------------------------------------- */
31671 +
31672 +static struct seq_file *au_seq(char *p, ssize_t len)
31673 +{
31674 +       struct seq_file *seq;
31675 +
31676 +       seq = kzalloc(sizeof(*seq), GFP_NOFS);
31677 +       if (seq) {
31678 +               /* mutex_init(&seq.lock); */
31679 +               seq->buf = p;
31680 +               seq->size = len;
31681 +               return seq; /* success */
31682 +       }
31683 +
31684 +       seq = ERR_PTR(-ENOMEM);
31685 +       return seq;
31686 +}
31687 +
31688 +#define SysaufsBr_PREFIX       "br"
31689 +#define SysaufsBrid_PREFIX     "brid"
31690 +
31691 +/* todo: file size may exceed PAGE_SIZE */
31692 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31693 +                       char *buf)
31694 +{
31695 +       ssize_t err;
31696 +       int idx;
31697 +       long l;
31698 +       aufs_bindex_t bbot;
31699 +       struct au_sbinfo *sbinfo;
31700 +       struct super_block *sb;
31701 +       struct seq_file *seq;
31702 +       char *name;
31703 +       struct attribute **cattr;
31704 +
31705 +       sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
31706 +       sb = sbinfo->si_sb;
31707 +
31708 +       /*
31709 +        * prevent a race condition between sysfs and aufs.
31710 +        * for instance, sysfs_file_read() calls sysfs_get_active_two() which
31711 +        * prohibits maintaining the sysfs entries.
31712 +        * hew we acquire read lock after sysfs_get_active_two().
31713 +        * on the other hand, the remount process may maintain the sysfs/aufs
31714 +        * entries after acquiring write lock.
31715 +        * it can cause a deadlock.
31716 +        * simply we gave up processing read here.
31717 +        */
31718 +       err = -EBUSY;
31719 +       if (unlikely(!si_noflush_read_trylock(sb)))
31720 +               goto out;
31721 +
31722 +       seq = au_seq(buf, PAGE_SIZE);
31723 +       err = PTR_ERR(seq);
31724 +       if (IS_ERR(seq))
31725 +               goto out_unlock;
31726 +
31727 +       name = (void *)attr->name;
31728 +       cattr = sysaufs_si_attrs;
31729 +       while (*cattr) {
31730 +               if (!strcmp(name, (*cattr)->name)) {
31731 +                       err = container_of(*cattr, struct sysaufs_si_attr, attr)
31732 +                               ->show(seq, sb);
31733 +                       goto out_seq;
31734 +               }
31735 +               cattr++;
31736 +       }
31737 +
31738 +       if (!strncmp(name, SysaufsBrid_PREFIX,
31739 +                    sizeof(SysaufsBrid_PREFIX) - 1)) {
31740 +               idx = AuBrSysfs_BRID;
31741 +               name += sizeof(SysaufsBrid_PREFIX) - 1;
31742 +       } else if (!strncmp(name, SysaufsBr_PREFIX,
31743 +                           sizeof(SysaufsBr_PREFIX) - 1)) {
31744 +               idx = AuBrSysfs_BR;
31745 +               name += sizeof(SysaufsBr_PREFIX) - 1;
31746 +       } else
31747 +                 BUG();
31748 +
31749 +       err = kstrtol(name, 10, &l);
31750 +       if (!err) {
31751 +               bbot = au_sbbot(sb);
31752 +               if (l <= bbot)
31753 +                       err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
31754 +               else
31755 +                       err = -ENOENT;
31756 +       }
31757 +
31758 +out_seq:
31759 +       if (!err) {
31760 +               err = seq->count;
31761 +               /* sysfs limit */
31762 +               if (unlikely(err == PAGE_SIZE))
31763 +                       err = -EFBIG;
31764 +       }
31765 +       au_kfree_rcu(seq);
31766 +out_unlock:
31767 +       si_read_unlock(sb);
31768 +out:
31769 +       return err;
31770 +}
31771 +
31772 +/* ---------------------------------------------------------------------- */
31773 +
31774 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
31775 +{
31776 +       int err;
31777 +       int16_t brid;
31778 +       aufs_bindex_t bindex, bbot;
31779 +       size_t sz;
31780 +       char *buf;
31781 +       struct seq_file *seq;
31782 +       struct au_branch *br;
31783 +
31784 +       si_read_lock(sb, AuLock_FLUSH);
31785 +       bbot = au_sbbot(sb);
31786 +       err = bbot + 1;
31787 +       if (!arg)
31788 +               goto out;
31789 +
31790 +       err = -ENOMEM;
31791 +       buf = (void *)__get_free_page(GFP_NOFS);
31792 +       if (unlikely(!buf))
31793 +               goto out;
31794 +
31795 +       seq = au_seq(buf, PAGE_SIZE);
31796 +       err = PTR_ERR(seq);
31797 +       if (IS_ERR(seq))
31798 +               goto out_buf;
31799 +
31800 +       sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
31801 +       for (bindex = 0; bindex <= bbot; bindex++, arg++) {
31802 +               /* VERIFY_WRITE */
31803 +               err = !access_ok(arg, sizeof(*arg));
31804 +               if (unlikely(err))
31805 +                       break;
31806 +
31807 +               br = au_sbr(sb, bindex);
31808 +               brid = br->br_id;
31809 +               BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
31810 +               err = __put_user(brid, &arg->id);
31811 +               if (unlikely(err))
31812 +                       break;
31813 +
31814 +               BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
31815 +               err = __put_user(br->br_perm, &arg->perm);
31816 +               if (unlikely(err))
31817 +                       break;
31818 +
31819 +               err = au_seq_path(seq, &br->br_path);
31820 +               if (unlikely(err))
31821 +                       break;
31822 +               seq_putc(seq, '\0');
31823 +               if (!seq_has_overflowed(seq)) {
31824 +                       err = copy_to_user(arg->path, seq->buf, seq->count);
31825 +                       seq->count = 0;
31826 +                       if (unlikely(err))
31827 +                               break;
31828 +               } else {
31829 +                       err = -E2BIG;
31830 +                       goto out_seq;
31831 +               }
31832 +       }
31833 +       if (unlikely(err))
31834 +               err = -EFAULT;
31835 +
31836 +out_seq:
31837 +       au_kfree_rcu(seq);
31838 +out_buf:
31839 +       free_page((unsigned long)buf);
31840 +out:
31841 +       si_read_unlock(sb);
31842 +       return err;
31843 +}
31844 +
31845 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
31846 +{
31847 +       return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
31848 +}
31849 +
31850 +#ifdef CONFIG_COMPAT
31851 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
31852 +{
31853 +       return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
31854 +}
31855 +#endif
31856 +
31857 +/* ---------------------------------------------------------------------- */
31858 +
31859 +void sysaufs_br_init(struct au_branch *br)
31860 +{
31861 +       int i;
31862 +       struct au_brsysfs *br_sysfs;
31863 +       struct attribute *attr;
31864 +
31865 +       br_sysfs = br->br_sysfs;
31866 +       for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31867 +               attr = &br_sysfs->attr;
31868 +               sysfs_attr_init(attr);
31869 +               attr->name = br_sysfs->name;
31870 +               attr->mode = 0444;
31871 +               br_sysfs++;
31872 +       }
31873 +}
31874 +
31875 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
31876 +{
31877 +       struct au_branch *br;
31878 +       struct kobject *kobj;
31879 +       struct au_brsysfs *br_sysfs;
31880 +       int i;
31881 +       aufs_bindex_t bbot;
31882 +
31883 +       if (!sysaufs_brs)
31884 +               return;
31885 +
31886 +       kobj = &au_sbi(sb)->si_kobj;
31887 +       bbot = au_sbbot(sb);
31888 +       for (; bindex <= bbot; bindex++) {
31889 +               br = au_sbr(sb, bindex);
31890 +               br_sysfs = br->br_sysfs;
31891 +               for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31892 +                       sysfs_remove_file(kobj, &br_sysfs->attr);
31893 +                       br_sysfs++;
31894 +               }
31895 +       }
31896 +}
31897 +
31898 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
31899 +{
31900 +       int err, i;
31901 +       aufs_bindex_t bbot;
31902 +       struct kobject *kobj;
31903 +       struct au_branch *br;
31904 +       struct au_brsysfs *br_sysfs;
31905 +
31906 +       if (!sysaufs_brs)
31907 +               return;
31908 +
31909 +       kobj = &au_sbi(sb)->si_kobj;
31910 +       bbot = au_sbbot(sb);
31911 +       for (; bindex <= bbot; bindex++) {
31912 +               br = au_sbr(sb, bindex);
31913 +               br_sysfs = br->br_sysfs;
31914 +               snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
31915 +                        SysaufsBr_PREFIX "%d", bindex);
31916 +               snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
31917 +                        SysaufsBrid_PREFIX "%d", bindex);
31918 +               for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31919 +                       err = sysfs_create_file(kobj, &br_sysfs->attr);
31920 +                       if (unlikely(err))
31921 +                               pr_warn("failed %s under sysfs(%d)\n",
31922 +                                       br_sysfs->name, err);
31923 +                       br_sysfs++;
31924 +               }
31925 +       }
31926 +}
31927 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
31928 --- /usr/share/empty/fs/aufs/sysrq.c    1970-01-01 01:00:00.000000000 +0100
31929 +++ linux/fs/aufs/sysrq.c       2022-11-05 23:02:18.969222617 +0100
31930 @@ -0,0 +1,149 @@
31931 +// SPDX-License-Identifier: GPL-2.0
31932 +/*
31933 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31934 + *
31935 + * This program is free software; you can redistribute it and/or modify
31936 + * it under the terms of the GNU General Public License as published by
31937 + * the Free Software Foundation; either version 2 of the License, or
31938 + * (at your option) any later version.
31939 + *
31940 + * This program is distributed in the hope that it will be useful,
31941 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31942 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31943 + * GNU General Public License for more details.
31944 + *
31945 + * You should have received a copy of the GNU General Public License
31946 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31947 + */
31948 +
31949 +/*
31950 + * magic sysrq handler
31951 + */
31952 +
31953 +/* #include <linux/sysrq.h> */
31954 +#include <linux/writeback.h>
31955 +#include "aufs.h"
31956 +
31957 +/* ---------------------------------------------------------------------- */
31958 +
31959 +static void sysrq_sb(struct super_block *sb)
31960 +{
31961 +       char *plevel;
31962 +       struct au_sbinfo *sbinfo;
31963 +       struct file *file;
31964 +       struct hlist_bl_head *files;
31965 +       struct hlist_bl_node *pos;
31966 +       struct au_finfo *finfo;
31967 +       struct inode *i;
31968 +
31969 +       plevel = au_plevel;
31970 +       au_plevel = KERN_WARNING;
31971 +
31972 +       /* since we define pr_fmt, call printk directly */
31973 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
31974 +
31975 +       sbinfo = au_sbi(sb);
31976 +       printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
31977 +       pr("superblock\n");
31978 +       au_dpri_sb(sb);
31979 +
31980 +#if 0 /* reserved */
31981 +       do {
31982 +               int err, i, j, ndentry;
31983 +               struct au_dcsub_pages dpages;
31984 +               struct au_dpage *dpage;
31985 +
31986 +               err = au_dpages_init(&dpages, GFP_ATOMIC);
31987 +               if (unlikely(err))
31988 +                       break;
31989 +               err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
31990 +               if (!err)
31991 +                       for (i = 0; i < dpages.ndpage; i++) {
31992 +                               dpage = dpages.dpages + i;
31993 +                               ndentry = dpage->ndentry;
31994 +                               for (j = 0; j < ndentry; j++)
31995 +                                       au_dpri_dentry(dpage->dentries[j]);
31996 +                       }
31997 +               au_dpages_free(&dpages);
31998 +       } while (0);
31999 +#endif
32000 +
32001 +       pr("isolated inode\n");
32002 +       spin_lock(&sb->s_inode_list_lock);
32003 +       list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
32004 +               spin_lock(&i->i_lock);
32005 +               if (hlist_empty(&i->i_dentry))
32006 +                       au_dpri_inode(i);
32007 +               spin_unlock(&i->i_lock);
32008 +       }
32009 +       spin_unlock(&sb->s_inode_list_lock);
32010 +
32011 +       pr("files\n");
32012 +       files = &au_sbi(sb)->si_files;
32013 +       hlist_bl_lock(files);
32014 +       hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
32015 +               umode_t mode;
32016 +
32017 +               file = finfo->fi_file;
32018 +               mode = file_inode(file)->i_mode;
32019 +               if (!special_file(mode))
32020 +                       au_dpri_file(file);
32021 +       }
32022 +       hlist_bl_unlock(files);
32023 +       pr("done\n");
32024 +
32025 +#undef pr
32026 +       au_plevel = plevel;
32027 +}
32028 +
32029 +/* ---------------------------------------------------------------------- */
32030 +
32031 +/* module parameter */
32032 +static char *aufs_sysrq_key = "a";
32033 +module_param_named(sysrq, aufs_sysrq_key, charp, 0444);
32034 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
32035 +
32036 +static void au_sysrq(int key __maybe_unused)
32037 +{
32038 +       struct au_sbinfo *sbinfo;
32039 +       struct hlist_bl_node *pos;
32040 +
32041 +       lockdep_off();
32042 +       au_sbilist_lock();
32043 +       hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
32044 +               sysrq_sb(sbinfo->si_sb);
32045 +       au_sbilist_unlock();
32046 +       lockdep_on();
32047 +}
32048 +
32049 +static struct sysrq_key_op au_sysrq_op = {
32050 +       .handler        = au_sysrq,
32051 +       .help_msg       = "Aufs",
32052 +       .action_msg     = "Aufs",
32053 +       .enable_mask    = SYSRQ_ENABLE_DUMP
32054 +};
32055 +
32056 +/* ---------------------------------------------------------------------- */
32057 +
32058 +int __init au_sysrq_init(void)
32059 +{
32060 +       int err;
32061 +       char key;
32062 +
32063 +       err = -1;
32064 +       key = *aufs_sysrq_key;
32065 +       if ('a' <= key && key <= 'z')
32066 +               err = register_sysrq_key(key, &au_sysrq_op);
32067 +       if (unlikely(err))
32068 +               pr_err("err %d, sysrq=%c\n", err, key);
32069 +       return err;
32070 +}
32071 +
32072 +void au_sysrq_fin(void)
32073 +{
32074 +       int err;
32075 +
32076 +       err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
32077 +       if (unlikely(err))
32078 +               pr_err("err %d (ignored)\n", err);
32079 +}
32080 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
32081 --- /usr/share/empty/fs/aufs/vdir.c     1970-01-01 01:00:00.000000000 +0100
32082 +++ linux/fs/aufs/vdir.c        2022-12-17 09:21:34.799855195 +0100
32083 @@ -0,0 +1,896 @@
32084 +// SPDX-License-Identifier: GPL-2.0
32085 +/*
32086 + * Copyright (C) 2005-2022 Junjiro R. Okajima
32087 + *
32088 + * This program is free software; you can redistribute it and/or modify
32089 + * it under the terms of the GNU General Public License as published by
32090 + * the Free Software Foundation; either version 2 of the License, or
32091 + * (at your option) any later version.
32092 + *
32093 + * This program is distributed in the hope that it will be useful,
32094 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32095 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
32096 + * GNU General Public License for more details.
32097 + *
32098 + * You should have received a copy of the GNU General Public License
32099 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
32100 + */
32101 +
32102 +/*
32103 + * virtual or vertical directory
32104 + */
32105 +
32106 +#include <linux/iversion.h>
32107 +#include "aufs.h"
32108 +
32109 +static unsigned int calc_size(int nlen)
32110 +{
32111 +       return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
32112 +}
32113 +
32114 +static int set_deblk_end(union au_vdir_deblk_p *p,
32115 +                        union au_vdir_deblk_p *deblk_end)
32116 +{
32117 +       if (calc_size(0) <= deblk_end->deblk - p->deblk) {
32118 +               p->de->de_str.len = 0;
32119 +               /* smp_mb(); */
32120 +               return 0;
32121 +       }
32122 +       return -1; /* error */
32123 +}
32124 +
32125 +/* returns true or false */
32126 +static int is_deblk_end(union au_vdir_deblk_p *p,
32127 +                       union au_vdir_deblk_p *deblk_end)
32128 +{
32129 +       if (calc_size(0) <= deblk_end->deblk - p->deblk)
32130 +               return !p->de->de_str.len;
32131 +       return 1;
32132 +}
32133 +
32134 +static unsigned char *last_deblk(struct au_vdir *vdir)
32135 +{
32136 +       return vdir->vd_deblk[vdir->vd_nblk - 1];
32137 +}
32138 +
32139 +/* ---------------------------------------------------------------------- */
32140 +
32141 +/* estimate the appropriate size for name hash table */
32142 +unsigned int au_rdhash_est(loff_t sz)
32143 +{
32144 +       unsigned int n;
32145 +
32146 +       n = UINT_MAX;
32147 +       sz >>= 10;
32148 +       if (sz < n)
32149 +               n = sz;
32150 +       if (sz < AUFS_RDHASH_DEF)
32151 +               n = AUFS_RDHASH_DEF;
32152 +       /* pr_info("n %u\n", n); */
32153 +       return n;
32154 +}
32155 +
32156 +/*
32157 + * the allocated memory has to be freed by
32158 + * au_nhash_wh_free() or au_nhash_de_free().
32159 + */
32160 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
32161 +{
32162 +       struct hlist_head *head;
32163 +       unsigned int u;
32164 +       size_t sz;
32165 +
32166 +       sz = sizeof(*nhash->nh_head) * num_hash;
32167 +       head = kmalloc(sz, gfp);
32168 +       if (head) {
32169 +               nhash->nh_num = num_hash;
32170 +               nhash->nh_head = head;
32171 +               for (u = 0; u < num_hash; u++)
32172 +                       INIT_HLIST_HEAD(head++);
32173 +               return 0; /* success */
32174 +       }
32175 +
32176 +       return -ENOMEM;
32177 +}
32178 +
32179 +static void nhash_count(struct hlist_head *head)
32180 +{
32181 +#if 0 /* debugging */
32182 +       unsigned long n;
32183 +       struct hlist_node *pos;
32184 +
32185 +       n = 0;
32186 +       hlist_for_each(pos, head)
32187 +               n++;
32188 +       pr_info("%lu\n", n);
32189 +#endif
32190 +}
32191 +
32192 +static void au_nhash_wh_do_free(struct hlist_head *head)
32193 +{
32194 +       struct au_vdir_wh *pos;
32195 +       struct hlist_node *node;
32196 +
32197 +       hlist_for_each_entry_safe(pos, node, head, wh_hash)
32198 +               au_kfree_rcu(pos);
32199 +}
32200 +
32201 +static void au_nhash_de_do_free(struct hlist_head *head)
32202 +{
32203 +       struct au_vdir_dehstr *pos;
32204 +       struct hlist_node *node;
32205 +
32206 +       hlist_for_each_entry_safe(pos, node, head, hash)
32207 +               au_cache_free_vdir_dehstr(pos);
32208 +}
32209 +
32210 +static void au_nhash_do_free(struct au_nhash *nhash,
32211 +                            void (*free)(struct hlist_head *head))
32212 +{
32213 +       unsigned int n;
32214 +       struct hlist_head *head;
32215 +
32216 +       n = nhash->nh_num;
32217 +       if (!n)
32218 +               return;
32219 +
32220 +       head = nhash->nh_head;
32221 +       while (n-- > 0) {
32222 +               nhash_count(head);
32223 +               free(head++);
32224 +       }
32225 +       au_kfree_try_rcu(nhash->nh_head);
32226 +}
32227 +
32228 +void au_nhash_wh_free(struct au_nhash *whlist)
32229 +{
32230 +       au_nhash_do_free(whlist, au_nhash_wh_do_free);
32231 +}
32232 +
32233 +static void au_nhash_de_free(struct au_nhash *delist)
32234 +{
32235 +       au_nhash_do_free(delist, au_nhash_de_do_free);
32236 +}
32237 +
32238 +/* ---------------------------------------------------------------------- */
32239 +
32240 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
32241 +                           int limit)
32242 +{
32243 +       int num;
32244 +       unsigned int u, n;
32245 +       struct hlist_head *head;
32246 +       struct au_vdir_wh *pos;
32247 +
32248 +       num = 0;
32249 +       n = whlist->nh_num;
32250 +       head = whlist->nh_head;
32251 +       for (u = 0; u < n; u++, head++)
32252 +               hlist_for_each_entry(pos, head, wh_hash)
32253 +                       if (pos->wh_bindex == btgt && ++num > limit)
32254 +                               return 1;
32255 +       return 0;
32256 +}
32257 +
32258 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
32259 +                                      unsigned char *name,
32260 +                                      unsigned int len)
32261 +{
32262 +       unsigned int v;
32263 +       /* const unsigned int magic_bit = 12; */
32264 +
32265 +       AuDebugOn(!nhash->nh_num || !nhash->nh_head);
32266 +
32267 +       v = 0;
32268 +       if (len > 8)
32269 +               len = 8;
32270 +       while (len--)
32271 +               v += *name++;
32272 +       /* v = hash_long(v, magic_bit); */
32273 +       v %= nhash->nh_num;
32274 +       return nhash->nh_head + v;
32275 +}
32276 +
32277 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
32278 +                             int nlen)
32279 +{
32280 +       return str->len == nlen && !memcmp(str->name, name, nlen);
32281 +}
32282 +
32283 +/* returns found or not */
32284 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
32285 +{
32286 +       struct hlist_head *head;
32287 +       struct au_vdir_wh *pos;
32288 +       struct au_vdir_destr *str;
32289 +
32290 +       head = au_name_hash(whlist, name, nlen);
32291 +       hlist_for_each_entry(pos, head, wh_hash) {
32292 +               str = &pos->wh_str;
32293 +               AuDbg("%.*s\n", str->len, str->name);
32294 +               if (au_nhash_test_name(str, name, nlen))
32295 +                       return 1;
32296 +       }
32297 +       return 0;
32298 +}
32299 +
32300 +/* returns found(true) or not */
32301 +static int test_known(struct au_nhash *delist, char *name, int nlen)
32302 +{
32303 +       struct hlist_head *head;
32304 +       struct au_vdir_dehstr *pos;
32305 +       struct au_vdir_destr *str;
32306 +
32307 +       head = au_name_hash(delist, name, nlen);
32308 +       hlist_for_each_entry(pos, head, hash) {
32309 +               str = pos->str;
32310 +               AuDbg("%.*s\n", str->len, str->name);
32311 +               if (au_nhash_test_name(str, name, nlen))
32312 +                       return 1;
32313 +       }
32314 +       return 0;
32315 +}
32316 +
32317 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
32318 +                           unsigned char d_type)
32319 +{
32320 +#ifdef CONFIG_AUFS_SHWH
32321 +       wh->wh_ino = ino;
32322 +       wh->wh_type = d_type;
32323 +#endif
32324 +}
32325 +
32326 +/* ---------------------------------------------------------------------- */
32327 +
32328 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
32329 +                      unsigned int d_type, aufs_bindex_t bindex,
32330 +                      unsigned char shwh)
32331 +{
32332 +       int err;
32333 +       struct au_vdir_destr *str;
32334 +       struct au_vdir_wh *wh;
32335 +
32336 +       AuDbg("%.*s\n", nlen, name);
32337 +       AuDebugOn(!whlist->nh_num || !whlist->nh_head);
32338 +
32339 +       err = -ENOMEM;
32340 +       wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
32341 +       if (unlikely(!wh))
32342 +               goto out;
32343 +
32344 +       err = 0;
32345 +       wh->wh_bindex = bindex;
32346 +       if (shwh)
32347 +               au_shwh_init_wh(wh, ino, d_type);
32348 +       str = &wh->wh_str;
32349 +       str->len = nlen;
32350 +       memcpy(str->name, name, nlen);
32351 +       hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
32352 +       /* smp_mb(); */
32353 +
32354 +out:
32355 +       return err;
32356 +}
32357 +
32358 +static int append_deblk(struct au_vdir *vdir)
32359 +{
32360 +       int err;
32361 +       unsigned long ul;
32362 +       const unsigned int deblk_sz = vdir->vd_deblk_sz;
32363 +       union au_vdir_deblk_p p, deblk_end;
32364 +       unsigned char **o;
32365 +
32366 +       err = -ENOMEM;
32367 +       o = au_krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
32368 +                       GFP_NOFS, /*may_shrink*/0);
32369 +       if (unlikely(!o))
32370 +               goto out;
32371 +
32372 +       vdir->vd_deblk = o;
32373 +       p.deblk = kmalloc(deblk_sz, GFP_NOFS);
32374 +       if (p.deblk) {
32375 +               ul = vdir->vd_nblk++;
32376 +               vdir->vd_deblk[ul] = p.deblk;
32377 +               vdir->vd_last.ul = ul;
32378 +               vdir->vd_last.p.deblk = p.deblk;
32379 +               deblk_end.deblk = p.deblk + deblk_sz;
32380 +               err = set_deblk_end(&p, &deblk_end);
32381 +       }
32382 +
32383 +out:
32384 +       return err;
32385 +}
32386 +
32387 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
32388 +                    unsigned int d_type, struct au_nhash *delist)
32389 +{
32390 +       int err;
32391 +       unsigned int sz;
32392 +       const unsigned int deblk_sz = vdir->vd_deblk_sz;
32393 +       union au_vdir_deblk_p p, *room, deblk_end;
32394 +       struct au_vdir_dehstr *dehstr;
32395 +
32396 +       p.deblk = last_deblk(vdir);
32397 +       deblk_end.deblk = p.deblk + deblk_sz;
32398 +       room = &vdir->vd_last.p;
32399 +       AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
32400 +                 || !is_deblk_end(room, &deblk_end));
32401 +
32402 +       sz = calc_size(nlen);
32403 +       if (unlikely(sz > deblk_end.deblk - room->deblk)) {
32404 +               err = append_deblk(vdir);
32405 +               if (unlikely(err))
32406 +                       goto out;
32407 +
32408 +               p.deblk = last_deblk(vdir);
32409 +               deblk_end.deblk = p.deblk + deblk_sz;
32410 +               /* smp_mb(); */
32411 +               AuDebugOn(room->deblk != p.deblk);
32412 +       }
32413 +
32414 +       err = -ENOMEM;
32415 +       dehstr = au_cache_alloc_vdir_dehstr();
32416 +       if (unlikely(!dehstr))
32417 +               goto out;
32418 +
32419 +       dehstr->str = &room->de->de_str;
32420 +       hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
32421 +       room->de->de_ino = ino;
32422 +       room->de->de_type = d_type;
32423 +       room->de->de_str.len = nlen;
32424 +       memcpy(room->de->de_str.name, name, nlen);
32425 +
32426 +       err = 0;
32427 +       room->deblk += sz;
32428 +       if (unlikely(set_deblk_end(room, &deblk_end)))
32429 +               err = append_deblk(vdir);
32430 +       /* smp_mb(); */
32431 +
32432 +out:
32433 +       return err;
32434 +}
32435 +
32436 +/* ---------------------------------------------------------------------- */
32437 +
32438 +void au_vdir_free(struct au_vdir *vdir)
32439 +{
32440 +       unsigned char **deblk;
32441 +
32442 +       deblk = vdir->vd_deblk;
32443 +       while (vdir->vd_nblk--)
32444 +               au_kfree_try_rcu(*deblk++);
32445 +       au_kfree_try_rcu(vdir->vd_deblk);
32446 +       au_cache_free_vdir(vdir);
32447 +}
32448 +
32449 +static struct au_vdir *alloc_vdir(struct file *file)
32450 +{
32451 +       struct au_vdir *vdir;
32452 +       struct super_block *sb;
32453 +       int err;
32454 +
32455 +       sb = file->f_path.dentry->d_sb;
32456 +       SiMustAnyLock(sb);
32457 +
32458 +       err = -ENOMEM;
32459 +       vdir = au_cache_alloc_vdir();
32460 +       if (unlikely(!vdir))
32461 +               goto out;
32462 +
32463 +       vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
32464 +       if (unlikely(!vdir->vd_deblk))
32465 +               goto out_free;
32466 +
32467 +       vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
32468 +       if (!vdir->vd_deblk_sz) {
32469 +               /* estimate the appropriate size for deblk */
32470 +               vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
32471 +               /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
32472 +       }
32473 +       vdir->vd_nblk = 0;
32474 +       vdir->vd_version = 0;
32475 +       vdir->vd_jiffy = 0;
32476 +       err = append_deblk(vdir);
32477 +       if (!err)
32478 +               return vdir; /* success */
32479 +
32480 +       au_kfree_try_rcu(vdir->vd_deblk);
32481 +
32482 +out_free:
32483 +       au_cache_free_vdir(vdir);
32484 +out:
32485 +       vdir = ERR_PTR(err);
32486 +       return vdir;
32487 +}
32488 +
32489 +static int reinit_vdir(struct au_vdir *vdir)
32490 +{
32491 +       int err;
32492 +       union au_vdir_deblk_p p, deblk_end;
32493 +
32494 +       while (vdir->vd_nblk > 1) {
32495 +               au_kfree_try_rcu(vdir->vd_deblk[vdir->vd_nblk - 1]);
32496 +               /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
32497 +               vdir->vd_nblk--;
32498 +       }
32499 +       p.deblk = vdir->vd_deblk[0];
32500 +       deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
32501 +       err = set_deblk_end(&p, &deblk_end);
32502 +       /* keep vd_dblk_sz */
32503 +       vdir->vd_last.ul = 0;
32504 +       vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32505 +       vdir->vd_version = 0;
32506 +       vdir->vd_jiffy = 0;
32507 +       /* smp_mb(); */
32508 +       return err;
32509 +}
32510 +
32511 +/* ---------------------------------------------------------------------- */
32512 +
32513 +#define AuFillVdir_CALLED      1
32514 +#define AuFillVdir_WHABLE      (1 << 1)
32515 +#define AuFillVdir_SHWH                (1 << 2)
32516 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
32517 +#define au_fset_fillvdir(flags, name) \
32518 +       do { (flags) |= AuFillVdir_##name; } while (0)
32519 +#define au_fclr_fillvdir(flags, name) \
32520 +       do { (flags) &= ~AuFillVdir_##name; } while (0)
32521 +
32522 +#ifndef CONFIG_AUFS_SHWH
32523 +#undef AuFillVdir_SHWH
32524 +#define AuFillVdir_SHWH                0
32525 +#endif
32526 +
32527 +struct fillvdir_arg {
32528 +       struct dir_context      ctx;
32529 +       struct file             *file;
32530 +       struct au_vdir          *vdir;
32531 +       struct au_nhash         delist;
32532 +       struct au_nhash         whlist;
32533 +       aufs_bindex_t           bindex;
32534 +       unsigned int            flags;
32535 +       int                     err;
32536 +};
32537 +
32538 +static bool fillvdir(struct dir_context *ctx, const char *__name, int nlen,
32539 +                   loff_t offset __maybe_unused, u64 h_ino,
32540 +                   unsigned int d_type)
32541 +{
32542 +       struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
32543 +       char *name = (void *)__name;
32544 +       struct super_block *sb;
32545 +       ino_t ino;
32546 +       const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
32547 +
32548 +       arg->err = 0;
32549 +       sb = arg->file->f_path.dentry->d_sb;
32550 +       au_fset_fillvdir(arg->flags, CALLED);
32551 +       /* smp_mb(); */
32552 +       if (nlen <= AUFS_WH_PFX_LEN
32553 +           || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
32554 +               if (test_known(&arg->delist, name, nlen)
32555 +                   || au_nhash_test_known_wh(&arg->whlist, name, nlen))
32556 +                       goto out; /* already exists or whiteouted */
32557 +
32558 +               arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
32559 +               if (!arg->err) {
32560 +                       if (unlikely(nlen > AUFS_MAX_NAMELEN))
32561 +                               d_type = DT_UNKNOWN;
32562 +                       arg->err = append_de(arg->vdir, name, nlen, ino,
32563 +                                            d_type, &arg->delist);
32564 +               }
32565 +       } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
32566 +               name += AUFS_WH_PFX_LEN;
32567 +               nlen -= AUFS_WH_PFX_LEN;
32568 +               if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
32569 +                       goto out; /* already whiteouted */
32570 +
32571 +               ino = 0; /* just to suppress a warning */
32572 +               if (shwh)
32573 +                       arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
32574 +                                            &ino);
32575 +               if (!arg->err) {
32576 +                       if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
32577 +                               d_type = DT_UNKNOWN;
32578 +                       arg->err = au_nhash_append_wh
32579 +                               (&arg->whlist, name, nlen, ino, d_type,
32580 +                                arg->bindex, shwh);
32581 +               }
32582 +       }
32583 +
32584 +out:
32585 +       if (!arg->err)
32586 +               arg->vdir->vd_jiffy = jiffies;
32587 +       /* smp_mb(); */
32588 +       AuTraceErr(arg->err);
32589 +       return !arg->err;
32590 +}
32591 +
32592 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
32593 +                         struct au_nhash *whlist, struct au_nhash *delist)
32594 +{
32595 +#ifdef CONFIG_AUFS_SHWH
32596 +       int err;
32597 +       unsigned int nh, u;
32598 +       struct hlist_head *head;
32599 +       struct au_vdir_wh *pos;
32600 +       struct hlist_node *n;
32601 +       char *p, *o;
32602 +       struct au_vdir_destr *destr;
32603 +
32604 +       AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
32605 +
32606 +       err = -ENOMEM;
32607 +       o = p = (void *)__get_free_page(GFP_NOFS);
32608 +       if (unlikely(!p))
32609 +               goto out;
32610 +
32611 +       err = 0;
32612 +       nh = whlist->nh_num;
32613 +       memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32614 +       p += AUFS_WH_PFX_LEN;
32615 +       for (u = 0; u < nh; u++) {
32616 +               head = whlist->nh_head + u;
32617 +               hlist_for_each_entry_safe(pos, n, head, wh_hash) {
32618 +                       destr = &pos->wh_str;
32619 +                       memcpy(p, destr->name, destr->len);
32620 +                       err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
32621 +                                       pos->wh_ino, pos->wh_type, delist);
32622 +                       if (unlikely(err))
32623 +                               break;
32624 +               }
32625 +       }
32626 +
32627 +       free_page((unsigned long)o);
32628 +
32629 +out:
32630 +       AuTraceErr(err);
32631 +       return err;
32632 +#else
32633 +       return 0;
32634 +#endif
32635 +}
32636 +
32637 +static int au_do_read_vdir(struct fillvdir_arg *arg)
32638 +{
32639 +       int err;
32640 +       unsigned int rdhash;
32641 +       loff_t offset;
32642 +       aufs_bindex_t bbot, bindex, btop;
32643 +       unsigned char shwh;
32644 +       struct file *hf, *file;
32645 +       struct super_block *sb;
32646 +
32647 +       file = arg->file;
32648 +       sb = file->f_path.dentry->d_sb;
32649 +       SiMustAnyLock(sb);
32650 +
32651 +       rdhash = au_sbi(sb)->si_rdhash;
32652 +       if (!rdhash)
32653 +               rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
32654 +       err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
32655 +       if (unlikely(err))
32656 +               goto out;
32657 +       err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
32658 +       if (unlikely(err))
32659 +               goto out_delist;
32660 +
32661 +       err = 0;
32662 +       arg->flags = 0;
32663 +       shwh = 0;
32664 +       if (au_opt_test(au_mntflags(sb), SHWH)) {
32665 +               shwh = 1;
32666 +               au_fset_fillvdir(arg->flags, SHWH);
32667 +       }
32668 +       btop = au_fbtop(file);
32669 +       bbot = au_fbbot_dir(file);
32670 +       for (bindex = btop; !err && bindex <= bbot; bindex++) {
32671 +               hf = au_hf_dir(file, bindex);
32672 +               if (!hf)
32673 +                       continue;
32674 +
32675 +               offset = vfsub_llseek(hf, 0, SEEK_SET);
32676 +               err = offset;
32677 +               if (unlikely(offset))
32678 +                       break;
32679 +
32680 +               arg->bindex = bindex;
32681 +               au_fclr_fillvdir(arg->flags, WHABLE);
32682 +               if (shwh
32683 +                   || (bindex != bbot
32684 +                       && au_br_whable(au_sbr_perm(sb, bindex))))
32685 +                       au_fset_fillvdir(arg->flags, WHABLE);
32686 +               do {
32687 +                       arg->err = 0;
32688 +                       au_fclr_fillvdir(arg->flags, CALLED);
32689 +                       /* smp_mb(); */
32690 +                       err = vfsub_iterate_dir(hf, &arg->ctx);
32691 +                       if (err >= 0)
32692 +                               err = arg->err;
32693 +               } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
32694 +
32695 +               /*
32696 +                * dir_relax() may be good for concurrency, but aufs should not
32697 +                * use it since it will cause a lockdep problem.
32698 +                */
32699 +       }
32700 +
32701 +       if (!err && shwh)
32702 +               err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
32703 +
32704 +       au_nhash_wh_free(&arg->whlist);
32705 +
32706 +out_delist:
32707 +       au_nhash_de_free(&arg->delist);
32708 +out:
32709 +       return err;
32710 +}
32711 +
32712 +static int read_vdir(struct file *file, int may_read)
32713 +{
32714 +       int err;
32715 +       unsigned long expire;
32716 +       unsigned char do_read;
32717 +       struct fillvdir_arg arg = {
32718 +               .ctx = {
32719 +                       .actor = fillvdir
32720 +               }
32721 +       };
32722 +       struct inode *inode;
32723 +       struct au_vdir *vdir, *allocated;
32724 +
32725 +       err = 0;
32726 +       inode = file_inode(file);
32727 +       IMustLock(inode);
32728 +       IiMustWriteLock(inode);
32729 +       SiMustAnyLock(inode->i_sb);
32730 +
32731 +       allocated = NULL;
32732 +       do_read = 0;
32733 +       expire = au_sbi(inode->i_sb)->si_rdcache;
32734 +       vdir = au_ivdir(inode);
32735 +       if (!vdir) {
32736 +               do_read = 1;
32737 +               vdir = alloc_vdir(file);
32738 +               err = PTR_ERR(vdir);
32739 +               if (IS_ERR(vdir))
32740 +                       goto out;
32741 +               err = 0;
32742 +               allocated = vdir;
32743 +       } else if (may_read
32744 +                  && (!inode_eq_iversion(inode, vdir->vd_version)
32745 +                      || time_after(jiffies, vdir->vd_jiffy + expire))) {
32746 +               do_read = 1;
32747 +               err = reinit_vdir(vdir);
32748 +               if (unlikely(err))
32749 +                       goto out;
32750 +       }
32751 +
32752 +       if (!do_read)
32753 +               return 0; /* success */
32754 +
32755 +       arg.file = file;
32756 +       arg.vdir = vdir;
32757 +       err = au_do_read_vdir(&arg);
32758 +       if (!err) {
32759 +               /* file->f_pos = 0; */ /* todo: ctx->pos? */
32760 +               vdir->vd_version = inode_query_iversion(inode);
32761 +               vdir->vd_last.ul = 0;
32762 +               vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32763 +               if (allocated)
32764 +                       au_set_ivdir(inode, allocated);
32765 +       } else if (allocated)
32766 +               au_vdir_free(allocated);
32767 +
32768 +out:
32769 +       return err;
32770 +}
32771 +
32772 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
32773 +{
32774 +       int err, rerr;
32775 +       unsigned long ul, n;
32776 +       const unsigned int deblk_sz = src->vd_deblk_sz;
32777 +
32778 +       AuDebugOn(tgt->vd_nblk != 1);
32779 +
32780 +       err = -ENOMEM;
32781 +       if (tgt->vd_nblk < src->vd_nblk) {
32782 +               unsigned char **p;
32783 +
32784 +               p = au_krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
32785 +                               GFP_NOFS, /*may_shrink*/0);
32786 +               if (unlikely(!p))
32787 +                       goto out;
32788 +               tgt->vd_deblk = p;
32789 +       }
32790 +
32791 +       if (tgt->vd_deblk_sz != deblk_sz) {
32792 +               unsigned char *p;
32793 +
32794 +               tgt->vd_deblk_sz = deblk_sz;
32795 +               p = au_krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS,
32796 +                               /*may_shrink*/1);
32797 +               if (unlikely(!p))
32798 +                       goto out;
32799 +               tgt->vd_deblk[0] = p;
32800 +       }
32801 +       memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
32802 +       tgt->vd_version = src->vd_version;
32803 +       tgt->vd_jiffy = src->vd_jiffy;
32804 +
32805 +       n = src->vd_nblk;
32806 +       for (ul = 1; ul < n; ul++) {
32807 +               tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
32808 +                                           GFP_NOFS);
32809 +               if (unlikely(!tgt->vd_deblk[ul]))
32810 +                       goto out;
32811 +               tgt->vd_nblk++;
32812 +       }
32813 +       tgt->vd_nblk = n;
32814 +       tgt->vd_last.ul = tgt->vd_last.ul;
32815 +       tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
32816 +       tgt->vd_last.p.deblk += src->vd_last.p.deblk
32817 +               - src->vd_deblk[src->vd_last.ul];
32818 +       /* smp_mb(); */
32819 +       return 0; /* success */
32820 +
32821 +out:
32822 +       rerr = reinit_vdir(tgt);
32823 +       BUG_ON(rerr);
32824 +       return err;
32825 +}
32826 +
32827 +int au_vdir_init(struct file *file)
32828 +{
32829 +       int err;
32830 +       struct inode *inode;
32831 +       struct au_vdir *vdir_cache, *allocated;
32832 +
32833 +       /* test file->f_pos here instead of ctx->pos */
32834 +       err = read_vdir(file, !file->f_pos);
32835 +       if (unlikely(err))
32836 +               goto out;
32837 +
32838 +       allocated = NULL;
32839 +       vdir_cache = au_fvdir_cache(file);
32840 +       if (!vdir_cache) {
32841 +               vdir_cache = alloc_vdir(file);
32842 +               err = PTR_ERR(vdir_cache);
32843 +               if (IS_ERR(vdir_cache))
32844 +                       goto out;
32845 +               allocated = vdir_cache;
32846 +       } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
32847 +               /* test file->f_pos here instead of ctx->pos */
32848 +               err = reinit_vdir(vdir_cache);
32849 +               if (unlikely(err))
32850 +                       goto out;
32851 +       } else
32852 +               return 0; /* success */
32853 +
32854 +       inode = file_inode(file);
32855 +       err = copy_vdir(vdir_cache, au_ivdir(inode));
32856 +       if (!err) {
32857 +               file->f_version = inode_query_iversion(inode);
32858 +               if (allocated)
32859 +                       au_set_fvdir_cache(file, allocated);
32860 +       } else if (allocated)
32861 +               au_vdir_free(allocated);
32862 +
32863 +out:
32864 +       return err;
32865 +}
32866 +
32867 +static loff_t calc_offset(struct au_vdir *vdir)
32868 +{
32869 +       loff_t offset;
32870 +       union au_vdir_deblk_p p;
32871 +
32872 +       p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
32873 +       offset = vdir->vd_last.p.deblk - p.deblk;
32874 +       offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
32875 +       return offset;
32876 +}
32877 +
32878 +/* returns true or false */
32879 +static int seek_vdir(struct file *file, struct dir_context *ctx)
32880 +{
32881 +       int valid;
32882 +       unsigned int deblk_sz;
32883 +       unsigned long ul, n;
32884 +       loff_t offset;
32885 +       union au_vdir_deblk_p p, deblk_end;
32886 +       struct au_vdir *vdir_cache;
32887 +
32888 +       valid = 1;
32889 +       vdir_cache = au_fvdir_cache(file);
32890 +       offset = calc_offset(vdir_cache);
32891 +       AuDbg("offset %lld\n", offset);
32892 +       if (ctx->pos == offset)
32893 +               goto out;
32894 +
32895 +       vdir_cache->vd_last.ul = 0;
32896 +       vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
32897 +       if (!ctx->pos)
32898 +               goto out;
32899 +
32900 +       valid = 0;
32901 +       deblk_sz = vdir_cache->vd_deblk_sz;
32902 +       ul = div64_u64(ctx->pos, deblk_sz);
32903 +       AuDbg("ul %lu\n", ul);
32904 +       if (ul >= vdir_cache->vd_nblk)
32905 +               goto out;
32906 +
32907 +       n = vdir_cache->vd_nblk;
32908 +       for (; ul < n; ul++) {
32909 +               p.deblk = vdir_cache->vd_deblk[ul];
32910 +               deblk_end.deblk = p.deblk + deblk_sz;
32911 +               offset = ul;
32912 +               offset *= deblk_sz;
32913 +               while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
32914 +                       unsigned int l;
32915 +
32916 +                       l = calc_size(p.de->de_str.len);
32917 +                       offset += l;
32918 +                       p.deblk += l;
32919 +               }
32920 +               if (!is_deblk_end(&p, &deblk_end)) {
32921 +                       valid = 1;
32922 +                       vdir_cache->vd_last.ul = ul;
32923 +                       vdir_cache->vd_last.p = p;
32924 +                       break;
32925 +               }
32926 +       }
32927 +
32928 +out:
32929 +       /* smp_mb(); */
32930 +       if (!valid)
32931 +               AuDbg("valid %d\n", !valid);
32932 +       return valid;
32933 +}
32934 +
32935 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
32936 +{
32937 +       unsigned int l, deblk_sz;
32938 +       union au_vdir_deblk_p deblk_end;
32939 +       struct au_vdir *vdir_cache;
32940 +       struct au_vdir_de *de;
32941 +
32942 +       if (!seek_vdir(file, ctx))
32943 +               return 0;
32944 +
32945 +       vdir_cache = au_fvdir_cache(file);
32946 +       deblk_sz = vdir_cache->vd_deblk_sz;
32947 +       while (1) {
32948 +               deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32949 +               deblk_end.deblk += deblk_sz;
32950 +               while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
32951 +                       de = vdir_cache->vd_last.p.de;
32952 +                       AuDbg("%.*s, off%lld, i%lu, dt%d\n",
32953 +                             de->de_str.len, de->de_str.name, ctx->pos,
32954 +                             (unsigned long)de->de_ino, de->de_type);
32955 +                       if (unlikely(!dir_emit(ctx, de->de_str.name,
32956 +                                              de->de_str.len, de->de_ino,
32957 +                                              de->de_type))) {
32958 +                               /* todo: ignore the error caused by udba? */
32959 +                               /* return err; */
32960 +                               return 0;
32961 +                       }
32962 +
32963 +                       l = calc_size(de->de_str.len);
32964 +                       vdir_cache->vd_last.p.deblk += l;
32965 +                       ctx->pos += l;
32966 +               }
32967 +               if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
32968 +                       vdir_cache->vd_last.ul++;
32969 +                       vdir_cache->vd_last.p.deblk
32970 +                               = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32971 +                       ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
32972 +                       continue;
32973 +               }
32974 +               break;
32975 +       }
32976 +
32977 +       /* smp_mb(); */
32978 +       return 0;
32979 +}
32980 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
32981 --- /usr/share/empty/fs/aufs/vfsub.c    1970-01-01 01:00:00.000000000 +0100
32982 +++ linux/fs/aufs/vfsub.c       2023-02-20 21:05:51.959693785 +0100
32983 @@ -0,0 +1,918 @@
32984 +// SPDX-License-Identifier: GPL-2.0
32985 +/*
32986 + * Copyright (C) 2005-2022 Junjiro R. Okajima
32987 + *
32988 + * This program is free software; you can redistribute it and/or modify
32989 + * it under the terms of the GNU General Public License as published by
32990 + * the Free Software Foundation; either version 2 of the License, or
32991 + * (at your option) any later version.
32992 + *
32993 + * This program is distributed in the hope that it will be useful,
32994 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32995 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
32996 + * GNU General Public License for more details.
32997 + *
32998 + * You should have received a copy of the GNU General Public License
32999 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
33000 + */
33001 +
33002 +/*
33003 + * sub-routines for VFS
33004 + */
33005 +
33006 +#include <linux/mnt_namespace.h>
33007 +#include <linux/nsproxy.h>
33008 +#include <linux/security.h>
33009 +#include <linux/splice.h>
33010 +#include "aufs.h"
33011 +
33012 +#ifdef CONFIG_AUFS_BR_FUSE
33013 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb)
33014 +{
33015 +       if (!au_test_fuse(h_sb) || !au_userns)
33016 +               return 0;
33017 +
33018 +       return is_current_mnt_ns(mnt) ? 0 : -EACCES;
33019 +}
33020 +#endif
33021 +
33022 +int vfsub_sync_filesystem(struct super_block *h_sb)
33023 +{
33024 +       int err;
33025 +
33026 +       lockdep_off();
33027 +       down_read(&h_sb->s_umount);
33028 +       err = sync_filesystem(h_sb);
33029 +       up_read(&h_sb->s_umount);
33030 +       lockdep_on();
33031 +
33032 +       return err;
33033 +}
33034 +
33035 +/* ---------------------------------------------------------------------- */
33036 +
33037 +int vfsub_update_h_iattr(struct path *h_path, int *did)
33038 +{
33039 +       int err;
33040 +       struct kstat st;
33041 +       struct super_block *h_sb;
33042 +
33043 +       /*
33044 +        * Always needs h_path->mnt for LSM or FUSE branch.
33045 +        */
33046 +       AuDebugOn(!h_path->mnt);
33047 +
33048 +       /* for remote fs, leave work for its getattr or d_revalidate */
33049 +       /* for bad i_attr fs, handle them in aufs_getattr() */
33050 +       /* still some fs may acquire i_mutex. we need to skip them */
33051 +       err = 0;
33052 +       if (!did)
33053 +               did = &err;
33054 +       h_sb = h_path->dentry->d_sb;
33055 +       *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
33056 +       if (*did)
33057 +               err = vfsub_getattr(h_path, &st);
33058 +
33059 +       return err;
33060 +}
33061 +
33062 +/* ---------------------------------------------------------------------- */
33063 +
33064 +struct file *vfsub_dentry_open(struct path *path, int flags)
33065 +{
33066 +       return dentry_open(path, flags /* | __FMODE_NONOTIFY */,
33067 +                          current_cred());
33068 +}
33069 +
33070 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
33071 +{
33072 +       struct file *file;
33073 +
33074 +       lockdep_off();
33075 +       file = filp_open(path,
33076 +                        oflags /* | __FMODE_NONOTIFY */,
33077 +                        mode);
33078 +       lockdep_on();
33079 +       if (IS_ERR(file))
33080 +               goto out;
33081 +       vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33082 +
33083 +out:
33084 +       return file;
33085 +}
33086 +
33087 +/*
33088 + * Ideally this function should call VFS:do_last() in order to keep all its
33089 + * checkings. But it is very hard for aufs to regenerate several VFS internal
33090 + * structure such as nameidata. This is a second (or third) best approach.
33091 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
33092 + */
33093 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33094 +                     struct vfsub_aopen_args *args)
33095 +{
33096 +       int err;
33097 +       struct au_branch *br = args->br;
33098 +       struct file *file = args->file;
33099 +       /* copied from linux/fs/namei.c:atomic_open() */
33100 +       struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
33101 +
33102 +       IMustLock(dir);
33103 +       AuDebugOn(!dir->i_op->atomic_open);
33104 +
33105 +       err = au_br_test_oflag(args->open_flag, br);
33106 +       if (unlikely(err))
33107 +               goto out;
33108 +
33109 +       au_lcnt_inc(&br->br_nfiles);
33110 +       file->f_path.dentry = DENTRY_NOT_SET;
33111 +       file->f_path.mnt = au_br_mnt(br);
33112 +       AuDbg("%ps\n", dir->i_op->atomic_open);
33113 +       err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
33114 +                                    args->create_mode);
33115 +       if (unlikely(err < 0)) {
33116 +               au_lcnt_dec(&br->br_nfiles);
33117 +               goto out;
33118 +       }
33119 +
33120 +       /* temporary workaround for nfsv4 branch */
33121 +       if (au_test_nfs(dir->i_sb))
33122 +               nfs_mark_for_revalidate(dir);
33123 +
33124 +       if (file->f_mode & FMODE_CREATED)
33125 +               fsnotify_create(dir, dentry);
33126 +       if (!(file->f_mode & FMODE_OPENED)) {
33127 +               au_lcnt_dec(&br->br_nfiles);
33128 +               goto out;
33129 +       }
33130 +
33131 +       /* todo: call VFS:may_open() here */
33132 +       /* todo: ima_file_check() too? */
33133 +       if (!err && (args->open_flag & __FMODE_EXEC))
33134 +               err = deny_write_access(file);
33135 +       if (!err)
33136 +               fsnotify_open(file);
33137 +       else
33138 +               au_lcnt_dec(&br->br_nfiles);
33139 +       /* note that the file is created and still opened */
33140 +
33141 +out:
33142 +       return err;
33143 +}
33144 +
33145 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
33146 +{
33147 +       int err;
33148 +
33149 +       err = kern_path(name, flags, path);
33150 +       if (!err && d_is_positive(path->dentry))
33151 +               vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33152 +       return err;
33153 +}
33154 +
33155 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33156 +                                            struct path *ppath, int len)
33157 +{
33158 +       struct path path;
33159 +
33160 +       path.dentry = lookup_one_len_unlocked(name, ppath->dentry, len);
33161 +       if (IS_ERR(path.dentry))
33162 +               goto out;
33163 +       if (d_is_positive(path.dentry)) {
33164 +               path.mnt = ppath->mnt;
33165 +               vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33166 +       }
33167 +
33168 +out:
33169 +       AuTraceErrPtr(path.dentry);
33170 +       return path.dentry;
33171 +}
33172 +
33173 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33174 +                                   int len)
33175 +{
33176 +       struct path path;
33177 +
33178 +       /* VFS checks it too, but by WARN_ON_ONCE() */
33179 +       IMustLock(d_inode(ppath->dentry));
33180 +
33181 +       path.dentry = lookup_one_len(name, ppath->dentry, len);
33182 +       if (IS_ERR(path.dentry))
33183 +               goto out;
33184 +       if (d_is_positive(path.dentry)) {
33185 +               path.mnt = ppath->mnt;
33186 +               vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33187 +       }
33188 +
33189 +out:
33190 +       AuTraceErrPtr(path.dentry);
33191 +       return path.dentry;
33192 +}
33193 +
33194 +void vfsub_call_lkup_one(void *args)
33195 +{
33196 +       struct vfsub_lkup_one_args *a = args;
33197 +       *a->errp = vfsub_lkup_one(a->name, a->ppath);
33198 +}
33199 +
33200 +/* ---------------------------------------------------------------------- */
33201 +
33202 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33203 +                                struct dentry *d2, struct au_hinode *hdir2)
33204 +{
33205 +       struct dentry *d;
33206 +
33207 +       lockdep_off();
33208 +       d = lock_rename(d1, d2);
33209 +       lockdep_on();
33210 +       au_hn_suspend(hdir1);
33211 +       if (hdir1 != hdir2)
33212 +               au_hn_suspend(hdir2);
33213 +
33214 +       return d;
33215 +}
33216 +
33217 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33218 +                        struct dentry *d2, struct au_hinode *hdir2)
33219 +{
33220 +       au_hn_resume(hdir1);
33221 +       if (hdir1 != hdir2)
33222 +               au_hn_resume(hdir2);
33223 +       lockdep_off();
33224 +       unlock_rename(d1, d2);
33225 +       lockdep_on();
33226 +}
33227 +
33228 +/* ---------------------------------------------------------------------- */
33229 +
33230 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
33231 +{
33232 +       int err;
33233 +       struct dentry *d;
33234 +       struct user_namespace *userns;
33235 +
33236 +       IMustLock(dir);
33237 +
33238 +       d = path->dentry;
33239 +       path->dentry = d->d_parent;
33240 +       err = security_path_mknod(path, d, mode, 0);
33241 +       path->dentry = d;
33242 +       if (unlikely(err))
33243 +               goto out;
33244 +       userns = mnt_user_ns(path->mnt);
33245 +
33246 +       lockdep_off();
33247 +       err = vfs_create(userns, dir, path->dentry, mode, want_excl);
33248 +       lockdep_on();
33249 +       if (!err) {
33250 +               struct path tmp = *path;
33251 +               int did;
33252 +
33253 +               vfsub_update_h_iattr(&tmp, &did);
33254 +               if (did) {
33255 +                       tmp.dentry = path->dentry->d_parent;
33256 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33257 +               }
33258 +               /*ignore*/
33259 +       }
33260 +
33261 +out:
33262 +       return err;
33263 +}
33264 +
33265 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
33266 +{
33267 +       int err;
33268 +       struct dentry *d;
33269 +       struct user_namespace *userns;
33270 +
33271 +       IMustLock(dir);
33272 +
33273 +       d = path->dentry;
33274 +       path->dentry = d->d_parent;
33275 +       err = security_path_symlink(path, d, symname);
33276 +       path->dentry = d;
33277 +       if (unlikely(err))
33278 +               goto out;
33279 +       userns = mnt_user_ns(path->mnt);
33280 +
33281 +       lockdep_off();
33282 +       err = vfs_symlink(userns, dir, path->dentry, symname);
33283 +       lockdep_on();
33284 +       if (!err) {
33285 +               struct path tmp = *path;
33286 +               int did;
33287 +
33288 +               vfsub_update_h_iattr(&tmp, &did);
33289 +               if (did) {
33290 +                       tmp.dentry = path->dentry->d_parent;
33291 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33292 +               }
33293 +               /*ignore*/
33294 +       }
33295 +
33296 +out:
33297 +       return err;
33298 +}
33299 +
33300 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
33301 +{
33302 +       int err;
33303 +       struct dentry *d;
33304 +       struct user_namespace *userns;
33305 +
33306 +       IMustLock(dir);
33307 +
33308 +       d = path->dentry;
33309 +       path->dentry = d->d_parent;
33310 +       err = security_path_mknod(path, d, mode, new_encode_dev(dev));
33311 +       path->dentry = d;
33312 +       if (unlikely(err))
33313 +               goto out;
33314 +       userns = mnt_user_ns(path->mnt);
33315 +
33316 +       lockdep_off();
33317 +       err = vfs_mknod(userns, dir, path->dentry, mode, dev);
33318 +       lockdep_on();
33319 +       if (!err) {
33320 +               struct path tmp = *path;
33321 +               int did;
33322 +
33323 +               vfsub_update_h_iattr(&tmp, &did);
33324 +               if (did) {
33325 +                       tmp.dentry = path->dentry->d_parent;
33326 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33327 +               }
33328 +               /*ignore*/
33329 +       }
33330 +
33331 +out:
33332 +       return err;
33333 +}
33334 +
33335 +static int au_test_nlink(struct inode *inode)
33336 +{
33337 +       const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
33338 +
33339 +       if (!au_test_fs_no_limit_nlink(inode->i_sb)
33340 +           || inode->i_nlink < link_max)
33341 +               return 0;
33342 +       return -EMLINK;
33343 +}
33344 +
33345 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
33346 +              struct inode **delegated_inode)
33347 +{
33348 +       int err;
33349 +       struct dentry *d;
33350 +       struct user_namespace *userns;
33351 +
33352 +       IMustLock(dir);
33353 +
33354 +       err = au_test_nlink(d_inode(src_dentry));
33355 +       if (unlikely(err))
33356 +               return err;
33357 +
33358 +       /* we don't call may_linkat() */
33359 +       d = path->dentry;
33360 +       path->dentry = d->d_parent;
33361 +       err = security_path_link(src_dentry, path, d);
33362 +       path->dentry = d;
33363 +       if (unlikely(err))
33364 +               goto out;
33365 +       userns = mnt_user_ns(path->mnt);
33366 +
33367 +       lockdep_off();
33368 +       err = vfs_link(src_dentry, userns, dir, path->dentry, delegated_inode);
33369 +       lockdep_on();
33370 +       if (!err) {
33371 +               struct path tmp = *path;
33372 +               int did;
33373 +
33374 +               /* fuse has different memory inode for the same inumber */
33375 +               vfsub_update_h_iattr(&tmp, &did);
33376 +               if (did) {
33377 +                       tmp.dentry = path->dentry->d_parent;
33378 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33379 +                       tmp.dentry = src_dentry;
33380 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33381 +               }
33382 +               /*ignore*/
33383 +       }
33384 +
33385 +out:
33386 +       return err;
33387 +}
33388 +
33389 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
33390 +                struct inode *dir, struct path *path,
33391 +                struct inode **delegated_inode, unsigned int flags)
33392 +{
33393 +       int err;
33394 +       struct renamedata rd;
33395 +       struct path tmp = {
33396 +               .mnt    = path->mnt
33397 +       };
33398 +       struct dentry *d;
33399 +
33400 +       IMustLock(dir);
33401 +       IMustLock(src_dir);
33402 +
33403 +       d = path->dentry;
33404 +       path->dentry = d->d_parent;
33405 +       tmp.dentry = src_dentry->d_parent;
33406 +       err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
33407 +       path->dentry = d;
33408 +       if (unlikely(err))
33409 +               goto out;
33410 +
33411 +       rd.old_mnt_userns = mnt_user_ns(path->mnt);
33412 +       rd.old_dir = src_dir;
33413 +       rd.old_dentry = src_dentry;
33414 +       rd.new_mnt_userns = rd.old_mnt_userns;
33415 +       rd.new_dir = dir;
33416 +       rd.new_dentry = path->dentry;
33417 +       rd.delegated_inode = delegated_inode;
33418 +       rd.flags = flags;
33419 +       lockdep_off();
33420 +       err = vfs_rename(&rd);
33421 +       lockdep_on();
33422 +       if (!err) {
33423 +               int did;
33424 +
33425 +               tmp.dentry = d->d_parent;
33426 +               vfsub_update_h_iattr(&tmp, &did);
33427 +               if (did) {
33428 +                       tmp.dentry = src_dentry;
33429 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33430 +                       tmp.dentry = src_dentry->d_parent;
33431 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33432 +               }
33433 +               /*ignore*/
33434 +       }
33435 +
33436 +out:
33437 +       return err;
33438 +}
33439 +
33440 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
33441 +{
33442 +       int err;
33443 +       struct dentry *d;
33444 +       struct user_namespace *userns;
33445 +
33446 +       IMustLock(dir);
33447 +
33448 +       d = path->dentry;
33449 +       path->dentry = d->d_parent;
33450 +       err = security_path_mkdir(path, d, mode);
33451 +       path->dentry = d;
33452 +       if (unlikely(err))
33453 +               goto out;
33454 +       userns = mnt_user_ns(path->mnt);
33455 +
33456 +       lockdep_off();
33457 +       err = vfs_mkdir(userns, dir, path->dentry, mode);
33458 +       lockdep_on();
33459 +       if (!err) {
33460 +               struct path tmp = *path;
33461 +               int did;
33462 +
33463 +               vfsub_update_h_iattr(&tmp, &did);
33464 +               if (did) {
33465 +                       tmp.dentry = path->dentry->d_parent;
33466 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33467 +               }
33468 +               /*ignore*/
33469 +       }
33470 +
33471 +out:
33472 +       return err;
33473 +}
33474 +
33475 +int vfsub_rmdir(struct inode *dir, struct path *path)
33476 +{
33477 +       int err;
33478 +       struct dentry *d;
33479 +       struct user_namespace *userns;
33480 +
33481 +       IMustLock(dir);
33482 +
33483 +       d = path->dentry;
33484 +       path->dentry = d->d_parent;
33485 +       err = security_path_rmdir(path, d);
33486 +       path->dentry = d;
33487 +       if (unlikely(err))
33488 +               goto out;
33489 +       userns = mnt_user_ns(path->mnt);
33490 +
33491 +       lockdep_off();
33492 +       err = vfs_rmdir(userns, dir, path->dentry);
33493 +       lockdep_on();
33494 +       if (!err) {
33495 +               struct path tmp = {
33496 +                       .dentry = path->dentry->d_parent,
33497 +                       .mnt    = path->mnt
33498 +               };
33499 +
33500 +               vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33501 +       }
33502 +
33503 +out:
33504 +       return err;
33505 +}
33506 +
33507 +/* ---------------------------------------------------------------------- */
33508 +
33509 +/* todo: support mmap_sem? */
33510 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
33511 +                    loff_t *ppos)
33512 +{
33513 +       ssize_t err;
33514 +
33515 +       lockdep_off();
33516 +       err = vfs_read(file, ubuf, count, ppos);
33517 +       lockdep_on();
33518 +       if (err >= 0)
33519 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33520 +       return err;
33521 +}
33522 +
33523 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
33524 +                    loff_t *ppos)
33525 +{
33526 +       ssize_t err;
33527 +
33528 +       lockdep_off();
33529 +       err = kernel_read(file, kbuf, count, ppos);
33530 +       lockdep_on();
33531 +       AuTraceErr(err);
33532 +       if (err >= 0)
33533 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33534 +       return err;
33535 +}
33536 +
33537 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
33538 +                     loff_t *ppos)
33539 +{
33540 +       ssize_t err;
33541 +
33542 +       lockdep_off();
33543 +       err = vfs_write(file, ubuf, count, ppos);
33544 +       lockdep_on();
33545 +       if (err >= 0)
33546 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33547 +       return err;
33548 +}
33549 +
33550 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
33551 +{
33552 +       ssize_t err;
33553 +
33554 +       lockdep_off();
33555 +       err = kernel_write(file, kbuf, count, ppos);
33556 +       lockdep_on();
33557 +       if (err >= 0)
33558 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33559 +       return err;
33560 +}
33561 +
33562 +int vfsub_flush(struct file *file, fl_owner_t id)
33563 +{
33564 +       int err;
33565 +
33566 +       err = 0;
33567 +       if (file->f_op->flush) {
33568 +               if (!au_test_nfs(file->f_path.dentry->d_sb))
33569 +                       err = file->f_op->flush(file, id);
33570 +               else {
33571 +                       lockdep_off();
33572 +                       err = file->f_op->flush(file, id);
33573 +                       lockdep_on();
33574 +               }
33575 +               if (!err)
33576 +                       vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
33577 +               /*ignore*/
33578 +       }
33579 +       return err;
33580 +}
33581 +
33582 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
33583 +{
33584 +       int err;
33585 +
33586 +       AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
33587 +
33588 +       lockdep_off();
33589 +       err = iterate_dir(file, ctx);
33590 +       lockdep_on();
33591 +       if (err >= 0)
33592 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33593 +
33594 +       return err;
33595 +}
33596 +
33597 +long vfsub_splice_to(struct file *in, loff_t *ppos,
33598 +                    struct pipe_inode_info *pipe, size_t len,
33599 +                    unsigned int flags)
33600 +{
33601 +       long err;
33602 +
33603 +       lockdep_off();
33604 +       err = do_splice_to(in, ppos, pipe, len, flags);
33605 +       lockdep_on();
33606 +       file_accessed(in);
33607 +       if (err >= 0)
33608 +               vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
33609 +       return err;
33610 +}
33611 +
33612 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
33613 +                      loff_t *ppos, size_t len, unsigned int flags)
33614 +{
33615 +       long err;
33616 +
33617 +       lockdep_off();
33618 +       err = do_splice_from(pipe, out, ppos, len, flags);
33619 +       lockdep_on();
33620 +       if (err >= 0)
33621 +               vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
33622 +       return err;
33623 +}
33624 +
33625 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
33626 +{
33627 +       int err;
33628 +
33629 +       /* file can be NULL */
33630 +       lockdep_off();
33631 +       err = vfs_fsync(file, datasync);
33632 +       lockdep_on();
33633 +       if (!err) {
33634 +               if (!path) {
33635 +                       AuDebugOn(!file);
33636 +                       path = &file->f_path;
33637 +               }
33638 +               vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33639 +       }
33640 +       return err;
33641 +}
33642 +
33643 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
33644 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33645 +               struct file *h_file)
33646 +{
33647 +       int err;
33648 +       struct inode *h_inode;
33649 +       struct super_block *h_sb;
33650 +       struct user_namespace *h_userns;
33651 +
33652 +       if (!h_file) {
33653 +               err = vfsub_truncate(h_path, length);
33654 +               goto out;
33655 +       }
33656 +
33657 +       h_inode = d_inode(h_path->dentry);
33658 +       h_sb = h_inode->i_sb;
33659 +       lockdep_off();
33660 +       sb_start_write(h_sb);
33661 +       lockdep_on();
33662 +       err = security_file_truncate(h_file);
33663 +       if (!err) {
33664 +               h_userns = mnt_user_ns(h_path->mnt);
33665 +               lockdep_off();
33666 +               err = do_truncate(h_userns, h_path->dentry, length, attr,
33667 +                                 h_file);
33668 +               lockdep_on();
33669 +       }
33670 +       lockdep_off();
33671 +       sb_end_write(h_sb);
33672 +       lockdep_on();
33673 +
33674 +out:
33675 +       return err;
33676 +}
33677 +
33678 +/* ---------------------------------------------------------------------- */
33679 +
33680 +struct au_vfsub_mkdir_args {
33681 +       int *errp;
33682 +       struct inode *dir;
33683 +       struct path *path;
33684 +       int mode;
33685 +};
33686 +
33687 +static void au_call_vfsub_mkdir(void *args)
33688 +{
33689 +       struct au_vfsub_mkdir_args *a = args;
33690 +       *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
33691 +}
33692 +
33693 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
33694 +{
33695 +       int err, do_sio, wkq_err;
33696 +       struct user_namespace *userns;
33697 +
33698 +       userns = mnt_user_ns(path->mnt);
33699 +       do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33700 +       if (!do_sio) {
33701 +               lockdep_off();
33702 +               err = vfsub_mkdir(dir, path, mode);
33703 +               lockdep_on();
33704 +       } else {
33705 +               struct au_vfsub_mkdir_args args = {
33706 +                       .errp   = &err,
33707 +                       .dir    = dir,
33708 +                       .path   = path,
33709 +                       .mode   = mode
33710 +               };
33711 +               wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
33712 +               if (unlikely(wkq_err))
33713 +                       err = wkq_err;
33714 +       }
33715 +
33716 +       return err;
33717 +}
33718 +
33719 +struct au_vfsub_rmdir_args {
33720 +       int *errp;
33721 +       struct inode *dir;
33722 +       struct path *path;
33723 +};
33724 +
33725 +static void au_call_vfsub_rmdir(void *args)
33726 +{
33727 +       struct au_vfsub_rmdir_args *a = args;
33728 +       *a->errp = vfsub_rmdir(a->dir, a->path);
33729 +}
33730 +
33731 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
33732 +{
33733 +       int err, do_sio, wkq_err;
33734 +       struct user_namespace *userns;
33735 +
33736 +       userns = mnt_user_ns(path->mnt);
33737 +       do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33738 +       if (!do_sio) {
33739 +               lockdep_off();
33740 +               err = vfsub_rmdir(dir, path);
33741 +               lockdep_on();
33742 +       } else {
33743 +               struct au_vfsub_rmdir_args args = {
33744 +                       .errp   = &err,
33745 +                       .dir    = dir,
33746 +                       .path   = path
33747 +               };
33748 +               wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
33749 +               if (unlikely(wkq_err))
33750 +                       err = wkq_err;
33751 +       }
33752 +
33753 +       return err;
33754 +}
33755 +
33756 +/* ---------------------------------------------------------------------- */
33757 +
33758 +struct notify_change_args {
33759 +       int *errp;
33760 +       struct path *path;
33761 +       struct iattr *ia;
33762 +       struct inode **delegated_inode;
33763 +};
33764 +
33765 +static void call_notify_change(void *args)
33766 +{
33767 +       struct notify_change_args *a = args;
33768 +       struct inode *h_inode;
33769 +       struct user_namespace *userns;
33770 +
33771 +       h_inode = d_inode(a->path->dentry);
33772 +       IMustLock(h_inode);
33773 +
33774 +       *a->errp = -EPERM;
33775 +       if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
33776 +               userns = mnt_user_ns(a->path->mnt);
33777 +               lockdep_off();
33778 +               *a->errp = notify_change(userns, a->path->dentry, a->ia,
33779 +                                        a->delegated_inode);
33780 +               lockdep_on();
33781 +               if (!*a->errp)
33782 +                       vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
33783 +       }
33784 +       AuTraceErr(*a->errp);
33785 +}
33786 +
33787 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33788 +                       struct inode **delegated_inode)
33789 +{
33790 +       int err;
33791 +       struct notify_change_args args = {
33792 +               .errp                   = &err,
33793 +               .path                   = path,
33794 +               .ia                     = ia,
33795 +               .delegated_inode        = delegated_inode
33796 +       };
33797 +
33798 +       call_notify_change(&args);
33799 +
33800 +       return err;
33801 +}
33802 +
33803 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33804 +                           struct inode **delegated_inode)
33805 +{
33806 +       int err, wkq_err;
33807 +       struct notify_change_args args = {
33808 +               .errp                   = &err,
33809 +               .path                   = path,
33810 +               .ia                     = ia,
33811 +               .delegated_inode        = delegated_inode
33812 +       };
33813 +
33814 +       wkq_err = au_wkq_wait(call_notify_change, &args);
33815 +       if (unlikely(wkq_err))
33816 +               err = wkq_err;
33817 +
33818 +       return err;
33819 +}
33820 +
33821 +/* ---------------------------------------------------------------------- */
33822 +
33823 +struct unlink_args {
33824 +       int *errp;
33825 +       struct inode *dir;
33826 +       struct path *path;
33827 +       struct inode **delegated_inode;
33828 +};
33829 +
33830 +static void call_unlink(void *args)
33831 +{
33832 +       struct unlink_args *a = args;
33833 +       struct dentry *d = a->path->dentry;
33834 +       struct inode *h_inode;
33835 +       struct user_namespace *userns;
33836 +       const int stop_sillyrename = (au_test_nfs(d->d_sb)
33837 +                                     && au_dcount(d) == 1);
33838 +
33839 +       IMustLock(a->dir);
33840 +
33841 +       a->path->dentry = d->d_parent;
33842 +       *a->errp = security_path_unlink(a->path, d);
33843 +       a->path->dentry = d;
33844 +       if (unlikely(*a->errp))
33845 +               return;
33846 +
33847 +       if (!stop_sillyrename)
33848 +               dget(d);
33849 +       h_inode = NULL;
33850 +       if (d_is_positive(d)) {
33851 +               h_inode = d_inode(d);
33852 +               ihold(h_inode);
33853 +       }
33854 +
33855 +       userns = mnt_user_ns(a->path->mnt);
33856 +       lockdep_off();
33857 +       *a->errp = vfs_unlink(userns, a->dir, d, a->delegated_inode);
33858 +       lockdep_on();
33859 +       if (!*a->errp) {
33860 +               struct path tmp = {
33861 +                       .dentry = d->d_parent,
33862 +                       .mnt    = a->path->mnt
33863 +               };
33864 +               vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33865 +       }
33866 +
33867 +       if (!stop_sillyrename)
33868 +               dput(d);
33869 +       if (h_inode)
33870 +               iput(h_inode);
33871 +
33872 +       AuTraceErr(*a->errp);
33873 +}
33874 +
33875 +/*
33876 + * @dir: must be locked.
33877 + * @dentry: target dentry.
33878 + */
33879 +int vfsub_unlink(struct inode *dir, struct path *path,
33880 +                struct inode **delegated_inode, int force)
33881 +{
33882 +       int err;
33883 +       struct unlink_args args = {
33884 +               .errp                   = &err,
33885 +               .dir                    = dir,
33886 +               .path                   = path,
33887 +               .delegated_inode        = delegated_inode
33888 +       };
33889 +
33890 +       if (!force)
33891 +               call_unlink(&args);
33892 +       else {
33893 +               int wkq_err;
33894 +
33895 +               wkq_err = au_wkq_wait(call_unlink, &args);
33896 +               if (unlikely(wkq_err))
33897 +                       err = wkq_err;
33898 +       }
33899 +
33900 +       return err;
33901 +}
33902 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
33903 --- /usr/share/empty/fs/aufs/vfsub.h    1970-01-01 01:00:00.000000000 +0100
33904 +++ linux/fs/aufs/vfsub.h       2023-02-20 21:05:51.959693785 +0100
33905 @@ -0,0 +1,390 @@
33906 +/* SPDX-License-Identifier: GPL-2.0 */
33907 +/*
33908 + * Copyright (C) 2005-2022 Junjiro R. Okajima
33909 + *
33910 + * This program is free software; you can redistribute it and/or modify
33911 + * it under the terms of the GNU General Public License as published by
33912 + * the Free Software Foundation; either version 2 of the License, or
33913 + * (at your option) any later version.
33914 + *
33915 + * This program is distributed in the hope that it will be useful,
33916 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33917 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
33918 + * GNU General Public License for more details.
33919 + *
33920 + * You should have received a copy of the GNU General Public License
33921 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
33922 + */
33923 +
33924 +/*
33925 + * sub-routines for VFS
33926 + */
33927 +
33928 +#ifndef __AUFS_VFSUB_H__
33929 +#define __AUFS_VFSUB_H__
33930 +
33931 +#ifdef __KERNEL__
33932 +
33933 +#include <linux/fs.h>
33934 +#include <linux/mount.h>
33935 +#include <linux/posix_acl.h>
33936 +#include <linux/xattr.h>
33937 +#include "debug.h"
33938 +
33939 +/* copied from linux/fs/internal.h */
33940 +/* todo: BAD approach!! */
33941 +extern void __mnt_drop_write(struct vfsmount *);
33942 +extern struct file *alloc_empty_file(int, const struct cred *);
33943 +
33944 +/* ---------------------------------------------------------------------- */
33945 +
33946 +/* lock subclass for lower inode */
33947 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
33948 +/* reduce? gave up. */
33949 +enum {
33950 +       AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
33951 +       AuLsc_I_PARENT,         /* lower inode, parent first */
33952 +       AuLsc_I_PARENT2,        /* copyup dirs */
33953 +       AuLsc_I_PARENT3,        /* copyup wh */
33954 +       AuLsc_I_CHILD,
33955 +       AuLsc_I_CHILD2,
33956 +       AuLsc_I_End
33957 +};
33958 +
33959 +/* to debug easier, do not make them inlined functions */
33960 +#define MtxMustLock(mtx)       AuDebugOn(!mutex_is_locked(mtx))
33961 +#define IMustLock(i)           AuDebugOn(!inode_is_locked(i))
33962 +
33963 +/* ---------------------------------------------------------------------- */
33964 +
33965 +static inline void vfsub_drop_nlink(struct inode *inode)
33966 +{
33967 +       AuDebugOn(!inode->i_nlink);
33968 +       drop_nlink(inode);
33969 +}
33970 +
33971 +static inline void vfsub_dead_dir(struct inode *inode)
33972 +{
33973 +       AuDebugOn(!S_ISDIR(inode->i_mode));
33974 +       inode->i_flags |= S_DEAD;
33975 +       clear_nlink(inode);
33976 +}
33977 +
33978 +static inline int vfsub_native_ro(struct inode *inode)
33979 +{
33980 +       return sb_rdonly(inode->i_sb)
33981 +               || IS_RDONLY(inode)
33982 +               /* || IS_APPEND(inode) */
33983 +               || IS_IMMUTABLE(inode);
33984 +}
33985 +
33986 +#ifdef CONFIG_AUFS_BR_FUSE
33987 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb);
33988 +#else
33989 +AuStubInt0(vfsub_test_mntns, struct vfsmount *mnt, struct super_block *h_sb);
33990 +#endif
33991 +
33992 +int vfsub_sync_filesystem(struct super_block *h_sb);
33993 +
33994 +/* ---------------------------------------------------------------------- */
33995 +
33996 +int vfsub_update_h_iattr(struct path *h_path, int *did);
33997 +struct file *vfsub_dentry_open(struct path *path, int flags);
33998 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
33999 +struct au_branch;
34000 +struct vfsub_aopen_args {
34001 +       struct file             *file;
34002 +       unsigned int            open_flag;
34003 +       umode_t                 create_mode;
34004 +       struct au_branch        *br;
34005 +};
34006 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
34007 +                     struct vfsub_aopen_args *args);
34008 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
34009 +
34010 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
34011 +                                            struct path *ppath, int len);
34012 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
34013 +                                   int len);
34014 +
34015 +struct vfsub_lkup_one_args {
34016 +       struct dentry **errp;
34017 +       struct qstr *name;
34018 +       struct path *ppath;
34019 +};
34020 +
34021 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
34022 +                                           struct path *ppath)
34023 +{
34024 +       return vfsub_lookup_one_len(name->name, ppath, name->len);
34025 +}
34026 +
34027 +void vfsub_call_lkup_one(void *args);
34028 +
34029 +/* ---------------------------------------------------------------------- */
34030 +
34031 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
34032 +{
34033 +       int err;
34034 +
34035 +       lockdep_off();
34036 +       err = mnt_want_write(mnt);
34037 +       lockdep_on();
34038 +       return err;
34039 +}
34040 +
34041 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
34042 +{
34043 +       lockdep_off();
34044 +       mnt_drop_write(mnt);
34045 +       lockdep_on();
34046 +}
34047 +
34048 +#if 0 /* reserved */
34049 +static inline void vfsub_mnt_drop_write_file(struct file *file)
34050 +{
34051 +       lockdep_off();
34052 +       mnt_drop_write_file(file);
34053 +       lockdep_on();
34054 +}
34055 +#endif
34056 +
34057 +/* ---------------------------------------------------------------------- */
34058 +
34059 +struct au_hinode;
34060 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
34061 +                                struct dentry *d2, struct au_hinode *hdir2);
34062 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
34063 +                        struct dentry *d2, struct au_hinode *hdir2);
34064 +
34065 +int vfsub_create(struct inode *dir, struct path *path, int mode,
34066 +                bool want_excl);
34067 +int vfsub_symlink(struct inode *dir, struct path *path,
34068 +                 const char *symname);
34069 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
34070 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
34071 +              struct path *path, struct inode **delegated_inode);
34072 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
34073 +                struct inode *hdir, struct path *path,
34074 +                struct inode **delegated_inode, unsigned int flags);
34075 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
34076 +int vfsub_rmdir(struct inode *dir, struct path *path);
34077 +
34078 +/* ---------------------------------------------------------------------- */
34079 +
34080 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
34081 +                    loff_t *ppos);
34082 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
34083 +                       loff_t *ppos);
34084 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
34085 +                     loff_t *ppos);
34086 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
34087 +                     loff_t *ppos);
34088 +int vfsub_flush(struct file *file, fl_owner_t id);
34089 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
34090 +
34091 +static inline loff_t vfsub_f_size_read(struct file *file)
34092 +{
34093 +       return i_size_read(file_inode(file));
34094 +}
34095 +
34096 +static inline unsigned int vfsub_file_flags(struct file *file)
34097 +{
34098 +       unsigned int flags;
34099 +
34100 +       spin_lock(&file->f_lock);
34101 +       flags = file->f_flags;
34102 +       spin_unlock(&file->f_lock);
34103 +
34104 +       return flags;
34105 +}
34106 +
34107 +static inline int vfsub_file_execed(struct file *file)
34108 +{
34109 +       /* todo: direct access f_flags */
34110 +       return !!(vfsub_file_flags(file) & __FMODE_EXEC);
34111 +}
34112 +
34113 +#if 0 /* reserved */
34114 +static inline void vfsub_file_accessed(struct file *h_file)
34115 +{
34116 +       file_accessed(h_file);
34117 +       vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
34118 +}
34119 +#endif
34120 +
34121 +#if 0 /* reserved */
34122 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
34123 +                                    struct dentry *h_dentry)
34124 +{
34125 +       struct path h_path = {
34126 +               .dentry = h_dentry,
34127 +               .mnt    = h_mnt
34128 +       };
34129 +       touch_atime(&h_path);
34130 +       vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
34131 +}
34132 +#endif
34133 +
34134 +static inline int vfsub_update_time(struct inode *h_inode,
34135 +                                   struct timespec64 *ts, int flags)
34136 +{
34137 +       return inode_update_time(h_inode, ts, flags);
34138 +       /* no vfsub_update_h_iattr() since we don't have struct path */
34139 +}
34140 +
34141 +#ifdef CONFIG_FS_POSIX_ACL
34142 +static inline int vfsub_acl_chmod(struct user_namespace *h_userns,
34143 +                                 struct dentry *h_dentry, umode_t h_mode)
34144 +{
34145 +       int err;
34146 +
34147 +       err = posix_acl_chmod(h_userns, h_dentry, h_mode);
34148 +       if (err == -EOPNOTSUPP)
34149 +               err = 0;
34150 +       return err;
34151 +}
34152 +#else
34153 +AuStubInt0(vfsub_acl_chmod, struct user_namespace *h_userns,
34154 +          struct dentry *h_dentry, umode_t h_mode);
34155 +#endif
34156 +
34157 +long vfsub_splice_to(struct file *in, loff_t *ppos,
34158 +                    struct pipe_inode_info *pipe, size_t len,
34159 +                    unsigned int flags);
34160 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
34161 +                      loff_t *ppos, size_t len, unsigned int flags);
34162 +
34163 +static inline long vfsub_truncate(struct path *path, loff_t length)
34164 +{
34165 +       long err;
34166 +
34167 +       lockdep_off();
34168 +       err = vfs_truncate(path, length);
34169 +       lockdep_on();
34170 +       return err;
34171 +}
34172 +
34173 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
34174 +               struct file *h_file);
34175 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
34176 +
34177 +/*
34178 + * re-use branch fs's ioctl(FICLONE) while aufs itself doesn't support such
34179 + * ioctl.
34180 + */
34181 +static inline loff_t vfsub_clone_file_range(struct file *src, struct file *dst,
34182 +                                           loff_t len)
34183 +{
34184 +       loff_t err;
34185 +
34186 +       lockdep_off();
34187 +       err = vfs_clone_file_range(src, 0, dst, 0, len, /*remap_flags*/0);
34188 +       lockdep_on();
34189 +
34190 +       return err;
34191 +}
34192 +
34193 +/* copy_file_range(2) is a systemcall */
34194 +static inline ssize_t vfsub_copy_file_range(struct file *src, loff_t src_pos,
34195 +                                           struct file *dst, loff_t dst_pos,
34196 +                                           size_t len, unsigned int flags)
34197 +{
34198 +       ssize_t ssz;
34199 +
34200 +       lockdep_off();
34201 +       ssz = vfs_copy_file_range(src, src_pos, dst, dst_pos, len, flags);
34202 +       lockdep_on();
34203 +
34204 +       return ssz;
34205 +}
34206 +
34207 +/* ---------------------------------------------------------------------- */
34208 +
34209 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
34210 +{
34211 +       loff_t err;
34212 +
34213 +       lockdep_off();
34214 +       err = vfs_llseek(file, offset, origin);
34215 +       lockdep_on();
34216 +       return err;
34217 +}
34218 +
34219 +/* ---------------------------------------------------------------------- */
34220 +
34221 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
34222 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
34223 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
34224 +                           struct inode **delegated_inode);
34225 +int vfsub_notify_change(struct path *path, struct iattr *ia,
34226 +                       struct inode **delegated_inode);
34227 +int vfsub_unlink(struct inode *dir, struct path *path,
34228 +                struct inode **delegated_inode, int force);
34229 +
34230 +static inline int vfsub_getattr(const struct path *path, struct kstat *st)
34231 +{
34232 +       return vfs_getattr(path, st, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
34233 +}
34234 +
34235 +/* ---------------------------------------------------------------------- */
34236 +
34237 +static inline int vfsub_setxattr(struct user_namespace *userns,
34238 +                                struct dentry *dentry, const char *name,
34239 +                                const void *value, size_t size, int flags)
34240 +{
34241 +       int err;
34242 +
34243 +       lockdep_off();
34244 +       err = vfs_setxattr(userns, dentry, name, value, size, flags);
34245 +       lockdep_on();
34246 +
34247 +       return err;
34248 +}
34249 +
34250 +static inline int vfsub_removexattr(struct user_namespace *userns,
34251 +                                   struct dentry *dentry, const char *name)
34252 +{
34253 +       int err;
34254 +
34255 +       lockdep_off();
34256 +       err = vfs_removexattr(userns, dentry, name);
34257 +       lockdep_on();
34258 +
34259 +       return err;
34260 +}
34261 +
34262 +#ifdef CONFIG_FS_POSIX_ACL
34263 +static inline int vfsub_set_acl(struct user_namespace *userns,
34264 +                               struct dentry *dentry, const char *name,
34265 +                               struct posix_acl *acl)
34266 +{
34267 +       int err;
34268 +
34269 +       lockdep_off();
34270 +       err = vfs_set_acl(userns, dentry, name, acl);
34271 +       lockdep_on();
34272 +
34273 +       return err;
34274 +}
34275 +
34276 +static inline int vfsub_remove_acl(struct user_namespace *userns,
34277 +                                  struct dentry *dentry, const char *name)
34278 +{
34279 +       int err;
34280 +
34281 +       lockdep_off();
34282 +       err = vfs_remove_acl(userns, dentry, name);
34283 +       lockdep_on();
34284 +
34285 +       return err;
34286 +}
34287 +#else
34288 +AuStubInt0(vfsub_set_acl, struct user_namespace *userns, struct dentry *dentry,
34289 +          const char *name, struct posix_acl *acl);
34290 +AuStubInt0(vfsub_remove_acl, struct user_namespace *userns,
34291 +          struct dentry *dentry, const char *name);
34292 +#endif
34293 +
34294 +#endif /* __KERNEL__ */
34295 +#endif /* __AUFS_VFSUB_H__ */
34296 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
34297 --- /usr/share/empty/fs/aufs/wbr_policy.c       1970-01-01 01:00:00.000000000 +0100
34298 +++ linux/fs/aufs/wbr_policy.c  2022-11-05 23:02:18.969222617 +0100
34299 @@ -0,0 +1,830 @@
34300 +// SPDX-License-Identifier: GPL-2.0
34301 +/*
34302 + * Copyright (C) 2005-2022 Junjiro R. Okajima
34303 + *
34304 + * This program is free software; you can redistribute it and/or modify
34305 + * it under the terms of the GNU General Public License as published by
34306 + * the Free Software Foundation; either version 2 of the License, or
34307 + * (at your option) any later version.
34308 + *
34309 + * This program is distributed in the hope that it will be useful,
34310 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
34311 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
34312 + * GNU General Public License for more details.
34313 + *
34314 + * You should have received a copy of the GNU General Public License
34315 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
34316 + */
34317 +
34318 +/*
34319 + * policies for selecting one among multiple writable branches
34320 + */
34321 +
34322 +#include <linux/statfs.h>
34323 +#include "aufs.h"
34324 +
34325 +/* subset of cpup_attr() */
34326 +static noinline_for_stack
34327 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
34328 +{
34329 +       int err, sbits;
34330 +       struct iattr ia;
34331 +       struct inode *h_isrc;
34332 +
34333 +       h_isrc = d_inode(h_src);
34334 +       ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
34335 +       ia.ia_mode = h_isrc->i_mode;
34336 +       ia.ia_uid = h_isrc->i_uid;
34337 +       ia.ia_gid = h_isrc->i_gid;
34338 +       sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
34339 +       au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
34340 +       /* no delegation since it is just created */
34341 +       err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34342 +
34343 +       /* is this nfs only? */
34344 +       if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
34345 +               ia.ia_valid = ATTR_FORCE | ATTR_MODE;
34346 +               ia.ia_mode = h_isrc->i_mode;
34347 +               err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34348 +       }
34349 +
34350 +       return err;
34351 +}
34352 +
34353 +#define AuCpdown_PARENT_OPQ    1
34354 +#define AuCpdown_WHED          (1 << 1)
34355 +#define AuCpdown_MADE_DIR      (1 << 2)
34356 +#define AuCpdown_DIROPQ                (1 << 3)
34357 +#define au_ftest_cpdown(flags, name)   ((flags) & AuCpdown_##name)
34358 +#define au_fset_cpdown(flags, name) \
34359 +       do { (flags) |= AuCpdown_##name; } while (0)
34360 +#define au_fclr_cpdown(flags, name) \
34361 +       do { (flags) &= ~AuCpdown_##name; } while (0)
34362 +
34363 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
34364 +                            unsigned int *flags)
34365 +{
34366 +       int err;
34367 +       struct dentry *opq_dentry;
34368 +
34369 +       opq_dentry = au_diropq_create(dentry, bdst);
34370 +       err = PTR_ERR(opq_dentry);
34371 +       if (IS_ERR(opq_dentry))
34372 +               goto out;
34373 +       dput(opq_dentry);
34374 +       au_fset_cpdown(*flags, DIROPQ);
34375 +
34376 +out:
34377 +       return err;
34378 +}
34379 +
34380 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
34381 +                           struct inode *dir, aufs_bindex_t bdst)
34382 +{
34383 +       int err;
34384 +       struct path h_path;
34385 +       struct au_branch *br;
34386 +
34387 +       br = au_sbr(dentry->d_sb, bdst);
34388 +       h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
34389 +       err = PTR_ERR(h_path.dentry);
34390 +       if (IS_ERR(h_path.dentry))
34391 +               goto out;
34392 +
34393 +       err = 0;
34394 +       if (d_is_positive(h_path.dentry)) {
34395 +               h_path.mnt = au_br_mnt(br);
34396 +               err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
34397 +                                         dentry);
34398 +       }
34399 +       dput(h_path.dentry);
34400 +
34401 +out:
34402 +       return err;
34403 +}
34404 +
34405 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
34406 +                        struct au_pin *pin,
34407 +                        struct dentry *h_parent, void *arg)
34408 +{
34409 +       int err, rerr;
34410 +       aufs_bindex_t bopq, btop;
34411 +       struct path h_path;
34412 +       struct dentry *parent;
34413 +       struct inode *h_dir, *h_inode, *inode, *dir;
34414 +       unsigned int *flags = arg;
34415 +
34416 +       btop = au_dbtop(dentry);
34417 +       /* dentry is di-locked */
34418 +       parent = dget_parent(dentry);
34419 +       dir = d_inode(parent);
34420 +       h_dir = d_inode(h_parent);
34421 +       AuDebugOn(h_dir != au_h_iptr(dir, bdst));
34422 +       IMustLock(h_dir);
34423 +
34424 +       err = au_lkup_neg(dentry, bdst, /*wh*/0);
34425 +       if (unlikely(err < 0))
34426 +               goto out;
34427 +       h_path.dentry = au_h_dptr(dentry, bdst);
34428 +       h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
34429 +       err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path, 0755);
34430 +       if (unlikely(err))
34431 +               goto out_put;
34432 +       au_fset_cpdown(*flags, MADE_DIR);
34433 +
34434 +       bopq = au_dbdiropq(dentry);
34435 +       au_fclr_cpdown(*flags, WHED);
34436 +       au_fclr_cpdown(*flags, DIROPQ);
34437 +       if (au_dbwh(dentry) == bdst)
34438 +               au_fset_cpdown(*flags, WHED);
34439 +       if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
34440 +               au_fset_cpdown(*flags, PARENT_OPQ);
34441 +       h_inode = d_inode(h_path.dentry);
34442 +       inode_lock_nested(h_inode, AuLsc_I_CHILD);
34443 +       if (au_ftest_cpdown(*flags, WHED)) {
34444 +               err = au_cpdown_dir_opq(dentry, bdst, flags);
34445 +               if (unlikely(err)) {
34446 +                       inode_unlock(h_inode);
34447 +                       goto out_dir;
34448 +               }
34449 +       }
34450 +
34451 +       err = au_cpdown_attr(&h_path, au_h_dptr(dentry, btop));
34452 +       inode_unlock(h_inode);
34453 +       if (unlikely(err))
34454 +               goto out_opq;
34455 +
34456 +       if (au_ftest_cpdown(*flags, WHED)) {
34457 +               err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
34458 +               if (unlikely(err))
34459 +                       goto out_opq;
34460 +       }
34461 +
34462 +       inode = d_inode(dentry);
34463 +       if (au_ibbot(inode) < bdst)
34464 +               au_set_ibbot(inode, bdst);
34465 +       au_set_h_iptr(inode, bdst, au_igrab(h_inode),
34466 +                     au_hi_flags(inode, /*isdir*/1));
34467 +       au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
34468 +       goto out; /* success */
34469 +
34470 +       /* revert */
34471 +out_opq:
34472 +       if (au_ftest_cpdown(*flags, DIROPQ)) {
34473 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
34474 +               rerr = au_diropq_remove(dentry, bdst);
34475 +               inode_unlock(h_inode);
34476 +               if (unlikely(rerr)) {
34477 +                       AuIOErr("failed removing diropq for %pd b%d (%d)\n",
34478 +                               dentry, bdst, rerr);
34479 +                       err = -EIO;
34480 +                       goto out;
34481 +               }
34482 +       }
34483 +out_dir:
34484 +       if (au_ftest_cpdown(*flags, MADE_DIR)) {
34485 +               rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
34486 +               if (unlikely(rerr)) {
34487 +                       AuIOErr("failed removing %pd b%d (%d)\n",
34488 +                               dentry, bdst, rerr);
34489 +                       err = -EIO;
34490 +               }
34491 +       }
34492 +out_put:
34493 +       au_set_h_dptr(dentry, bdst, NULL);
34494 +       if (au_dbbot(dentry) == bdst)
34495 +               au_update_dbbot(dentry);
34496 +out:
34497 +       dput(parent);
34498 +       return err;
34499 +}
34500 +
34501 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
34502 +{
34503 +       int err;
34504 +       unsigned int flags;
34505 +
34506 +       flags = 0;
34507 +       err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
34508 +
34509 +       return err;
34510 +}
34511 +
34512 +/* ---------------------------------------------------------------------- */
34513 +
34514 +/* policies for create */
34515 +
34516 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
34517 +{
34518 +       int err, i, j, ndentry;
34519 +       aufs_bindex_t bopq;
34520 +       struct au_dcsub_pages dpages;
34521 +       struct au_dpage *dpage;
34522 +       struct dentry **dentries, *parent, *d;
34523 +
34524 +       err = au_dpages_init(&dpages, GFP_NOFS);
34525 +       if (unlikely(err))
34526 +               goto out;
34527 +       parent = dget_parent(dentry);
34528 +       err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
34529 +       if (unlikely(err))
34530 +               goto out_free;
34531 +
34532 +       err = bindex;
34533 +       for (i = 0; i < dpages.ndpage; i++) {
34534 +               dpage = dpages.dpages + i;
34535 +               dentries = dpage->dentries;
34536 +               ndentry = dpage->ndentry;
34537 +               for (j = 0; j < ndentry; j++) {
34538 +                       d = dentries[j];
34539 +                       di_read_lock_parent2(d, !AuLock_IR);
34540 +                       bopq = au_dbdiropq(d);
34541 +                       di_read_unlock(d, !AuLock_IR);
34542 +                       if (bopq >= 0 && bopq < err)
34543 +                               err = bopq;
34544 +               }
34545 +       }
34546 +
34547 +out_free:
34548 +       dput(parent);
34549 +       au_dpages_free(&dpages);
34550 +out:
34551 +       return err;
34552 +}
34553 +
34554 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
34555 +{
34556 +       for (; bindex >= 0; bindex--)
34557 +               if (!au_br_rdonly(au_sbr(sb, bindex)))
34558 +                       return bindex;
34559 +       return -EROFS;
34560 +}
34561 +
34562 +/* top down parent */
34563 +static int au_wbr_create_tdp(struct dentry *dentry,
34564 +                            unsigned int flags __maybe_unused)
34565 +{
34566 +       int err;
34567 +       aufs_bindex_t btop, bindex;
34568 +       struct super_block *sb;
34569 +       struct dentry *parent, *h_parent;
34570 +
34571 +       sb = dentry->d_sb;
34572 +       btop = au_dbtop(dentry);
34573 +       err = btop;
34574 +       if (!au_br_rdonly(au_sbr(sb, btop)))
34575 +               goto out;
34576 +
34577 +       err = -EROFS;
34578 +       parent = dget_parent(dentry);
34579 +       for (bindex = au_dbtop(parent); bindex < btop; bindex++) {
34580 +               h_parent = au_h_dptr(parent, bindex);
34581 +               if (!h_parent || d_is_negative(h_parent))
34582 +                       continue;
34583 +
34584 +               if (!au_br_rdonly(au_sbr(sb, bindex))) {
34585 +                       err = bindex;
34586 +                       break;
34587 +               }
34588 +       }
34589 +       dput(parent);
34590 +
34591 +       /* bottom up here */
34592 +       if (unlikely(err < 0)) {
34593 +               err = au_wbr_bu(sb, btop - 1);
34594 +               if (err >= 0)
34595 +                       err = au_wbr_nonopq(dentry, err);
34596 +       }
34597 +
34598 +out:
34599 +       AuDbg("b%d\n", err);
34600 +       return err;
34601 +}
34602 +
34603 +/* ---------------------------------------------------------------------- */
34604 +
34605 +/* an exception for the policy other than tdp */
34606 +static int au_wbr_create_exp(struct dentry *dentry)
34607 +{
34608 +       int err;
34609 +       aufs_bindex_t bwh, bdiropq;
34610 +       struct dentry *parent;
34611 +
34612 +       err = -1;
34613 +       bwh = au_dbwh(dentry);
34614 +       parent = dget_parent(dentry);
34615 +       bdiropq = au_dbdiropq(parent);
34616 +       if (bwh >= 0) {
34617 +               if (bdiropq >= 0)
34618 +                       err = min(bdiropq, bwh);
34619 +               else
34620 +                       err = bwh;
34621 +               AuDbg("%d\n", err);
34622 +       } else if (bdiropq >= 0) {
34623 +               err = bdiropq;
34624 +               AuDbg("%d\n", err);
34625 +       }
34626 +       dput(parent);
34627 +
34628 +       if (err >= 0)
34629 +               err = au_wbr_nonopq(dentry, err);
34630 +
34631 +       if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
34632 +               err = -1;
34633 +
34634 +       AuDbg("%d\n", err);
34635 +       return err;
34636 +}
34637 +
34638 +/* ---------------------------------------------------------------------- */
34639 +
34640 +/* round robin */
34641 +static int au_wbr_create_init_rr(struct super_block *sb)
34642 +{
34643 +       int err;
34644 +
34645 +       err = au_wbr_bu(sb, au_sbbot(sb));
34646 +       atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
34647 +       /* smp_mb(); */
34648 +
34649 +       AuDbg("b%d\n", err);
34650 +       return err;
34651 +}
34652 +
34653 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
34654 +{
34655 +       int err, nbr;
34656 +       unsigned int u;
34657 +       aufs_bindex_t bindex, bbot;
34658 +       struct super_block *sb;
34659 +       atomic_t *next;
34660 +
34661 +       err = au_wbr_create_exp(dentry);
34662 +       if (err >= 0)
34663 +               goto out;
34664 +
34665 +       sb = dentry->d_sb;
34666 +       next = &au_sbi(sb)->si_wbr_rr_next;
34667 +       bbot = au_sbbot(sb);
34668 +       nbr = bbot + 1;
34669 +       for (bindex = 0; bindex <= bbot; bindex++) {
34670 +               if (!au_ftest_wbr(flags, DIR)) {
34671 +                       err = atomic_dec_return(next) + 1;
34672 +                       /* modulo for 0 is meaningless */
34673 +                       if (unlikely(!err))
34674 +                               err = atomic_dec_return(next) + 1;
34675 +               } else
34676 +                       err = atomic_read(next);
34677 +               AuDbg("%d\n", err);
34678 +               u = err;
34679 +               err = u % nbr;
34680 +               AuDbg("%d\n", err);
34681 +               if (!au_br_rdonly(au_sbr(sb, err)))
34682 +                       break;
34683 +               err = -EROFS;
34684 +       }
34685 +
34686 +       if (err >= 0)
34687 +               err = au_wbr_nonopq(dentry, err);
34688 +
34689 +out:
34690 +       AuDbg("%d\n", err);
34691 +       return err;
34692 +}
34693 +
34694 +/* ---------------------------------------------------------------------- */
34695 +
34696 +/* most free space */
34697 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
34698 +{
34699 +       struct super_block *sb;
34700 +       struct au_branch *br;
34701 +       struct au_wbr_mfs *mfs;
34702 +       struct dentry *h_parent;
34703 +       aufs_bindex_t bindex, bbot;
34704 +       int err;
34705 +       unsigned long long b, bavail;
34706 +       struct path h_path;
34707 +       /* reduce the stack usage */
34708 +       struct kstatfs *st;
34709 +
34710 +       st = kmalloc(sizeof(*st), GFP_NOFS);
34711 +       if (unlikely(!st)) {
34712 +               AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
34713 +               return;
34714 +       }
34715 +
34716 +       bavail = 0;
34717 +       sb = dentry->d_sb;
34718 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34719 +       MtxMustLock(&mfs->mfs_lock);
34720 +       mfs->mfs_bindex = -EROFS;
34721 +       mfs->mfsrr_bytes = 0;
34722 +       if (!parent) {
34723 +               bindex = 0;
34724 +               bbot = au_sbbot(sb);
34725 +       } else {
34726 +               bindex = au_dbtop(parent);
34727 +               bbot = au_dbtaildir(parent);
34728 +       }
34729 +
34730 +       for (; bindex <= bbot; bindex++) {
34731 +               if (parent) {
34732 +                       h_parent = au_h_dptr(parent, bindex);
34733 +                       if (!h_parent || d_is_negative(h_parent))
34734 +                               continue;
34735 +               }
34736 +               br = au_sbr(sb, bindex);
34737 +               if (au_br_rdonly(br))
34738 +                       continue;
34739 +
34740 +               /* sb->s_root for NFS is unreliable */
34741 +               h_path.mnt = au_br_mnt(br);
34742 +               h_path.dentry = h_path.mnt->mnt_root;
34743 +               err = vfs_statfs(&h_path, st);
34744 +               if (unlikely(err)) {
34745 +                       AuWarn1("failed statfs, b%d, %d\n", bindex, err);
34746 +                       continue;
34747 +               }
34748 +
34749 +               /* when the available size is equal, select the lower one */
34750 +               BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
34751 +                            || sizeof(b) < sizeof(st->f_bsize));
34752 +               b = st->f_bavail * st->f_bsize;
34753 +               br->br_wbr->wbr_bytes = b;
34754 +               if (b >= bavail) {
34755 +                       bavail = b;
34756 +                       mfs->mfs_bindex = bindex;
34757 +                       mfs->mfs_jiffy = jiffies;
34758 +               }
34759 +       }
34760 +
34761 +       mfs->mfsrr_bytes = bavail;
34762 +       AuDbg("b%d\n", mfs->mfs_bindex);
34763 +       au_kfree_rcu(st);
34764 +}
34765 +
34766 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
34767 +{
34768 +       int err;
34769 +       struct dentry *parent;
34770 +       struct super_block *sb;
34771 +       struct au_wbr_mfs *mfs;
34772 +
34773 +       err = au_wbr_create_exp(dentry);
34774 +       if (err >= 0)
34775 +               goto out;
34776 +
34777 +       sb = dentry->d_sb;
34778 +       parent = NULL;
34779 +       if (au_ftest_wbr(flags, PARENT))
34780 +               parent = dget_parent(dentry);
34781 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34782 +       mutex_lock(&mfs->mfs_lock);
34783 +       if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34784 +           || mfs->mfs_bindex < 0
34785 +           || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
34786 +               au_mfs(dentry, parent);
34787 +       mutex_unlock(&mfs->mfs_lock);
34788 +       err = mfs->mfs_bindex;
34789 +       dput(parent);
34790 +
34791 +       if (err >= 0)
34792 +               err = au_wbr_nonopq(dentry, err);
34793 +
34794 +out:
34795 +       AuDbg("b%d\n", err);
34796 +       return err;
34797 +}
34798 +
34799 +static int au_wbr_create_init_mfs(struct super_block *sb)
34800 +{
34801 +       struct au_wbr_mfs *mfs;
34802 +
34803 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34804 +       mutex_init(&mfs->mfs_lock);
34805 +       mfs->mfs_jiffy = 0;
34806 +       mfs->mfs_bindex = -EROFS;
34807 +
34808 +       return 0;
34809 +}
34810 +
34811 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
34812 +{
34813 +       mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
34814 +       return 0;
34815 +}
34816 +
34817 +/* ---------------------------------------------------------------------- */
34818 +
34819 +/* top down regardless parent, and then mfs */
34820 +static int au_wbr_create_tdmfs(struct dentry *dentry,
34821 +                              unsigned int flags __maybe_unused)
34822 +{
34823 +       int err;
34824 +       aufs_bindex_t bwh, btail, bindex, bfound, bmfs;
34825 +       unsigned long long watermark;
34826 +       struct super_block *sb;
34827 +       struct au_wbr_mfs *mfs;
34828 +       struct au_branch *br;
34829 +       struct dentry *parent;
34830 +
34831 +       sb = dentry->d_sb;
34832 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34833 +       mutex_lock(&mfs->mfs_lock);
34834 +       if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34835 +           || mfs->mfs_bindex < 0)
34836 +               au_mfs(dentry, /*parent*/NULL);
34837 +       watermark = mfs->mfsrr_watermark;
34838 +       bmfs = mfs->mfs_bindex;
34839 +       mutex_unlock(&mfs->mfs_lock);
34840 +
34841 +       /* another style of au_wbr_create_exp() */
34842 +       bwh = au_dbwh(dentry);
34843 +       parent = dget_parent(dentry);
34844 +       btail = au_dbtaildir(parent);
34845 +       if (bwh >= 0 && bwh < btail)
34846 +               btail = bwh;
34847 +
34848 +       err = au_wbr_nonopq(dentry, btail);
34849 +       if (unlikely(err < 0))
34850 +               goto out;
34851 +       btail = err;
34852 +       bfound = -1;
34853 +       for (bindex = 0; bindex <= btail; bindex++) {
34854 +               br = au_sbr(sb, bindex);
34855 +               if (au_br_rdonly(br))
34856 +                       continue;
34857 +               if (br->br_wbr->wbr_bytes > watermark) {
34858 +                       bfound = bindex;
34859 +                       break;
34860 +               }
34861 +       }
34862 +       err = bfound;
34863 +       if (err < 0)
34864 +               err = bmfs;
34865 +
34866 +out:
34867 +       dput(parent);
34868 +       AuDbg("b%d\n", err);
34869 +       return err;
34870 +}
34871 +
34872 +/* ---------------------------------------------------------------------- */
34873 +
34874 +/* most free space and then round robin */
34875 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
34876 +{
34877 +       int err;
34878 +       struct au_wbr_mfs *mfs;
34879 +
34880 +       err = au_wbr_create_mfs(dentry, flags);
34881 +       if (err >= 0) {
34882 +               mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
34883 +               mutex_lock(&mfs->mfs_lock);
34884 +               if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
34885 +                       err = au_wbr_create_rr(dentry, flags);
34886 +               mutex_unlock(&mfs->mfs_lock);
34887 +       }
34888 +
34889 +       AuDbg("b%d\n", err);
34890 +       return err;
34891 +}
34892 +
34893 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
34894 +{
34895 +       int err;
34896 +
34897 +       au_wbr_create_init_mfs(sb); /* ignore */
34898 +       err = au_wbr_create_init_rr(sb);
34899 +
34900 +       return err;
34901 +}
34902 +
34903 +/* ---------------------------------------------------------------------- */
34904 +
34905 +/* top down parent and most free space */
34906 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
34907 +{
34908 +       int err, e2;
34909 +       unsigned long long b;
34910 +       aufs_bindex_t bindex, btop, bbot;
34911 +       struct super_block *sb;
34912 +       struct dentry *parent, *h_parent;
34913 +       struct au_branch *br;
34914 +
34915 +       err = au_wbr_create_tdp(dentry, flags);
34916 +       if (unlikely(err < 0))
34917 +               goto out;
34918 +       parent = dget_parent(dentry);
34919 +       btop = au_dbtop(parent);
34920 +       bbot = au_dbtaildir(parent);
34921 +       if (btop == bbot)
34922 +               goto out_parent; /* success */
34923 +
34924 +       e2 = au_wbr_create_mfs(dentry, flags);
34925 +       if (e2 < 0)
34926 +               goto out_parent; /* success */
34927 +
34928 +       /* when the available size is equal, select upper one */
34929 +       sb = dentry->d_sb;
34930 +       br = au_sbr(sb, err);
34931 +       b = br->br_wbr->wbr_bytes;
34932 +       AuDbg("b%d, %llu\n", err, b);
34933 +
34934 +       for (bindex = btop; bindex <= bbot; bindex++) {
34935 +               h_parent = au_h_dptr(parent, bindex);
34936 +               if (!h_parent || d_is_negative(h_parent))
34937 +                       continue;
34938 +
34939 +               br = au_sbr(sb, bindex);
34940 +               if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
34941 +                       b = br->br_wbr->wbr_bytes;
34942 +                       err = bindex;
34943 +                       AuDbg("b%d, %llu\n", err, b);
34944 +               }
34945 +       }
34946 +
34947 +       if (err >= 0)
34948 +               err = au_wbr_nonopq(dentry, err);
34949 +
34950 +out_parent:
34951 +       dput(parent);
34952 +out:
34953 +       AuDbg("b%d\n", err);
34954 +       return err;
34955 +}
34956 +
34957 +/* ---------------------------------------------------------------------- */
34958 +
34959 +/*
34960 + * - top down parent
34961 + * - most free space with parent
34962 + * - most free space round-robin regardless parent
34963 + */
34964 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
34965 +{
34966 +       int err;
34967 +       unsigned long long watermark;
34968 +       struct super_block *sb;
34969 +       struct au_branch *br;
34970 +       struct au_wbr_mfs *mfs;
34971 +
34972 +       err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
34973 +       if (unlikely(err < 0))
34974 +               goto out;
34975 +
34976 +       sb = dentry->d_sb;
34977 +       br = au_sbr(sb, err);
34978 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34979 +       mutex_lock(&mfs->mfs_lock);
34980 +       watermark = mfs->mfsrr_watermark;
34981 +       mutex_unlock(&mfs->mfs_lock);
34982 +       if (br->br_wbr->wbr_bytes < watermark)
34983 +               /* regardless the parent dir */
34984 +               err = au_wbr_create_mfsrr(dentry, flags);
34985 +
34986 +out:
34987 +       AuDbg("b%d\n", err);
34988 +       return err;
34989 +}
34990 +
34991 +/* ---------------------------------------------------------------------- */
34992 +
34993 +/* policies for copyup */
34994 +
34995 +/* top down parent */
34996 +static int au_wbr_copyup_tdp(struct dentry *dentry)
34997 +{
34998 +       return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
34999 +}
35000 +
35001 +/* bottom up parent */
35002 +static int au_wbr_copyup_bup(struct dentry *dentry)
35003 +{
35004 +       int err;
35005 +       aufs_bindex_t bindex, btop;
35006 +       struct dentry *parent, *h_parent;
35007 +       struct super_block *sb;
35008 +
35009 +       err = -EROFS;
35010 +       sb = dentry->d_sb;
35011 +       parent = dget_parent(dentry);
35012 +       btop = au_dbtop(parent);
35013 +       for (bindex = au_dbtop(dentry); bindex >= btop; bindex--) {
35014 +               h_parent = au_h_dptr(parent, bindex);
35015 +               if (!h_parent || d_is_negative(h_parent))
35016 +                       continue;
35017 +
35018 +               if (!au_br_rdonly(au_sbr(sb, bindex))) {
35019 +                       err = bindex;
35020 +                       break;
35021 +               }
35022 +       }
35023 +       dput(parent);
35024 +
35025 +       /* bottom up here */
35026 +       if (unlikely(err < 0))
35027 +               err = au_wbr_bu(sb, btop - 1);
35028 +
35029 +       AuDbg("b%d\n", err);
35030 +       return err;
35031 +}
35032 +
35033 +/* bottom up */
35034 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop)
35035 +{
35036 +       int err;
35037 +
35038 +       err = au_wbr_bu(dentry->d_sb, btop);
35039 +       AuDbg("b%d\n", err);
35040 +       if (err > btop)
35041 +               err = au_wbr_nonopq(dentry, err);
35042 +
35043 +       AuDbg("b%d\n", err);
35044 +       return err;
35045 +}
35046 +
35047 +static int au_wbr_copyup_bu(struct dentry *dentry)
35048 +{
35049 +       int err;
35050 +       aufs_bindex_t btop;
35051 +
35052 +       btop = au_dbtop(dentry);
35053 +       err = au_wbr_do_copyup_bu(dentry, btop);
35054 +       return err;
35055 +}
35056 +
35057 +/* ---------------------------------------------------------------------- */
35058 +
35059 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
35060 +       [AuWbrCopyup_TDP] = {
35061 +               .copyup = au_wbr_copyup_tdp
35062 +       },
35063 +       [AuWbrCopyup_BUP] = {
35064 +               .copyup = au_wbr_copyup_bup
35065 +       },
35066 +       [AuWbrCopyup_BU] = {
35067 +               .copyup = au_wbr_copyup_bu
35068 +       }
35069 +};
35070 +
35071 +struct au_wbr_create_operations au_wbr_create_ops[] = {
35072 +       [AuWbrCreate_TDP] = {
35073 +               .create = au_wbr_create_tdp
35074 +       },
35075 +       [AuWbrCreate_RR] = {
35076 +               .create = au_wbr_create_rr,
35077 +               .init   = au_wbr_create_init_rr
35078 +       },
35079 +       [AuWbrCreate_MFS] = {
35080 +               .create = au_wbr_create_mfs,
35081 +               .init   = au_wbr_create_init_mfs,
35082 +               .fin    = au_wbr_create_fin_mfs
35083 +       },
35084 +       [AuWbrCreate_MFSV] = {
35085 +               .create = au_wbr_create_mfs,
35086 +               .init   = au_wbr_create_init_mfs,
35087 +               .fin    = au_wbr_create_fin_mfs
35088 +       },
35089 +       [AuWbrCreate_MFSRR] = {
35090 +               .create = au_wbr_create_mfsrr,
35091 +               .init   = au_wbr_create_init_mfsrr,
35092 +               .fin    = au_wbr_create_fin_mfs
35093 +       },
35094 +       [AuWbrCreate_MFSRRV] = {
35095 +               .create = au_wbr_create_mfsrr,
35096 +               .init   = au_wbr_create_init_mfsrr,
35097 +               .fin    = au_wbr_create_fin_mfs
35098 +       },
35099 +       [AuWbrCreate_TDMFS] = {
35100 +               .create = au_wbr_create_tdmfs,
35101 +               .init   = au_wbr_create_init_mfs,
35102 +               .fin    = au_wbr_create_fin_mfs
35103 +       },
35104 +       [AuWbrCreate_TDMFSV] = {
35105 +               .create = au_wbr_create_tdmfs,
35106 +               .init   = au_wbr_create_init_mfs,
35107 +               .fin    = au_wbr_create_fin_mfs
35108 +       },
35109 +       [AuWbrCreate_PMFS] = {
35110 +               .create = au_wbr_create_pmfs,
35111 +               .init   = au_wbr_create_init_mfs,
35112 +               .fin    = au_wbr_create_fin_mfs
35113 +       },
35114 +       [AuWbrCreate_PMFSV] = {
35115 +               .create = au_wbr_create_pmfs,
35116 +               .init   = au_wbr_create_init_mfs,
35117 +               .fin    = au_wbr_create_fin_mfs
35118 +       },
35119 +       [AuWbrCreate_PMFSRR] = {
35120 +               .create = au_wbr_create_pmfsrr,
35121 +               .init   = au_wbr_create_init_mfsrr,
35122 +               .fin    = au_wbr_create_fin_mfs
35123 +       },
35124 +       [AuWbrCreate_PMFSRRV] = {
35125 +               .create = au_wbr_create_pmfsrr,
35126 +               .init   = au_wbr_create_init_mfsrr,
35127 +               .fin    = au_wbr_create_fin_mfs
35128 +       }
35129 +};
35130 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
35131 --- /usr/share/empty/fs/aufs/whout.c    1970-01-01 01:00:00.000000000 +0100
35132 +++ linux/fs/aufs/whout.c       2022-11-05 23:02:18.972555950 +0100
35133 @@ -0,0 +1,1072 @@
35134 +// SPDX-License-Identifier: GPL-2.0
35135 +/*
35136 + * Copyright (C) 2005-2022 Junjiro R. Okajima
35137 + *
35138 + * This program is free software; you can redistribute it and/or modify
35139 + * it under the terms of the GNU General Public License as published by
35140 + * the Free Software Foundation; either version 2 of the License, or
35141 + * (at your option) any later version.
35142 + *
35143 + * This program is distributed in the hope that it will be useful,
35144 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35145 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
35146 + * GNU General Public License for more details.
35147 + *
35148 + * You should have received a copy of the GNU General Public License
35149 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
35150 + */
35151 +
35152 +/*
35153 + * whiteout for logical deletion and opaque directory
35154 + */
35155 +
35156 +#include "aufs.h"
35157 +
35158 +#define WH_MASK                        0444
35159 +
35160 +/*
35161 + * If a directory contains this file, then it is opaque.  We start with the
35162 + * .wh. flag so that it is blocked by lookup.
35163 + */
35164 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
35165 +                                          sizeof(AUFS_WH_DIROPQ) - 1);
35166 +
35167 +/*
35168 + * generate whiteout name, which is NOT terminated by NULL.
35169 + * @name: original d_name.name
35170 + * @len: original d_name.len
35171 + * @wh: whiteout qstr
35172 + * returns zero when succeeds, otherwise error.
35173 + * succeeded value as wh->name should be freed by kfree().
35174 + */
35175 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
35176 +{
35177 +       char *p;
35178 +
35179 +       if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
35180 +               return -ENAMETOOLONG;
35181 +
35182 +       wh->len = name->len + AUFS_WH_PFX_LEN;
35183 +       p = kmalloc(wh->len, GFP_NOFS);
35184 +       wh->name = p;
35185 +       if (p) {
35186 +               memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35187 +               memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
35188 +               /* smp_mb(); */
35189 +               return 0;
35190 +       }
35191 +       return -ENOMEM;
35192 +}
35193 +
35194 +/* ---------------------------------------------------------------------- */
35195 +
35196 +/*
35197 + * test if the @wh_name exists under @h_ppath.
35198 + * @try_sio specifies the necessary of super-io.
35199 + */
35200 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
35201 +              struct qstr *wh_name, int try_sio)
35202 +{
35203 +       int err;
35204 +       struct dentry *wh_dentry;
35205 +
35206 +       if (!try_sio)
35207 +               wh_dentry = vfsub_lkup_one(wh_name, h_ppath);
35208 +       else
35209 +               wh_dentry = au_sio_lkup_one(h_userns, wh_name, h_ppath);
35210 +       err = PTR_ERR(wh_dentry);
35211 +       if (IS_ERR(wh_dentry)) {
35212 +               if (err == -ENAMETOOLONG)
35213 +                       err = 0;
35214 +               goto out;
35215 +       }
35216 +
35217 +       err = 0;
35218 +       if (d_is_negative(wh_dentry))
35219 +               goto out_wh; /* success */
35220 +
35221 +       err = 1;
35222 +       if (d_is_reg(wh_dentry))
35223 +               goto out_wh; /* success */
35224 +
35225 +       err = -EIO;
35226 +       AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
35227 +               wh_dentry, d_inode(wh_dentry)->i_mode);
35228 +
35229 +out_wh:
35230 +       dput(wh_dentry);
35231 +out:
35232 +       return err;
35233 +}
35234 +
35235 +/*
35236 + * test if the @h_path->dentry sets opaque or not.
35237 + */
35238 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path)
35239 +{
35240 +       int err;
35241 +       struct inode *h_dir;
35242 +
35243 +       h_dir = d_inode(h_path->dentry);
35244 +       err = au_wh_test(h_userns, h_path, &diropq_name,
35245 +                        au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC));
35246 +       return err;
35247 +}
35248 +
35249 +/*
35250 + * returns a negative dentry whose name is unique and temporary.
35251 + */
35252 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
35253 +                            struct qstr *prefix)
35254 +{
35255 +       struct dentry *dentry;
35256 +       int i;
35257 +       char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
35258 +               *name, *p;
35259 +       /* strict atomic_t is unnecessary here */
35260 +       static unsigned short cnt;
35261 +       struct qstr qs;
35262 +       struct path h_ppath;
35263 +       struct user_namespace *h_userns;
35264 +
35265 +       BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
35266 +
35267 +       name = defname;
35268 +       qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
35269 +       if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
35270 +               dentry = ERR_PTR(-ENAMETOOLONG);
35271 +               if (unlikely(qs.len > NAME_MAX))
35272 +                       goto out;
35273 +               dentry = ERR_PTR(-ENOMEM);
35274 +               name = kmalloc(qs.len + 1, GFP_NOFS);
35275 +               if (unlikely(!name))
35276 +                       goto out;
35277 +       }
35278 +
35279 +       /* doubly whiteout-ed */
35280 +       memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
35281 +       p = name + AUFS_WH_PFX_LEN * 2;
35282 +       memcpy(p, prefix->name, prefix->len);
35283 +       p += prefix->len;
35284 +       *p++ = '.';
35285 +       AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
35286 +
35287 +       h_ppath.dentry = h_parent;
35288 +       h_ppath.mnt = au_br_mnt(br);
35289 +       h_userns = au_br_userns(br);
35290 +       qs.name = name;
35291 +       for (i = 0; i < 3; i++) {
35292 +               sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
35293 +               dentry = au_sio_lkup_one(h_userns, &qs, &h_ppath);
35294 +               if (IS_ERR(dentry) || d_is_negative(dentry))
35295 +                       goto out_name;
35296 +               dput(dentry);
35297 +       }
35298 +       /* pr_warn("could not get random name\n"); */
35299 +       dentry = ERR_PTR(-EEXIST);
35300 +       AuDbg("%.*s\n", AuLNPair(&qs));
35301 +       BUG();
35302 +
35303 +out_name:
35304 +       if (name != defname)
35305 +               au_kfree_try_rcu(name);
35306 +out:
35307 +       AuTraceErrPtr(dentry);
35308 +       return dentry;
35309 +}
35310 +
35311 +/*
35312 + * rename the @h_dentry on @br to the whiteouted temporary name.
35313 + */
35314 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
35315 +{
35316 +       int err;
35317 +       struct path h_path = {
35318 +               .mnt = au_br_mnt(br)
35319 +       };
35320 +       struct inode *h_dir, *delegated;
35321 +       struct dentry *h_parent;
35322 +
35323 +       h_parent = h_dentry->d_parent; /* dir inode is locked */
35324 +       h_dir = d_inode(h_parent);
35325 +       IMustLock(h_dir);
35326 +
35327 +       h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
35328 +       err = PTR_ERR(h_path.dentry);
35329 +       if (IS_ERR(h_path.dentry))
35330 +               goto out;
35331 +
35332 +       /* under the same dir, no need to lock_rename() */
35333 +       delegated = NULL;
35334 +       err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated,
35335 +                          /*flags*/0);
35336 +       AuTraceErr(err);
35337 +       if (unlikely(err == -EWOULDBLOCK)) {
35338 +               pr_warn("cannot retry for NFSv4 delegation"
35339 +                       " for an internal rename\n");
35340 +               iput(delegated);
35341 +       }
35342 +       dput(h_path.dentry);
35343 +
35344 +out:
35345 +       AuTraceErr(err);
35346 +       return err;
35347 +}
35348 +
35349 +/* ---------------------------------------------------------------------- */
35350 +/*
35351 + * functions for removing a whiteout
35352 + */
35353 +
35354 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
35355 +{
35356 +       int err, force;
35357 +       struct inode *delegated;
35358 +
35359 +       /*
35360 +        * forces superio when the dir has a sticky bit.
35361 +        * this may be a violation of unix fs semantics.
35362 +        */
35363 +       force = (h_dir->i_mode & S_ISVTX)
35364 +               && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
35365 +       delegated = NULL;
35366 +       err = vfsub_unlink(h_dir, h_path, &delegated, force);
35367 +       if (unlikely(err == -EWOULDBLOCK)) {
35368 +               pr_warn("cannot retry for NFSv4 delegation"
35369 +                       " for an internal unlink\n");
35370 +               iput(delegated);
35371 +       }
35372 +       return err;
35373 +}
35374 +
35375 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
35376 +                       struct dentry *dentry)
35377 +{
35378 +       int err;
35379 +
35380 +       err = do_unlink_wh(h_dir, h_path);
35381 +       if (!err && dentry)
35382 +               au_set_dbwh(dentry, -1);
35383 +
35384 +       return err;
35385 +}
35386 +
35387 +static int unlink_wh_name(struct path *h_ppath, struct qstr *wh)
35388 +{
35389 +       int err;
35390 +       struct path h_path;
35391 +
35392 +       err = 0;
35393 +       h_path.dentry = vfsub_lkup_one(wh, h_ppath);
35394 +       if (IS_ERR(h_path.dentry))
35395 +               err = PTR_ERR(h_path.dentry);
35396 +       else {
35397 +               if (d_is_reg(h_path.dentry)) {
35398 +                       h_path.mnt = h_ppath->mnt;
35399 +                       err = do_unlink_wh(d_inode(h_ppath->dentry), &h_path);
35400 +               }
35401 +               dput(h_path.dentry);
35402 +       }
35403 +
35404 +       return err;
35405 +}
35406 +
35407 +/* ---------------------------------------------------------------------- */
35408 +/*
35409 + * initialize/clean whiteout for a branch
35410 + */
35411 +
35412 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
35413 +                       const int isdir)
35414 +{
35415 +       int err;
35416 +       struct inode *delegated;
35417 +
35418 +       if (d_is_negative(whpath->dentry))
35419 +               return;
35420 +
35421 +       if (isdir)
35422 +               err = vfsub_rmdir(h_dir, whpath);
35423 +       else {
35424 +               delegated = NULL;
35425 +               err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
35426 +               if (unlikely(err == -EWOULDBLOCK)) {
35427 +                       pr_warn("cannot retry for NFSv4 delegation"
35428 +                               " for an internal unlink\n");
35429 +                       iput(delegated);
35430 +               }
35431 +       }
35432 +       if (unlikely(err))
35433 +               pr_warn("failed removing %pd (%d), ignored.\n",
35434 +                       whpath->dentry, err);
35435 +}
35436 +
35437 +static int test_linkable(struct dentry *h_root)
35438 +{
35439 +       struct inode *h_dir = d_inode(h_root);
35440 +
35441 +       if (h_dir->i_op->link)
35442 +               return 0;
35443 +
35444 +       pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
35445 +              h_root, au_sbtype(h_root->d_sb));
35446 +       return -ENOSYS; /* the branch doesn't have its ->link() */
35447 +}
35448 +
35449 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
35450 +static int au_whdir(struct inode *h_dir, struct path *path)
35451 +{
35452 +       int err;
35453 +
35454 +       err = -EEXIST;
35455 +       if (d_is_negative(path->dentry)) {
35456 +               int mode = 0700;
35457 +
35458 +               if (au_test_nfs(path->dentry->d_sb))
35459 +                       mode |= 0111;
35460 +               err = vfsub_mkdir(h_dir, path, mode);
35461 +       } else if (d_is_dir(path->dentry))
35462 +               err = 0;
35463 +       else
35464 +               pr_err("unknown %pd exists\n", path->dentry);
35465 +
35466 +       return err;
35467 +}
35468 +
35469 +struct au_wh_base {
35470 +       const struct qstr *name;
35471 +       struct dentry *dentry;
35472 +};
35473 +
35474 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
35475 +                         struct path *h_path)
35476 +{
35477 +       h_path->dentry = base[AuBrWh_BASE].dentry;
35478 +       au_wh_clean(h_dir, h_path, /*isdir*/0);
35479 +       h_path->dentry = base[AuBrWh_PLINK].dentry;
35480 +       au_wh_clean(h_dir, h_path, /*isdir*/1);
35481 +       h_path->dentry = base[AuBrWh_ORPH].dentry;
35482 +       au_wh_clean(h_dir, h_path, /*isdir*/1);
35483 +}
35484 +
35485 +/*
35486 + * returns tri-state,
35487 + * minus: error, caller should print the message
35488 + * zero: success
35489 + * plus: error, caller should NOT print the message
35490 + */
35491 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
35492 +                               int do_plink, struct au_wh_base base[],
35493 +                               struct path *h_path)
35494 +{
35495 +       int err;
35496 +       struct inode *h_dir;
35497 +
35498 +       h_dir = d_inode(h_root);
35499 +       h_path->dentry = base[AuBrWh_BASE].dentry;
35500 +       au_wh_clean(h_dir, h_path, /*isdir*/0);
35501 +       h_path->dentry = base[AuBrWh_PLINK].dentry;
35502 +       if (do_plink) {
35503 +               err = test_linkable(h_root);
35504 +               if (unlikely(err)) {
35505 +                       err = 1;
35506 +                       goto out;
35507 +               }
35508 +
35509 +               err = au_whdir(h_dir, h_path);
35510 +               if (unlikely(err))
35511 +                       goto out;
35512 +               wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35513 +       } else
35514 +               au_wh_clean(h_dir, h_path, /*isdir*/1);
35515 +       h_path->dentry = base[AuBrWh_ORPH].dentry;
35516 +       err = au_whdir(h_dir, h_path);
35517 +       if (unlikely(err))
35518 +               goto out;
35519 +       wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35520 +
35521 +out:
35522 +       return err;
35523 +}
35524 +
35525 +/*
35526 + * for the moment, aufs supports the branch filesystem which does not support
35527 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
35528 + * copyup failed. finally, such filesystem will not be used as the writable
35529 + * branch.
35530 + *
35531 + * returns tri-state, see above.
35532 + */
35533 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
35534 +                        int do_plink, struct au_wh_base base[],
35535 +                        struct path *h_path)
35536 +{
35537 +       int err;
35538 +       struct inode *h_dir;
35539 +
35540 +       WbrWhMustWriteLock(wbr);
35541 +
35542 +       err = test_linkable(h_root);
35543 +       if (unlikely(err)) {
35544 +               err = 1;
35545 +               goto out;
35546 +       }
35547 +
35548 +       /*
35549 +        * todo: should this create be done in /sbin/mount.aufs helper?
35550 +        */
35551 +       err = -EEXIST;
35552 +       h_dir = d_inode(h_root);
35553 +       if (d_is_negative(base[AuBrWh_BASE].dentry)) {
35554 +               h_path->dentry = base[AuBrWh_BASE].dentry;
35555 +               err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
35556 +       } else if (d_is_reg(base[AuBrWh_BASE].dentry))
35557 +               err = 0;
35558 +       else
35559 +               pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
35560 +       if (unlikely(err))
35561 +               goto out;
35562 +
35563 +       h_path->dentry = base[AuBrWh_PLINK].dentry;
35564 +       if (do_plink) {
35565 +               err = au_whdir(h_dir, h_path);
35566 +               if (unlikely(err))
35567 +                       goto out;
35568 +               wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35569 +       } else
35570 +               au_wh_clean(h_dir, h_path, /*isdir*/1);
35571 +       wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
35572 +
35573 +       h_path->dentry = base[AuBrWh_ORPH].dentry;
35574 +       err = au_whdir(h_dir, h_path);
35575 +       if (unlikely(err))
35576 +               goto out;
35577 +       wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35578 +
35579 +out:
35580 +       return err;
35581 +}
35582 +
35583 +/*
35584 + * initialize the whiteout base file/dir for @br.
35585 + */
35586 +int au_wh_init(struct au_branch *br, struct super_block *sb)
35587 +{
35588 +       int err, i;
35589 +       const unsigned char do_plink
35590 +               = !!au_opt_test(au_mntflags(sb), PLINK);
35591 +       struct inode *h_dir;
35592 +       struct path path = br->br_path;
35593 +       struct dentry *h_root = path.dentry;
35594 +       struct au_wbr *wbr = br->br_wbr;
35595 +       static const struct qstr base_name[] = {
35596 +               [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
35597 +                                         sizeof(AUFS_BASE_NAME) - 1),
35598 +               [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
35599 +                                          sizeof(AUFS_PLINKDIR_NAME) - 1),
35600 +               [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
35601 +                                         sizeof(AUFS_ORPHDIR_NAME) - 1)
35602 +       };
35603 +       struct au_wh_base base[] = {
35604 +               [AuBrWh_BASE] = {
35605 +                       .name   = base_name + AuBrWh_BASE,
35606 +                       .dentry = NULL
35607 +               },
35608 +               [AuBrWh_PLINK] = {
35609 +                       .name   = base_name + AuBrWh_PLINK,
35610 +                       .dentry = NULL
35611 +               },
35612 +               [AuBrWh_ORPH] = {
35613 +                       .name   = base_name + AuBrWh_ORPH,
35614 +                       .dentry = NULL
35615 +               }
35616 +       };
35617 +
35618 +       if (wbr)
35619 +               WbrWhMustWriteLock(wbr);
35620 +
35621 +       for (i = 0; i < AuBrWh_Last; i++) {
35622 +               /* doubly whiteouted */
35623 +               struct dentry *d;
35624 +
35625 +               d = au_wh_lkup(h_root, (void *)base[i].name, br);
35626 +               err = PTR_ERR(d);
35627 +               if (IS_ERR(d))
35628 +                       goto out;
35629 +
35630 +               base[i].dentry = d;
35631 +               AuDebugOn(wbr
35632 +                         && wbr->wbr_wh[i]
35633 +                         && wbr->wbr_wh[i] != base[i].dentry);
35634 +       }
35635 +
35636 +       if (wbr)
35637 +               for (i = 0; i < AuBrWh_Last; i++) {
35638 +                       dput(wbr->wbr_wh[i]);
35639 +                       wbr->wbr_wh[i] = NULL;
35640 +               }
35641 +
35642 +       err = 0;
35643 +       if (!au_br_writable(br->br_perm)) {
35644 +               h_dir = d_inode(h_root);
35645 +               au_wh_init_ro(h_dir, base, &path);
35646 +       } else if (!au_br_wh_linkable(br->br_perm)) {
35647 +               err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
35648 +               if (err > 0)
35649 +                       goto out;
35650 +               else if (err)
35651 +                       goto out_err;
35652 +       } else {
35653 +               err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
35654 +               if (err > 0)
35655 +                       goto out;
35656 +               else if (err)
35657 +                       goto out_err;
35658 +       }
35659 +       goto out; /* success */
35660 +
35661 +out_err:
35662 +       pr_err("an error(%d) on the writable branch %pd(%s)\n",
35663 +              err, h_root, au_sbtype(h_root->d_sb));
35664 +out:
35665 +       for (i = 0; i < AuBrWh_Last; i++)
35666 +               dput(base[i].dentry);
35667 +       return err;
35668 +}
35669 +
35670 +/* ---------------------------------------------------------------------- */
35671 +/*
35672 + * whiteouts are all hard-linked usually.
35673 + * when its link count reaches a ceiling, we create a new whiteout base
35674 + * asynchronously.
35675 + */
35676 +
35677 +struct reinit_br_wh {
35678 +       struct super_block *sb;
35679 +       struct au_branch *br;
35680 +};
35681 +
35682 +static void reinit_br_wh(void *arg)
35683 +{
35684 +       int err;
35685 +       aufs_bindex_t bindex;
35686 +       struct path h_path;
35687 +       struct reinit_br_wh *a = arg;
35688 +       struct au_wbr *wbr;
35689 +       struct inode *dir, *delegated;
35690 +       struct dentry *h_root;
35691 +       struct au_hinode *hdir;
35692 +
35693 +       err = 0;
35694 +       wbr = a->br->br_wbr;
35695 +       /* big aufs lock */
35696 +       si_noflush_write_lock(a->sb);
35697 +       if (!au_br_writable(a->br->br_perm))
35698 +               goto out;
35699 +       bindex = au_br_index(a->sb, a->br->br_id);
35700 +       if (unlikely(bindex < 0))
35701 +               goto out;
35702 +
35703 +       di_read_lock_parent(a->sb->s_root, AuLock_IR);
35704 +       dir = d_inode(a->sb->s_root);
35705 +       hdir = au_hi(dir, bindex);
35706 +       h_root = au_h_dptr(a->sb->s_root, bindex);
35707 +       AuDebugOn(h_root != au_br_dentry(a->br));
35708 +
35709 +       au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35710 +       wbr_wh_write_lock(wbr);
35711 +       err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
35712 +                         h_root, a->br);
35713 +       if (!err) {
35714 +               h_path.dentry = wbr->wbr_whbase;
35715 +               h_path.mnt = au_br_mnt(a->br);
35716 +               delegated = NULL;
35717 +               err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
35718 +                                  /*force*/0);
35719 +               if (unlikely(err == -EWOULDBLOCK)) {
35720 +                       pr_warn("cannot retry for NFSv4 delegation"
35721 +                               " for an internal unlink\n");
35722 +                       iput(delegated);
35723 +               }
35724 +       } else {
35725 +               pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
35726 +               err = 0;
35727 +       }
35728 +       dput(wbr->wbr_whbase);
35729 +       wbr->wbr_whbase = NULL;
35730 +       if (!err)
35731 +               err = au_wh_init(a->br, a->sb);
35732 +       wbr_wh_write_unlock(wbr);
35733 +       au_hn_inode_unlock(hdir);
35734 +       di_read_unlock(a->sb->s_root, AuLock_IR);
35735 +       if (!err)
35736 +               au_fhsm_wrote(a->sb, bindex, /*force*/0);
35737 +
35738 +out:
35739 +       if (wbr)
35740 +               atomic_dec(&wbr->wbr_wh_running);
35741 +       au_lcnt_dec(&a->br->br_count);
35742 +       si_write_unlock(a->sb);
35743 +       au_nwt_done(&au_sbi(a->sb)->si_nowait);
35744 +       au_kfree_rcu(a);
35745 +       if (unlikely(err))
35746 +               AuIOErr("err %d\n", err);
35747 +}
35748 +
35749 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
35750 +{
35751 +       int do_dec, wkq_err;
35752 +       struct reinit_br_wh *arg;
35753 +
35754 +       do_dec = 1;
35755 +       if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
35756 +               goto out;
35757 +
35758 +       /* ignore ENOMEM */
35759 +       arg = kmalloc(sizeof(*arg), GFP_NOFS);
35760 +       if (arg) {
35761 +               /*
35762 +                * dec(wh_running), kfree(arg) and dec(br_count)
35763 +                * in reinit function
35764 +                */
35765 +               arg->sb = sb;
35766 +               arg->br = br;
35767 +               au_lcnt_inc(&br->br_count);
35768 +               wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
35769 +               if (unlikely(wkq_err)) {
35770 +                       atomic_dec(&br->br_wbr->wbr_wh_running);
35771 +                       au_lcnt_dec(&br->br_count);
35772 +                       au_kfree_rcu(arg);
35773 +               }
35774 +               do_dec = 0;
35775 +       }
35776 +
35777 +out:
35778 +       if (do_dec)
35779 +               atomic_dec(&br->br_wbr->wbr_wh_running);
35780 +}
35781 +
35782 +/* ---------------------------------------------------------------------- */
35783 +
35784 +/*
35785 + * create the whiteout @wh.
35786 + */
35787 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
35788 +                            struct dentry *wh)
35789 +{
35790 +       int err;
35791 +       struct path h_path = {
35792 +               .dentry = wh
35793 +       };
35794 +       struct au_branch *br;
35795 +       struct au_wbr *wbr;
35796 +       struct dentry *h_parent;
35797 +       struct inode *h_dir, *delegated;
35798 +
35799 +       h_parent = wh->d_parent; /* dir inode is locked */
35800 +       h_dir = d_inode(h_parent);
35801 +       IMustLock(h_dir);
35802 +
35803 +       br = au_sbr(sb, bindex);
35804 +       h_path.mnt = au_br_mnt(br);
35805 +       wbr = br->br_wbr;
35806 +       wbr_wh_read_lock(wbr);
35807 +       if (wbr->wbr_whbase) {
35808 +               delegated = NULL;
35809 +               err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
35810 +               if (unlikely(err == -EWOULDBLOCK)) {
35811 +                       pr_warn("cannot retry for NFSv4 delegation"
35812 +                               " for an internal link\n");
35813 +                       iput(delegated);
35814 +               }
35815 +               if (!err || err != -EMLINK)
35816 +                       goto out;
35817 +
35818 +               /* link count full. re-initialize br_whbase. */
35819 +               kick_reinit_br_wh(sb, br);
35820 +       }
35821 +
35822 +       /* return this error in this context */
35823 +       err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
35824 +       if (!err)
35825 +               au_fhsm_wrote(sb, bindex, /*force*/0);
35826 +
35827 +out:
35828 +       wbr_wh_read_unlock(wbr);
35829 +       return err;
35830 +}
35831 +
35832 +/* ---------------------------------------------------------------------- */
35833 +
35834 +/*
35835 + * create or remove the diropq.
35836 + */
35837 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
35838 +                               unsigned int flags)
35839 +{
35840 +       struct dentry *opq_dentry;
35841 +       struct super_block *sb;
35842 +       struct au_branch *br;
35843 +       struct path h_path;
35844 +       int err;
35845 +
35846 +       sb = dentry->d_sb;
35847 +       br = au_sbr(sb, bindex);
35848 +       h_path.dentry = au_h_dptr(dentry, bindex);
35849 +       h_path.mnt = au_br_mnt(br);
35850 +       opq_dentry = vfsub_lkup_one(&diropq_name, &h_path);
35851 +       if (IS_ERR(opq_dentry))
35852 +               goto out;
35853 +
35854 +       if (au_ftest_diropq(flags, CREATE)) {
35855 +               err = link_or_create_wh(sb, bindex, opq_dentry);
35856 +               if (!err) {
35857 +                       au_set_dbdiropq(dentry, bindex);
35858 +                       goto out; /* success */
35859 +               }
35860 +       } else {
35861 +               h_path.dentry = opq_dentry;
35862 +               err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &h_path);
35863 +               if (!err)
35864 +                       au_set_dbdiropq(dentry, -1);
35865 +       }
35866 +       dput(opq_dentry);
35867 +       opq_dentry = ERR_PTR(err);
35868 +
35869 +out:
35870 +       return opq_dentry;
35871 +}
35872 +
35873 +struct do_diropq_args {
35874 +       struct dentry **errp;
35875 +       struct dentry *dentry;
35876 +       aufs_bindex_t bindex;
35877 +       unsigned int flags;
35878 +};
35879 +
35880 +static void call_do_diropq(void *args)
35881 +{
35882 +       struct do_diropq_args *a = args;
35883 +       *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
35884 +}
35885 +
35886 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35887 +                            unsigned int flags)
35888 +{
35889 +       struct dentry *diropq, *h_dentry;
35890 +       struct user_namespace *h_userns;
35891 +
35892 +       h_userns = au_sbr_userns(dentry->d_sb, bindex);
35893 +       h_dentry = au_h_dptr(dentry, bindex);
35894 +       if (!au_test_h_perm_sio(h_userns, d_inode(h_dentry),
35895 +                               MAY_EXEC | MAY_WRITE))
35896 +               diropq = do_diropq(dentry, bindex, flags);
35897 +       else {
35898 +               int wkq_err;
35899 +               struct do_diropq_args args = {
35900 +                       .errp           = &diropq,
35901 +                       .dentry         = dentry,
35902 +                       .bindex         = bindex,
35903 +                       .flags          = flags
35904 +               };
35905 +
35906 +               wkq_err = au_wkq_wait(call_do_diropq, &args);
35907 +               if (unlikely(wkq_err))
35908 +                       diropq = ERR_PTR(wkq_err);
35909 +       }
35910 +
35911 +       return diropq;
35912 +}
35913 +
35914 +/* ---------------------------------------------------------------------- */
35915 +
35916 +/*
35917 + * lookup whiteout dentry.
35918 + * @h_parent: lower parent dentry which must exist and be locked
35919 + * @base_name: name of dentry which will be whiteouted
35920 + * returns dentry for whiteout.
35921 + */
35922 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35923 +                         struct au_branch *br)
35924 +{
35925 +       int err;
35926 +       struct qstr wh_name;
35927 +       struct dentry *wh_dentry;
35928 +       struct path h_path;
35929 +
35930 +       err = au_wh_name_alloc(&wh_name, base_name);
35931 +       wh_dentry = ERR_PTR(err);
35932 +       if (!err) {
35933 +               h_path.dentry = h_parent;
35934 +               h_path.mnt = au_br_mnt(br);
35935 +               wh_dentry = vfsub_lkup_one(&wh_name, &h_path);
35936 +               au_kfree_try_rcu(wh_name.name);
35937 +       }
35938 +       return wh_dentry;
35939 +}
35940 +
35941 +/*
35942 + * link/create a whiteout for @dentry on @bindex.
35943 + */
35944 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35945 +                           struct dentry *h_parent)
35946 +{
35947 +       struct dentry *wh_dentry;
35948 +       struct super_block *sb;
35949 +       int err;
35950 +
35951 +       sb = dentry->d_sb;
35952 +       wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
35953 +       if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
35954 +               err = link_or_create_wh(sb, bindex, wh_dentry);
35955 +               if (!err) {
35956 +                       au_set_dbwh(dentry, bindex);
35957 +                       au_fhsm_wrote(sb, bindex, /*force*/0);
35958 +               } else {
35959 +                       dput(wh_dentry);
35960 +                       wh_dentry = ERR_PTR(err);
35961 +               }
35962 +       }
35963 +
35964 +       return wh_dentry;
35965 +}
35966 +
35967 +/* ---------------------------------------------------------------------- */
35968 +
35969 +/* Delete all whiteouts in this directory on branch bindex. */
35970 +static int del_wh_children(struct path *h_path, struct au_nhash *whlist,
35971 +                          aufs_bindex_t bindex)
35972 +{
35973 +       int err;
35974 +       unsigned long ul, n;
35975 +       struct qstr wh_name;
35976 +       char *p;
35977 +       struct hlist_head *head;
35978 +       struct au_vdir_wh *pos;
35979 +       struct au_vdir_destr *str;
35980 +
35981 +       err = -ENOMEM;
35982 +       p = (void *)__get_free_page(GFP_NOFS);
35983 +       wh_name.name = p;
35984 +       if (unlikely(!wh_name.name))
35985 +               goto out;
35986 +
35987 +       err = 0;
35988 +       memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35989 +       p += AUFS_WH_PFX_LEN;
35990 +       n = whlist->nh_num;
35991 +       head = whlist->nh_head;
35992 +       for (ul = 0; !err && ul < n; ul++, head++) {
35993 +               hlist_for_each_entry(pos, head, wh_hash) {
35994 +                       if (pos->wh_bindex != bindex)
35995 +                               continue;
35996 +
35997 +                       str = &pos->wh_str;
35998 +                       if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
35999 +                               memcpy(p, str->name, str->len);
36000 +                               wh_name.len = AUFS_WH_PFX_LEN + str->len;
36001 +                               err = unlink_wh_name(h_path, &wh_name);
36002 +                               if (!err)
36003 +                                       continue;
36004 +                               break;
36005 +                       }
36006 +                       AuIOErr("whiteout name too long %.*s\n",
36007 +                               str->len, str->name);
36008 +                       err = -EIO;
36009 +                       break;
36010 +               }
36011 +       }
36012 +       free_page((unsigned long)wh_name.name);
36013 +
36014 +out:
36015 +       return err;
36016 +}
36017 +
36018 +struct del_wh_children_args {
36019 +       int *errp;
36020 +       struct path *h_path;
36021 +       struct au_nhash *whlist;
36022 +       aufs_bindex_t bindex;
36023 +};
36024 +
36025 +static void call_del_wh_children(void *args)
36026 +{
36027 +       struct del_wh_children_args *a = args;
36028 +       *a->errp = del_wh_children(a->h_path, a->whlist, a->bindex);
36029 +}
36030 +
36031 +/* ---------------------------------------------------------------------- */
36032 +
36033 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
36034 +{
36035 +       struct au_whtmp_rmdir *whtmp;
36036 +       int err;
36037 +       unsigned int rdhash;
36038 +
36039 +       SiMustAnyLock(sb);
36040 +
36041 +       whtmp = kzalloc(sizeof(*whtmp), gfp);
36042 +       if (unlikely(!whtmp)) {
36043 +               whtmp = ERR_PTR(-ENOMEM);
36044 +               goto out;
36045 +       }
36046 +
36047 +       /* no estimation for dir size */
36048 +       rdhash = au_sbi(sb)->si_rdhash;
36049 +       if (!rdhash)
36050 +               rdhash = AUFS_RDHASH_DEF;
36051 +       err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
36052 +       if (unlikely(err)) {
36053 +               au_kfree_rcu(whtmp);
36054 +               whtmp = ERR_PTR(err);
36055 +       }
36056 +
36057 +out:
36058 +       return whtmp;
36059 +}
36060 +
36061 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
36062 +{
36063 +       if (whtmp->br)
36064 +               au_lcnt_dec(&whtmp->br->br_count);
36065 +       dput(whtmp->wh_dentry);
36066 +       iput(whtmp->dir);
36067 +       au_nhash_wh_free(&whtmp->whlist);
36068 +       au_kfree_rcu(whtmp);
36069 +}
36070 +
36071 +/*
36072 + * rmdir the whiteouted temporary named dir @h_dentry.
36073 + * @whlist: whiteouted children.
36074 + */
36075 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36076 +                  struct dentry *wh_dentry, struct au_nhash *whlist)
36077 +{
36078 +       int err;
36079 +       unsigned int h_nlink;
36080 +       struct path wh_path;
36081 +       struct inode *wh_inode, *h_dir;
36082 +       struct au_branch *br;
36083 +       struct user_namespace *h_userns;
36084 +
36085 +       h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
36086 +       IMustLock(h_dir);
36087 +
36088 +       br = au_sbr(dir->i_sb, bindex);
36089 +       wh_path.dentry = wh_dentry;
36090 +       wh_path.mnt = au_br_mnt(br);
36091 +       h_userns = au_br_userns(br);
36092 +       wh_inode = d_inode(wh_dentry);
36093 +       inode_lock_nested(wh_inode, AuLsc_I_CHILD);
36094 +
36095 +       /*
36096 +        * someone else might change some whiteouts while we were sleeping.
36097 +        * it means this whlist may have an obsoleted entry.
36098 +        */
36099 +       if (!au_test_h_perm_sio(h_userns, wh_inode, MAY_EXEC | MAY_WRITE))
36100 +               err = del_wh_children(&wh_path, whlist, bindex);
36101 +       else {
36102 +               int wkq_err;
36103 +               struct del_wh_children_args args = {
36104 +                       .errp           = &err,
36105 +                       .h_path         = &wh_path,
36106 +                       .whlist         = whlist,
36107 +                       .bindex         = bindex
36108 +               };
36109 +
36110 +               wkq_err = au_wkq_wait(call_del_wh_children, &args);
36111 +               if (unlikely(wkq_err))
36112 +                       err = wkq_err;
36113 +       }
36114 +       inode_unlock(wh_inode);
36115 +
36116 +       if (!err) {
36117 +               h_nlink = h_dir->i_nlink;
36118 +               err = vfsub_rmdir(h_dir, &wh_path);
36119 +               /* some fs doesn't change the parent nlink in some cases */
36120 +               h_nlink -= h_dir->i_nlink;
36121 +       }
36122 +
36123 +       if (!err) {
36124 +               if (au_ibtop(dir) == bindex) {
36125 +                       /* todo: dir->i_mutex is necessary */
36126 +                       au_cpup_attr_timesizes(dir);
36127 +                       if (h_nlink)
36128 +                               vfsub_drop_nlink(dir);
36129 +               }
36130 +               return 0; /* success */
36131 +       }
36132 +
36133 +       pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
36134 +       return err;
36135 +}
36136 +
36137 +static void call_rmdir_whtmp(void *args)
36138 +{
36139 +       int err;
36140 +       aufs_bindex_t bindex;
36141 +       struct au_whtmp_rmdir *a = args;
36142 +       struct super_block *sb;
36143 +       struct dentry *h_parent;
36144 +       struct inode *h_dir;
36145 +       struct au_hinode *hdir;
36146 +
36147 +       /* rmdir by nfsd may cause deadlock with this i_mutex */
36148 +       /* inode_lock(a->dir); */
36149 +       err = -EROFS;
36150 +       sb = a->dir->i_sb;
36151 +       si_read_lock(sb, !AuLock_FLUSH);
36152 +       if (!au_br_writable(a->br->br_perm))
36153 +               goto out;
36154 +       bindex = au_br_index(sb, a->br->br_id);
36155 +       if (unlikely(bindex < 0))
36156 +               goto out;
36157 +
36158 +       err = -EIO;
36159 +       ii_write_lock_parent(a->dir);
36160 +       h_parent = dget_parent(a->wh_dentry);
36161 +       h_dir = d_inode(h_parent);
36162 +       hdir = au_hi(a->dir, bindex);
36163 +       err = vfsub_mnt_want_write(au_br_mnt(a->br));
36164 +       if (unlikely(err))
36165 +               goto out_mnt;
36166 +       au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
36167 +       err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
36168 +                         a->br);
36169 +       if (!err)
36170 +               err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
36171 +       au_hn_inode_unlock(hdir);
36172 +       vfsub_mnt_drop_write(au_br_mnt(a->br));
36173 +
36174 +out_mnt:
36175 +       dput(h_parent);
36176 +       ii_write_unlock(a->dir);
36177 +out:
36178 +       /* inode_unlock(a->dir); */
36179 +       au_whtmp_rmdir_free(a);
36180 +       si_read_unlock(sb);
36181 +       au_nwt_done(&au_sbi(sb)->si_nowait);
36182 +       if (unlikely(err))
36183 +               AuIOErr("err %d\n", err);
36184 +}
36185 +
36186 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36187 +                        struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
36188 +{
36189 +       int wkq_err;
36190 +       struct super_block *sb;
36191 +
36192 +       IMustLock(dir);
36193 +
36194 +       /* all post-process will be done in do_rmdir_whtmp(). */
36195 +       sb = dir->i_sb;
36196 +       args->dir = au_igrab(dir);
36197 +       args->br = au_sbr(sb, bindex);
36198 +       au_lcnt_inc(&args->br->br_count);
36199 +       args->wh_dentry = dget(wh_dentry);
36200 +       wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
36201 +       if (unlikely(wkq_err)) {
36202 +               pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
36203 +               au_whtmp_rmdir_free(args);
36204 +       }
36205 +}
36206 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
36207 --- /usr/share/empty/fs/aufs/whout.h    1970-01-01 01:00:00.000000000 +0100
36208 +++ linux/fs/aufs/whout.h       2022-11-05 23:02:18.972555950 +0100
36209 @@ -0,0 +1,87 @@
36210 +/* SPDX-License-Identifier: GPL-2.0 */
36211 +/*
36212 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36213 + *
36214 + * This program is free software; you can redistribute it and/or modify
36215 + * it under the terms of the GNU General Public License as published by
36216 + * the Free Software Foundation; either version 2 of the License, or
36217 + * (at your option) any later version.
36218 + *
36219 + * This program is distributed in the hope that it will be useful,
36220 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36221 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36222 + * GNU General Public License for more details.
36223 + *
36224 + * You should have received a copy of the GNU General Public License
36225 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36226 + */
36227 +
36228 +/*
36229 + * whiteout for logical deletion and opaque directory
36230 + */
36231 +
36232 +#ifndef __AUFS_WHOUT_H__
36233 +#define __AUFS_WHOUT_H__
36234 +
36235 +#ifdef __KERNEL__
36236 +
36237 +#include "dir.h"
36238 +
36239 +/* whout.c */
36240 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
36241 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
36242 +              struct qstr *wh_name, int try_sio);
36243 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path);
36244 +struct au_branch;
36245 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
36246 +                            struct qstr *prefix);
36247 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
36248 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
36249 +                       struct dentry *dentry);
36250 +int au_wh_init(struct au_branch *br, struct super_block *sb);
36251 +
36252 +/* diropq flags */
36253 +#define AuDiropq_CREATE        1
36254 +#define au_ftest_diropq(flags, name)   ((flags) & AuDiropq_##name)
36255 +#define au_fset_diropq(flags, name) \
36256 +       do { (flags) |= AuDiropq_##name; } while (0)
36257 +#define au_fclr_diropq(flags, name) \
36258 +       do { (flags) &= ~AuDiropq_##name; } while (0)
36259 +
36260 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
36261 +                            unsigned int flags);
36262 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
36263 +                         struct au_branch *br);
36264 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
36265 +                           struct dentry *h_parent);
36266 +
36267 +/* real rmdir for the whiteout-ed dir */
36268 +struct au_whtmp_rmdir {
36269 +       struct inode *dir;
36270 +       struct au_branch *br;
36271 +       struct dentry *wh_dentry;
36272 +       struct au_nhash whlist;
36273 +};
36274 +
36275 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
36276 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
36277 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36278 +                  struct dentry *wh_dentry, struct au_nhash *whlist);
36279 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36280 +                        struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
36281 +
36282 +/* ---------------------------------------------------------------------- */
36283 +
36284 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
36285 +                                             aufs_bindex_t bindex)
36286 +{
36287 +       return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
36288 +}
36289 +
36290 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
36291 +{
36292 +       return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
36293 +}
36294 +
36295 +#endif /* __KERNEL__ */
36296 +#endif /* __AUFS_WHOUT_H__ */
36297 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
36298 --- /usr/share/empty/fs/aufs/wkq.c      1970-01-01 01:00:00.000000000 +0100
36299 +++ linux/fs/aufs/wkq.c 2022-11-05 23:02:18.972555950 +0100
36300 @@ -0,0 +1,372 @@
36301 +// SPDX-License-Identifier: GPL-2.0
36302 +/*
36303 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36304 + *
36305 + * This program is free software; you can redistribute it and/or modify
36306 + * it under the terms of the GNU General Public License as published by
36307 + * the Free Software Foundation; either version 2 of the License, or
36308 + * (at your option) any later version.
36309 + *
36310 + * This program is distributed in the hope that it will be useful,
36311 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36312 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36313 + * GNU General Public License for more details.
36314 + *
36315 + * You should have received a copy of the GNU General Public License
36316 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36317 + */
36318 +
36319 +/*
36320 + * workqueue for asynchronous/super-io operations
36321 + * todo: try new credential scheme
36322 + */
36323 +
36324 +#include <linux/module.h>
36325 +#include "aufs.h"
36326 +
36327 +/* internal workqueue named AUFS_WKQ_NAME */
36328 +
36329 +static struct workqueue_struct *au_wkq;
36330 +
36331 +struct au_wkinfo {
36332 +       struct work_struct wk;
36333 +       struct kobject *kobj;
36334 +
36335 +       unsigned int flags; /* see wkq.h */
36336 +
36337 +       au_wkq_func_t func;
36338 +       void *args;
36339 +
36340 +#ifdef CONFIG_LOCKDEP
36341 +       int dont_check;
36342 +       struct held_lock **hlock;
36343 +#endif
36344 +
36345 +       struct completion *comp;
36346 +};
36347 +
36348 +/* ---------------------------------------------------------------------- */
36349 +/*
36350 + * Aufs passes some operations to the workqueue such as the internal copyup.
36351 + * This scheme looks rather unnatural for LOCKDEP debugging feature, since the
36352 + * job run by workqueue depends upon the locks acquired in the other task.
36353 + * Delegating a small operation to the workqueue, aufs passes its lockdep
36354 + * information too. And the job in the workqueue restores the info in order to
36355 + * pretend as if it acquired those locks. This is just to make LOCKDEP work
36356 + * correctly and expectedly.
36357 + */
36358 +
36359 +#ifndef CONFIG_LOCKDEP
36360 +AuStubInt0(au_wkq_lockdep_alloc, struct au_wkinfo *wkinfo);
36361 +AuStubVoid(au_wkq_lockdep_free, struct au_wkinfo *wkinfo);
36362 +AuStubVoid(au_wkq_lockdep_pre, struct au_wkinfo *wkinfo);
36363 +AuStubVoid(au_wkq_lockdep_post, struct au_wkinfo *wkinfo);
36364 +AuStubVoid(au_wkq_lockdep_init, struct au_wkinfo *wkinfo);
36365 +#else
36366 +static void au_wkq_lockdep_init(struct au_wkinfo *wkinfo)
36367 +{
36368 +       wkinfo->hlock = NULL;
36369 +       wkinfo->dont_check = 0;
36370 +}
36371 +
36372 +/*
36373 + * 1: matched
36374 + * 0: unmatched
36375 + */
36376 +static int au_wkq_lockdep_test(struct lock_class_key *key, const char *name)
36377 +{
36378 +       static DEFINE_SPINLOCK(spin);
36379 +       static struct {
36380 +               char *name;
36381 +               struct lock_class_key *key;
36382 +       } a[] = {
36383 +               { .name = "&sbinfo->si_rwsem" },
36384 +               { .name = "&finfo->fi_rwsem" },
36385 +               { .name = "&dinfo->di_rwsem" },
36386 +               { .name = "&iinfo->ii_rwsem" }
36387 +       };
36388 +       static int set;
36389 +       int i;
36390 +
36391 +       /* lockless read from 'set.' see below */
36392 +       if (set == ARRAY_SIZE(a)) {
36393 +               for (i = 0; i < ARRAY_SIZE(a); i++)
36394 +                       if (a[i].key == key)
36395 +                               goto match;
36396 +               goto unmatch;
36397 +       }
36398 +
36399 +       spin_lock(&spin);
36400 +       if (set)
36401 +               for (i = 0; i < ARRAY_SIZE(a); i++)
36402 +                       if (a[i].key == key) {
36403 +                               spin_unlock(&spin);
36404 +                               goto match;
36405 +                       }
36406 +       for (i = 0; i < ARRAY_SIZE(a); i++) {
36407 +               if (a[i].key) {
36408 +                       if (unlikely(a[i].key == key)) { /* rare but possible */
36409 +                               spin_unlock(&spin);
36410 +                               goto match;
36411 +                       } else
36412 +                               continue;
36413 +               }
36414 +               if (strstr(a[i].name, name)) {
36415 +                       /*
36416 +                        * the order of these three lines is important for the
36417 +                        * lockless read above.
36418 +                        */
36419 +                       a[i].key = key;
36420 +                       spin_unlock(&spin);
36421 +                       set++;
36422 +                       /* AuDbg("%d, %s\n", set, name); */
36423 +                       goto match;
36424 +               }
36425 +       }
36426 +       spin_unlock(&spin);
36427 +       goto unmatch;
36428 +
36429 +match:
36430 +       return 1;
36431 +unmatch:
36432 +       return 0;
36433 +}
36434 +
36435 +static int au_wkq_lockdep_alloc(struct au_wkinfo *wkinfo)
36436 +{
36437 +       int err, n;
36438 +       struct task_struct *curr;
36439 +       struct held_lock **hl, *held_locks, *p;
36440 +
36441 +       err = 0;
36442 +       curr = current;
36443 +       wkinfo->dont_check = lockdep_recursing(curr);
36444 +       if (wkinfo->dont_check)
36445 +               goto out;
36446 +       n = curr->lockdep_depth;
36447 +       if (!n)
36448 +               goto out;
36449 +
36450 +       err = -ENOMEM;
36451 +       wkinfo->hlock = kmalloc_array(n + 1, sizeof(*wkinfo->hlock), GFP_NOFS);
36452 +       if (unlikely(!wkinfo->hlock))
36453 +               goto out;
36454 +
36455 +       err = 0;
36456 +#if 0 /* left for debugging */
36457 +       if (0 && au_debug_test())
36458 +               lockdep_print_held_locks(curr);
36459 +#endif
36460 +       held_locks = curr->held_locks;
36461 +       hl = wkinfo->hlock;
36462 +       while (n--) {
36463 +               p = held_locks++;
36464 +               if (au_wkq_lockdep_test(p->instance->key, p->instance->name))
36465 +                       *hl++ = p;
36466 +       }
36467 +       *hl = NULL;
36468 +
36469 +out:
36470 +       return err;
36471 +}
36472 +
36473 +static void au_wkq_lockdep_free(struct au_wkinfo *wkinfo)
36474 +{
36475 +       au_kfree_try_rcu(wkinfo->hlock);
36476 +}
36477 +
36478 +static void au_wkq_lockdep_pre(struct au_wkinfo *wkinfo)
36479 +{
36480 +       struct held_lock *p, **hl = wkinfo->hlock;
36481 +       int subclass;
36482 +
36483 +       if (wkinfo->dont_check)
36484 +               lockdep_off();
36485 +       if (!hl)
36486 +               return;
36487 +       while ((p = *hl++)) { /* assignment */
36488 +               subclass = lockdep_hlock_class(p)->subclass;
36489 +               /* AuDbg("%s, %d\n", p->instance->name, subclass); */
36490 +               if (p->read)
36491 +                       rwsem_acquire_read(p->instance, subclass, 0,
36492 +                                          /*p->acquire_ip*/_RET_IP_);
36493 +               else
36494 +                       rwsem_acquire(p->instance, subclass, 0,
36495 +                                     /*p->acquire_ip*/_RET_IP_);
36496 +       }
36497 +}
36498 +
36499 +static void au_wkq_lockdep_post(struct au_wkinfo *wkinfo)
36500 +{
36501 +       struct held_lock *p, **hl = wkinfo->hlock;
36502 +
36503 +       if (wkinfo->dont_check)
36504 +               lockdep_on();
36505 +       if (!hl)
36506 +               return;
36507 +       while ((p = *hl++)) /* assignment */
36508 +               rwsem_release(p->instance, /*p->acquire_ip*/_RET_IP_);
36509 +}
36510 +#endif
36511 +
36512 +static void wkq_func(struct work_struct *wk)
36513 +{
36514 +       struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
36515 +
36516 +       AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
36517 +       AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
36518 +
36519 +       au_wkq_lockdep_pre(wkinfo);
36520 +       wkinfo->func(wkinfo->args);
36521 +       au_wkq_lockdep_post(wkinfo);
36522 +       if (au_ftest_wkq(wkinfo->flags, WAIT))
36523 +               complete(wkinfo->comp);
36524 +       else {
36525 +               kobject_put(wkinfo->kobj);
36526 +               module_put(THIS_MODULE); /* todo: ?? */
36527 +               au_kfree_rcu(wkinfo);
36528 +       }
36529 +}
36530 +
36531 +/*
36532 + * Since struct completion is large, try allocating it dynamically.
36533 + */
36534 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
36535 +
36536 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
36537 +{
36538 +       *comp = kmalloc(sizeof(**comp), GFP_NOFS);
36539 +       if (*comp) {
36540 +               init_completion(*comp);
36541 +               wkinfo->comp = *comp;
36542 +               return 0;
36543 +       }
36544 +       return -ENOMEM;
36545 +}
36546 +
36547 +static void au_wkq_comp_free(struct completion *comp)
36548 +{
36549 +       au_kfree_rcu(comp);
36550 +}
36551 +
36552 +static void au_wkq_run(struct au_wkinfo *wkinfo)
36553 +{
36554 +       if (au_ftest_wkq(wkinfo->flags, NEST)) {
36555 +               if (au_wkq_test()) {
36556 +                       AuWarn1("wkq from wkq, unless silly-rename on NFS,"
36557 +                               " due to a dead dir by UDBA,"
36558 +                               " or async xino write?\n");
36559 +                       AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
36560 +               }
36561 +       } else
36562 +               au_dbg_verify_kthread();
36563 +
36564 +       if (au_ftest_wkq(wkinfo->flags, WAIT)) {
36565 +               INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
36566 +               queue_work(au_wkq, &wkinfo->wk);
36567 +       } else {
36568 +               INIT_WORK(&wkinfo->wk, wkq_func);
36569 +               schedule_work(&wkinfo->wk);
36570 +       }
36571 +}
36572 +
36573 +/*
36574 + * Be careful. It is easy to make deadlock happen.
36575 + * processA: lock, wkq and wait
36576 + * processB: wkq and wait, lock in wkq
36577 + * --> deadlock
36578 + */
36579 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
36580 +{
36581 +       int err;
36582 +       AuWkqCompDeclare(comp);
36583 +       struct au_wkinfo wkinfo = {
36584 +               .flags  = flags,
36585 +               .func   = func,
36586 +               .args   = args
36587 +       };
36588 +
36589 +       err = au_wkq_comp_alloc(&wkinfo, &comp);
36590 +       if (unlikely(err))
36591 +               goto out;
36592 +       err = au_wkq_lockdep_alloc(&wkinfo);
36593 +       if (unlikely(err))
36594 +               goto out_comp;
36595 +       if (!err) {
36596 +               au_wkq_run(&wkinfo);
36597 +               /* no timeout, no interrupt */
36598 +               wait_for_completion(wkinfo.comp);
36599 +       }
36600 +       au_wkq_lockdep_free(&wkinfo);
36601 +
36602 +out_comp:
36603 +       au_wkq_comp_free(comp);
36604 +out:
36605 +       destroy_work_on_stack(&wkinfo.wk);
36606 +       return err;
36607 +}
36608 +
36609 +/*
36610 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
36611 + * problem in a concurrent umounting.
36612 + */
36613 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36614 +                 unsigned int flags)
36615 +{
36616 +       int err;
36617 +       struct au_wkinfo *wkinfo;
36618 +
36619 +       atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
36620 +
36621 +       /*
36622 +        * wkq_func() must free this wkinfo.
36623 +        * it highly depends upon the implementation of workqueue.
36624 +        */
36625 +       err = 0;
36626 +       wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
36627 +       if (wkinfo) {
36628 +               wkinfo->kobj = &au_sbi(sb)->si_kobj;
36629 +               wkinfo->flags = flags & ~AuWkq_WAIT;
36630 +               wkinfo->func = func;
36631 +               wkinfo->args = args;
36632 +               wkinfo->comp = NULL;
36633 +               au_wkq_lockdep_init(wkinfo);
36634 +               kobject_get(wkinfo->kobj);
36635 +               __module_get(THIS_MODULE); /* todo: ?? */
36636 +
36637 +               au_wkq_run(wkinfo);
36638 +       } else {
36639 +               err = -ENOMEM;
36640 +               au_nwt_done(&au_sbi(sb)->si_nowait);
36641 +       }
36642 +
36643 +       return err;
36644 +}
36645 +
36646 +/* ---------------------------------------------------------------------- */
36647 +
36648 +void au_nwt_init(struct au_nowait_tasks *nwt)
36649 +{
36650 +       atomic_set(&nwt->nw_len, 0);
36651 +       /* smp_mb(); */ /* atomic_set */
36652 +       init_waitqueue_head(&nwt->nw_wq);
36653 +}
36654 +
36655 +void au_wkq_fin(void)
36656 +{
36657 +       destroy_workqueue(au_wkq);
36658 +}
36659 +
36660 +int __init au_wkq_init(void)
36661 +{
36662 +       int err;
36663 +
36664 +       err = 0;
36665 +       au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
36666 +       if (IS_ERR(au_wkq))
36667 +               err = PTR_ERR(au_wkq);
36668 +       else if (!au_wkq)
36669 +               err = -ENOMEM;
36670 +
36671 +       return err;
36672 +}
36673 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
36674 --- /usr/share/empty/fs/aufs/wkq.h      1970-01-01 01:00:00.000000000 +0100
36675 +++ linux/fs/aufs/wkq.h 2022-11-05 23:02:18.972555950 +0100
36676 @@ -0,0 +1,89 @@
36677 +/* SPDX-License-Identifier: GPL-2.0 */
36678 +/*
36679 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36680 + *
36681 + * This program is free software; you can redistribute it and/or modify
36682 + * it under the terms of the GNU General Public License as published by
36683 + * the Free Software Foundation; either version 2 of the License, or
36684 + * (at your option) any later version.
36685 + *
36686 + * This program is distributed in the hope that it will be useful,
36687 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36688 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36689 + * GNU General Public License for more details.
36690 + *
36691 + * You should have received a copy of the GNU General Public License
36692 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36693 + */
36694 +
36695 +/*
36696 + * workqueue for asynchronous/super-io operations
36697 + * todo: try new credentials management scheme
36698 + */
36699 +
36700 +#ifndef __AUFS_WKQ_H__
36701 +#define __AUFS_WKQ_H__
36702 +
36703 +#ifdef __KERNEL__
36704 +
36705 +#include <linux/wait.h>
36706 +
36707 +struct super_block;
36708 +
36709 +/* ---------------------------------------------------------------------- */
36710 +
36711 +/*
36712 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
36713 + */
36714 +struct au_nowait_tasks {
36715 +       atomic_t                nw_len;
36716 +       wait_queue_head_t       nw_wq;
36717 +};
36718 +
36719 +/* ---------------------------------------------------------------------- */
36720 +
36721 +typedef void (*au_wkq_func_t)(void *args);
36722 +
36723 +/* wkq flags */
36724 +#define AuWkq_WAIT     1
36725 +#define AuWkq_NEST     (1 << 1)
36726 +#define au_ftest_wkq(flags, name)      ((flags) & AuWkq_##name)
36727 +#define au_fset_wkq(flags, name) \
36728 +       do { (flags) |= AuWkq_##name; } while (0)
36729 +#define au_fclr_wkq(flags, name) \
36730 +       do { (flags) &= ~AuWkq_##name; } while (0)
36731 +
36732 +/* wkq.c */
36733 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
36734 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36735 +                 unsigned int flags);
36736 +void au_nwt_init(struct au_nowait_tasks *nwt);
36737 +int __init au_wkq_init(void);
36738 +void au_wkq_fin(void);
36739 +
36740 +/* ---------------------------------------------------------------------- */
36741 +
36742 +static inline int au_wkq_test(void)
36743 +{
36744 +       return current->flags & PF_WQ_WORKER;
36745 +}
36746 +
36747 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
36748 +{
36749 +       return au_wkq_do_wait(AuWkq_WAIT, func, args);
36750 +}
36751 +
36752 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
36753 +{
36754 +       if (atomic_dec_and_test(&nwt->nw_len))
36755 +               wake_up_all(&nwt->nw_wq);
36756 +}
36757 +
36758 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
36759 +{
36760 +       wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
36761 +       return 0;
36762 +}
36763 +
36764 +#endif /* __KERNEL__ */
36765 +#endif /* __AUFS_WKQ_H__ */
36766 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
36767 --- /usr/share/empty/fs/aufs/xattr.c    1970-01-01 01:00:00.000000000 +0100
36768 +++ linux/fs/aufs/xattr.c       2023-02-20 21:05:51.959693785 +0100
36769 @@ -0,0 +1,365 @@
36770 +// SPDX-License-Identifier: GPL-2.0
36771 +/*
36772 + * Copyright (C) 2014-2022 Junjiro R. Okajima
36773 + *
36774 + * This program is free software; you can redistribute it and/or modify
36775 + * it under the terms of the GNU General Public License as published by
36776 + * the Free Software Foundation; either version 2 of the License, or
36777 + * (at your option) any later version.
36778 + *
36779 + * This program is distributed in the hope that it will be useful,
36780 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36781 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36782 + * GNU General Public License for more details.
36783 + *
36784 + * You should have received a copy of the GNU General Public License
36785 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36786 + */
36787 +
36788 +/*
36789 + * handling xattr functions
36790 + */
36791 +
36792 +#include <linux/fs.h>
36793 +#include <linux/posix_acl_xattr.h>
36794 +#include <linux/xattr.h>
36795 +#include "aufs.h"
36796 +
36797 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
36798 +{
36799 +       if (!ignore_flags)
36800 +               goto out;
36801 +       switch (err) {
36802 +       case -ENOMEM:
36803 +       case -EDQUOT:
36804 +               goto out;
36805 +       }
36806 +
36807 +       if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
36808 +               err = 0;
36809 +               goto out;
36810 +       }
36811 +
36812 +#define cmp(brattr, prefix) do {                                       \
36813 +               if (!strncmp(name, XATTR_##prefix##_PREFIX,             \
36814 +                            XATTR_##prefix##_PREFIX_LEN)) {            \
36815 +                       if (ignore_flags & AuBrAttr_ICEX_##brattr)      \
36816 +                               err = 0;                                \
36817 +                       goto out;                                       \
36818 +               }                                                       \
36819 +       } while (0)
36820 +
36821 +       cmp(SEC, SECURITY);
36822 +       cmp(SYS, SYSTEM);
36823 +       cmp(TR, TRUSTED);
36824 +       cmp(USR, USER);
36825 +#undef cmp
36826 +
36827 +       if (ignore_flags & AuBrAttr_ICEX_OTH)
36828 +               err = 0;
36829 +
36830 +out:
36831 +       return err;
36832 +}
36833 +
36834 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
36835 +
36836 +static int au_do_cpup_xattr(struct path *h_dst, struct path *h_src,
36837 +                           char *name, char **buf, unsigned int ignore_flags,
36838 +                           unsigned int verbose)
36839 +{
36840 +       int err, is_acl;
36841 +       ssize_t ssz;
36842 +       struct inode *h_idst;
36843 +       struct dentry *h_dst_dentry, *h_src_dentry;
36844 +       struct user_namespace *h_dst_userns, *h_src_userns;
36845 +       struct posix_acl *acl;
36846 +
36847 +       is_acl = !!is_posix_acl_xattr(name);
36848 +       h_src_userns = mnt_user_ns(h_src->mnt);
36849 +       h_src_dentry = h_src->dentry;
36850 +       if (is_acl) {
36851 +               acl = vfs_get_acl(h_src_userns, h_src_dentry, name);
36852 +               AuDebugOn(!acl);
36853 +               if (unlikely(IS_ERR(acl))) {
36854 +                       err = PTR_ERR(acl);
36855 +                       if (err == -ENODATA)
36856 +                               err = 0;
36857 +                       else if (err == -EOPNOTSUPP
36858 +                                && au_test_nfs_noacl(d_inode(h_src_dentry)))
36859 +                               err = 0;
36860 +                       else if (verbose || au_debug_test())
36861 +                               pr_err("%s, err %d\n", name, err);
36862 +                       goto out;
36863 +               }
36864 +       } else {
36865 +               ssz = vfs_getxattr_alloc(h_src_userns, h_src_dentry, name, buf,
36866 +                                        0, GFP_NOFS);
36867 +               if (unlikely(ssz <= 0)) {
36868 +                       err = ssz;
36869 +                       if (err == -ENODATA)
36870 +                               err = 0;
36871 +                       else if (err == -EOPNOTSUPP
36872 +                                && (ignore_flags & au_xattr_out_of_list))
36873 +                                err = 0;
36874 +                       else if (err && (verbose || au_debug_test()))
36875 +                               pr_err("%s, err %d\n", name, err);
36876 +                       goto out;
36877 +               }
36878 +       }
36879 +
36880 +       /* unlock it temporary */
36881 +       h_dst_userns = mnt_user_ns(h_dst->mnt);
36882 +       h_dst_dentry = h_dst->dentry;
36883 +       h_idst = d_inode(h_dst_dentry);
36884 +       inode_unlock(h_idst);
36885 +       if (is_acl) {
36886 +               err = vfsub_set_acl(h_dst_userns, h_dst_dentry, name, acl);
36887 +               posix_acl_release(acl);
36888 +       } else
36889 +               err = vfsub_setxattr(h_dst_userns, h_dst_dentry, name, *buf,
36890 +                                    ssz, /*flags*/0);
36891 +       inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36892 +       if (unlikely(err)) {
36893 +               if (verbose || au_debug_test())
36894 +                       pr_err("%s, err %d\n", name, err);
36895 +               err = au_xattr_ignore(err, name, ignore_flags);
36896 +       }
36897 +
36898 +out:
36899 +       return err;
36900 +}
36901 +
36902 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
36903 +                 unsigned int verbose)
36904 +{
36905 +       int err, unlocked;
36906 +       ssize_t ssz;
36907 +       struct dentry *h_dst_dentry, *h_src_dentry;
36908 +       struct inode *h_isrc, *h_idst;
36909 +       char *value, *p, *o, *e;
36910 +
36911 +       /* try stopping to update the source inode while we are referencing */
36912 +       /* there should not be the parent-child relationship between them */
36913 +       h_dst_dentry = h_dst->dentry;
36914 +       h_idst = d_inode(h_dst_dentry);
36915 +       h_src_dentry = h_src->dentry;
36916 +       h_isrc = d_inode(h_src_dentry);
36917 +       inode_unlock(h_idst);
36918 +       inode_lock_shared_nested(h_isrc, AuLsc_I_CHILD);
36919 +       inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36920 +       unlocked = 0;
36921 +
36922 +       /* some filesystems don't list POSIX ACL, for example tmpfs */
36923 +       ssz = vfs_listxattr(h_src_dentry, NULL, 0);
36924 +       err = ssz;
36925 +       if (unlikely(err < 0)) {
36926 +               AuTraceErr(err);
36927 +               if (err == -ENODATA
36928 +                   || err == -EOPNOTSUPP)
36929 +                       err = 0;        /* ignore */
36930 +               goto out;
36931 +       }
36932 +
36933 +       err = 0;
36934 +       p = NULL;
36935 +       o = NULL;
36936 +       if (ssz) {
36937 +               err = -ENOMEM;
36938 +               p = kmalloc(ssz, GFP_NOFS);
36939 +               o = p;
36940 +               if (unlikely(!p))
36941 +                       goto out;
36942 +               err = vfs_listxattr(h_src_dentry, p, ssz);
36943 +       }
36944 +       inode_unlock_shared(h_isrc);
36945 +       unlocked = 1;
36946 +       AuDbg("err %d, ssz %zd\n", err, ssz);
36947 +       if (unlikely(err < 0))
36948 +               goto out_free;
36949 +
36950 +       err = 0;
36951 +       e = p + ssz;
36952 +       value = NULL;
36953 +       ignore_flags |= au_xattr_out_of_list;
36954 +       while (!err && p < e) {
36955 +               err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
36956 +                                      verbose);
36957 +               p += strlen(p) + 1;
36958 +       }
36959 +       au_kfree_try_rcu(value);
36960 +
36961 +out_free:
36962 +       au_kfree_try_rcu(o);
36963 +out:
36964 +       if (!unlocked)
36965 +               inode_unlock_shared(h_isrc);
36966 +       AuTraceErr(err);
36967 +       return err;
36968 +}
36969 +
36970 +/* ---------------------------------------------------------------------- */
36971 +
36972 +static int au_smack_reentering(struct super_block *sb)
36973 +{
36974 +#if IS_ENABLED(CONFIG_SECURITY_SMACK) || IS_ENABLED(CONFIG_SECURITY_SELINUX)
36975 +       /*
36976 +        * as a part of lookup, smack_d_instantiate() is called, and it calls
36977 +        * i_op->getxattr(). ouch.
36978 +        */
36979 +       return si_pid_test(sb);
36980 +#else
36981 +       return 0;
36982 +#endif
36983 +}
36984 +
36985 +enum {
36986 +       AU_XATTR_LIST,
36987 +       AU_XATTR_GET
36988 +};
36989 +
36990 +struct au_lgxattr {
36991 +       int type;
36992 +       union {
36993 +               struct {
36994 +                       char    *list;
36995 +                       size_t  size;
36996 +               } list;
36997 +               struct {
36998 +                       const char      *name;
36999 +                       void            *value;
37000 +                       size_t          size;
37001 +               } get;
37002 +       } u;
37003 +};
37004 +
37005 +static ssize_t au_lgxattr(struct dentry *dentry, struct inode *inode,
37006 +                         struct au_lgxattr *arg)
37007 +{
37008 +       ssize_t err;
37009 +       int reenter;
37010 +       struct path h_path;
37011 +       struct super_block *sb;
37012 +
37013 +       sb = dentry->d_sb;
37014 +       reenter = au_smack_reentering(sb);
37015 +       if (!reenter) {
37016 +               err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
37017 +               if (unlikely(err))
37018 +                       goto out;
37019 +       }
37020 +       err = au_h_path_getattr(dentry, inode, /*force*/1, &h_path, reenter);
37021 +       if (unlikely(err))
37022 +               goto out_si;
37023 +       if (unlikely(!h_path.dentry))
37024 +               /* illegally overlapped or something */
37025 +               goto out_di; /* pretending success */
37026 +
37027 +       /* always topmost entry only */
37028 +       switch (arg->type) {
37029 +       case AU_XATTR_LIST:
37030 +               err = vfs_listxattr(h_path.dentry,
37031 +                                   arg->u.list.list, arg->u.list.size);
37032 +               break;
37033 +       case AU_XATTR_GET:
37034 +               AuDebugOn(d_is_negative(h_path.dentry));
37035 +               err = vfs_getxattr(mnt_user_ns(h_path.mnt), h_path.dentry,
37036 +                                  arg->u.get.name, arg->u.get.value,
37037 +                                  arg->u.get.size);
37038 +               break;
37039 +       }
37040 +
37041 +out_di:
37042 +       if (!reenter)
37043 +               di_read_unlock(dentry, AuLock_IR);
37044 +out_si:
37045 +       if (!reenter)
37046 +               si_read_unlock(sb);
37047 +out:
37048 +       AuTraceErr(err);
37049 +       return err;
37050 +}
37051 +
37052 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
37053 +{
37054 +       struct au_lgxattr arg = {
37055 +               .type = AU_XATTR_LIST,
37056 +               .u.list = {
37057 +                       .list   = list,
37058 +                       .size   = size
37059 +               },
37060 +       };
37061 +
37062 +       return au_lgxattr(dentry, /*inode*/NULL, &arg);
37063 +}
37064 +
37065 +static ssize_t au_getxattr(struct dentry *dentry, struct inode *inode,
37066 +                          const char *name, void *value, size_t size)
37067 +{
37068 +       struct au_lgxattr arg = {
37069 +               .type = AU_XATTR_GET,
37070 +               .u.get = {
37071 +                       .name   = name,
37072 +                       .value  = value,
37073 +                       .size   = size
37074 +               },
37075 +       };
37076 +
37077 +       return au_lgxattr(dentry, inode, &arg);
37078 +}
37079 +
37080 +static int au_setxattr(struct dentry *dentry, struct inode *inode,
37081 +                      const char *name, const void *value, size_t size,
37082 +                      int flags)
37083 +{
37084 +       struct au_sxattr arg = {
37085 +               .type = AU_XATTR_SET,
37086 +               .u.set = {
37087 +                       .name   = name,
37088 +                       .value  = value,
37089 +                       .size   = size,
37090 +                       .flags  = flags
37091 +               },
37092 +       };
37093 +
37094 +       return au_sxattr(dentry, inode, &arg);
37095 +}
37096 +
37097 +/* ---------------------------------------------------------------------- */
37098 +
37099 +static int au_xattr_get(const struct xattr_handler *handler,
37100 +                       struct dentry *dentry, struct inode *inode,
37101 +                       const char *name, void *buffer, size_t size)
37102 +{
37103 +       return au_getxattr(dentry, inode, name, buffer, size);
37104 +}
37105 +
37106 +static int au_xattr_set(const struct xattr_handler *handler,
37107 +                       struct user_namespace *userns,
37108 +                       struct dentry *dentry, struct inode *inode,
37109 +                       const char *name, const void *value, size_t size,
37110 +                       int flags)
37111 +{
37112 +       return au_setxattr(dentry, inode, name, value, size, flags);
37113 +}
37114 +
37115 +static const struct xattr_handler au_xattr_handler = {
37116 +       .name   = "",
37117 +       .prefix = "",
37118 +       .get    = au_xattr_get,
37119 +       .set    = au_xattr_set
37120 +};
37121 +
37122 +static const struct xattr_handler *au_xattr_handlers[] = {
37123 +#ifdef CONFIG_FS_POSIX_ACL
37124 +       &posix_acl_access_xattr_handler,
37125 +       &posix_acl_default_xattr_handler,
37126 +#endif
37127 +       &au_xattr_handler, /* must be last */
37128 +       NULL
37129 +};
37130 +
37131 +void au_xattr_init(struct super_block *sb)
37132 +{
37133 +       sb->s_xattr = au_xattr_handlers;
37134 +}
37135 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
37136 --- /usr/share/empty/fs/aufs/xino.c     1970-01-01 01:00:00.000000000 +0100
37137 +++ linux/fs/aufs/xino.c        2022-11-05 23:02:18.972555950 +0100
37138 @@ -0,0 +1,1926 @@
37139 +// SPDX-License-Identifier: GPL-2.0
37140 +/*
37141 + * Copyright (C) 2005-2022 Junjiro R. Okajima
37142 + *
37143 + * This program is free software; you can redistribute it and/or modify
37144 + * it under the terms of the GNU General Public License as published by
37145 + * the Free Software Foundation; either version 2 of the License, or
37146 + * (at your option) any later version.
37147 + *
37148 + * This program is distributed in the hope that it will be useful,
37149 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
37150 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
37151 + * GNU General Public License for more details.
37152 + *
37153 + * You should have received a copy of the GNU General Public License
37154 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
37155 + */
37156 +
37157 +/*
37158 + * external inode number translation table and bitmap
37159 + *
37160 + * things to consider
37161 + * - the lifetime
37162 + *   + au_xino object
37163 + *   + XINO files (xino, xib, xigen)
37164 + *   + dynamic debugfs entries (xiN)
37165 + *   + static debugfs entries (xib, xigen)
37166 + *   + static sysfs entry (xi_path)
37167 + * - several entry points to handle them.
37168 + *   + mount(2) without xino option (default)
37169 + *   + mount(2) with xino option
37170 + *   + mount(2) with noxino option
37171 + *   + umount(2)
37172 + *   + remount with add/del branches
37173 + *   + remount with xino/noxino options
37174 + */
37175 +
37176 +#include <linux/seq_file.h>
37177 +#include <linux/statfs.h>
37178 +#include "aufs.h"
37179 +
37180 +static aufs_bindex_t sbr_find_shared(struct super_block *sb, aufs_bindex_t btop,
37181 +                                    aufs_bindex_t bbot,
37182 +                                    struct super_block *h_sb)
37183 +{
37184 +       /* todo: try binary-search if the branches are many */
37185 +       for (; btop <= bbot; btop++)
37186 +               if (h_sb == au_sbr_sb(sb, btop))
37187 +                       return btop;
37188 +       return -1;
37189 +}
37190 +
37191 +/*
37192 + * find another branch who is on the same filesystem of the specified
37193 + * branch{@btgt}. search until @bbot.
37194 + */
37195 +static aufs_bindex_t is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
37196 +                                 aufs_bindex_t bbot)
37197 +{
37198 +       aufs_bindex_t bindex;
37199 +       struct super_block *tgt_sb;
37200 +
37201 +       tgt_sb = au_sbr_sb(sb, btgt);
37202 +       bindex = sbr_find_shared(sb, /*btop*/0, btgt - 1, tgt_sb);
37203 +       if (bindex < 0)
37204 +               bindex = sbr_find_shared(sb, btgt + 1, bbot, tgt_sb);
37205 +
37206 +       return bindex;
37207 +}
37208 +
37209 +/* ---------------------------------------------------------------------- */
37210 +
37211 +/*
37212 + * stop unnecessary notify events at creating xino files
37213 + */
37214 +
37215 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry)
37216 +{
37217 +       aufs_bindex_t bfound, bindex, bbot;
37218 +       struct dentry *parent;
37219 +       struct au_branch *br;
37220 +
37221 +       bfound = -1;
37222 +       parent = dentry->d_parent; /* safe d_parent access */
37223 +       bbot = au_sbbot(sb);
37224 +       for (bindex = 0; bindex <= bbot; bindex++) {
37225 +               br = au_sbr(sb, bindex);
37226 +               if (au_br_dentry(br) == parent) {
37227 +                       bfound = bindex;
37228 +                       break;
37229 +               }
37230 +       }
37231 +
37232 +       AuDbg("bfound b%d\n", bfound);
37233 +       return bfound;
37234 +}
37235 +
37236 +struct au_xino_lock_dir {
37237 +       struct au_hinode *hdir;
37238 +       struct dentry *parent;
37239 +       struct inode *dir;
37240 +};
37241 +
37242 +static struct dentry *au_dget_parent_lock(struct dentry *dentry,
37243 +                                         unsigned int lsc)
37244 +{
37245 +       struct dentry *parent;
37246 +       struct inode *dir;
37247 +
37248 +       parent = dget_parent(dentry);
37249 +       dir = d_inode(parent);
37250 +       inode_lock_nested(dir, lsc);
37251 +#if 0 /* it should not happen */
37252 +       spin_lock(&dentry->d_lock);
37253 +       if (unlikely(dentry->d_parent != parent)) {
37254 +               spin_unlock(&dentry->d_lock);
37255 +               inode_unlock(dir);
37256 +               dput(parent);
37257 +               parent = NULL;
37258 +               goto out;
37259 +       }
37260 +       spin_unlock(&dentry->d_lock);
37261 +
37262 +out:
37263 +#endif
37264 +       return parent;
37265 +}
37266 +
37267 +static void au_xino_lock_dir(struct super_block *sb, struct path *xipath,
37268 +                            struct au_xino_lock_dir *ldir)
37269 +{
37270 +       aufs_bindex_t bindex;
37271 +
37272 +       ldir->hdir = NULL;
37273 +       bindex = au_xi_root(sb, xipath->dentry);
37274 +       if (bindex >= 0) {
37275 +               /* rw branch root */
37276 +               ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
37277 +               au_hn_inode_lock_nested(ldir->hdir, AuLsc_I_PARENT);
37278 +       } else {
37279 +               /* other */
37280 +               ldir->parent = au_dget_parent_lock(xipath->dentry,
37281 +                                                  AuLsc_I_PARENT);
37282 +               ldir->dir = d_inode(ldir->parent);
37283 +       }
37284 +}
37285 +
37286 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
37287 +{
37288 +       if (ldir->hdir)
37289 +               au_hn_inode_unlock(ldir->hdir);
37290 +       else {
37291 +               inode_unlock(ldir->dir);
37292 +               dput(ldir->parent);
37293 +       }
37294 +}
37295 +
37296 +/* ---------------------------------------------------------------------- */
37297 +
37298 +/*
37299 + * create and set a new xino file
37300 + */
37301 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
37302 +                           int wbrtop)
37303 +{
37304 +       struct file *file;
37305 +       struct dentry *h_parent, *d;
37306 +       struct inode *h_dir, *inode;
37307 +       int err;
37308 +       static DEFINE_MUTEX(mtx);
37309 +
37310 +       /*
37311 +        * at mount-time, and the xino file is the default path,
37312 +        * hnotify is disabled so we have no notify events to ignore.
37313 +        * when a user specified the xino, we cannot get au_hdir to be ignored.
37314 +        */
37315 +       if (!wbrtop)
37316 +               mutex_lock(&mtx);
37317 +       file = vfsub_filp_open(fpath, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37318 +                              /* | __FMODE_NONOTIFY */,
37319 +                              0666);
37320 +       if (IS_ERR(file)) {
37321 +               if (!wbrtop)
37322 +                       mutex_unlock(&mtx);
37323 +               if (!silent)
37324 +                       pr_err("open %s(%ld)\n", fpath, PTR_ERR(file));
37325 +               return file;
37326 +       }
37327 +
37328 +       /* keep file count */
37329 +       err = 0;
37330 +       d = file->f_path.dentry;
37331 +       h_parent = au_dget_parent_lock(d, AuLsc_I_PARENT);
37332 +       if (!wbrtop)
37333 +               mutex_unlock(&mtx);
37334 +       /* mnt_want_write() is unnecessary here */
37335 +       h_dir = d_inode(h_parent);
37336 +       inode = file_inode(file);
37337 +       /* no delegation since it is just created */
37338 +       if (inode->i_nlink)
37339 +               err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
37340 +                                  /*force*/0);
37341 +       inode_unlock(h_dir);
37342 +       dput(h_parent);
37343 +       if (unlikely(err)) {
37344 +               if (!silent)
37345 +                       pr_err("unlink %s(%d)\n", fpath, err);
37346 +               goto out;
37347 +       }
37348 +
37349 +       err = -EINVAL;
37350 +       if (unlikely(sb && sb == d->d_sb)) {
37351 +               if (!silent)
37352 +                       pr_err("%s must be outside\n", fpath);
37353 +               goto out;
37354 +       }
37355 +       if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
37356 +               if (!silent)
37357 +                       pr_err("xino doesn't support %s(%s)\n",
37358 +                              fpath, au_sbtype(d->d_sb));
37359 +               goto out;
37360 +       }
37361 +       return file; /* success */
37362 +
37363 +out:
37364 +       fput(file);
37365 +       file = ERR_PTR(err);
37366 +       return file;
37367 +}
37368 +
37369 +/*
37370 + * create a new xinofile at the same place/path as @base.
37371 + */
37372 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
37373 +                            struct file *copy_src)
37374 +{
37375 +       struct file *file;
37376 +       struct dentry *dentry;
37377 +       struct inode *dir, *delegated;
37378 +       struct qstr *name;
37379 +       struct path ppath, path;
37380 +       int err, do_unlock;
37381 +       struct au_xino_lock_dir ldir;
37382 +
37383 +       do_unlock = 1;
37384 +       au_xino_lock_dir(sb, base, &ldir);
37385 +       dentry = base->dentry;
37386 +       ppath.dentry = dentry->d_parent; /* dir inode is locked */
37387 +       ppath.mnt = base->mnt;
37388 +       dir = d_inode(ppath.dentry);
37389 +       IMustLock(dir);
37390 +
37391 +       name = &dentry->d_name;
37392 +       path.dentry = vfsub_lookup_one_len(name->name, &ppath, name->len);
37393 +       if (IS_ERR(path.dentry)) {
37394 +               file = (void *)path.dentry;
37395 +               pr_err("%pd lookup err %ld\n", dentry, PTR_ERR(path.dentry));
37396 +               goto out;
37397 +       }
37398 +
37399 +       /* no need to mnt_want_write() since we call dentry_open() later */
37400 +       err = vfs_create(mnt_user_ns(base->mnt), dir, path.dentry, 0666, NULL);
37401 +       if (unlikely(err)) {
37402 +               file = ERR_PTR(err);
37403 +               pr_err("%pd create err %d\n", dentry, err);
37404 +               goto out_dput;
37405 +       }
37406 +
37407 +       path.mnt = base->mnt;
37408 +       file = vfsub_dentry_open(&path,
37409 +                                O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37410 +                                /* | __FMODE_NONOTIFY */);
37411 +       if (IS_ERR(file)) {
37412 +               pr_err("%pd open err %ld\n", dentry, PTR_ERR(file));
37413 +               goto out_dput;
37414 +       }
37415 +
37416 +       delegated = NULL;
37417 +       err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
37418 +       au_xino_unlock_dir(&ldir);
37419 +       do_unlock = 0;
37420 +       if (unlikely(err == -EWOULDBLOCK)) {
37421 +               pr_warn("cannot retry for NFSv4 delegation"
37422 +                       " for an internal unlink\n");
37423 +               iput(delegated);
37424 +       }
37425 +       if (unlikely(err)) {
37426 +               pr_err("%pd unlink err %d\n", dentry, err);
37427 +               goto out_fput;
37428 +       }
37429 +
37430 +       if (copy_src) {
37431 +               /* no one can touch copy_src xino */
37432 +               err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
37433 +               if (unlikely(err)) {
37434 +                       pr_err("%pd copy err %d\n", dentry, err);
37435 +                       goto out_fput;
37436 +               }
37437 +       }
37438 +       goto out_dput; /* success */
37439 +
37440 +out_fput:
37441 +       fput(file);
37442 +       file = ERR_PTR(err);
37443 +out_dput:
37444 +       dput(path.dentry);
37445 +out:
37446 +       if (do_unlock)
37447 +               au_xino_unlock_dir(&ldir);
37448 +       return file;
37449 +}
37450 +
37451 +struct file *au_xino_file1(struct au_xino *xi)
37452 +{
37453 +       struct file *file;
37454 +       unsigned int u, nfile;
37455 +
37456 +       file = NULL;
37457 +       nfile = xi->xi_nfile;
37458 +       for (u = 0; u < nfile; u++) {
37459 +               file = xi->xi_file[u];
37460 +               if (file)
37461 +                       break;
37462 +       }
37463 +
37464 +       return file;
37465 +}
37466 +
37467 +static int au_xino_file_set(struct au_xino *xi, int idx, struct file *file)
37468 +{
37469 +       int err;
37470 +       struct file *f;
37471 +       void *p;
37472 +
37473 +       if (file)
37474 +               get_file(file);
37475 +
37476 +       err = 0;
37477 +       f = NULL;
37478 +       if (idx < xi->xi_nfile) {
37479 +               f = xi->xi_file[idx];
37480 +               if (f)
37481 +                       fput(f);
37482 +       } else {
37483 +               p = au_kzrealloc(xi->xi_file,
37484 +                                sizeof(*xi->xi_file) * xi->xi_nfile,
37485 +                                sizeof(*xi->xi_file) * (idx + 1),
37486 +                                GFP_NOFS, /*may_shrink*/0);
37487 +               if (p) {
37488 +                       MtxMustLock(&xi->xi_mtx);
37489 +                       xi->xi_file = p;
37490 +                       xi->xi_nfile = idx + 1;
37491 +               } else {
37492 +                       err = -ENOMEM;
37493 +                       if (file)
37494 +                               fput(file);
37495 +                       goto out;
37496 +               }
37497 +       }
37498 +       xi->xi_file[idx] = file;
37499 +
37500 +out:
37501 +       return err;
37502 +}
37503 +
37504 +/*
37505 + * if @xinew->xi is not set, then create new xigen file.
37506 + */
37507 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew)
37508 +{
37509 +       struct file *file;
37510 +       int err;
37511 +
37512 +       SiMustAnyLock(sb);
37513 +
37514 +       file = au_xino_create2(sb, xinew->base, xinew->copy_src);
37515 +       if (IS_ERR(file)) {
37516 +               err = PTR_ERR(file);
37517 +               pr_err("%s[%d], err %d\n",
37518 +                      xinew->xi ? "xino" : "xigen",
37519 +                      xinew->idx, err);
37520 +               goto out;
37521 +       }
37522 +
37523 +       if (xinew->xi)
37524 +               err = au_xino_file_set(xinew->xi, xinew->idx, file);
37525 +       else {
37526 +               BUG();
37527 +               /* todo: make xigen file an array */
37528 +               /* err = au_xigen_file_set(sb, xinew->idx, file); */
37529 +       }
37530 +       fput(file);
37531 +       if (unlikely(err))
37532 +               file = ERR_PTR(err);
37533 +
37534 +out:
37535 +       return file;
37536 +}
37537 +
37538 +/* ---------------------------------------------------------------------- */
37539 +
37540 +/*
37541 + * truncate xino files
37542 + */
37543 +static int au_xino_do_trunc(struct super_block *sb, aufs_bindex_t bindex,
37544 +                           int idx, struct kstatfs *st)
37545 +{
37546 +       int err;
37547 +       blkcnt_t blocks;
37548 +       struct file *file, *new_xino;
37549 +       struct au_xi_new xinew = {
37550 +               .idx = idx
37551 +       };
37552 +
37553 +       err = 0;
37554 +       xinew.xi = au_sbr(sb, bindex)->br_xino;
37555 +       file = au_xino_file(xinew.xi, idx);
37556 +       if (!file)
37557 +               goto out;
37558 +
37559 +       xinew.base = &file->f_path;
37560 +       err = vfs_statfs(xinew.base, st);
37561 +       if (unlikely(err)) {
37562 +               AuErr1("statfs err %d, ignored\n", err);
37563 +               err = 0;
37564 +               goto out;
37565 +       }
37566 +
37567 +       blocks = file_inode(file)->i_blocks;
37568 +       pr_info("begin truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37569 +               bindex, idx, (u64)blocks, st->f_bfree, st->f_blocks);
37570 +
37571 +       xinew.copy_src = file;
37572 +       new_xino = au_xi_new(sb, &xinew);
37573 +       if (IS_ERR(new_xino)) {
37574 +               err = PTR_ERR(new_xino);
37575 +               pr_err("xino(b%d-%d), err %d, ignored\n", bindex, idx, err);
37576 +               goto out;
37577 +       }
37578 +
37579 +       err = vfs_statfs(&new_xino->f_path, st);
37580 +       if (!err)
37581 +               pr_info("end truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37582 +                       bindex, idx, (u64)file_inode(new_xino)->i_blocks,
37583 +                       st->f_bfree, st->f_blocks);
37584 +       else {
37585 +               AuErr1("statfs err %d, ignored\n", err);
37586 +               err = 0;
37587 +       }
37588 +
37589 +out:
37590 +       return err;
37591 +}
37592 +
37593 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin)
37594 +{
37595 +       int err, i;
37596 +       unsigned long jiffy;
37597 +       aufs_bindex_t bbot;
37598 +       struct kstatfs *st;
37599 +       struct au_branch *br;
37600 +       struct au_xino *xi;
37601 +
37602 +       err = -ENOMEM;
37603 +       st = kmalloc(sizeof(*st), GFP_NOFS);
37604 +       if (unlikely(!st))
37605 +               goto out;
37606 +
37607 +       err = -EINVAL;
37608 +       bbot = au_sbbot(sb);
37609 +       if (unlikely(bindex < 0 || bbot < bindex))
37610 +               goto out_st;
37611 +
37612 +       err = 0;
37613 +       jiffy = jiffies;
37614 +       br = au_sbr(sb, bindex);
37615 +       xi = br->br_xino;
37616 +       for (i = idx_begin; !err && i < xi->xi_nfile; i++)
37617 +               err = au_xino_do_trunc(sb, bindex, i, st);
37618 +       if (!err)
37619 +               au_sbi(sb)->si_xino_jiffy = jiffy;
37620 +
37621 +out_st:
37622 +       au_kfree_rcu(st);
37623 +out:
37624 +       return err;
37625 +}
37626 +
37627 +struct xino_do_trunc_args {
37628 +       struct super_block *sb;
37629 +       struct au_branch *br;
37630 +       int idx;
37631 +};
37632 +
37633 +static void xino_do_trunc(void *_args)
37634 +{
37635 +       struct xino_do_trunc_args *args = _args;
37636 +       struct super_block *sb;
37637 +       struct au_branch *br;
37638 +       struct inode *dir;
37639 +       int err, idx;
37640 +       aufs_bindex_t bindex;
37641 +
37642 +       err = 0;
37643 +       sb = args->sb;
37644 +       dir = d_inode(sb->s_root);
37645 +       br = args->br;
37646 +       idx = args->idx;
37647 +
37648 +       si_noflush_write_lock(sb);
37649 +       ii_read_lock_parent(dir);
37650 +       bindex = au_br_index(sb, br->br_id);
37651 +       err = au_xino_trunc(sb, bindex, idx);
37652 +       ii_read_unlock(dir);
37653 +       if (unlikely(err))
37654 +               pr_warn("err b%d, (%d)\n", bindex, err);
37655 +       atomic_dec(&br->br_xino->xi_truncating);
37656 +       au_lcnt_dec(&br->br_count);
37657 +       si_write_unlock(sb);
37658 +       au_nwt_done(&au_sbi(sb)->si_nowait);
37659 +       au_kfree_rcu(args);
37660 +}
37661 +
37662 +/*
37663 + * returns the index in the xi_file array whose corresponding file is necessary
37664 + * to truncate, or -1 which means no need to truncate.
37665 + */
37666 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
37667 +{
37668 +       int err;
37669 +       unsigned int u;
37670 +       struct kstatfs st;
37671 +       struct au_sbinfo *sbinfo;
37672 +       struct au_xino *xi;
37673 +       struct file *file;
37674 +
37675 +       /* todo: si_xino_expire and the ratio should be customizable */
37676 +       sbinfo = au_sbi(sb);
37677 +       if (time_before(jiffies,
37678 +                       sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
37679 +               return -1;
37680 +
37681 +       /* truncation border */
37682 +       xi = br->br_xino;
37683 +       for (u = 0; u < xi->xi_nfile; u++) {
37684 +               file = au_xino_file(xi, u);
37685 +               if (!file)
37686 +                       continue;
37687 +
37688 +               err = vfs_statfs(&file->f_path, &st);
37689 +               if (unlikely(err)) {
37690 +                       AuErr1("statfs err %d, ignored\n", err);
37691 +                       return -1;
37692 +               }
37693 +               if (div64_u64(st.f_bfree * 100, st.f_blocks)
37694 +                   >= AUFS_XINO_DEF_TRUNC)
37695 +                       return u;
37696 +       }
37697 +
37698 +       return -1;
37699 +}
37700 +
37701 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
37702 +{
37703 +       int idx;
37704 +       struct xino_do_trunc_args *args;
37705 +       int wkq_err;
37706 +
37707 +       idx = xino_trunc_test(sb, br);
37708 +       if (idx < 0)
37709 +               return;
37710 +
37711 +       if (atomic_inc_return(&br->br_xino->xi_truncating) > 1)
37712 +               goto out;
37713 +
37714 +       /* lock and kfree() will be called in trunc_xino() */
37715 +       args = kmalloc(sizeof(*args), GFP_NOFS);
37716 +       if (unlikely(!args)) {
37717 +               AuErr1("no memory\n");
37718 +               goto out;
37719 +       }
37720 +
37721 +       au_lcnt_inc(&br->br_count);
37722 +       args->sb = sb;
37723 +       args->br = br;
37724 +       args->idx = idx;
37725 +       wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
37726 +       if (!wkq_err)
37727 +               return; /* success */
37728 +
37729 +       pr_err("wkq %d\n", wkq_err);
37730 +       au_lcnt_dec(&br->br_count);
37731 +       au_kfree_rcu(args);
37732 +
37733 +out:
37734 +       atomic_dec(&br->br_xino->xi_truncating);
37735 +}
37736 +
37737 +/* ---------------------------------------------------------------------- */
37738 +
37739 +struct au_xi_calc {
37740 +       int idx;
37741 +       loff_t pos;
37742 +};
37743 +
37744 +static void au_xi_calc(struct super_block *sb, ino_t h_ino,
37745 +                      struct au_xi_calc *calc)
37746 +{
37747 +       loff_t maxent;
37748 +
37749 +       maxent = au_xi_maxent(sb);
37750 +       calc->idx = div64_u64_rem(h_ino, maxent, &calc->pos);
37751 +       calc->pos *= sizeof(ino_t);
37752 +}
37753 +
37754 +static int au_xino_do_new_async(struct super_block *sb, struct au_branch *br,
37755 +                               struct au_xi_calc *calc)
37756 +{
37757 +       int err;
37758 +       struct file *file;
37759 +       struct au_xino *xi = br->br_xino;
37760 +       struct au_xi_new xinew = {
37761 +               .xi = xi
37762 +       };
37763 +
37764 +       SiMustAnyLock(sb);
37765 +
37766 +       err = 0;
37767 +       if (!xi)
37768 +               goto out;
37769 +
37770 +       mutex_lock(&xi->xi_mtx);
37771 +       file = au_xino_file(xi, calc->idx);
37772 +       if (file)
37773 +               goto out_mtx;
37774 +
37775 +       file = au_xino_file(xi, /*idx*/-1);
37776 +       AuDebugOn(!file);
37777 +       xinew.idx = calc->idx;
37778 +       xinew.base = &file->f_path;
37779 +       /* xinew.copy_src = NULL; */
37780 +       file = au_xi_new(sb, &xinew);
37781 +       if (IS_ERR(file))
37782 +               err = PTR_ERR(file);
37783 +
37784 +out_mtx:
37785 +       mutex_unlock(&xi->xi_mtx);
37786 +out:
37787 +       return err;
37788 +}
37789 +
37790 +struct au_xino_do_new_async_args {
37791 +       struct super_block *sb;
37792 +       struct au_branch *br;
37793 +       struct au_xi_calc calc;
37794 +       ino_t ino;
37795 +};
37796 +
37797 +struct au_xi_writing {
37798 +       struct hlist_bl_node node;
37799 +       ino_t h_ino, ino;
37800 +};
37801 +
37802 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37803 +                           ino_t ino);
37804 +
37805 +static void au_xino_call_do_new_async(void *args)
37806 +{
37807 +       struct au_xino_do_new_async_args *a = args;
37808 +       struct au_branch *br;
37809 +       struct super_block *sb;
37810 +       struct au_sbinfo *sbi;
37811 +       struct inode *root;
37812 +       struct file *file;
37813 +       struct au_xi_writing *del, *p;
37814 +       struct hlist_bl_head *hbl;
37815 +       struct hlist_bl_node *pos;
37816 +       int err;
37817 +
37818 +       br = a->br;
37819 +       sb = a->sb;
37820 +       sbi = au_sbi(sb);
37821 +       si_noflush_read_lock(sb);
37822 +       root = d_inode(sb->s_root);
37823 +       ii_read_lock_child(root);
37824 +       err = au_xino_do_new_async(sb, br, &a->calc);
37825 +       if (unlikely(err)) {
37826 +               AuIOErr("err %d\n", err);
37827 +               goto out;
37828 +       }
37829 +
37830 +       file = au_xino_file(br->br_xino, a->calc.idx);
37831 +       AuDebugOn(!file);
37832 +       err = au_xino_do_write(file, &a->calc, a->ino);
37833 +       if (unlikely(err)) {
37834 +               AuIOErr("err %d\n", err);
37835 +               goto out;
37836 +       }
37837 +
37838 +       del = NULL;
37839 +       hbl = &br->br_xino->xi_writing;
37840 +       hlist_bl_lock(hbl);
37841 +       au_hbl_for_each(pos, hbl) {
37842 +               p = container_of(pos, struct au_xi_writing, node);
37843 +               if (p->ino == a->ino) {
37844 +                       del = p;
37845 +                       hlist_bl_del(&p->node);
37846 +                       break;
37847 +               }
37848 +       }
37849 +       hlist_bl_unlock(hbl);
37850 +       au_kfree_rcu(del);
37851 +
37852 +out:
37853 +       au_lcnt_dec(&br->br_count);
37854 +       ii_read_unlock(root);
37855 +       si_read_unlock(sb);
37856 +       au_nwt_done(&sbi->si_nowait);
37857 +       au_kfree_rcu(a);
37858 +}
37859 +
37860 +/*
37861 + * create a new xino file asynchronously
37862 + */
37863 +static int au_xino_new_async(struct super_block *sb, struct au_branch *br,
37864 +                            struct au_xi_calc *calc, ino_t ino)
37865 +{
37866 +       int err;
37867 +       struct au_xino_do_new_async_args *arg;
37868 +
37869 +       err = -ENOMEM;
37870 +       arg = kmalloc(sizeof(*arg), GFP_NOFS);
37871 +       if (unlikely(!arg))
37872 +               goto out;
37873 +
37874 +       arg->sb = sb;
37875 +       arg->br = br;
37876 +       arg->calc = *calc;
37877 +       arg->ino = ino;
37878 +       au_lcnt_inc(&br->br_count);
37879 +       err = au_wkq_nowait(au_xino_call_do_new_async, arg, sb, AuWkq_NEST);
37880 +       if (unlikely(err)) {
37881 +               pr_err("wkq %d\n", err);
37882 +               au_lcnt_dec(&br->br_count);
37883 +               au_kfree_rcu(arg);
37884 +       }
37885 +
37886 +out:
37887 +       return err;
37888 +}
37889 +
37890 +/*
37891 + * read @ino from xinofile for the specified branch{@sb, @bindex}
37892 + * at the position of @h_ino.
37893 + */
37894 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37895 +                ino_t *ino)
37896 +{
37897 +       int err;
37898 +       ssize_t sz;
37899 +       struct au_xi_calc calc;
37900 +       struct au_sbinfo *sbinfo;
37901 +       struct file *file;
37902 +       struct au_xino *xi;
37903 +       struct hlist_bl_head *hbl;
37904 +       struct hlist_bl_node *pos;
37905 +       struct au_xi_writing *p;
37906 +
37907 +       *ino = 0;
37908 +       if (!au_opt_test(au_mntflags(sb), XINO))
37909 +               return 0; /* no xino */
37910 +
37911 +       err = 0;
37912 +       au_xi_calc(sb, h_ino, &calc);
37913 +       xi = au_sbr(sb, bindex)->br_xino;
37914 +       file = au_xino_file(xi, calc.idx);
37915 +       if (!file) {
37916 +               hbl = &xi->xi_writing;
37917 +               hlist_bl_lock(hbl);
37918 +               au_hbl_for_each(pos, hbl) {
37919 +                       p = container_of(pos, struct au_xi_writing, node);
37920 +                       if (p->h_ino == h_ino) {
37921 +                               AuDbg("hi%llu, i%llu, found\n",
37922 +                                     (u64)p->h_ino, (u64)p->ino);
37923 +                               *ino = p->ino;
37924 +                               break;
37925 +                       }
37926 +               }
37927 +               hlist_bl_unlock(hbl);
37928 +               return 0;
37929 +       } else if (vfsub_f_size_read(file) < calc.pos + sizeof(*ino))
37930 +               return 0; /* no xino */
37931 +
37932 +       sbinfo = au_sbi(sb);
37933 +       sz = xino_fread(file, ino, sizeof(*ino), &calc.pos);
37934 +       if (sz == sizeof(*ino))
37935 +               return 0; /* success */
37936 +
37937 +       err = sz;
37938 +       if (unlikely(sz >= 0)) {
37939 +               err = -EIO;
37940 +               AuIOErr("xino read error (%zd)\n", sz);
37941 +       }
37942 +       return err;
37943 +}
37944 +
37945 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37946 +                           ino_t ino)
37947 +{
37948 +       ssize_t sz;
37949 +
37950 +       sz = xino_fwrite(file, &ino, sizeof(ino), &calc->pos);
37951 +       if (sz == sizeof(ino))
37952 +               return 0; /* success */
37953 +
37954 +       AuIOErr("write failed (%zd)\n", sz);
37955 +       return -EIO;
37956 +}
37957 +
37958 +/*
37959 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
37960 + * at the position of @h_ino.
37961 + * even if @ino is zero, it is written to the xinofile and means no entry.
37962 + * if the size of the xino file on a specific filesystem exceeds the watermark,
37963 + * try truncating it.
37964 + */
37965 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37966 +                 ino_t ino)
37967 +{
37968 +       int err;
37969 +       unsigned int mnt_flags;
37970 +       struct au_xi_calc calc;
37971 +       struct file *file;
37972 +       struct au_branch *br;
37973 +       struct au_xino *xi;
37974 +       struct au_xi_writing *p;
37975 +
37976 +       SiMustAnyLock(sb);
37977 +
37978 +       mnt_flags = au_mntflags(sb);
37979 +       if (!au_opt_test(mnt_flags, XINO))
37980 +               return 0;
37981 +
37982 +       au_xi_calc(sb, h_ino, &calc);
37983 +       br = au_sbr(sb, bindex);
37984 +       xi = br->br_xino;
37985 +       file = au_xino_file(xi, calc.idx);
37986 +       if (!file) {
37987 +               /* store the inum pair into the list */
37988 +               p = kmalloc(sizeof(*p), GFP_NOFS | __GFP_NOFAIL);
37989 +               p->h_ino = h_ino;
37990 +               p->ino = ino;
37991 +               au_hbl_add(&p->node, &xi->xi_writing);
37992 +
37993 +               /* create and write a new xino file asynchronously */
37994 +               err = au_xino_new_async(sb, br, &calc, ino);
37995 +               if (!err)
37996 +                       return 0; /* success */
37997 +               goto out;
37998 +       }
37999 +
38000 +       err = au_xino_do_write(file, &calc, ino);
38001 +       if (!err) {
38002 +               br = au_sbr(sb, bindex);
38003 +               if (au_opt_test(mnt_flags, TRUNC_XINO)
38004 +                   && au_test_fs_trunc_xino(au_br_sb(br)))
38005 +                       xino_try_trunc(sb, br);
38006 +               return 0; /* success */
38007 +       }
38008 +
38009 +out:
38010 +       AuIOErr("write failed (%d)\n", err);
38011 +       return -EIO;
38012 +}
38013 +
38014 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
38015 +                             loff_t *pos);
38016 +
38017 +/* todo: unnecessary to support mmap_sem since kernel-space? */
38018 +ssize_t xino_fread(struct file *file, void *kbuf, size_t size, loff_t *pos)
38019 +{
38020 +       ssize_t err;
38021 +       int i;
38022 +       const int prevent_endless = 10;
38023 +
38024 +       i = 0;
38025 +       do {
38026 +               err = vfsub_read_k(file, kbuf, size, pos);
38027 +               if (err == -EINTR
38028 +                   && !au_wkq_test()
38029 +                   && fatal_signal_pending(current)) {
38030 +                       err = xino_fread_wkq(file, kbuf, size, pos);
38031 +                       BUG_ON(err == -EINTR);
38032 +               }
38033 +       } while (i++ < prevent_endless
38034 +                && (err == -EAGAIN || err == -EINTR));
38035 +
38036 +#if 0 /* reserved for future use */
38037 +       if (err > 0)
38038 +               fsnotify_access(file->f_path.dentry);
38039 +#endif
38040 +
38041 +       return err;
38042 +}
38043 +
38044 +struct xino_fread_args {
38045 +       ssize_t *errp;
38046 +       struct file *file;
38047 +       void *buf;
38048 +       size_t size;
38049 +       loff_t *pos;
38050 +};
38051 +
38052 +static void call_xino_fread(void *args)
38053 +{
38054 +       struct xino_fread_args *a = args;
38055 +       *a->errp = xino_fread(a->file, a->buf, a->size, a->pos);
38056 +}
38057 +
38058 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
38059 +                             loff_t *pos)
38060 +{
38061 +       ssize_t err;
38062 +       int wkq_err;
38063 +       struct xino_fread_args args = {
38064 +               .errp   = &err,
38065 +               .file   = file,
38066 +               .buf    = buf,
38067 +               .size   = size,
38068 +               .pos    = pos
38069 +       };
38070 +
38071 +       wkq_err = au_wkq_wait(call_xino_fread, &args);
38072 +       if (unlikely(wkq_err))
38073 +               err = wkq_err;
38074 +
38075 +       return err;
38076 +}
38077 +
38078 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38079 +                              loff_t *pos);
38080 +
38081 +static ssize_t do_xino_fwrite(struct file *file, void *kbuf, size_t size,
38082 +                             loff_t *pos)
38083 +{
38084 +       ssize_t err;
38085 +       int i;
38086 +       const int prevent_endless = 10;
38087 +
38088 +       i = 0;
38089 +       do {
38090 +               err = vfsub_write_k(file, kbuf, size, pos);
38091 +               if (err == -EINTR
38092 +                   && !au_wkq_test()
38093 +                   && fatal_signal_pending(current)) {
38094 +                       err = xino_fwrite_wkq(file, kbuf, size, pos);
38095 +                       BUG_ON(err == -EINTR);
38096 +               }
38097 +       } while (i++ < prevent_endless
38098 +                && (err == -EAGAIN || err == -EINTR));
38099 +
38100 +#if 0 /* reserved for future use */
38101 +       if (err > 0)
38102 +               fsnotify_modify(file->f_path.dentry);
38103 +#endif
38104 +
38105 +       return err;
38106 +}
38107 +
38108 +struct do_xino_fwrite_args {
38109 +       ssize_t *errp;
38110 +       struct file *file;
38111 +       void *buf;
38112 +       size_t size;
38113 +       loff_t *pos;
38114 +};
38115 +
38116 +static void call_do_xino_fwrite(void *args)
38117 +{
38118 +       struct do_xino_fwrite_args *a = args;
38119 +       *a->errp = do_xino_fwrite(a->file, a->buf, a->size, a->pos);
38120 +}
38121 +
38122 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38123 +                              loff_t *pos)
38124 +{
38125 +       ssize_t err;
38126 +       int wkq_err;
38127 +       struct do_xino_fwrite_args args = {
38128 +               .errp   = &err,
38129 +               .file   = file,
38130 +               .buf    = buf,
38131 +               .size   = size,
38132 +               .pos    = pos
38133 +       };
38134 +
38135 +       /*
38136 +        * it breaks RLIMIT_FSIZE and normal user's limit,
38137 +        * users should care about quota and real 'filesystem full.'
38138 +        */
38139 +       wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
38140 +       if (unlikely(wkq_err))
38141 +               err = wkq_err;
38142 +
38143 +       return err;
38144 +}
38145 +
38146 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos)
38147 +{
38148 +       ssize_t err;
38149 +
38150 +       if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
38151 +               lockdep_off();
38152 +               err = do_xino_fwrite(file, buf, size, pos);
38153 +               lockdep_on();
38154 +       } else {
38155 +               lockdep_off();
38156 +               err = xino_fwrite_wkq(file, buf, size, pos);
38157 +               lockdep_on();
38158 +       }
38159 +
38160 +       return err;
38161 +}
38162 +
38163 +/* ---------------------------------------------------------------------- */
38164 +
38165 +/*
38166 + * inode number bitmap
38167 + */
38168 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
38169 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
38170 +{
38171 +       ino_t ino;
38172 +
38173 +       AuDebugOn(bit < 0 || page_bits <= bit);
38174 +       ino = AUFS_FIRST_INO + pindex * page_bits + bit;
38175 +       return ino;
38176 +}
38177 +
38178 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
38179 +{
38180 +       AuDebugOn(ino < AUFS_FIRST_INO);
38181 +       ino -= AUFS_FIRST_INO;
38182 +       *pindex = ino / page_bits;
38183 +       *bit = ino % page_bits;
38184 +}
38185 +
38186 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
38187 +{
38188 +       int err;
38189 +       loff_t pos;
38190 +       ssize_t sz;
38191 +       struct au_sbinfo *sbinfo;
38192 +       struct file *xib;
38193 +       unsigned long *p;
38194 +
38195 +       sbinfo = au_sbi(sb);
38196 +       MtxMustLock(&sbinfo->si_xib_mtx);
38197 +       AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
38198 +                 || !au_opt_test(sbinfo->si_mntflags, XINO));
38199 +
38200 +       if (pindex == sbinfo->si_xib_last_pindex)
38201 +               return 0;
38202 +
38203 +       xib = sbinfo->si_xib;
38204 +       p = sbinfo->si_xib_buf;
38205 +       pos = sbinfo->si_xib_last_pindex;
38206 +       pos *= PAGE_SIZE;
38207 +       sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38208 +       if (unlikely(sz != PAGE_SIZE))
38209 +               goto out;
38210 +
38211 +       pos = pindex;
38212 +       pos *= PAGE_SIZE;
38213 +       if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
38214 +               sz = xino_fread(xib, p, PAGE_SIZE, &pos);
38215 +       else {
38216 +               memset(p, 0, PAGE_SIZE);
38217 +               sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38218 +       }
38219 +       if (sz == PAGE_SIZE) {
38220 +               sbinfo->si_xib_last_pindex = pindex;
38221 +               return 0; /* success */
38222 +       }
38223 +
38224 +out:
38225 +       AuIOErr1("write failed (%zd)\n", sz);
38226 +       err = sz;
38227 +       if (sz >= 0)
38228 +               err = -EIO;
38229 +       return err;
38230 +}
38231 +
38232 +static void au_xib_clear_bit(struct inode *inode)
38233 +{
38234 +       int err, bit;
38235 +       unsigned long pindex;
38236 +       struct super_block *sb;
38237 +       struct au_sbinfo *sbinfo;
38238 +
38239 +       AuDebugOn(inode->i_nlink);
38240 +
38241 +       sb = inode->i_sb;
38242 +       xib_calc_bit(inode->i_ino, &pindex, &bit);
38243 +       AuDebugOn(page_bits <= bit);
38244 +       sbinfo = au_sbi(sb);
38245 +       mutex_lock(&sbinfo->si_xib_mtx);
38246 +       err = xib_pindex(sb, pindex);
38247 +       if (!err) {
38248 +               clear_bit(bit, sbinfo->si_xib_buf);
38249 +               sbinfo->si_xib_next_bit = bit;
38250 +       }
38251 +       mutex_unlock(&sbinfo->si_xib_mtx);
38252 +}
38253 +
38254 +/* ---------------------------------------------------------------------- */
38255 +
38256 +/*
38257 + * truncate a xino bitmap file
38258 + */
38259 +
38260 +/* todo: slow */
38261 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
38262 +{
38263 +       int err, bit;
38264 +       ssize_t sz;
38265 +       unsigned long pindex;
38266 +       loff_t pos, pend;
38267 +       struct au_sbinfo *sbinfo;
38268 +       ino_t *ino;
38269 +       unsigned long *p;
38270 +
38271 +       err = 0;
38272 +       sbinfo = au_sbi(sb);
38273 +       MtxMustLock(&sbinfo->si_xib_mtx);
38274 +       p = sbinfo->si_xib_buf;
38275 +       pend = vfsub_f_size_read(file);
38276 +       pos = 0;
38277 +       while (pos < pend) {
38278 +               sz = xino_fread(file, page, PAGE_SIZE, &pos);
38279 +               err = sz;
38280 +               if (unlikely(sz <= 0))
38281 +                       goto out;
38282 +
38283 +               err = 0;
38284 +               for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
38285 +                       if (unlikely(*ino < AUFS_FIRST_INO))
38286 +                               continue;
38287 +
38288 +                       xib_calc_bit(*ino, &pindex, &bit);
38289 +                       AuDebugOn(page_bits <= bit);
38290 +                       err = xib_pindex(sb, pindex);
38291 +                       if (!err)
38292 +                               set_bit(bit, p);
38293 +                       else
38294 +                               goto out;
38295 +               }
38296 +       }
38297 +
38298 +out:
38299 +       return err;
38300 +}
38301 +
38302 +static int xib_restore(struct super_block *sb)
38303 +{
38304 +       int err, i;
38305 +       unsigned int nfile;
38306 +       aufs_bindex_t bindex, bbot;
38307 +       void *page;
38308 +       struct au_branch *br;
38309 +       struct au_xino *xi;
38310 +       struct file *file;
38311 +
38312 +       err = -ENOMEM;
38313 +       page = (void *)__get_free_page(GFP_NOFS);
38314 +       if (unlikely(!page))
38315 +               goto out;
38316 +
38317 +       err = 0;
38318 +       bbot = au_sbbot(sb);
38319 +       for (bindex = 0; !err && bindex <= bbot; bindex++)
38320 +               if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0) {
38321 +                       br = au_sbr(sb, bindex);
38322 +                       xi = br->br_xino;
38323 +                       nfile = xi->xi_nfile;
38324 +                       for (i = 0; i < nfile; i++) {
38325 +                               file = au_xino_file(xi, i);
38326 +                               if (file)
38327 +                                       err = do_xib_restore(sb, file, page);
38328 +                       }
38329 +               } else
38330 +                       AuDbg("skip shared b%d\n", bindex);
38331 +       free_page((unsigned long)page);
38332 +
38333 +out:
38334 +       return err;
38335 +}
38336 +
38337 +int au_xib_trunc(struct super_block *sb)
38338 +{
38339 +       int err;
38340 +       ssize_t sz;
38341 +       loff_t pos;
38342 +       struct au_sbinfo *sbinfo;
38343 +       unsigned long *p;
38344 +       struct file *file;
38345 +
38346 +       SiMustWriteLock(sb);
38347 +
38348 +       err = 0;
38349 +       sbinfo = au_sbi(sb);
38350 +       if (!au_opt_test(sbinfo->si_mntflags, XINO))
38351 +               goto out;
38352 +
38353 +       file = sbinfo->si_xib;
38354 +       if (vfsub_f_size_read(file) <= PAGE_SIZE)
38355 +               goto out;
38356 +
38357 +       file = au_xino_create2(sb, &sbinfo->si_xib->f_path, NULL);
38358 +       err = PTR_ERR(file);
38359 +       if (IS_ERR(file))
38360 +               goto out;
38361 +       fput(sbinfo->si_xib);
38362 +       sbinfo->si_xib = file;
38363 +
38364 +       p = sbinfo->si_xib_buf;
38365 +       memset(p, 0, PAGE_SIZE);
38366 +       pos = 0;
38367 +       sz = xino_fwrite(sbinfo->si_xib, p, PAGE_SIZE, &pos);
38368 +       if (unlikely(sz != PAGE_SIZE)) {
38369 +               err = sz;
38370 +               AuIOErr("err %d\n", err);
38371 +               if (sz >= 0)
38372 +                       err = -EIO;
38373 +               goto out;
38374 +       }
38375 +
38376 +       mutex_lock(&sbinfo->si_xib_mtx);
38377 +       /* mnt_want_write() is unnecessary here */
38378 +       err = xib_restore(sb);
38379 +       mutex_unlock(&sbinfo->si_xib_mtx);
38380 +
38381 +out:
38382 +       return err;
38383 +}
38384 +
38385 +/* ---------------------------------------------------------------------- */
38386 +
38387 +struct au_xino *au_xino_alloc(unsigned int nfile)
38388 +{
38389 +       struct au_xino *xi;
38390 +
38391 +       xi = kzalloc(sizeof(*xi), GFP_NOFS);
38392 +       if (unlikely(!xi))
38393 +               goto out;
38394 +       xi->xi_nfile = nfile;
38395 +       xi->xi_file = kcalloc(nfile, sizeof(*xi->xi_file), GFP_NOFS);
38396 +       if (unlikely(!xi->xi_file))
38397 +               goto out_free;
38398 +
38399 +       xi->xi_nondir.total = 8; /* initial size */
38400 +       xi->xi_nondir.array = kcalloc(xi->xi_nondir.total, sizeof(ino_t),
38401 +                                     GFP_NOFS);
38402 +       if (unlikely(!xi->xi_nondir.array))
38403 +               goto out_file;
38404 +
38405 +       spin_lock_init(&xi->xi_nondir.spin);
38406 +       init_waitqueue_head(&xi->xi_nondir.wqh);
38407 +       mutex_init(&xi->xi_mtx);
38408 +       INIT_HLIST_BL_HEAD(&xi->xi_writing);
38409 +       atomic_set(&xi->xi_truncating, 0);
38410 +       kref_init(&xi->xi_kref);
38411 +       goto out; /* success */
38412 +
38413 +out_file:
38414 +       au_kfree_try_rcu(xi->xi_file);
38415 +out_free:
38416 +       au_kfree_rcu(xi);
38417 +       xi = NULL;
38418 +out:
38419 +       return xi;
38420 +}
38421 +
38422 +static int au_xino_init(struct au_branch *br, int idx, struct file *file)
38423 +{
38424 +       int err;
38425 +       struct au_xino *xi;
38426 +
38427 +       err = 0;
38428 +       xi = au_xino_alloc(idx + 1);
38429 +       if (unlikely(!xi)) {
38430 +               err = -ENOMEM;
38431 +               goto out;
38432 +       }
38433 +
38434 +       if (file)
38435 +               get_file(file);
38436 +       xi->xi_file[idx] = file;
38437 +       AuDebugOn(br->br_xino);
38438 +       br->br_xino = xi;
38439 +
38440 +out:
38441 +       return err;
38442 +}
38443 +
38444 +static void au_xino_release(struct kref *kref)
38445 +{
38446 +       struct au_xino *xi;
38447 +       int i;
38448 +       unsigned long ul;
38449 +       struct hlist_bl_head *hbl;
38450 +       struct hlist_bl_node *pos, *n;
38451 +       struct au_xi_writing *p;
38452 +
38453 +       xi = container_of(kref, struct au_xino, xi_kref);
38454 +       for (i = 0; i < xi->xi_nfile; i++)
38455 +               if (xi->xi_file[i])
38456 +                       fput(xi->xi_file[i]);
38457 +       for (i = xi->xi_nondir.total - 1; i >= 0; i--)
38458 +               AuDebugOn(xi->xi_nondir.array[i]);
38459 +       mutex_destroy(&xi->xi_mtx);
38460 +       hbl = &xi->xi_writing;
38461 +       ul = au_hbl_count(hbl);
38462 +       if (unlikely(ul)) {
38463 +               pr_warn("xi_writing %lu\n", ul);
38464 +               hlist_bl_lock(hbl);
38465 +               hlist_bl_for_each_entry_safe(p, pos, n, hbl, node) {
38466 +                       hlist_bl_del(&p->node);
38467 +                       /* kmemleak reported au_kfree_rcu() doesn't free it */
38468 +                       kfree(p);
38469 +               }
38470 +               hlist_bl_unlock(hbl);
38471 +       }
38472 +       au_kfree_try_rcu(xi->xi_file);
38473 +       au_kfree_try_rcu(xi->xi_nondir.array);
38474 +       au_kfree_rcu(xi);
38475 +}
38476 +
38477 +int au_xino_put(struct au_branch *br)
38478 +{
38479 +       int ret;
38480 +       struct au_xino *xi;
38481 +
38482 +       ret = 0;
38483 +       xi = br->br_xino;
38484 +       if (xi) {
38485 +               br->br_xino = NULL;
38486 +               ret = kref_put(&xi->xi_kref, au_xino_release);
38487 +       }
38488 +
38489 +       return ret;
38490 +}
38491 +
38492 +/* ---------------------------------------------------------------------- */
38493 +
38494 +/*
38495 + * xino mount option handlers
38496 + */
38497 +
38498 +/* xino bitmap */
38499 +static void xino_clear_xib(struct super_block *sb)
38500 +{
38501 +       struct au_sbinfo *sbinfo;
38502 +
38503 +       SiMustWriteLock(sb);
38504 +
38505 +       sbinfo = au_sbi(sb);
38506 +       if (sbinfo->si_xib)
38507 +               fput(sbinfo->si_xib);
38508 +       sbinfo->si_xib = NULL;
38509 +       if (sbinfo->si_xib_buf)
38510 +               free_page((unsigned long)sbinfo->si_xib_buf);
38511 +       sbinfo->si_xib_buf = NULL;
38512 +}
38513 +
38514 +static int au_xino_set_xib(struct super_block *sb, struct path *path)
38515 +{
38516 +       int err;
38517 +       loff_t pos;
38518 +       struct au_sbinfo *sbinfo;
38519 +       struct file *file;
38520 +       struct super_block *xi_sb;
38521 +
38522 +       SiMustWriteLock(sb);
38523 +
38524 +       sbinfo = au_sbi(sb);
38525 +       file = au_xino_create2(sb, path, sbinfo->si_xib);
38526 +       err = PTR_ERR(file);
38527 +       if (IS_ERR(file))
38528 +               goto out;
38529 +       if (sbinfo->si_xib)
38530 +               fput(sbinfo->si_xib);
38531 +       sbinfo->si_xib = file;
38532 +       xi_sb = file_inode(file)->i_sb;
38533 +       sbinfo->si_ximaxent = xi_sb->s_maxbytes;
38534 +       if (unlikely(sbinfo->si_ximaxent < PAGE_SIZE)) {
38535 +               err = -EIO;
38536 +               pr_err("s_maxbytes(%llu) on %s is too small\n",
38537 +                      (u64)sbinfo->si_ximaxent, au_sbtype(xi_sb));
38538 +               goto out_unset;
38539 +       }
38540 +       sbinfo->si_ximaxent /= sizeof(ino_t);
38541 +
38542 +       err = -ENOMEM;
38543 +       if (!sbinfo->si_xib_buf)
38544 +               sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
38545 +       if (unlikely(!sbinfo->si_xib_buf))
38546 +               goto out_unset;
38547 +
38548 +       sbinfo->si_xib_last_pindex = 0;
38549 +       sbinfo->si_xib_next_bit = 0;
38550 +       if (vfsub_f_size_read(file) < PAGE_SIZE) {
38551 +               pos = 0;
38552 +               err = xino_fwrite(file, sbinfo->si_xib_buf, PAGE_SIZE, &pos);
38553 +               if (unlikely(err != PAGE_SIZE))
38554 +                       goto out_free;
38555 +       }
38556 +       err = 0;
38557 +       goto out; /* success */
38558 +
38559 +out_free:
38560 +       if (sbinfo->si_xib_buf)
38561 +               free_page((unsigned long)sbinfo->si_xib_buf);
38562 +       sbinfo->si_xib_buf = NULL;
38563 +       if (err >= 0)
38564 +               err = -EIO;
38565 +out_unset:
38566 +       fput(sbinfo->si_xib);
38567 +       sbinfo->si_xib = NULL;
38568 +out:
38569 +       AuTraceErr(err);
38570 +       return err;
38571 +}
38572 +
38573 +/* xino for each branch */
38574 +static void xino_clear_br(struct super_block *sb)
38575 +{
38576 +       aufs_bindex_t bindex, bbot;
38577 +       struct au_branch *br;
38578 +
38579 +       bbot = au_sbbot(sb);
38580 +       for (bindex = 0; bindex <= bbot; bindex++) {
38581 +               br = au_sbr(sb, bindex);
38582 +               AuDebugOn(!br);
38583 +               au_xino_put(br);
38584 +       }
38585 +}
38586 +
38587 +static void au_xino_set_br_shared(struct super_block *sb, struct au_branch *br,
38588 +                                 aufs_bindex_t bshared)
38589 +{
38590 +       struct au_branch *brshared;
38591 +
38592 +       brshared = au_sbr(sb, bshared);
38593 +       AuDebugOn(!brshared->br_xino);
38594 +       AuDebugOn(!brshared->br_xino->xi_file);
38595 +       if (br->br_xino != brshared->br_xino) {
38596 +               au_xino_get(brshared);
38597 +               au_xino_put(br);
38598 +               br->br_xino = brshared->br_xino;
38599 +       }
38600 +}
38601 +
38602 +struct au_xino_do_set_br {
38603 +       struct au_branch *br;
38604 +       ino_t h_ino;
38605 +       aufs_bindex_t bshared;
38606 +};
38607 +
38608 +static int au_xino_do_set_br(struct super_block *sb, struct path *path,
38609 +                            struct au_xino_do_set_br *args)
38610 +{
38611 +       int err;
38612 +       struct au_xi_calc calc;
38613 +       struct file *file;
38614 +       struct au_branch *br;
38615 +       struct au_xi_new xinew = {
38616 +               .base = path
38617 +       };
38618 +
38619 +       br = args->br;
38620 +       xinew.xi = br->br_xino;
38621 +       au_xi_calc(sb, args->h_ino, &calc);
38622 +       xinew.copy_src = au_xino_file(xinew.xi, calc.idx);
38623 +       if (args->bshared >= 0)
38624 +               /* shared xino */
38625 +               au_xino_set_br_shared(sb, br, args->bshared);
38626 +       else if (!xinew.xi) {
38627 +               /* new xino */
38628 +               err = au_xino_init(br, calc.idx, xinew.copy_src);
38629 +               if (unlikely(err))
38630 +                       goto out;
38631 +       }
38632 +
38633 +       /* force re-creating */
38634 +       xinew.xi = br->br_xino;
38635 +       xinew.idx = calc.idx;
38636 +       mutex_lock(&xinew.xi->xi_mtx);
38637 +       file = au_xi_new(sb, &xinew);
38638 +       mutex_unlock(&xinew.xi->xi_mtx);
38639 +       err = PTR_ERR(file);
38640 +       if (IS_ERR(file))
38641 +               goto out;
38642 +       AuDebugOn(!file);
38643 +
38644 +       err = au_xino_do_write(file, &calc, AUFS_ROOT_INO);
38645 +       if (unlikely(err))
38646 +               au_xino_put(br);
38647 +
38648 +out:
38649 +       AuTraceErr(err);
38650 +       return err;
38651 +}
38652 +
38653 +static int au_xino_set_br(struct super_block *sb, struct path *path)
38654 +{
38655 +       int err;
38656 +       aufs_bindex_t bindex, bbot;
38657 +       struct au_xino_do_set_br args;
38658 +       struct inode *inode;
38659 +
38660 +       SiMustWriteLock(sb);
38661 +
38662 +       bbot = au_sbbot(sb);
38663 +       inode = d_inode(sb->s_root);
38664 +       for (bindex = 0; bindex <= bbot; bindex++) {
38665 +               args.h_ino = au_h_iptr(inode, bindex)->i_ino;
38666 +               args.br = au_sbr(sb, bindex);
38667 +               args.bshared = is_sb_shared(sb, bindex, bindex - 1);
38668 +               err = au_xino_do_set_br(sb, path, &args);
38669 +               if (unlikely(err))
38670 +                       break;
38671 +       }
38672 +
38673 +       AuTraceErr(err);
38674 +       return err;
38675 +}
38676 +
38677 +void au_xino_clr(struct super_block *sb)
38678 +{
38679 +       struct au_sbinfo *sbinfo;
38680 +
38681 +       au_xigen_clr(sb);
38682 +       xino_clear_xib(sb);
38683 +       xino_clear_br(sb);
38684 +       dbgaufs_brs_del(sb, 0);
38685 +       sbinfo = au_sbi(sb);
38686 +       /* lvalue, do not call au_mntflags() */
38687 +       au_opt_clr(sbinfo->si_mntflags, XINO);
38688 +}
38689 +
38690 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount)
38691 +{
38692 +       int err, skip;
38693 +       struct dentry *dentry, *parent, *cur_dentry, *cur_parent;
38694 +       struct qstr *dname, *cur_name;
38695 +       struct file *cur_xino;
38696 +       struct au_sbinfo *sbinfo;
38697 +       struct path *path, *cur_path;
38698 +
38699 +       SiMustWriteLock(sb);
38700 +
38701 +       err = 0;
38702 +       sbinfo = au_sbi(sb);
38703 +       path = &xiopt->file->f_path;
38704 +       dentry = path->dentry;
38705 +       parent = dget_parent(dentry);
38706 +       if (remount) {
38707 +               skip = 0;
38708 +               cur_xino = sbinfo->si_xib;
38709 +               if (cur_xino) {
38710 +                       cur_path = &cur_xino->f_path;
38711 +                       cur_dentry = cur_path->dentry;
38712 +                       cur_parent = dget_parent(cur_dentry);
38713 +                       cur_name = &cur_dentry->d_name;
38714 +                       dname = &dentry->d_name;
38715 +                       skip = (cur_parent == parent
38716 +                               && au_qstreq(dname, cur_name));
38717 +                       dput(cur_parent);
38718 +               }
38719 +               if (skip)
38720 +                       goto out;
38721 +       }
38722 +
38723 +       au_opt_set(sbinfo->si_mntflags, XINO);
38724 +       err = au_xino_set_xib(sb, path);
38725 +       /* si_x{read,write} are set */
38726 +       if (!err)
38727 +               err = au_xigen_set(sb, path);
38728 +       if (!err)
38729 +               err = au_xino_set_br(sb, path);
38730 +       if (!err) {
38731 +               dbgaufs_brs_add(sb, 0, /*topdown*/1);
38732 +               goto out; /* success */
38733 +       }
38734 +
38735 +       /* reset all */
38736 +       AuIOErr("failed setting xino(%d).\n", err);
38737 +       au_xino_clr(sb);
38738 +
38739 +out:
38740 +       dput(parent);
38741 +       return err;
38742 +}
38743 +
38744 +/*
38745 + * create a xinofile at the default place/path.
38746 + */
38747 +struct file *au_xino_def(struct super_block *sb)
38748 +{
38749 +       struct file *file;
38750 +       char *page, *p;
38751 +       struct au_branch *br;
38752 +       struct super_block *h_sb;
38753 +       struct path path;
38754 +       aufs_bindex_t bbot, bindex, bwr;
38755 +
38756 +       br = NULL;
38757 +       bbot = au_sbbot(sb);
38758 +       bwr = -1;
38759 +       for (bindex = 0; bindex <= bbot; bindex++) {
38760 +               br = au_sbr(sb, bindex);
38761 +               if (au_br_writable(br->br_perm)
38762 +                   && !au_test_fs_bad_xino(au_br_sb(br))) {
38763 +                       bwr = bindex;
38764 +                       break;
38765 +               }
38766 +       }
38767 +
38768 +       if (bwr >= 0) {
38769 +               file = ERR_PTR(-ENOMEM);
38770 +               page = (void *)__get_free_page(GFP_NOFS);
38771 +               if (unlikely(!page))
38772 +                       goto out;
38773 +               path.mnt = au_br_mnt(br);
38774 +               path.dentry = au_h_dptr(sb->s_root, bwr);
38775 +               p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
38776 +               file = (void *)p;
38777 +               if (!IS_ERR(p)) {
38778 +                       strcat(p, "/" AUFS_XINO_FNAME);
38779 +                       AuDbg("%s\n", p);
38780 +                       file = au_xino_create(sb, p, /*silent*/0, /*wbrtop*/1);
38781 +               }
38782 +               free_page((unsigned long)page);
38783 +       } else {
38784 +               file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0,
38785 +                                     /*wbrtop*/0);
38786 +               if (IS_ERR(file))
38787 +                       goto out;
38788 +               h_sb = file->f_path.dentry->d_sb;
38789 +               if (unlikely(au_test_fs_bad_xino(h_sb))) {
38790 +                       pr_err("xino doesn't support %s(%s)\n",
38791 +                              AUFS_XINO_DEFPATH, au_sbtype(h_sb));
38792 +                       fput(file);
38793 +                       file = ERR_PTR(-EINVAL);
38794 +               }
38795 +       }
38796 +
38797 +out:
38798 +       return file;
38799 +}
38800 +
38801 +/* ---------------------------------------------------------------------- */
38802 +
38803 +/*
38804 + * initialize the xinofile for the specified branch @br
38805 + * at the place/path where @base_file indicates.
38806 + * test whether another branch is on the same filesystem or not,
38807 + * if found then share the xinofile with another branch.
38808 + */
38809 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
38810 +                   struct path *base)
38811 +{
38812 +       int err;
38813 +       struct au_xino_do_set_br args = {
38814 +               .h_ino  = h_ino,
38815 +               .br     = br
38816 +       };
38817 +
38818 +       args.bshared = sbr_find_shared(sb, /*btop*/0, au_sbbot(sb),
38819 +                                      au_br_sb(br));
38820 +       err = au_xino_do_set_br(sb, base, &args);
38821 +       if (unlikely(err))
38822 +               au_xino_put(br);
38823 +
38824 +       return err;
38825 +}
38826 +
38827 +/* ---------------------------------------------------------------------- */
38828 +
38829 +/*
38830 + * get an unused inode number from bitmap
38831 + */
38832 +ino_t au_xino_new_ino(struct super_block *sb)
38833 +{
38834 +       ino_t ino;
38835 +       unsigned long *p, pindex, ul, pend;
38836 +       struct au_sbinfo *sbinfo;
38837 +       struct file *file;
38838 +       int free_bit, err;
38839 +
38840 +       if (!au_opt_test(au_mntflags(sb), XINO))
38841 +               return iunique(sb, AUFS_FIRST_INO);
38842 +
38843 +       sbinfo = au_sbi(sb);
38844 +       mutex_lock(&sbinfo->si_xib_mtx);
38845 +       p = sbinfo->si_xib_buf;
38846 +       free_bit = sbinfo->si_xib_next_bit;
38847 +       if (free_bit < page_bits && !test_bit(free_bit, p))
38848 +               goto out; /* success */
38849 +       free_bit = find_first_zero_bit(p, page_bits);
38850 +       if (free_bit < page_bits)
38851 +               goto out; /* success */
38852 +
38853 +       pindex = sbinfo->si_xib_last_pindex;
38854 +       for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
38855 +               err = xib_pindex(sb, ul);
38856 +               if (unlikely(err))
38857 +                       goto out_err;
38858 +               free_bit = find_first_zero_bit(p, page_bits);
38859 +               if (free_bit < page_bits)
38860 +                       goto out; /* success */
38861 +       }
38862 +
38863 +       file = sbinfo->si_xib;
38864 +       pend = vfsub_f_size_read(file) / PAGE_SIZE;
38865 +       for (ul = pindex + 1; ul <= pend; ul++) {
38866 +               err = xib_pindex(sb, ul);
38867 +               if (unlikely(err))
38868 +                       goto out_err;
38869 +               free_bit = find_first_zero_bit(p, page_bits);
38870 +               if (free_bit < page_bits)
38871 +                       goto out; /* success */
38872 +       }
38873 +       BUG();
38874 +
38875 +out:
38876 +       set_bit(free_bit, p);
38877 +       sbinfo->si_xib_next_bit = free_bit + 1;
38878 +       pindex = sbinfo->si_xib_last_pindex;
38879 +       mutex_unlock(&sbinfo->si_xib_mtx);
38880 +       ino = xib_calc_ino(pindex, free_bit);
38881 +       AuDbg("i%lu\n", (unsigned long)ino);
38882 +       return ino;
38883 +out_err:
38884 +       mutex_unlock(&sbinfo->si_xib_mtx);
38885 +       AuDbg("i0\n");
38886 +       return 0;
38887 +}
38888 +
38889 +/* for s_op->delete_inode() */
38890 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
38891 +{
38892 +       int err;
38893 +       unsigned int mnt_flags;
38894 +       aufs_bindex_t bindex, bbot, bi;
38895 +       unsigned char try_trunc;
38896 +       struct au_iinfo *iinfo;
38897 +       struct super_block *sb;
38898 +       struct au_hinode *hi;
38899 +       struct inode *h_inode;
38900 +       struct au_branch *br;
38901 +       struct au_xi_calc calc;
38902 +       struct file *file;
38903 +
38904 +       AuDebugOn(au_is_bad_inode(inode));
38905 +
38906 +       sb = inode->i_sb;
38907 +       mnt_flags = au_mntflags(sb);
38908 +       if (!au_opt_test(mnt_flags, XINO)
38909 +           || inode->i_ino == AUFS_ROOT_INO)
38910 +               return;
38911 +
38912 +       if (unlinked) {
38913 +               au_xigen_inc(inode);
38914 +               au_xib_clear_bit(inode);
38915 +       }
38916 +
38917 +       iinfo = au_ii(inode);
38918 +       bindex = iinfo->ii_btop;
38919 +       if (bindex < 0)
38920 +               return;
38921 +
38922 +       try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
38923 +       hi = au_hinode(iinfo, bindex);
38924 +       bbot = iinfo->ii_bbot;
38925 +       for (; bindex <= bbot; bindex++, hi++) {
38926 +               h_inode = hi->hi_inode;
38927 +               if (!h_inode
38928 +                   || (!unlinked && h_inode->i_nlink))
38929 +                       continue;
38930 +
38931 +               /* inode may not be revalidated */
38932 +               bi = au_br_index(sb, hi->hi_id);
38933 +               if (bi < 0)
38934 +                       continue;
38935 +
38936 +               br = au_sbr(sb, bi);
38937 +               au_xi_calc(sb, h_inode->i_ino, &calc);
38938 +               file = au_xino_file(br->br_xino, calc.idx);
38939 +               if (IS_ERR_OR_NULL(file))
38940 +                       continue;
38941 +
38942 +               err = au_xino_do_write(file, &calc, /*ino*/0);
38943 +               if (!err && try_trunc
38944 +                   && au_test_fs_trunc_xino(au_br_sb(br)))
38945 +                       xino_try_trunc(sb, br);
38946 +       }
38947 +}
38948 +
38949 +/* ---------------------------------------------------------------------- */
38950 +
38951 +static int au_xinondir_find(struct au_xino *xi, ino_t h_ino)
38952 +{
38953 +       int found, total, i;
38954 +
38955 +       found = -1;
38956 +       total = xi->xi_nondir.total;
38957 +       for (i = 0; i < total; i++) {
38958 +               if (xi->xi_nondir.array[i] != h_ino)
38959 +                       continue;
38960 +               found = i;
38961 +               break;
38962 +       }
38963 +
38964 +       return found;
38965 +}
38966 +
38967 +static int au_xinondir_expand(struct au_xino *xi)
38968 +{
38969 +       int err, sz;
38970 +       ino_t *p;
38971 +
38972 +       BUILD_BUG_ON(KMALLOC_MAX_SIZE > INT_MAX);
38973 +
38974 +       err = -ENOMEM;
38975 +       sz = xi->xi_nondir.total * sizeof(ino_t);
38976 +       if (unlikely(sz > KMALLOC_MAX_SIZE / 2))
38977 +               goto out;
38978 +       p = au_kzrealloc(xi->xi_nondir.array, sz, sz << 1, GFP_ATOMIC,
38979 +                        /*may_shrink*/0);
38980 +       if (p) {
38981 +               xi->xi_nondir.array = p;
38982 +               xi->xi_nondir.total <<= 1;
38983 +               AuDbg("xi_nondir.total %d\n", xi->xi_nondir.total);
38984 +               err = 0;
38985 +       }
38986 +
38987 +out:
38988 +       return err;
38989 +}
38990 +
38991 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
38992 +                      ino_t h_ino, int idx)
38993 +{
38994 +       struct au_xino *xi;
38995 +
38996 +       AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
38997 +       xi = au_sbr(sb, bindex)->br_xino;
38998 +       AuDebugOn(idx < 0 || xi->xi_nondir.total <= idx);
38999 +
39000 +       spin_lock(&xi->xi_nondir.spin);
39001 +       AuDebugOn(xi->xi_nondir.array[idx] != h_ino);
39002 +       xi->xi_nondir.array[idx] = 0;
39003 +       spin_unlock(&xi->xi_nondir.spin);
39004 +       wake_up_all(&xi->xi_nondir.wqh);
39005 +}
39006 +
39007 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
39008 +                     int *idx)
39009 +{
39010 +       int err, found, empty;
39011 +       struct au_xino *xi;
39012 +
39013 +       err = 0;
39014 +       *idx = -1;
39015 +       if (!au_opt_test(au_mntflags(sb), XINO))
39016 +               goto out; /* no xino */
39017 +
39018 +       xi = au_sbr(sb, bindex)->br_xino;
39019 +
39020 +again:
39021 +       spin_lock(&xi->xi_nondir.spin);
39022 +       found = au_xinondir_find(xi, h_ino);
39023 +       if (found == -1) {
39024 +               empty = au_xinondir_find(xi, /*h_ino*/0);
39025 +               if (empty == -1) {
39026 +                       empty = xi->xi_nondir.total;
39027 +                       err = au_xinondir_expand(xi);
39028 +                       if (unlikely(err))
39029 +                               goto out_unlock;
39030 +               }
39031 +               xi->xi_nondir.array[empty] = h_ino;
39032 +               *idx = empty;
39033 +       } else {
39034 +               spin_unlock(&xi->xi_nondir.spin);
39035 +               wait_event(xi->xi_nondir.wqh,
39036 +                          xi->xi_nondir.array[found] != h_ino);
39037 +               goto again;
39038 +       }
39039 +
39040 +out_unlock:
39041 +       spin_unlock(&xi->xi_nondir.spin);
39042 +out:
39043 +       return err;
39044 +}
39045 +
39046 +/* ---------------------------------------------------------------------- */
39047 +
39048 +int au_xino_path(struct seq_file *seq, struct file *file)
39049 +{
39050 +       int err;
39051 +
39052 +       err = au_seq_path(seq, &file->f_path);
39053 +       if (unlikely(err))
39054 +               goto out;
39055 +
39056 +#define Deleted "\\040(deleted)"
39057 +       seq->count -= sizeof(Deleted) - 1;
39058 +       AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
39059 +                        sizeof(Deleted) - 1));
39060 +#undef Deleted
39061 +
39062 +out:
39063 +       return err;
39064 +}
39065 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
39066 --- /usr/share/empty/include/uapi/linux/aufs_type.h     1970-01-01 01:00:00.000000000 +0100
39067 +++ linux/include/uapi/linux/aufs_type.h        2023-02-20 21:05:51.959693785 +0100
39068 @@ -0,0 +1,452 @@
39069 +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
39070 +/*
39071 + * Copyright (C) 2005-2022 Junjiro R. Okajima
39072 + *
39073 + * This program is free software; you can redistribute it and/or modify
39074 + * it under the terms of the GNU General Public License as published by
39075 + * the Free Software Foundation; either version 2 of the License, or
39076 + * (at your option) any later version.
39077 + *
39078 + * This program is distributed in the hope that it will be useful,
39079 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
39080 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
39081 + * GNU General Public License for more details.
39082 + *
39083 + * You should have received a copy of the GNU General Public License
39084 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
39085 + */
39086 +
39087 +#ifndef __AUFS_TYPE_H__
39088 +#define __AUFS_TYPE_H__
39089 +
39090 +#define AUFS_NAME      "aufs"
39091 +
39092 +#ifdef __KERNEL__
39093 +/*
39094 + * define it before including all other headers.
39095 + * sched.h may use pr_* macros before defining "current", so define the
39096 + * no-current version first, and re-define later.
39097 + */
39098 +#define pr_fmt(fmt)    AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
39099 +#include <linux/sched.h>
39100 +#undef pr_fmt
39101 +#define pr_fmt(fmt) \
39102 +               AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
39103 +               (int)sizeof(current->comm), current->comm, current->pid
39104 +#include <linux/limits.h>
39105 +#else
39106 +#include <stdint.h>
39107 +#include <sys/types.h>
39108 +#include <limits.h>
39109 +#endif /* __KERNEL__ */
39110 +
39111 +#define AUFS_VERSION   "6.x-rcN-20230109"
39112 +
39113 +/* todo? move this to linux-2.6.19/include/magic.h */
39114 +#define AUFS_SUPER_MAGIC       ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
39115 +
39116 +/* ---------------------------------------------------------------------- */
39117 +
39118 +#ifdef __KERNEL__
39119 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
39120 +typedef int8_t aufs_bindex_t;
39121 +#define AUFS_BRANCH_MAX 127
39122 +#else
39123 +typedef int16_t aufs_bindex_t;
39124 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
39125 +#define AUFS_BRANCH_MAX 511
39126 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
39127 +#define AUFS_BRANCH_MAX 1023
39128 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
39129 +#define AUFS_BRANCH_MAX 32767
39130 +#endif
39131 +#endif
39132 +
39133 +#ifndef AUFS_BRANCH_MAX
39134 +#error unknown CONFIG_AUFS_BRANCH_MAX value
39135 +#endif
39136 +#endif /* __KERNEL__ */
39137 +
39138 +/* ---------------------------------------------------------------------- */
39139 +
39140 +#define AUFS_FSTYPE            AUFS_NAME
39141 +
39142 +#define AUFS_ROOT_INO          2
39143 +#define AUFS_FIRST_INO         11
39144 +
39145 +#define AUFS_WH_PFX            ".wh."
39146 +#define AUFS_WH_PFX_LEN                ((int)sizeof(AUFS_WH_PFX) - 1)
39147 +#define AUFS_WH_TMP_LEN                4
39148 +/* a limit for rmdir/rename a dir and copyup */
39149 +#define AUFS_MAX_NAMELEN       (NAME_MAX \
39150 +                               - AUFS_WH_PFX_LEN * 2   /* doubly whiteouted */\
39151 +                               - 1                     /* dot */\
39152 +                               - AUFS_WH_TMP_LEN)      /* hex */
39153 +#define AUFS_XINO_FNAME                "." AUFS_NAME ".xino"
39154 +#define AUFS_XINO_DEFPATH      "/tmp/" AUFS_XINO_FNAME
39155 +#define AUFS_XINO_DEF_SEC      30 /* seconds */
39156 +#define AUFS_XINO_DEF_TRUNC    45 /* percentage */
39157 +#define AUFS_DIRWH_DEF         3
39158 +#define AUFS_RDCACHE_DEF       10 /* seconds */
39159 +#define AUFS_RDCACHE_MAX       3600 /* seconds */
39160 +#define AUFS_RDBLK_DEF         512 /* bytes */
39161 +#define AUFS_RDHASH_DEF                32
39162 +#define AUFS_WKQ_NAME          AUFS_NAME "d"
39163 +#define AUFS_MFS_DEF_SEC       30 /* seconds */
39164 +#define AUFS_MFS_MAX_SEC       3600 /* seconds */
39165 +#define AUFS_FHSM_CACHE_DEF_SEC        30 /* seconds */
39166 +#define AUFS_PLINK_WARN                50 /* number of plinks in a single bucket */
39167 +
39168 +/* pseudo-link maintenace under /proc */
39169 +#define AUFS_PLINK_MAINT_NAME  "plink_maint"
39170 +#define AUFS_PLINK_MAINT_DIR   "fs/" AUFS_NAME
39171 +#define AUFS_PLINK_MAINT_PATH  AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
39172 +
39173 +/* dirren, renamed dir */
39174 +#define AUFS_DR_INFO_PFX       AUFS_WH_PFX ".dr."
39175 +#define AUFS_DR_BRHINO_NAME    AUFS_WH_PFX "hino"
39176 +/* whiteouted doubly */
39177 +#define AUFS_WH_DR_INFO_PFX    AUFS_WH_PFX AUFS_DR_INFO_PFX
39178 +#define AUFS_WH_DR_BRHINO      AUFS_WH_PFX AUFS_DR_BRHINO_NAME
39179 +
39180 +#define AUFS_DIROPQ_NAME       AUFS_WH_PFX ".opq" /* whiteouted doubly */
39181 +#define AUFS_WH_DIROPQ         AUFS_WH_PFX AUFS_DIROPQ_NAME
39182 +
39183 +#define AUFS_BASE_NAME         AUFS_WH_PFX AUFS_NAME
39184 +#define AUFS_PLINKDIR_NAME     AUFS_WH_PFX "plnk"
39185 +#define AUFS_ORPHDIR_NAME      AUFS_WH_PFX "orph"
39186 +
39187 +/* doubly whiteouted */
39188 +#define AUFS_WH_BASE           AUFS_WH_PFX AUFS_BASE_NAME
39189 +#define AUFS_WH_PLINKDIR       AUFS_WH_PFX AUFS_PLINKDIR_NAME
39190 +#define AUFS_WH_ORPHDIR                AUFS_WH_PFX AUFS_ORPHDIR_NAME
39191 +
39192 +/* branch permissions and attributes */
39193 +#define AUFS_BRPERM_RW         "rw"
39194 +#define AUFS_BRPERM_RO         "ro"
39195 +#define AUFS_BRPERM_RR         "rr"
39196 +#define AUFS_BRATTR_COO_REG    "coo_reg"
39197 +#define AUFS_BRATTR_COO_ALL    "coo_all"
39198 +#define AUFS_BRATTR_FHSM       "fhsm"
39199 +#define AUFS_BRATTR_UNPIN      "unpin"
39200 +#define AUFS_BRATTR_ICEX       "icex"
39201 +#define AUFS_BRATTR_ICEX_SEC   "icexsec"
39202 +#define AUFS_BRATTR_ICEX_SYS   "icexsys"
39203 +#define AUFS_BRATTR_ICEX_TR    "icextr"
39204 +#define AUFS_BRATTR_ICEX_USR   "icexusr"
39205 +#define AUFS_BRATTR_ICEX_OTH   "icexoth"
39206 +#define AUFS_BRRATTR_WH                "wh"
39207 +#define AUFS_BRWATTR_NLWH      "nolwh"
39208 +#define AUFS_BRWATTR_MOO       "moo"
39209 +
39210 +#define AuBrPerm_RW            1               /* writable, hardlinkable wh */
39211 +#define AuBrPerm_RO            (1 << 1)        /* readonly */
39212 +#define AuBrPerm_RR            (1 << 2)        /* natively readonly */
39213 +#define AuBrPerm_Mask          (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
39214 +
39215 +#define AuBrAttr_COO_REG       (1 << 3)        /* copy-up on open */
39216 +#define AuBrAttr_COO_ALL       (1 << 4)
39217 +#define AuBrAttr_COO_Mask      (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
39218 +
39219 +#define AuBrAttr_FHSM          (1 << 5)        /* file-based hsm */
39220 +#define AuBrAttr_UNPIN         (1 << 6)        /* rename-able top dir of
39221 +                                                  branch. meaningless since
39222 +                                                  linux-3.18-rc1 */
39223 +
39224 +/* ignore error in copying XATTR */
39225 +#define AuBrAttr_ICEX_SEC      (1 << 7)
39226 +#define AuBrAttr_ICEX_SYS      (1 << 8)
39227 +#define AuBrAttr_ICEX_TR       (1 << 9)
39228 +#define AuBrAttr_ICEX_USR      (1 << 10)
39229 +#define AuBrAttr_ICEX_OTH      (1 << 11)
39230 +#define AuBrAttr_ICEX          (AuBrAttr_ICEX_SEC      \
39231 +                                | AuBrAttr_ICEX_SYS    \
39232 +                                | AuBrAttr_ICEX_TR     \
39233 +                                | AuBrAttr_ICEX_USR    \
39234 +                                | AuBrAttr_ICEX_OTH)
39235 +
39236 +#define AuBrRAttr_WH           (1 << 12)       /* whiteout-able */
39237 +#define AuBrRAttr_Mask         AuBrRAttr_WH
39238 +
39239 +#define AuBrWAttr_NoLinkWH     (1 << 13)       /* un-hardlinkable whiteouts */
39240 +#define AuBrWAttr_MOO          (1 << 14)       /* move-up on open */
39241 +#define AuBrWAttr_Mask         (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
39242 +
39243 +#define AuBrAttr_CMOO_Mask     (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
39244 +
39245 +/* #warning test userspace */
39246 +#ifdef __KERNEL__
39247 +#ifndef CONFIG_AUFS_FHSM
39248 +#undef AuBrAttr_FHSM
39249 +#define AuBrAttr_FHSM          0
39250 +#endif
39251 +#ifndef CONFIG_AUFS_XATTR
39252 +#undef AuBrAttr_ICEX
39253 +#define AuBrAttr_ICEX          0
39254 +#undef AuBrAttr_ICEX_SEC
39255 +#define AuBrAttr_ICEX_SEC      0
39256 +#undef AuBrAttr_ICEX_SYS
39257 +#define AuBrAttr_ICEX_SYS      0
39258 +#undef AuBrAttr_ICEX_TR
39259 +#define AuBrAttr_ICEX_TR       0
39260 +#undef AuBrAttr_ICEX_USR
39261 +#define AuBrAttr_ICEX_USR      0
39262 +#undef AuBrAttr_ICEX_OTH
39263 +#define AuBrAttr_ICEX_OTH      0
39264 +#endif
39265 +#endif
39266 +
39267 +/* the longest combination */
39268 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
39269 +#define AuBrPermStrSz  sizeof(AUFS_BRPERM_RW                   \
39270 +                              "+" AUFS_BRATTR_COO_REG          \
39271 +                              "+" AUFS_BRATTR_FHSM             \
39272 +                              "+" AUFS_BRATTR_UNPIN            \
39273 +                              "+" AUFS_BRATTR_ICEX_SEC         \
39274 +                              "+" AUFS_BRATTR_ICEX_SYS         \
39275 +                              "+" AUFS_BRATTR_ICEX_USR         \
39276 +                              "+" AUFS_BRATTR_ICEX_OTH         \
39277 +                              "+" AUFS_BRWATTR_NLWH)
39278 +
39279 +typedef struct {
39280 +       char a[AuBrPermStrSz];
39281 +} au_br_perm_str_t;
39282 +
39283 +static inline int au_br_writable(int brperm)
39284 +{
39285 +       return brperm & AuBrPerm_RW;
39286 +}
39287 +
39288 +static inline int au_br_whable(int brperm)
39289 +{
39290 +       return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
39291 +}
39292 +
39293 +static inline int au_br_wh_linkable(int brperm)
39294 +{
39295 +       return !(brperm & AuBrWAttr_NoLinkWH);
39296 +}
39297 +
39298 +static inline int au_br_cmoo(int brperm)
39299 +{
39300 +       return brperm & AuBrAttr_CMOO_Mask;
39301 +}
39302 +
39303 +static inline int au_br_fhsm(int brperm)
39304 +{
39305 +       return brperm & AuBrAttr_FHSM;
39306 +}
39307 +
39308 +/* ---------------------------------------------------------------------- */
39309 +
39310 +/* ioctl */
39311 +enum {
39312 +       /* readdir in userspace */
39313 +       AuCtl_RDU,
39314 +       AuCtl_RDU_INO,
39315 +
39316 +       AuCtl_WBR_FD,   /* pathconf wrapper */
39317 +       AuCtl_IBUSY,    /* busy inode */
39318 +       AuCtl_MVDOWN,   /* move-down */
39319 +       AuCtl_BR,       /* info about branches */
39320 +       AuCtl_FHSM_FD   /* connection for fhsm */
39321 +};
39322 +
39323 +/* borrowed from linux/include/linux/kernel.h */
39324 +#ifndef ALIGN
39325 +#ifdef _GNU_SOURCE
39326 +#define ALIGN(x, a)            __ALIGN_MASK(x, (typeof(x))(a)-1)
39327 +#else
39328 +#define ALIGN(x, a)            (((x) + (a) - 1) & ~((a) - 1))
39329 +#endif
39330 +#define __ALIGN_MASK(x, mask)  (((x)+(mask))&~(mask))
39331 +#endif
39332 +
39333 +/* borrowed from linux/include/linux/compiler-gcc3.h */
39334 +#ifndef __aligned
39335 +#define __aligned(x)                   __attribute__((aligned(x)))
39336 +#endif
39337 +
39338 +#ifdef __KERNEL__
39339 +#ifndef __packed
39340 +#define __packed                       __attribute__((packed))
39341 +#endif
39342 +#endif
39343 +
39344 +struct au_rdu_cookie {
39345 +       uint64_t        h_pos;
39346 +       int16_t         bindex;
39347 +       uint8_t         flags;
39348 +       uint8_t         pad;
39349 +       uint32_t        generation;
39350 +} __aligned(8);
39351 +
39352 +struct au_rdu_ent {
39353 +       uint64_t        ino;
39354 +       int16_t         bindex;
39355 +       uint8_t         type;
39356 +       uint8_t         nlen;
39357 +       uint8_t         wh;
39358 +       char            name[];
39359 +} __aligned(8);
39360 +
39361 +static inline int au_rdu_len(int nlen)
39362 +{
39363 +       /* include the terminating NULL */
39364 +       return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
39365 +                    sizeof(uint64_t));
39366 +}
39367 +
39368 +union au_rdu_ent_ul {
39369 +       struct au_rdu_ent __user        *e;
39370 +       uint64_t                        ul;
39371 +};
39372 +
39373 +enum {
39374 +       AufsCtlRduV_SZ,
39375 +       AufsCtlRduV_End
39376 +};
39377 +
39378 +struct aufs_rdu {
39379 +       /* input */
39380 +       union {
39381 +               uint64_t        sz;     /* AuCtl_RDU */
39382 +               uint64_t        nent;   /* AuCtl_RDU_INO */
39383 +       };
39384 +       union au_rdu_ent_ul     ent;
39385 +       uint16_t                verify[AufsCtlRduV_End];
39386 +
39387 +       /* input/output */
39388 +       uint32_t                blk;
39389 +
39390 +       /* output */
39391 +       union au_rdu_ent_ul     tail;
39392 +       /* number of entries which were added in a single call */
39393 +       uint64_t                rent;
39394 +       uint8_t                 full;
39395 +       uint8_t                 shwh;
39396 +
39397 +       struct au_rdu_cookie    cookie;
39398 +} __aligned(8);
39399 +
39400 +/* ---------------------------------------------------------------------- */
39401 +
39402 +/* dirren. the branch is identified by the filename who contains this */
39403 +struct au_drinfo {
39404 +       uint64_t ino;
39405 +       union {
39406 +               uint8_t oldnamelen;
39407 +               uint64_t _padding;
39408 +       };
39409 +       uint8_t oldname[];
39410 +} __aligned(8);
39411 +
39412 +struct au_drinfo_fdata {
39413 +       uint32_t magic;
39414 +       struct au_drinfo drinfo;
39415 +} __aligned(8);
39416 +
39417 +#define AUFS_DRINFO_MAGIC_V1   ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x01)
39418 +/* future */
39419 +#define AUFS_DRINFO_MAGIC_V2   ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x02)
39420 +
39421 +/* ---------------------------------------------------------------------- */
39422 +
39423 +struct aufs_wbr_fd {
39424 +       uint32_t        oflags;
39425 +       int16_t         brid;
39426 +} __aligned(8);
39427 +
39428 +/* ---------------------------------------------------------------------- */
39429 +
39430 +struct aufs_ibusy {
39431 +       uint64_t        ino, h_ino;
39432 +       int16_t         bindex;
39433 +} __aligned(8);
39434 +
39435 +/* ---------------------------------------------------------------------- */
39436 +
39437 +/* error code for move-down */
39438 +/* the actual message strings are implemented in aufs-util.git */
39439 +enum {
39440 +       EAU_MVDOWN_OPAQUE = 1,
39441 +       EAU_MVDOWN_WHITEOUT,
39442 +       EAU_MVDOWN_UPPER,
39443 +       EAU_MVDOWN_BOTTOM,
39444 +       EAU_MVDOWN_NOUPPER,
39445 +       EAU_MVDOWN_NOLOWERBR,
39446 +       EAU_Last
39447 +};
39448 +
39449 +/* flags for move-down */
39450 +#define AUFS_MVDOWN_DMSG       1
39451 +#define AUFS_MVDOWN_OWLOWER    (1 << 1)        /* overwrite lower */
39452 +#define AUFS_MVDOWN_KUPPER     (1 << 2)        /* keep upper */
39453 +#define AUFS_MVDOWN_ROLOWER    (1 << 3)        /* do even if lower is RO */
39454 +#define AUFS_MVDOWN_ROLOWER_R  (1 << 4)        /* did on lower RO */
39455 +#define AUFS_MVDOWN_ROUPPER    (1 << 5)        /* do even if upper is RO */
39456 +#define AUFS_MVDOWN_ROUPPER_R  (1 << 6)        /* did on upper RO */
39457 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7)        /* upper brid */
39458 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8)        /* lower brid */
39459 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9)        /* find fhsm attr for lower */
39460 +#define AUFS_MVDOWN_STFS       (1 << 10)       /* req. stfs */
39461 +#define AUFS_MVDOWN_STFS_FAILED        (1 << 11)       /* output: stfs is unusable */
39462 +#define AUFS_MVDOWN_BOTTOM     (1 << 12)       /* output: no more lowers */
39463 +
39464 +/* index for move-down */
39465 +enum {
39466 +       AUFS_MVDOWN_UPPER,
39467 +       AUFS_MVDOWN_LOWER,
39468 +       AUFS_MVDOWN_NARRAY
39469 +};
39470 +
39471 +/*
39472 + * additional info of move-down
39473 + * number of free blocks and inodes.
39474 + * subset of struct kstatfs, but smaller and always 64bit.
39475 + */
39476 +struct aufs_stfs {
39477 +       uint64_t        f_blocks;
39478 +       uint64_t        f_bavail;
39479 +       uint64_t        f_files;
39480 +       uint64_t        f_ffree;
39481 +};
39482 +
39483 +struct aufs_stbr {
39484 +       int16_t                 brid;   /* optional input */
39485 +       int16_t                 bindex; /* output */
39486 +       struct aufs_stfs        stfs;   /* output when AUFS_MVDOWN_STFS set */
39487 +} __aligned(8);
39488 +
39489 +struct aufs_mvdown {
39490 +       uint32_t                flags;                  /* input/output */
39491 +       struct aufs_stbr        stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
39492 +       int8_t                  au_errno;               /* output */
39493 +} __aligned(8);
39494 +
39495 +/* ---------------------------------------------------------------------- */
39496 +
39497 +union aufs_brinfo {
39498 +       /* PATH_MAX may differ between kernel-space and user-space */
39499 +       char    _spacer[4096];
39500 +       struct {
39501 +               int16_t id;
39502 +               int     perm;
39503 +               char    path[];
39504 +       };
39505 +} __aligned(8);
39506 +
39507 +/* ---------------------------------------------------------------------- */
39508 +
39509 +#define AuCtlType              'A'
39510 +#define AUFS_CTL_RDU           _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
39511 +#define AUFS_CTL_RDU_INO       _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
39512 +#define AUFS_CTL_WBR_FD                _IOW(AuCtlType, AuCtl_WBR_FD, \
39513 +                                    struct aufs_wbr_fd)
39514 +#define AUFS_CTL_IBUSY         _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
39515 +#define AUFS_CTL_MVDOWN                _IOWR(AuCtlType, AuCtl_MVDOWN, \
39516 +                                     struct aufs_mvdown)
39517 +#define AUFS_CTL_BRINFO                _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
39518 +#define AUFS_CTL_FHSM_FD       _IOW(AuCtlType, AuCtl_FHSM_FD, int)
39519 +
39520 +#endif /* __AUFS_TYPE_H__ */
39521 SPDX-License-Identifier: GPL-2.0
39522 aufs6.x-rcN loopback patch
39523
39524 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
39525 index 7a9928c6db9d..0bac486deea2 100644
39526 --- a/drivers/block/loop.c
39527 +++ b/drivers/block/loop.c
39528 @@ -54,7 +54,7 @@ struct loop_device {
39529         int             lo_flags;
39530         char            lo_file_name[LO_NAME_SIZE];
39531
39532 -       struct file *   lo_backing_file;
39533 +       struct file     *lo_backing_file, *lo_backing_virt_file;
39534         struct block_device *lo_device;
39535
39536         gfp_t           old_gfp_mask;
39537 @@ -510,6 +510,15 @@ static inline void loop_update_dio(struct loop_device *lo)
39538                                 lo->use_dio);
39539  }
39540
39541 +static struct file *loop_real_file(struct file *file)
39542 +{
39543 +       struct file *f = NULL;
39544 +
39545 +       if (file->f_path.dentry->d_sb->s_op->real_loop)
39546 +               f = file->f_path.dentry->d_sb->s_op->real_loop(file);
39547 +       return f;
39548 +}
39549 +
39550  static void loop_reread_partitions(struct loop_device *lo)
39551  {
39552         int rc;
39553 @@ -567,6 +576,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39554  {
39555         struct file *file = fget(arg);
39556         struct file *old_file;
39557 +       struct file *f, *virt_file = NULL, *old_virt_file;
39558         int error;
39559         bool partscan;
39560         bool is_loop;
39561 @@ -590,11 +600,19 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39562         if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
39563                 goto out_err;
39564
39565 +       f = loop_real_file(file);
39566 +       if (f) {
39567 +               virt_file = file;
39568 +               file = f;
39569 +               get_file(file);
39570 +       }
39571 +
39572         error = loop_validate_file(file, bdev);
39573         if (error)
39574                 goto out_err;
39575
39576         old_file = lo->lo_backing_file;
39577 +       old_virt_file = lo->lo_backing_virt_file;
39578
39579         error = -EINVAL;
39580
39581 @@ -607,6 +625,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39582         blk_mq_freeze_queue(lo->lo_queue);
39583         mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
39584         lo->lo_backing_file = file;
39585 +       lo->lo_backing_virt_file = virt_file;
39586         lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
39587         mapping_set_gfp_mask(file->f_mapping,
39588                              lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39589 @@ -629,6 +648,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39590          * dependency.
39591          */
39592         fput(old_file);
39593 +       if (old_virt_file)
39594 +               fput(old_virt_file);
39595         if (partscan)
39596                 loop_reread_partitions(lo);
39597
39598 @@ -642,6 +663,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39599         loop_global_unlock(lo, is_loop);
39600  out_putf:
39601         fput(file);
39602 +       if (virt_file)
39603 +               fput(virt_file);
39604         goto done;
39605  }
39606
39607 @@ -1013,6 +1036,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39608                           const struct loop_config *config)
39609  {
39610         struct file *file = fget(config->fd);
39611 +       struct file *f, *virt_file = NULL;
39612         struct inode *inode;
39613         struct address_space *mapping;
39614         int error;
39615 @@ -1031,6 +1055,13 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39616         /* suppress uevents while reconfiguring the device */
39617         dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 1);
39618
39619 +       f = loop_real_file(file);
39620 +       if (f) {
39621 +               virt_file = file;
39622 +               file = f;
39623 +               get_file(file);
39624 +       }
39625 +
39626         /*
39627          * If we don't hold exclusive handle for the device, upgrade to it
39628          * here to avoid changing device under exclusive owner.
39629 @@ -1091,6 +1122,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39630         lo->use_dio = lo->lo_flags & LO_FLAGS_DIRECT_IO;
39631         lo->lo_device = bdev;
39632         lo->lo_backing_file = file;
39633 +       lo->lo_backing_virt_file = virt_file;
39634         lo->old_gfp_mask = mapping_gfp_mask(mapping);
39635         mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39636
39637 @@ -1146,6 +1178,8 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39638                 bd_abort_claiming(bdev, loop_configure);
39639  out_putf:
39640         fput(file);
39641 +       if (virt_file)
39642 +               fput(virt_file);
39643         /* This is safe: open() is still holding a reference. */
39644         module_put(THIS_MODULE);
39645         goto done;
39646 @@ -1154,6 +1188,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39647  static void __loop_clr_fd(struct loop_device *lo, bool release)
39648  {
39649         struct file *filp;
39650 +       struct file *virt_filp = lo->lo_backing_virt_file;
39651         gfp_t gfp = lo->old_gfp_mask;
39652
39653         if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags))
39654 @@ -1170,6 +1205,7 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39655         spin_lock_irq(&lo->lo_lock);
39656         filp = lo->lo_backing_file;
39657         lo->lo_backing_file = NULL;
39658 +       lo->lo_backing_virt_file = NULL;
39659         spin_unlock_irq(&lo->lo_lock);
39660
39661         lo->lo_device = NULL;
39662 @@ -1232,6 +1268,8 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39663          * fput can take open_mutex which is usually taken before lo_mutex.
39664          */
39665         fput(filp);
39666 +       if (virt_filp)
39667 +               fput(virt_filp);
39668  }
39669
39670  static int loop_clr_fd(struct loop_device *lo)
39671 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
39672 index 1ed300047a41..ea63191d6473 100644
39673 --- a/fs/aufs/f_op.c
39674 +++ b/fs/aufs/f_op.c
39675 @@ -309,7 +309,7 @@ static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
39676         if (IS_ERR(h_file))
39677                 goto out;
39678
39679 -       if (au_test_loopback_kthread()) {
39680 +       if (0 && au_test_loopback_kthread()) {
39681                 au_warn_loopback(h_file->f_path.dentry->d_sb);
39682                 if (file->f_mapping != h_file->f_mapping) {
39683                         file->f_mapping = h_file->f_mapping;
39684 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
39685 index 58043e31e5f3..e2bfae6f9d59 100644
39686 --- a/fs/aufs/loop.c
39687 +++ b/fs/aufs/loop.c
39688 @@ -133,3 +133,19 @@ void au_loopback_fin(void)
39689                 symbol_put(loop_backing_file);
39690         au_kfree_try_rcu(au_warn_loopback_array);
39691  }
39692 +
39693 +/* ---------------------------------------------------------------------- */
39694 +
39695 +/* support the loopback block device insude aufs */
39696 +
39697 +struct file *aufs_real_loop(struct file *file)
39698 +{
39699 +       struct file *f;
39700 +
39701 +       BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
39702 +       fi_read_lock(file);
39703 +       f = au_hf_top(file);
39704 +       fi_read_unlock(file);
39705 +       AuDebugOn(!f);
39706 +       return f;
39707 +}
39708 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
39709 index 03d4908a6c03..34d356e181d5 100644
39710 --- a/fs/aufs/loop.h
39711 +++ b/fs/aufs/loop.h
39712 @@ -26,6 +26,8 @@ void au_warn_loopback(struct super_block *h_sb);
39713
39714  int au_loopback_init(void);
39715  void au_loopback_fin(void);
39716 +
39717 +struct file *aufs_real_loop(struct file *file);
39718  #else
39719  AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
39720
39721 @@ -36,6 +38,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
39722
39723  AuStubInt0(au_loopback_init, void)
39724  AuStubVoid(au_loopback_fin, void)
39725 +
39726 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
39727  #endif /* BLK_DEV_LOOP */
39728
39729  #endif /* __KERNEL__ */
39730 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
39731 index 81922d4faf54..c8a62c267d72 100644
39732 --- a/fs/aufs/super.c
39733 +++ b/fs/aufs/super.c
39734 @@ -758,7 +758,10 @@ const struct super_operations aufs_sop = {
39735         .show_options   = aufs_show_options,
39736         .statfs         = aufs_statfs,
39737         .put_super      = aufs_put_super,
39738 -       .sync_fs        = aufs_sync_fs
39739 +       .sync_fs        = aufs_sync_fs,
39740 +#ifdef CONFIG_AUFS_BDEV_LOOP
39741 +       .real_loop      = aufs_real_loop
39742 +#endif
39743  };
39744
39745  /* ---------------------------------------------------------------------- */
39746 diff --git a/include/linux/fs.h b/include/linux/fs.h
39747 index 2d30def9a580..84a83c480302 100644
39748 --- a/include/linux/fs.h
39749 +++ b/include/linux/fs.h
39750 @@ -2252,6 +2252,10 @@ struct super_operations {
39751                                   struct shrink_control *);
39752         long (*free_cached_objects)(struct super_block *,
39753                                     struct shrink_control *);
39754 +#if IS_ENABLED(CONFIG_BLK_DEV_LOOP) || IS_ENABLED(CONFIG_BLK_DEV_LOOP_MODULE)
39755 +       /* and aufs */
39756 +       struct file *(*real_loop)(struct file *);
39757 +#endif
39758  };
39759
39760  /*