kernel-aufs.patch

   1 SPDX-License-Identifier: GPL-2.0
   2 aufs6.0 kbuild patch
   3
   4 diff --git a/fs/Kconfig b/fs/Kconfig
   5 index a547307c1ae8..919ca5404804 100644
   6 --- a/fs/Kconfig
   7 +++ b/fs/Kconfig
   8 @@ -318,6 +318,7 @@ 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 93b80529f8e8..a26c58d09435 100644
  18 --- a/fs/Makefile
  19 +++ b/fs/Makefile
  20 @@ -138,3 +138,4 @@ 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 aufs6.0 base patch
  27
  28 diff --git a/MAINTAINERS b/MAINTAINERS
  29 index 72b9654f764c..53fc0e89d1ea 100644
  30 --- a/MAINTAINERS
  31 +++ b/MAINTAINERS
  32 @@ -3425,6 +3425,19 @@ 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 ad92192c7d61..8ca56da3ca69 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 bb0c4d0038db..d753ad7d77e6 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 df137ba19d37..faae7f8644ac 100644
 118 --- a/fs/namespace.c
 119 +++ b/fs/namespace.c
 120 @@ -852,6 +852,12 @@ 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 0878b852b355..d83f9e5cddbc 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 9eced4cc286e..9c5a33fce3a8 100644
 163 --- a/include/linux/fs.h
 164 +++ b/include/linux/fs.h
 165 @@ -1364,6 +1364,7 @@ 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 @@ -2115,6 +2116,7 @@ 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 64a13eb56078..c9f30e12aa32 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 aufs6.0 mmap patch
 252
 253 diff --git a/fs/proc/base.c b/fs/proc/base.c
 254 index 93f7e3d971e4..23aa1626020a 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 4e0023643f8b..09394cf6e349 100644
 284 --- a/fs/proc/task_mmu.c
 285 +++ b/fs/proc/task_mmu.c
 286 @@ -281,7 +281,10 @@ 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 @@ -1912,7 +1915,7 @@ 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 a6d21fc0033c..02c2de31196e 100644
 309 --- a/fs/proc/task_nommu.c
 310 +++ b/fs/proc/task_nommu.c
 311 @@ -155,7 +155,10 @@ 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 21f8b27bd9fd..de9572fe212b 100644
 325 --- a/include/linux/mm.h
 326 +++ b/include/linux/mm.h
 327 @@ -1858,6 +1858,43 @@ 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 cf97f3884fda..a783130a669d 100644
 373 --- a/include/linux/mm_types.h
 374 +++ b/include/linux/mm_types.h
 375 @@ -372,6 +372,9 @@ 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 @@ -468,6 +471,9 @@ 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_SWAP
 395 diff --git a/kernel/fork.c b/kernel/fork.c
 396 index 2b6bd511c6ed..944adfb5189a 100644
 397 --- a/kernel/fork.c
 398 +++ b/kernel/fork.c
 399 @@ -662,7 +662,7 @@ 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 9a564f836403..86a4be7f34a8 100644
 410 --- a/mm/Makefile
 411 +++ b/mm/Makefile
 412 @@ -134,3 +134,4 @@ 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 15800334147b..2c22abb4a14d 100644
 419 --- a/mm/filemap.c
 420 +++ b/mm/filemap.c
 421 @@ -3403,7 +3403,7 @@ 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 9d780f415be3..e744ba276454 100644
 432 --- a/mm/mmap.c
 433 +++ b/mm/mmap.c
 434 @@ -142,7 +142,7 @@ static struct vm_area_struct *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         return next;
 443 @@ -911,7 +911,7 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 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 @@ -1845,7 +1845,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
 453         if (vma->vm_ops && vma->vm_ops->close)
 454                 vma->vm_ops->close(vma);
 455  unmap_and_free_vma:
 456 -       fput(vma->vm_file);
 457 +       vma_fput(vma);
 458         vma->vm_file = NULL;
 459
 460         /* Undo any partial mapping done by a device driver. */
 461 @@ -2705,7 +2705,7 @@ 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 @@ -2724,7 +2724,7 @@ 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 @@ -2897,6 +2897,9 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
 480         unsigned long populate = 0;
 481         unsigned long ret = -EINVAL;
 482         struct file *file;
 483 +#if IS_ENABLED(CONFIG_AUFS_FS)
 484 +       struct file *prfile;
 485 +#endif
 486
 487         pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/mm/remap_file_pages.rst.\n",
 488                      current->comm, current->pid);
 489 @@ -2952,10 +2955,34 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
 490         if (vma->vm_flags & VM_LOCKED)
 491                 flags |= MAP_LOCKED;
 492
 493 +#if IS_ENABLED(CONFIG_AUFS_FS)
 494 +       vma_get_file(vma);
 495 +       file = vma->vm_file;
 496 +       prfile = vma->vm_prfile;
 497 +       ret = do_mmap(vma->vm_file, start, size,
 498 +                       prot, flags, pgoff, &populate, NULL);
 499 +       if (!IS_ERR_VALUE(ret) && file && prfile) {
 500 +               struct vm_area_struct *new_vma;
 501 +
 502 +               new_vma = find_vma(mm, ret);
 503 +               if (!new_vma->vm_prfile)
 504 +                       new_vma->vm_prfile = prfile;
 505 +               if (new_vma != vma)
 506 +                       get_file(prfile);
 507 +       }
 508 +       /*
 509 +        * two fput()s instead of vma_fput(vma),
 510 +        * coz vma may not be available anymore.
 511 +        */
 512 +       fput(file);
 513 +       if (prfile)
 514 +               fput(prfile);
 515 +#else
 516         file = get_file(vma->vm_file);
 517         ret = do_mmap(vma->vm_file, start, size,
 518                         prot, flags, pgoff, &populate, NULL);
 519         fput(file);
 520 +#endif /* CONFIG_AUFS_FS */
 521  out:
 522         mmap_write_unlock(mm);
 523         if (populate)
 524 @@ -3229,7 +3256,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
 525                 if (anon_vma_clone(new_vma, vma))
 526                         goto out_free_mempol;
 527                 if (new_vma->vm_file)
 528 -                       get_file(new_vma->vm_file);
 529 +                       vma_get_file(new_vma);
 530                 if (new_vma->vm_ops && new_vma->vm_ops->open)
 531                         new_vma->vm_ops->open(new_vma);
 532                 vma_link(mm, new_vma, prev, rb_link, rb_parent);
 533 diff --git a/mm/nommu.c b/mm/nommu.c
 534 index e819cbc21b39..dc7d40fb197f 100644
 535 --- a/mm/nommu.c
 536 +++ b/mm/nommu.c
 537 @@ -524,7 +524,7 @@ static void __put_nommu_region(struct vm_region *region)
 538                 up_write(&nommu_region_sem);
 539
 540                 if (region->vm_file)
 541 -                       fput(region->vm_file);
 542 +                       vmr_fput(region);
 543
 544                 /* IO memory and memory shared directly out of the pagecache
 545                  * from ramfs/tmpfs mustn't be released here */
 546 @@ -656,7 +656,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
 547         if (vma->vm_ops && vma->vm_ops->close)
 548                 vma->vm_ops->close(vma);
 549         if (vma->vm_file)
 550 -               fput(vma->vm_file);
 551 +               vma_fput(vma);
 552         put_nommu_region(vma->vm_region);
 553         vm_area_free(vma);
 554  }
 555 @@ -1176,7 +1176,7 @@ unsigned long do_mmap(struct file *file,
 556                                         goto error_just_free;
 557                                 }
 558                         }
 559 -                       fput(region->vm_file);
 560 +                       vmr_fput(region);
 561                         kmem_cache_free(vm_region_jar, region);
 562                         region = pregion;
 563                         result = start;
 564 @@ -1253,10 +1253,10 @@ unsigned long do_mmap(struct file *file,
 565         up_write(&nommu_region_sem);
 566  error:
 567         if (region->vm_file)
 568 -               fput(region->vm_file);
 569 +               vmr_fput(region);
 570         kmem_cache_free(vm_region_jar, region);
 571         if (vma->vm_file)
 572 -               fput(vma->vm_file);
 573 +               vma_fput(vma);
 574         vm_area_free(vma);
 575         return ret;
 576
 577 diff --git a/mm/prfile.c b/mm/prfile.c
 578 new file mode 100644
 579 index 000000000000..8f820a235364
 580 --- /dev/null
 581 +++ b/mm/prfile.c
 582 @@ -0,0 +1,86 @@
 583 +// SPDX-License-Identifier: GPL-2.0
 584 +/*
 585 + * Mainly for aufs which mmap(2) different file and wants to print different
 586 + * path in /proc/PID/maps.
 587 + * Call these functions via macros defined in linux/mm.h.
 588 + *
 589 + * See Documentation/filesystems/aufs/design/06mmap.txt
 590 + *
 591 + * Copyright (c) 2014-2022 Junjro R. Okajima
 592 + * Copyright (c) 2014 Ian Campbell
 593 + */
 594 +
 595 +#include <linux/mm.h>
 596 +#include <linux/file.h>
 597 +#include <linux/fs.h>
 598 +
 599 +/* #define PRFILE_TRACE */
 600 +static inline void prfile_trace(struct file *f, struct file *pr,
 601 +                             const char func[], int line, const char func2[])
 602 +{
 603 +#ifdef PRFILE_TRACE
 604 +       if (pr)
 605 +               pr_info("%s:%d: %s, %pD2\n", func, line, func2, f);
 606 +#endif
 607 +}
 608 +
 609 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
 610 +                            int line)
 611 +{
 612 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 613 +
 614 +       prfile_trace(f, pr, func, line, __func__);
 615 +       file_update_time(f);
 616 +       if (f && pr)
 617 +               file_update_time(pr);
 618 +}
 619 +
 620 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
 621 +                              int line)
 622 +{
 623 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 624 +
 625 +       prfile_trace(f, pr, func, line, __func__);
 626 +       return (f && pr) ? pr : f;
 627 +}
 628 +
 629 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
 630 +{
 631 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 632 +
 633 +       prfile_trace(f, pr, func, line, __func__);
 634 +       get_file(f);
 635 +       if (f && pr)
 636 +               get_file(pr);
 637 +}
 638 +
 639 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
 640 +{
 641 +       struct file *f = vma->vm_file, *pr = vma->vm_prfile;
 642 +
 643 +       prfile_trace(f, pr, func, line, __func__);
 644 +       fput(f);
 645 +       if (f && pr)
 646 +               fput(pr);
 647 +}
 648 +
 649 +#ifndef CONFIG_MMU
 650 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
 651 +                              int line)
 652 +{
 653 +       struct file *f = region->vm_file, *pr = region->vm_prfile;
 654 +
 655 +       prfile_trace(f, pr, func, line, __func__);
 656 +       return (f && pr) ? pr : f;
 657 +}
 658 +
 659 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
 660 +{
 661 +       struct file *f = region->vm_file, *pr = region->vm_prfile;
 662 +
 663 +       prfile_trace(f, pr, func, line, __func__);
 664 +       fput(f);
 665 +       if (f && pr)
 666 +               fput(pr);
 667 +}
 668 +#endif /* !CONFIG_MMU */
 669 SPDX-License-Identifier: GPL-2.0
 670 aufs6.0 standalone patch
 671
 672 diff --git a/fs/dcache.c b/fs/dcache.c
 673 index d753ad7d77e6..1a3b40ebc12b 100644
 674 --- a/fs/dcache.c
 675 +++ b/fs/dcache.c
 676 @@ -1450,6 +1450,7 @@ void d_walk(struct dentry *parent, void *data,
 677         seq = 1;
 678         goto again;
 679  }
 680 +EXPORT_SYMBOL_GPL(d_walk);
 681
 682  struct check_mount {
 683         struct vfsmount *mnt;
 684 @@ -3061,6 +3062,7 @@ void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
 685
 686         write_sequnlock(&rename_lock);
 687  }
 688 +EXPORT_SYMBOL_GPL(d_exchange);
 689
 690  /**
 691   * d_ancestor - search for an ancestor
 692 diff --git a/fs/exec.c b/fs/exec.c
 693 index d046dbb9cbd0..5b9023e75de0 100644
 694 --- a/fs/exec.c
 695 +++ b/fs/exec.c
 696 @@ -111,6 +111,7 @@ bool path_noexec(const struct path *path)
 697         return (path->mnt->mnt_flags & MNT_NOEXEC) ||
 698                (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
 699  }
 700 +EXPORT_SYMBOL_GPL(path_noexec);
 701
 702  #ifdef CONFIG_USELIB
 703  /*
 704 diff --git a/fs/fcntl.c b/fs/fcntl.c
 705 index c980a0524118..40087d11a974 100644
 706 --- a/fs/fcntl.c
 707 +++ b/fs/fcntl.c
 708 @@ -86,6 +86,7 @@ int setfl(int fd, struct file *filp, unsigned long arg)
 709   out:
 710         return error;
 711  }
 712 +EXPORT_SYMBOL_GPL(setfl);
 713
 714  static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
 715                       int force)
 716 diff --git a/fs/file_table.c b/fs/file_table.c
 717 index 99c6796c9f28..ad533dcedfec 100644
 718 --- a/fs/file_table.c
 719 +++ b/fs/file_table.c
 720 @@ -198,6 +198,7 @@ struct file *alloc_empty_file(int flags, const struct cred *cred)
 721         }
 722         return ERR_PTR(-ENFILE);
 723  }
 724 +EXPORT_SYMBOL_GPL(alloc_empty_file);
 725
 726  /*
 727   * Variant of alloc_empty_file() that doesn't check and modify nr_files.
 728 diff --git a/fs/namespace.c b/fs/namespace.c
 729 index faae7f8644ac..0310c99a83dd 100644
 730 --- a/fs/namespace.c
 731 +++ b/fs/namespace.c
 732 @@ -456,6 +456,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
 733         mnt_dec_writers(real_mount(mnt));
 734         preempt_enable();
 735  }
 736 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
 737
 738  /**
 739   * mnt_drop_write - give up write access to a mount
 740 @@ -857,6 +858,7 @@ int is_current_mnt_ns(struct vfsmount *mnt)
 741  {
 742         return check_mnt(real_mount(mnt));
 743  }
 744 +EXPORT_SYMBOL_GPL(is_current_mnt_ns);
 745
 746  /*
 747   * vfsmount lock must be held for write
 748 @@ -2060,6 +2062,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
 749         }
 750         return 0;
 751  }
 752 +EXPORT_SYMBOL_GPL(iterate_mounts);
 753
 754  static void lock_mnt_tree(struct mount *mnt)
 755  {
 756 diff --git a/fs/notify/group.c b/fs/notify/group.c
 757 index 1de6631a3925..3008eb37a18d 100644
 758 --- a/fs/notify/group.c
 759 +++ b/fs/notify/group.c
 760 @@ -100,6 +100,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
 761  {
 762         refcount_inc(&group->refcnt);
 763  }
 764 +EXPORT_SYMBOL_GPL(fsnotify_get_group);
 765
 766  /*
 767   * Drop a reference to a group.  Free it if it's through.
 768 diff --git a/fs/open.c b/fs/open.c
 769 index cf7e5c350a54..c99195ee46ba 100644
 770 --- a/fs/open.c
 771 +++ b/fs/open.c
 772 @@ -66,6 +66,7 @@ int do_truncate(struct user_namespace *mnt_userns, struct dentry *dentry,
 773         inode_unlock(dentry->d_inode);
 774         return ret;
 775  }
 776 +EXPORT_SYMBOL_GPL(do_truncate);
 777
 778  long vfs_truncate(const struct path *path, loff_t length)
 779  {
 780 diff --git a/fs/read_write.c b/fs/read_write.c
 781 index 328ce8cf9a85..59931848f794 100644
 782 --- a/fs/read_write.c
 783 +++ b/fs/read_write.c
 784 @@ -477,6 +477,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
 785         inc_syscr(current);
 786         return ret;
 787  }
 788 +EXPORT_SYMBOL_GPL(vfs_read);
 789
 790  static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
 791  {
 792 @@ -592,6 +593,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
 793         file_end_write(file);
 794         return ret;
 795  }
 796 +EXPORT_SYMBOL_GPL(vfs_write);
 797
 798  /* file_ppos returns &file->f_pos or NULL if file is stream */
 799  static inline loff_t *file_ppos(struct file *file)
 800 diff --git a/fs/splice.c b/fs/splice.c
 801 index d83f9e5cddbc..ac355ebeb9f0 100644
 802 --- a/fs/splice.c
 803 +++ b/fs/splice.c
 804 @@ -763,6 +763,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
 805                 return warn_unsupported(out, "write");
 806         return out->f_op->splice_write(pipe, out, ppos, len, flags);
 807  }
 808 +EXPORT_SYMBOL_GPL(do_splice_from);
 809
 810  /*
 811   * Attempt to initiate a splice from a file to a pipe.
 812 @@ -792,6 +793,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
 813                 return warn_unsupported(in, "read");
 814         return in->f_op->splice_read(in, ppos, pipe, len, flags);
 815  }
 816 +EXPORT_SYMBOL_GPL(do_splice_to);
 817
 818  /**
 819   * splice_direct_to_actor - splices data directly between two non-pipes
 820 diff --git a/fs/xattr.c b/fs/xattr.c
 821 index a1f4998bc6be..284d4ca54893 100644
 822 --- a/fs/xattr.c
 823 +++ b/fs/xattr.c
 824 @@ -396,6 +396,7 @@ vfs_getxattr_alloc(struct user_namespace *mnt_userns, struct dentry *dentry,
 825         *xattr_value = value;
 826         return error;
 827  }
 828 +EXPORT_SYMBOL_GPL(vfs_getxattr_alloc);
 829
 830  ssize_t
 831  __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
 832 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
 833 index c9f30e12aa32..d04fedf1e4d9 100644
 834 --- a/kernel/locking/lockdep.c
 835 +++ b/kernel/locking/lockdep.c
 836 @@ -238,6 +238,7 @@ inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
 837          */
 838         return lock_classes + class_idx;
 839  }
 840 +EXPORT_SYMBOL_GPL(lockdep_hlock_class);
 841  #define hlock_class(hlock) lockdep_hlock_class(hlock)
 842
 843  #ifdef CONFIG_LOCK_STAT
 844 diff --git a/kernel/task_work.c b/kernel/task_work.c
 845 index dff75bcde151..30b814618535 100644
 846 --- a/kernel/task_work.c
 847 +++ b/kernel/task_work.c
 848 @@ -180,3 +180,4 @@ void task_work_run(void)
 849                 } while (work);
 850         }
 851  }
 852 +EXPORT_SYMBOL_GPL(task_work_run);
 853 diff --git a/security/security.c b/security/security.c
 854 index 4b95de24bc8d..139e1443c49b 100644
 855 --- a/security/security.c
 856 +++ b/security/security.c
 857 @@ -1165,6 +1165,7 @@ int security_path_rmdir(const struct path *dir, struct dentry *dentry)
 858                 return 0;
 859         return call_int_hook(path_rmdir, 0, dir, dentry);
 860  }
 861 +EXPORT_SYMBOL_GPL(security_path_rmdir);
 862
 863  int security_path_unlink(const struct path *dir, struct dentry *dentry)
 864  {
 865 @@ -1181,6 +1182,7 @@ int security_path_symlink(const struct path *dir, struct dentry *dentry,
 866                 return 0;
 867         return call_int_hook(path_symlink, 0, dir, dentry, old_name);
 868  }
 869 +EXPORT_SYMBOL_GPL(security_path_symlink);
 870
 871  int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
 872                        struct dentry *new_dentry)
 873 @@ -1189,6 +1191,7 @@ int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
 874                 return 0;
 875         return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
 876  }
 877 +EXPORT_SYMBOL_GPL(security_path_link);
 878
 879  int security_path_rename(const struct path *old_dir, struct dentry *old_dentry,
 880                          const struct path *new_dir, struct dentry *new_dentry,
 881 @@ -1209,6 +1212,7 @@ int security_path_truncate(const struct path *path)
 882                 return 0;
 883         return call_int_hook(path_truncate, 0, path);
 884  }
 885 +EXPORT_SYMBOL_GPL(security_path_truncate);
 886
 887  int security_path_chmod(const struct path *path, umode_t mode)
 888  {
 889 @@ -1216,6 +1220,7 @@ int security_path_chmod(const struct path *path, umode_t mode)
 890                 return 0;
 891         return call_int_hook(path_chmod, 0, path, mode);
 892  }
 893 +EXPORT_SYMBOL_GPL(security_path_chmod);
 894
 895  int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
 896  {
 897 @@ -1223,6 +1228,7 @@ int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
 898                 return 0;
 899         return call_int_hook(path_chown, 0, path, uid, gid);
 900  }
 901 +EXPORT_SYMBOL_GPL(security_path_chown);
 902
 903  int security_path_chroot(const struct path *path)
 904  {
 905 @@ -1323,6 +1329,7 @@ int security_inode_permission(struct inode *inode, int mask)
 906                 return 0;
 907         return call_int_hook(inode_permission, 0, inode, mask);
 908  }
 909 +EXPORT_SYMBOL_GPL(security_inode_permission);
 910
 911  int security_inode_setattr(struct user_namespace *mnt_userns,
 912                            struct dentry *dentry, struct iattr *attr)
 913 @@ -1521,6 +1528,7 @@ int security_file_permission(struct file *file, int mask)
 914
 915         return fsnotify_perm(file, mask);
 916  }
 917 +EXPORT_SYMBOL_GPL(security_file_permission);
 918
 919  int security_file_alloc(struct file *file)
 920  {
 921 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
 922 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs     1970-01-01 01:00:00.000000000 +0100
 923 +++ linux/Documentation/ABI/testing/debugfs-aufs        2022-11-05 23:02:18.955889283 +0100
 924 @@ -0,0 +1,55 @@
 925 +What:          /debug/aufs/si_<id>/
 926 +Date:          March 2009
 927 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 928 +Description:
 929 +               Under /debug/aufs, a directory named si_<id> is created
 930 +               per aufs mount, where <id> is a unique id generated
 931 +               internally.
 932 +
 933 +What:          /debug/aufs/si_<id>/plink
 934 +Date:          Apr 2013
 935 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 936 +Description:
 937 +               It has three lines and shows the information about the
 938 +               pseudo-link. The first line is a single number
 939 +               representing a number of buckets. The second line is a
 940 +               number of pseudo-links per buckets (separated by a
 941 +               blank). The last line is a single number representing a
 942 +               total number of psedo-links.
 943 +               When the aufs mount option 'noplink' is specified, it
 944 +               will show "1\n0\n0\n".
 945 +
 946 +What:          /debug/aufs/si_<id>/xib
 947 +Date:          March 2009
 948 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 949 +Description:
 950 +               It shows the consumed blocks by xib (External Inode Number
 951 +               Bitmap), its block size and file size.
 952 +               When the aufs mount option 'noxino' is specified, it
 953 +               will be empty. About XINO files, see the aufs manual.
 954 +
 955 +What:          /debug/aufs/si_<id>/xi<branch-index>
 956 +Date:          March 2009
 957 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 958 +Description:
 959 +               It shows the consumed blocks by xino (External Inode Number
 960 +               Translation Table), its link count, block size and file
 961 +               size.
 962 +               Due to the file size limit, there may exist multiple
 963 +               xino files per branch.  In this case, "-N" is added to
 964 +               the filename and it corresponds to the index of the
 965 +               internal xino array.  "-0" is omitted.
 966 +               When the aufs mount option 'noxino' is specified, Those
 967 +               entries won't exist.  About XINO files, see the aufs
 968 +               manual.
 969 +
 970 +What:          /debug/aufs/si_<id>/xigen
 971 +Date:          March 2009
 972 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 973 +Description:
 974 +               It shows the consumed blocks by xigen (External Inode
 975 +               Generation Table), its block size and file size.
 976 +               If CONFIG_AUFS_EXPORT is disabled, this entry will not
 977 +               be created.
 978 +               When the aufs mount option 'noxino' is specified, it
 979 +               will be empty. About XINO files, see the aufs manual.
 980 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
 981 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs       1970-01-01 01:00:00.000000000 +0100
 982 +++ linux/Documentation/ABI/testing/sysfs-aufs  2022-11-05 23:02:18.955889283 +0100
 983 @@ -0,0 +1,31 @@
 984 +What:          /sys/fs/aufs/si_<id>/
 985 +Date:          March 2009
 986 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 987 +Description:
 988 +               Under /sys/fs/aufs, a directory named si_<id> is created
 989 +               per aufs mount, where <id> is a unique id generated
 990 +               internally.
 991 +
 992 +What:          /sys/fs/aufs/si_<id>/br<idx>
 993 +Date:          March 2009
 994 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
 995 +Description:
 996 +               It shows the abolute path of a member directory (which
 997 +               is called branch) in aufs, and its permission.
 998 +
 999 +What:          /sys/fs/aufs/si_<id>/brid<idx>
1000 +Date:          July 2013
1001 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
1002 +Description:
1003 +               It shows the id of a member directory (which is called
1004 +               branch) in aufs.
1005 +
1006 +What:          /sys/fs/aufs/si_<id>/xi_path
1007 +Date:          March 2009
1008 +Contact:       J. R. Okajima <hooanon05g@gmail.com>
1009 +Description:
1010 +               It shows the abolute path of XINO (External Inode Number
1011 +               Bitmap, Translation Table and Generation Table) file
1012 +               even if it is the default path.
1013 +               When the aufs mount option 'noxino' is specified, it
1014 +               will be empty. About XINO files, see the aufs manual.
1015 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1016 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt  1970-01-01 01:00:00.000000000 +0100
1017 +++ linux/Documentation/filesystems/aufs/design/01intro.txt     2022-11-05 23:02:18.955889283 +0100
1018 @@ -0,0 +1,171 @@
1019 +
1020 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1021 +#
1022 +# This program is free software; you can redistribute it and/or modify
1023 +# it under the terms of the GNU General Public License as published by
1024 +# the Free Software Foundation; either version 2 of the License, or
1025 +# (at your option) any later version.
1026 +#
1027 +# This program is distributed in the hope that it will be useful,
1028 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1029 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1030 +# GNU General Public License for more details.
1031 +#
1032 +# You should have received a copy of the GNU General Public License
1033 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1034 +
1035 +Introduction
1036 +----------------------------------------
1037 +
1038 +aufs [ei ju: ef es] | /ey-yoo-ef-es/ | [a u f s]
1039 +1. abbrev. for "advanced multi-layered unification filesystem".
1040 +2. abbrev. for "another unionfs".
1041 +3. abbrev. for "auf das" in German which means "on the" in English.
1042 +   Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1043 +   But "Filesystem aufs Filesystem" is hard to understand.
1044 +4. abbrev. for "African Urban Fashion Show".
1045 +
1046 +AUFS is a filesystem with features:
1047 +- multi layered stackable unification filesystem, the member directory
1048 +  is called as a branch.
1049 +- branch permission and attribute, 'readonly', 'real-readonly',
1050 +  'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1051 +  combination.
1052 +- internal "file copy-on-write".
1053 +- logical deletion, whiteout.
1054 +- dynamic branch manipulation, adding, deleting and changing permission.
1055 +- allow bypassing aufs, user's direct branch access.
1056 +- external inode number translation table and bitmap which maintains the
1057 +  persistent aufs inode number.
1058 +- seekable directory, including NFS readdir.
1059 +- file mapping, mmap and sharing pages.
1060 +- pseudo-link, hardlink over branches.
1061 +- loopback mounted filesystem as a branch.
1062 +- several policies to select one among multiple writable branches.
1063 +- revert a single systemcall when an error occurs in aufs.
1064 +- and more...
1065 +
1066 +
1067 +Multi Layered Stackable Unification Filesystem
1068 +----------------------------------------------------------------------
1069 +Most people already knows what it is.
1070 +It is a filesystem which unifies several directories and provides a
1071 +merged single directory. When users access a file, the access will be
1072 +passed/re-directed/converted (sorry, I am not sure which English word is
1073 +correct) to the real file on the member filesystem. The member
1074 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1075 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1076 +readonly branch is handled by creating 'whiteout' on the upper writable
1077 +branch.
1078 +
1079 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1080 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1081 +different approaches to implement the merged-view.
1082 +The former tries putting it into VFS, and the latter implements as a
1083 +separate filesystem.
1084 +(If I misunderstand about these implementations, please let me know and
1085 +I shall correct it. Because it is a long time ago when I read their
1086 +source files last time).
1087 +
1088 +UnionMount's approach will be able to small, but may be hard to share
1089 +branches between several UnionMount since the whiteout in it is
1090 +implemented in the inode on branch filesystem and always
1091 +shared. According to Bharata's post, readdir does not seems to be
1092 +finished yet.
1093 +There are several missing features known in this implementations such as
1094 +- for users, the inode number may change silently. eg. copy-up.
1095 +- link(2) may break by copy-up.
1096 +- read(2) may get an obsoleted filedata (fstat(2) too).
1097 +- fcntl(F_SETLK) may be broken by copy-up.
1098 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1099 +  open(O_RDWR).
1100 +
1101 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1102 +merged into mainline. This is another implementation of UnionMount as a
1103 +separated filesystem. All the limitations and known problems which
1104 +UnionMount are equally inherited to "overlay" filesystem.
1105 +
1106 +Unionfs has a longer history. When I started implementing a stackable
1107 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1108 +inode, dentry and file objects and they have an array pointing lower
1109 +same kind objects. After contributing many patches for Unionfs, I
1110 +re-started my project AUFS (Jun 2006).
1111 +
1112 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1113 +implemented my own ideas, approaches and enhancements and it became
1114 +totally different one.
1115 +
1116 +Comparing DM snapshot and fs based implementation
1117 +- the number of bytes to be copied between devices is much smaller.
1118 +- the type of filesystem must be one and only.
1119 +- the fs must be writable, no readonly fs, even for the lower original
1120 +  device. so the compression fs will not be usable. but if we use
1121 +  loopback mount, we may address this issue.
1122 +  for instance,
1123 +       mount /cdrom/squashfs.img /sq
1124 +       losetup /sq/ext2.img
1125 +       losetup /somewhere/cow
1126 +       dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1127 +- it will be difficult (or needs more operations) to extract the
1128 +  difference between the original device and COW.
1129 +- DM snapshot-merge may help a lot when users try merging. in the
1130 +  fs-layer union, users will use rsync(1).
1131 +
1132 +You may want to read my old paper "Filesystems in LiveCD"
1133 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1134 +
1135 +
1136 +Several characters/aspects/persona of aufs
1137 +----------------------------------------------------------------------
1138 +
1139 +Aufs has several characters, aspects or persona.
1140 +1. a filesystem, callee of VFS helper
1141 +2. sub-VFS, caller of VFS helper for branches
1142 +3. a virtual filesystem which maintains persistent inode number
1143 +4. reader/writer of files on branches such like an application
1144 +
1145 +1. Callee of VFS Helper
1146 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1147 +unlink(2) from an application reaches sys_unlink() kernel function and
1148 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1149 +calls filesystem specific unlink operation. Actually aufs implements the
1150 +unlink operation but it behaves like a redirector.
1151 +
1152 +2. Caller of VFS Helper for Branches
1153 +aufs_unlink() passes the unlink request to the branch filesystem as if
1154 +it were called from VFS. So the called unlink operation of the branch
1155 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1156 +every necessary pre/post operation for the branch filesystem.
1157 +- acquire the lock for the parent dir on a branch
1158 +- lookup in a branch
1159 +- revalidate dentry on a branch
1160 +- mnt_want_write() for a branch
1161 +- vfs_unlink() for a branch
1162 +- mnt_drop_write() for a branch
1163 +- release the lock on a branch
1164 +
1165 +3. Persistent Inode Number
1166 +One of the most important issue for a filesystem is to maintain inode
1167 +numbers. This is particularly important to support exporting a
1168 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1169 +backend block device for its own. But some storage is necessary to
1170 +keep and maintain the inode numbers. It may be a large space and may not
1171 +suit to keep in memory. Aufs rents some space from its first writable
1172 +branch filesystem (by default) and creates file(s) on it. These files
1173 +are created by aufs internally and removed soon (currently) keeping
1174 +opened.
1175 +Note: Because these files are removed, they are totally gone after
1176 +      unmounting aufs. It means the inode numbers are not persistent
1177 +      across unmount or reboot. I have a plan to make them really
1178 +      persistent which will be important for aufs on NFS server.
1179 +
1180 +4. Read/Write Files Internally (copy-on-write)
1181 +Because a branch can be readonly, when you write a file on it, aufs will
1182 +"copy-up" it to the upper writable branch internally. And then write the
1183 +originally requested thing to the file. Generally kernel doesn't
1184 +open/read/write file actively. In aufs, even a single write may cause a
1185 +internal "file copy". This behaviour is very similar to cp(1) command.
1186 +
1187 +Some people may think it is better to pass such work to user space
1188 +helper, instead of doing in kernel space. Actually I am still thinking
1189 +about it. But currently I have implemented it in kernel space.
1190 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1191 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1192 +++ linux/Documentation/filesystems/aufs/design/02struct.txt    2022-11-05 23:02:18.955889283 +0100
1193 @@ -0,0 +1,258 @@
1194 +
1195 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1196 +#
1197 +# This program is free software; you can redistribute it and/or modify
1198 +# it under the terms of the GNU General Public License as published by
1199 +# the Free Software Foundation; either version 2 of the License, or
1200 +# (at your option) any later version.
1201 +#
1202 +# This program is distributed in the hope that it will be useful,
1203 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1204 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1205 +# GNU General Public License for more details.
1206 +#
1207 +# You should have received a copy of the GNU General Public License
1208 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1209 +
1210 +Basic Aufs Internal Structure
1211 +
1212 +Superblock/Inode/Dentry/File Objects
1213 +----------------------------------------------------------------------
1214 +As like an ordinary filesystem, aufs has its own
1215 +superblock/inode/dentry/file objects. All these objects have a
1216 +dynamically allocated array and store the same kind of pointers to the
1217 +lower filesystem, branch.
1218 +For example, when you build a union with one readwrite branch and one
1219 +readonly, mounted /au, /rw and /ro respectively.
1220 +- /au = /rw + /ro
1221 +- /ro/fileA exists but /rw/fileA
1222 +
1223 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1224 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1225 +- [0] = NULL (because /rw/fileA doesn't exist)
1226 +- [1] = /ro/fileA
1227 +
1228 +This style of an array is essentially same to the aufs
1229 +superblock/inode/dentry/file objects.
1230 +
1231 +Because aufs supports manipulating branches, ie. add/delete/change
1232 +branches dynamically, these objects has its own generation. When
1233 +branches are changed, the generation in aufs superblock is
1234 +incremented. And a generation in other object are compared when it is
1235 +accessed. When a generation in other objects are obsoleted, aufs
1236 +refreshes the internal array.
1237 +
1238 +
1239 +Superblock
1240 +----------------------------------------------------------------------
1241 +Additionally aufs superblock has some data for policies to select one
1242 +among multiple writable branches, XIB files, pseudo-links and kobject.
1243 +See below in detail.
1244 +About the policies which supports copy-down a directory, see
1245 +wbr_policy.txt too.
1246 +
1247 +
1248 +Branch and XINO(External Inode Number Translation Table)
1249 +----------------------------------------------------------------------
1250 +Every branch has its own xino (external inode number translation table)
1251 +file. The xino file is created and unlinked by aufs internally. When two
1252 +members of a union exist on the same filesystem, they share the single
1253 +xino file.
1254 +The struct of a xino file is simple, just a sequence of aufs inode
1255 +numbers which is indexed by the lower inode number.
1256 +In the above sample, assume the inode number of /ro/fileA is i111 and
1257 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1258 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1259 +
1260 +When the inode numbers are not contiguous, the xino file will be sparse
1261 +which has a hole in it and doesn't consume as much disk space as it
1262 +might appear. If your branch filesystem consumes disk space for such
1263 +holes, then you should specify 'xino=' option at mounting aufs.
1264 +
1265 +Aufs has a mount option to free the disk blocks for such holes in XINO
1266 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1267 +meet a problem of disk shortage due to XINO files, then you should try
1268 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1269 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1270 +the holes in XINO files.
1271 +
1272 +Also a writable branch has three kinds of "whiteout bases". All these
1273 +are existed when the branch is joined to aufs, and their names are
1274 +whiteout-ed doubly, so that users will never see their names in aufs
1275 +hierarchy.
1276 +1. a regular file which will be hardlinked to all whiteouts.
1277 +2. a directory to store a pseudo-link.
1278 +3. a directory to store an "orphan"-ed file temporary.
1279 +
1280 +1. Whiteout Base
1281 +   When you remove a file on a readonly branch, aufs handles it as a
1282 +   logical deletion and creates a whiteout on the upper writable branch
1283 +   as a hardlink of this file in order not to consume inode on the
1284 +   writable branch.
1285 +2. Pseudo-link Dir
1286 +   See below, Pseudo-link.
1287 +3. Step-Parent Dir
1288 +   When "fileC" exists on the lower readonly branch only and it is
1289 +   opened and removed with its parent dir, and then user writes
1290 +   something into it, then aufs copies-up fileC to this
1291 +   directory. Because there is no other dir to store fileC. After
1292 +   creating a file under this dir, the file is unlinked.
1293 +
1294 +Because aufs supports manipulating branches, ie. add/delete/change
1295 +dynamically, a branch has its own id. When the branch order changes,
1296 +aufs finds the new index by searching the branch id.
1297 +
1298 +
1299 +Pseudo-link
1300 +----------------------------------------------------------------------
1301 +Assume "fileA" exists on the lower readonly branch only and it is
1302 +hardlinked to "fileB" on the branch. When you write something to fileA,
1303 +aufs copies-up it to the upper writable branch. Additionally aufs
1304 +creates a hardlink under the Pseudo-link Directory of the writable
1305 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1306 +simple list. If fileB is read after unlinking fileA, aufs returns
1307 +filedata from the pseudo-link instead of the lower readonly
1308 +branch. Because the pseudo-link is based upon the inode, to keep the
1309 +inode number by xino (see above) is essentially necessary.
1310 +
1311 +All the hardlinks under the Pseudo-link Directory of the writable branch
1312 +should be restored in a proper location later. Aufs provides a utility
1313 +to do this. The userspace helpers executed at remounting and unmounting
1314 +aufs by default.
1315 +During this utility is running, it puts aufs into the pseudo-link
1316 +maintenance mode. In this mode, only the process which began the
1317 +maintenance mode (and its child processes) is allowed to operate in
1318 +aufs. Some other processes which are not related to the pseudo-link will
1319 +be allowed to run too, but the rest have to return an error or wait
1320 +until the maintenance mode ends. If a process already acquires an inode
1321 +mutex (in VFS), it has to return an error.
1322 +
1323 +
1324 +XIB(external inode number bitmap)
1325 +----------------------------------------------------------------------
1326 +Addition to the xino file per a branch, aufs has an external inode number
1327 +bitmap in a superblock object. It is also an internal file such like a
1328 +xino file.
1329 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1330 +not.
1331 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1332 +
1333 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1334 +reduce the number of consumed disk blocks for these files.
1335 +
1336 +
1337 +Virtual or Vertical Dir, and Readdir in Userspace
1338 +----------------------------------------------------------------------
1339 +In order to support multiple layers (branches), aufs readdir operation
1340 +constructs a virtual dir block on memory. For readdir, aufs calls
1341 +vfs_readdir() internally for each dir on branches, merges their entries
1342 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1343 +object. So the file object has its entry list until it is closed. The
1344 +entry list will be updated when the file position is zero and becomes
1345 +obsoleted. This decision is made in aufs automatically.
1346 +
1347 +The dynamically allocated memory block for the name of entries has a
1348 +unit of 512 bytes (by default) and stores the names contiguously (no
1349 +padding). Another block for each entry is handled by kmem_cache too.
1350 +During building dir blocks, aufs creates hash list and judging whether
1351 +the entry is whiteouted by its upper branch or already listed.
1352 +The merged result is cached in the corresponding inode object and
1353 +maintained by a customizable life-time option.
1354 +
1355 +Some people may call it can be a security hole or invite DoS attack
1356 +since the opened and once readdir-ed dir (file object) holds its entry
1357 +list and becomes a pressure for system memory. But I'd say it is similar
1358 +to files under /proc or /sys. The virtual files in them also holds a
1359 +memory page (generally) while they are opened. When an idea to reduce
1360 +memory for them is introduced, it will be applied to aufs too.
1361 +For those who really hate this situation, I've developed readdir(3)
1362 +library which operates this merging in userspace. You just need to set
1363 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1364 +kernel space for readdir(3).
1365 +
1366 +
1367 +Workqueue
1368 +----------------------------------------------------------------------
1369 +Aufs sometimes requires privilege access to a branch. For instance,
1370 +in copy-up/down operation. When a user process is going to make changes
1371 +to a file which exists in the lower readonly branch only, and the mode
1372 +of one of ancestor directories may not be writable by a user
1373 +process. Here aufs copy-up the file with its ancestors and they may
1374 +require privilege to set its owner/group/mode/etc.
1375 +This is a typical case of a application character of aufs (see
1376 +Introduction).
1377 +
1378 +Aufs uses workqueue synchronously for this case. It creates its own
1379 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1380 +passes the request to call mkdir or write (for example), and wait for
1381 +its completion. This approach solves a problem of a signal handler
1382 +simply.
1383 +If aufs didn't adopt the workqueue and changed the privilege of the
1384 +process, then the process may receive the unexpected SIGXFSZ or other
1385 +signals.
1386 +
1387 +Also aufs uses the system global workqueue ("events" kernel thread) too
1388 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1389 +whiteout base and etc. This is unrelated to a privilege.
1390 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1391 +superblock at the beginning, at the same time waits for the completion
1392 +of all queued asynchronous tasks.
1393 +
1394 +
1395 +Whiteout
1396 +----------------------------------------------------------------------
1397 +The whiteout in aufs is very similar to Unionfs's. That is represented
1398 +by its filename. UnionMount takes an approach of a file mode, but I am
1399 +afraid several utilities (find(1) or something) will have to support it.
1400 +
1401 +Basically the whiteout represents "logical deletion" which stops aufs to
1402 +lookup further, but also it represents "dir is opaque" which also stop
1403 +further lookup.
1404 +
1405 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1406 +In order to make several functions in a single systemcall to be
1407 +revertible, aufs adopts an approach to rename a directory to a temporary
1408 +unique whiteouted name.
1409 +For example, in rename(2) dir where the target dir already existed, aufs
1410 +renames the target dir to a temporary unique whiteouted name before the
1411 +actual rename on a branch, and then handles other actions (make it opaque,
1412 +update the attributes, etc). If an error happens in these actions, aufs
1413 +simply renames the whiteouted name back and returns an error. If all are
1414 +succeeded, aufs registers a function to remove the whiteouted unique
1415 +temporary name completely and asynchronously to the system global
1416 +workqueue.
1417 +
1418 +
1419 +Copy-up
1420 +----------------------------------------------------------------------
1421 +It is a well-known feature or concept.
1422 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1423 +internally and makes change to the new file on the upper writable branch.
1424 +When the trigger systemcall does not update the timestamps of the parent
1425 +dir, aufs reverts it after copy-up.
1426 +
1427 +
1428 +Move-down (aufs3.9 and later)
1429 +----------------------------------------------------------------------
1430 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1431 +the lower readonly branch to the upper writable branch when a user
1432 +changes something about the file.
1433 +"Move-down" is an opposite action of copy-up. Basically this action is
1434 +ran manually instead of automatically and internally.
1435 +For desgin and implementation, aufs has to consider these issues.
1436 +- whiteout for the file may exist on the lower branch.
1437 +- ancestor directories may not exist on the lower branch.
1438 +- diropq for the ancestor directories may exist on the upper branch.
1439 +- free space on the lower branch will reduce.
1440 +- another access to the file may happen during moving-down, including
1441 +  UDBA (see "Revalidate Dentry and UDBA").
1442 +- the file should not be hard-linked nor pseudo-linked. they should be
1443 +  handled by auplink utility later.
1444 +
1445 +Sometimes users want to move-down a file from the upper writable branch
1446 +to the lower readonly or writable branch. For instance,
1447 +- the free space of the upper writable branch is going to run out.
1448 +- create a new intermediate branch between the upper and lower branch.
1449 +- etc.
1450 +
1451 +For this purpose, use "aumvdown" command in aufs-util.git.
1452 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1453 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt    1970-01-01 01:00:00.000000000 +0100
1454 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt       2022-11-05 23:02:18.955889283 +0100
1455 @@ -0,0 +1,85 @@
1456 +
1457 +# Copyright (C) 2015-2022 Junjiro R. Okajima
1458 +#
1459 +# This program is free software; you can redistribute it and/or modify
1460 +# it under the terms of the GNU General Public License as published by
1461 +# the Free Software Foundation; either version 2 of the License, or
1462 +# (at your option) any later version.
1463 +#
1464 +# This program is distributed in the hope that it will be useful,
1465 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1466 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1467 +# GNU General Public License for more details.
1468 +#
1469 +# You should have received a copy of the GNU General Public License
1470 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1471 +
1472 +Support for a branch who has its ->atomic_open()
1473 +----------------------------------------------------------------------
1474 +The filesystems who implement its ->atomic_open() are not majority. For
1475 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1476 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1477 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1478 +sure whether all filesystems who have ->atomic_open() behave like this,
1479 +but NFSv4 surely returns the error.
1480 +
1481 +In order to support ->atomic_open() for aufs, there are a few
1482 +approaches.
1483 +
1484 +A. Introduce aufs_atomic_open()
1485 +   - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1486 +     branch fs.
1487 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1488 +   an aufs user Pip Cet's approach
1489 +   - calls aufs_create(), VFS finish_open() and notify_change().
1490 +   - pass fake-mode to finish_open(), and then correct the mode by
1491 +     notify_change().
1492 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1493 +   - no aufs_atomic_open().
1494 +   - aufs_lookup() registers the TID to an aufs internal object.
1495 +   - aufs_create() does nothing when the matching TID is registered, but
1496 +     registers the mode.
1497 +   - aufs_open() calls branch fs's ->atomic_open() when the matching
1498 +     TID is registered.
1499 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1500 +   credential
1501 +   - no aufs_atomic_open().
1502 +   - aufs_create() registers the TID to an internal object. this info
1503 +     represents "this process created this file just now."
1504 +   - when aufs gets EACCES from branch fs's ->open(), then confirm the
1505 +     registered TID and re-try open() with superuser's credential.
1506 +
1507 +Pros and cons for each approach.
1508 +
1509 +A.
1510 +   - straightforward but highly depends upon VFS internal.
1511 +   - the atomic behavaiour is kept.
1512 +   - some of parameters such as nameidata are hard to reproduce for
1513 +     branch fs.
1514 +   - large overhead.
1515 +B.
1516 +   - easy to implement.
1517 +   - the atomic behavaiour is lost.
1518 +C.
1519 +   - the atomic behavaiour is kept.
1520 +   - dirty and tricky.
1521 +   - VFS checks whether the file is created correctly after calling
1522 +     ->create(), which means this approach doesn't work.
1523 +D.
1524 +   - easy to implement.
1525 +   - the atomic behavaiour is lost.
1526 +   - to open a file with superuser's credential and give it to a user
1527 +     process is a bad idea, since the file object keeps the credential
1528 +     in it. It may affect LSM or something. This approach doesn't work
1529 +     either.
1530 +
1531 +The approach A is ideal, but it hard to implement. So here is a
1532 +variation of A, which is to be implemented.
1533 +
1534 +A-1. Introduce aufs_atomic_open()
1535 +     - calls branch fs ->atomic_open() if exists. otherwise calls
1536 +       vfs_create() and finish_open().
1537 +     - the demerit is that the several checks after branch fs
1538 +       ->atomic_open() are lost. in the ordinary case, the checks are
1539 +       done by VFS:do_last(), lookup_open() and atomic_open(). some can
1540 +       be implemented in aufs, but not all I am afraid.
1541 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1542 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1543 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt    2022-11-05 23:02:18.959222617 +0100
1544 @@ -0,0 +1,113 @@
1545 +
1546 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1547 +#
1548 +# This program is free software; you can redistribute it and/or modify
1549 +# it under the terms of the GNU General Public License as published by
1550 +# the Free Software Foundation; either version 2 of the License, or
1551 +# (at your option) any later version.
1552 +#
1553 +# This program is distributed in the hope that it will be useful,
1554 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1555 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1556 +# GNU General Public License for more details.
1557 +#
1558 +# You should have received a copy of the GNU General Public License
1559 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1560 +
1561 +Lookup in a Branch
1562 +----------------------------------------------------------------------
1563 +Since aufs has a character of sub-VFS (see Introduction), it operates
1564 +lookup for branches as VFS does. It may be a heavy work. But almost all
1565 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1566 +directly connected to its parent. Digging down the directory hierarchy
1567 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1568 +aufs calls it.
1569 +
1570 +When a branch is a remote filesystem, aufs basically relies upon its
1571 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1572 +them.
1573 +For d_revalidate, aufs implements three levels of revalidate tests. See
1574 +"Revalidate Dentry and UDBA" in detail.
1575 +
1576 +
1577 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1578 +----------------------------------------------------------------------
1579 +Let's try case study.
1580 +- aufs has two branches, upper readwrite and lower readonly.
1581 +  /au = /rw + /ro
1582 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1583 +- user invoked "chmod a+rx /au/dirA"
1584 +- the internal copy-up is activated and "/rw/dirA" is created and its
1585 +  permission bits are set to world readable.
1586 +- then "/au/dirA" becomes world readable?
1587 +
1588 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1589 +or it may be a natively readonly filesystem. If aufs respects the lower
1590 +branch, it should not respond readdir request from other users. But user
1591 +allowed it by chmod. Should really aufs rejects showing the entries
1592 +under /ro/dirA?
1593 +
1594 +To be honest, I don't have a good solution for this case. So aufs
1595 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1596 +users.
1597 +When dirperm1 is specified, aufs checks only the highest one for the
1598 +directory permission, and shows the entries. Otherwise, as usual, checks
1599 +every dir existing on all branches and rejects the request.
1600 +
1601 +As a side effect, dirperm1 option improves the performance of aufs
1602 +because the number of permission check is reduced when the number of
1603 +branch is many.
1604 +
1605 +
1606 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1607 +----------------------------------------------------------------------
1608 +Generally VFS helpers re-validate a dentry as a part of lookup.
1609 +0. digging down the directory hierarchy.
1610 +1. lock the parent dir by its i_mutex.
1611 +2. lookup the final (child) entry.
1612 +3. revalidate it.
1613 +4. call the actual operation (create, unlink, etc.)
1614 +5. unlock the parent dir
1615 +
1616 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1617 +called. Actually aufs implements it and checks the dentry on a branch is
1618 +still valid.
1619 +But it is not enough. Because aufs has to release the lock for the
1620 +parent dir on a branch at the end of ->lookup() (step 2) and
1621 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1622 +held by VFS.
1623 +If the file on a branch is changed directly, eg. bypassing aufs, after
1624 +aufs released the lock, then the subsequent operation may cause
1625 +something unpleasant result.
1626 +
1627 +This situation is a result of VFS architecture, ->lookup() and
1628 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1629 +design from VFS's point of view. It is just not suitable for sub-VFS
1630 +character in aufs.
1631 +
1632 +Aufs supports such case by three level of revalidation which is
1633 +selectable by user.
1634 +1. Simple Revalidate
1635 +   Addition to the native flow in VFS's, confirm the child-parent
1636 +   relationship on the branch just after locking the parent dir on the
1637 +   branch in the "actual operation" (step 4). When this validation
1638 +   fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1639 +   checks the validation of the dentry on branches.
1640 +2. Monitor Changes Internally by Inotify/Fsnotify
1641 +   Addition to above, in the "actual operation" (step 4) aufs re-lookup
1642 +   the dentry on the branch, and returns EBUSY if it finds different
1643 +   dentry.
1644 +   Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1645 +   during it is in cache. When the event is notified, aufs registers a
1646 +   function to kernel 'events' thread by schedule_work(). And the
1647 +   function sets some special status to the cached aufs dentry and inode
1648 +   private data. If they are not cached, then aufs has nothing to
1649 +   do. When the same file is accessed through aufs (step 0-3) later,
1650 +   aufs will detect the status and refresh all necessary data.
1651 +   In this mode, aufs has to ignore the event which is fired by aufs
1652 +   itself.
1653 +3. No Extra Validation
1654 +   This is the simplest test and doesn't add any additional revalidation
1655 +   test, and skip the revalidation in step 4. It is useful and improves
1656 +   aufs performance when system surely hide the aufs branches from user,
1657 +   by over-mounting something (or another method).
1658 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1659 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1660 +++ linux/Documentation/filesystems/aufs/design/04branch.txt    2022-11-05 23:02:18.959222617 +0100
1661 @@ -0,0 +1,74 @@
1662 +
1663 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1664 +#
1665 +# This program is free software; you can redistribute it and/or modify
1666 +# it under the terms of the GNU General Public License as published by
1667 +# the Free Software Foundation; either version 2 of the License, or
1668 +# (at your option) any later version.
1669 +#
1670 +# This program is distributed in the hope that it will be useful,
1671 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1672 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1673 +# GNU General Public License for more details.
1674 +#
1675 +# You should have received a copy of the GNU General Public License
1676 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1677 +
1678 +Branch Manipulation
1679 +
1680 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1681 +and changing its permission/attribute, there are a lot of works to do.
1682 +
1683 +
1684 +Add a Branch
1685 +----------------------------------------------------------------------
1686 +o Confirm the adding dir exists outside of aufs, including loopback
1687 +  mount, and its various attributes.
1688 +o Initialize the xino file and whiteout bases if necessary.
1689 +  See struct.txt.
1690 +
1691 +o Check the owner/group/mode of the directory
1692 +  When the owner/group/mode of the adding directory differs from the
1693 +  existing branch, aufs issues a warning because it may impose a
1694 +  security risk.
1695 +  For example, when a upper writable branch has a world writable empty
1696 +  top directory, a malicious user can create any files on the writable
1697 +  branch directly, like copy-up and modify manually. If something like
1698 +  /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1699 +  writable branch, and the writable branch is world-writable, then a
1700 +  malicious guy may create /etc/passwd on the writable branch directly
1701 +  and the infected file will be valid in aufs.
1702 +  I am afraid it can be a security issue, but aufs can do nothing except
1703 +  producing a warning.
1704 +
1705 +
1706 +Delete a Branch
1707 +----------------------------------------------------------------------
1708 +o Confirm the deleting branch is not busy
1709 +  To be general, there is one merit to adopt "remount" interface to
1710 +  manipulate branches. It is to discard caches. At deleting a branch,
1711 +  aufs checks the still cached (and connected) dentries and inodes. If
1712 +  there are any, then they are all in-use. An inode without its
1713 +  corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1714 +
1715 +  For the cached one, aufs checks whether the same named entry exists on
1716 +  other branches.
1717 +  If the cached one is a directory, because aufs provides a merged view
1718 +  to users, as long as one dir is left on any branch aufs can show the
1719 +  dir to users. In this case, the branch can be removed from aufs.
1720 +  Otherwise aufs rejects deleting the branch.
1721 +
1722 +  If any file on the deleting branch is opened by aufs, then aufs
1723 +  rejects deleting.
1724 +
1725 +
1726 +Modify the Permission of a Branch
1727 +----------------------------------------------------------------------
1728 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1729 +  See struct.txt.
1730 +
1731 +o rw --> ro: Confirm the modifying branch is not busy
1732 +  Aufs rejects the request if any of these conditions are true.
1733 +  - a file on the branch is mmap-ed.
1734 +  - a regular file on the branch is opened for write and there is no
1735 +    same named entry on the upper branch.
1736 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1737 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt     1970-01-01 01:00:00.000000000 +0100
1738 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt        2022-11-05 23:02:18.959222617 +0100
1739 @@ -0,0 +1,64 @@
1740 +
1741 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1742 +#
1743 +# This program is free software; you can redistribute it and/or modify
1744 +# it under the terms of the GNU General Public License as published by
1745 +# the Free Software Foundation; either version 2 of the License, or
1746 +# (at your option) any later version.
1747 +#
1748 +# This program is distributed in the hope that it will be useful,
1749 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1750 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1751 +# GNU General Public License for more details.
1752 +#
1753 +# You should have received a copy of the GNU General Public License
1754 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1755 +
1756 +Policies to Select One among Multiple Writable Branches
1757 +----------------------------------------------------------------------
1758 +When the number of writable branch is more than one, aufs has to decide
1759 +the target branch for file creation or copy-up. By default, the highest
1760 +writable branch which has the parent (or ancestor) dir of the target
1761 +file is chosen (top-down-parent policy).
1762 +By user's request, aufs implements some other policies to select the
1763 +writable branch, for file creation several policies, round-robin,
1764 +most-free-space, and other policies. For copy-up, top-down-parent,
1765 +bottom-up-parent, bottom-up and others.
1766 +
1767 +As expected, the round-robin policy selects the branch in circular. When
1768 +you have two writable branches and creates 10 new files, 5 files will be
1769 +created for each branch. mkdir(2) systemcall is an exception. When you
1770 +create 10 new directories, all will be created on the same branch.
1771 +And the most-free-space policy selects the one which has most free
1772 +space among the writable branches. The amount of free space will be
1773 +checked by aufs internally, and users can specify its time interval.
1774 +
1775 +The policies for copy-up is more simple,
1776 +top-down-parent is equivalent to the same named on in create policy,
1777 +bottom-up-parent selects the writable branch where the parent dir
1778 +exists and the nearest upper one from the copyup-source,
1779 +bottom-up selects the nearest upper writable branch from the
1780 +copyup-source, regardless the existence of the parent dir.
1781 +
1782 +There are some rules or exceptions to apply these policies.
1783 +- If there is a readonly branch above the policy-selected branch and
1784 +  the parent dir is marked as opaque (a variation of whiteout), or the
1785 +  target (creating) file is whiteout-ed on the upper readonly branch,
1786 +  then the result of the policy is ignored and the target file will be
1787 +  created on the nearest upper writable branch than the readonly branch.
1788 +- If there is a writable branch above the policy-selected branch and
1789 +  the parent dir is marked as opaque or the target file is whiteouted
1790 +  on the branch, then the result of the policy is ignored and the target
1791 +  file will be created on the highest one among the upper writable
1792 +  branches who has diropq or whiteout. In case of whiteout, aufs removes
1793 +  it as usual.
1794 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1795 +  They try selecting the branch where the source exists as possible
1796 +  since copyup a large file will take long time. If it can't be,
1797 +  ie. the branch where the source exists is readonly, then they will
1798 +  follow the copyup policy.
1799 +- There is an exception for rename(2) when the target exists.
1800 +  If the rename target exists, aufs compares the index of the branches
1801 +  where the source and the target exists and selects the higher
1802 +  one. If the selected branch is readonly, then aufs follows the
1803 +  copyup policy.
1804 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot linux/Documentation/filesystems/aufs/design/06dirren.dot
1805 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot 1970-01-01 01:00:00.000000000 +0100
1806 +++ linux/Documentation/filesystems/aufs/design/06dirren.dot    2022-11-05 23:02:18.959222617 +0100
1807 @@ -0,0 +1,44 @@
1808 +
1809 +// to view this graph, run dot(1) command in GRAPHVIZ.
1810 +//
1811 +// This program is free software; you can redistribute it and/or modify
1812 +// it under the terms of the GNU General Public License as published by
1813 +// the Free Software Foundation; either version 2 of the License, or
1814 +// (at your option) any later version.
1815 +//
1816 +// This program is distributed in the hope that it will be useful,
1817 +// but WITHOUT ANY WARRANTY; without even the implied warranty of
1818 +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1819 +// GNU General Public License for more details.
1820 +//
1821 +// You should have received a copy of the GNU General Public License
1822 +// along with this program.  If not, see <http://www.gnu.org/licenses/>.
1823 +
1824 +digraph G {
1825 +node [shape=box];
1826 +whinfo [label="detailed info file\n(lower_brid_root-hinum, h_inum, namelen, old name)"];
1827 +
1828 +node [shape=oval];
1829 +
1830 +aufs_rename -> whinfo [label="store/remove"];
1831 +
1832 +node [shape=oval];
1833 +inode_list [label="h_inum list in branch\ncache"];
1834 +
1835 +node [shape=box];
1836 +whinode [label="h_inum list file"];
1837 +
1838 +node [shape=oval];
1839 +brmgmt [label="br_add/del/mod/umount"];
1840 +
1841 +brmgmt -> inode_list [label="create/remove"];
1842 +brmgmt -> whinode [label="load/store"];
1843 +
1844 +inode_list -> whinode [style=dashed,dir=both];
1845 +
1846 +aufs_rename -> inode_list [label="add/del"];
1847 +
1848 +aufs_lookup -> inode_list [label="search"];
1849 +
1850 +aufs_lookup -> whinfo [label="load/remove"];
1851 +}
1852 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt linux/Documentation/filesystems/aufs/design/06dirren.txt
1853 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt 1970-01-01 01:00:00.000000000 +0100
1854 +++ linux/Documentation/filesystems/aufs/design/06dirren.txt    2022-11-05 23:02:18.959222617 +0100
1855 @@ -0,0 +1,102 @@
1856 +
1857 +# Copyright (C) 2017-2022 Junjiro R. Okajima
1858 +#
1859 +# This program is free software; you can redistribute it and/or modify
1860 +# it under the terms of the GNU General Public License as published by
1861 +# the Free Software Foundation; either version 2 of the License, or
1862 +# (at your option) any later version.
1863 +#
1864 +# This program is distributed in the hope that it will be useful,
1865 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1866 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1867 +# GNU General Public License for more details.
1868 +#
1869 +# You should have received a copy of the GNU General Public License
1870 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1871 +
1872 +Special handling for renaming a directory (DIRREN)
1873 +----------------------------------------------------------------------
1874 +First, let's assume we have a simple usecase.
1875 +
1876 +- /u = /rw + /ro
1877 +- /rw/dirA exists
1878 +- /ro/dirA and /ro/dirA/file exist too
1879 +- there is no dirB on both branches
1880 +- a user issues rename("dirA", "dirB")
1881 +
1882 +Now, what should aufs behave against this rename(2)?
1883 +There are a few possible cases.
1884 +
1885 +A. returns EROFS.
1886 +   since dirA exists on a readonly branch which cannot be renamed.
1887 +B. returns EXDEV.
1888 +   it is possible to copy-up dirA (only the dir itself), but the child
1889 +   entries ("file" in this case) should not be. it must be a bad
1890 +   approach to copy-up recursively.
1891 +C. returns a success.
1892 +   even the branch /ro is readonly, aufs tries renaming it. Obviously it
1893 +   is a violation of aufs' policy.
1894 +D. construct an extra information which indicates that /ro/dirA should
1895 +   be handled as the name of dirB.
1896 +   overlayfs has a similar feature called REDIRECT.
1897 +
1898 +Until now, aufs implements the case B only which returns EXDEV, and
1899 +expects the userspace application behaves like mv(1) which tries
1900 +issueing rename(2) recursively.
1901 +
1902 +A new aufs feature called DIRREN is introduced which implements the case
1903 +D. There are several "extra information" added.
1904 +
1905 +1. detailed info per renamed directory
1906 +   path: /rw/dirB/$AUFS_WH_DR_INFO_PFX.<lower branch-id>
1907 +2. the inode-number list of directories on a branch
1908 +   path: /rw/dirB/$AUFS_WH_DR_BRHINO
1909 +
1910 +The filename of "detailed info per directory" represents the lower
1911 +branch, and its format is
1912 +- a type of the branch id
1913 +  one of these.
1914 +  + uuid (not implemented yet)
1915 +  + fsid
1916 +  + dev
1917 +- the inode-number of the branch root dir
1918 +
1919 +And it contains these info in a single regular file.
1920 +- magic number
1921 +- branch's inode-number of the logically renamed dir
1922 +- the name of the before-renamed dir
1923 +
1924 +The "detailed info per directory" file is created in aufs rename(2), and
1925 +loaded in any lookup.
1926 +The info is considered in lookup for the matching case only. Here
1927 +"matching" means that the root of branch (in the info filename) is same
1928 +to the current looking-up branch. After looking-up the before-renamed
1929 +name, the inode-number is compared. And the matched dentry is used.
1930 +
1931 +The "inode-number list of directories" is a regular file which contains
1932 +simply the inode-numbers on the branch. The file is created or updated
1933 +in removing the branch, and loaded in adding the branch. Its lifetime is
1934 +equal to the branch.
1935 +The list is referred in lookup, and when the current target inode is
1936 +found in the list, the aufs tries loading the "detailed info per
1937 +directory" and get the changed and valid name of the dir.
1938 +
1939 +Theoretically these "extra informaiton" may be able to be put into XATTR
1940 +in the dir inode. But aufs doesn't choose this way because
1941 +1. XATTR may not be supported by the branch (or its configuration)
1942 +2. XATTR may have its size limit.
1943 +3. XATTR may be less easy to convert than a regular file, when the
1944 +   format of the info is changed in the future.
1945 +At the same time, I agree that the regular file approach is much slower
1946 +than XATTR approach. So, in the future, aufs may take the XATTR or other
1947 +better approach.
1948 +
1949 +This DIRREN feature is enabled by aufs configuration, and is activated
1950 +by a new mount option.
1951 +
1952 +For the more complicated case, there is a work with UDBA option, which
1953 +is to dected the direct access to the branches (by-passing aufs) and to
1954 +maintain the cashes in aufs. Since a single cached aufs dentry may
1955 +contains two names, before- and after-rename, the name comparision in
1956 +UDBA handler may not work correctly. In this case, the behaviour will be
1957 +equivalen to udba=reval case.
1958 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1959 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt   1970-01-01 01:00:00.000000000 +0100
1960 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt      2022-11-05 23:02:18.959222617 +0100
1961 @@ -0,0 +1,118 @@
1962 +
1963 +# Copyright (C) 2011-2022 Junjiro R. Okajima
1964 +#
1965 +# This program is free software; you can redistribute it and/or modify
1966 +# it under the terms of the GNU General Public License as published by
1967 +# the Free Software Foundation; either version 2 of the License, or
1968 +# (at your option) any later version.
1969 +#
1970 +# This program is distributed in the hope that it will be useful,
1971 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1972 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1973 +# GNU General Public License for more details.
1974 +#
1975 +# You should have received a copy of the GNU General Public License
1976 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
1977 +
1978 +File-based Hierarchical Storage Management (FHSM)
1979 +----------------------------------------------------------------------
1980 +Hierarchical Storage Management (or HSM) is a well-known feature in the
1981 +storage world. Aufs provides this feature as file-based with multiple
1982 +writable branches, based upon the principle of "Colder, the Lower".
1983 +Here the word "colder" means that the less used files, and "lower" means
1984 +that the position in the order of the stacked branches vertically.
1985 +These multiple writable branches are prioritized, ie. the topmost one
1986 +should be the fastest drive and be used heavily.
1987 +
1988 +o Characters in aufs FHSM story
1989 +- aufs itself and a new branch attribute.
1990 +- a new ioctl interface to move-down and to establish a connection with
1991 +  the daemon ("move-down" is a converse of "copy-up").
1992 +- userspace tool and daemon.
1993 +
1994 +The userspace daemon establishes a connection with aufs and waits for
1995 +the notification. The notified information is very similar to struct
1996 +statfs containing the number of consumed blocks and inodes.
1997 +When the consumed blocks/inodes of a branch exceeds the user-specified
1998 +upper watermark, the daemon activates its move-down process until the
1999 +consumed blocks/inodes reaches the user-specified lower watermark.
2000 +
2001 +The actual move-down is done by aufs based upon the request from
2002 +user-space since we need to maintain the inode number and the internal
2003 +pointer arrays in aufs.
2004 +
2005 +Currently aufs FHSM handles the regular files only. Additionally they
2006 +must not be hard-linked nor pseudo-linked.
2007 +
2008 +
2009 +o Cowork of aufs and the user-space daemon
2010 +  During the userspace daemon established the connection, aufs sends a
2011 +  small notification to it whenever aufs writes something into the
2012 +  writable branch. But it may cost high since aufs issues statfs(2)
2013 +  internally. So user can specify a new option to cache the
2014 +  info. Actually the notification is controlled by these factors.
2015 +  + the specified cache time.
2016 +  + classified as "force" by aufs internally.
2017 +  Until the specified time expires, aufs doesn't send the info
2018 +  except the forced cases. When aufs decide forcing, the info is always
2019 +  notified to userspace.
2020 +  For example, the number of free inodes is generally large enough and
2021 +  the shortage of it happens rarely. So aufs doesn't force the
2022 +  notification when creating a new file, directory and others. This is
2023 +  the typical case which aufs doesn't force.
2024 +  When aufs writes the actual filedata and the files consumes any of new
2025 +  blocks, the aufs forces notifying.
2026 +
2027 +
2028 +o Interfaces in aufs
2029 +- New branch attribute.
2030 +  + fhsm
2031 +    Specifies that the branch is managed by FHSM feature. In other word,
2032 +    participant in the FHSM.
2033 +    When nofhsm is set to the branch, it will not be the source/target
2034 +    branch of the move-down operation. This attribute is set
2035 +    independently from coo and moo attributes, and if you want full
2036 +    FHSM, you should specify them as well.
2037 +- New mount option.
2038 +  + fhsm_sec
2039 +    Specifies a second to suppress many less important info to be
2040 +    notified.
2041 +- New ioctl.
2042 +  + AUFS_CTL_FHSM_FD
2043 +    create a new file descriptor which userspace can read the notification
2044 +    (a subset of struct statfs) from aufs.
2045 +- Module parameter 'brs'
2046 +  It has to be set to 1. Otherwise the new mount option 'fhsm' will not
2047 +  be set.
2048 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
2049 +  When there are two or more branches with fhsm attributes,
2050 +  /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
2051 +  terminates it. As a result of remounting and branch-manipulation, the
2052 +  number of branches with fhsm attribute can be one. In this case,
2053 +  /sbin/mount.aufs will terminate the user-space daemon.
2054 +
2055 +
2056 +Finally the operation is done as these steps in kernel-space.
2057 +- make sure that,
2058 +  + no one else is using the file.
2059 +  + the file is not hard-linked.
2060 +  + the file is not pseudo-linked.
2061 +  + the file is a regular file.
2062 +  + the parent dir is not opaqued.
2063 +- find the target writable branch.
2064 +- make sure the file is not whiteout-ed by the upper (than the target)
2065 +  branch.
2066 +- make the parent dir on the target branch.
2067 +- mutex lock the inode on the branch.
2068 +- unlink the whiteout on the target branch (if exists).
2069 +- lookup and create the whiteout-ed temporary name on the target branch.
2070 +- copy the file as the whiteout-ed temporary name on the target branch.
2071 +- rename the whiteout-ed temporary name to the original name.
2072 +- unlink the file on the source branch.
2073 +- maintain the internal pointer array and the external inode number
2074 +  table (XINO).
2075 +- maintain the timestamps and other attributes of the parent dir and the
2076 +  file.
2077 +
2078 +And of course, in every step, an error may happen. So the operation
2079 +should restore the original file state after an error happens.
2080 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
2081 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt   1970-01-01 01:00:00.000000000 +0100
2082 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt      2022-11-05 23:02:18.959222617 +0100
2083 @@ -0,0 +1,72 @@
2084 +
2085 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2086 +#
2087 +# This program is free software; you can redistribute it and/or modify
2088 +# it under the terms of the GNU General Public License as published by
2089 +# the Free Software Foundation; either version 2 of the License, or
2090 +# (at your option) any later version.
2091 +#
2092 +# This program is distributed in the hope that it will be useful,
2093 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2094 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2095 +# GNU General Public License for more details.
2096 +#
2097 +# You should have received a copy of the GNU General Public License
2098 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2099 +
2100 +mmap(2) -- File Memory Mapping
2101 +----------------------------------------------------------------------
2102 +In aufs, the file-mapped pages are handled by a branch fs directly, no
2103 +interaction with aufs. It means aufs_mmap() calls the branch fs's
2104 +->mmap().
2105 +This approach is simple and good, but there is one problem.
2106 +Under /proc, several entries show the mmapped files by its path (with
2107 +device and inode number), and the printed path will be the path on the
2108 +branch fs's instead of virtual aufs's.
2109 +This is not a problem in most cases, but some utilities lsof(1) (and its
2110 +user) may expect the path on aufs.
2111 +
2112 +To address this issue, aufs adds a new member called vm_prfile in struct
2113 +vm_area_struct (and struct vm_region). The original vm_file points to
2114 +the file on the branch fs in order to handle everything correctly as
2115 +usual. The new vm_prfile points to a virtual file in aufs, and the
2116 +show-functions in procfs refers to vm_prfile if it is set.
2117 +Also we need to maintain several other places where touching vm_file
2118 +such like
2119 +- fork()/clone() copies vma and the reference count of vm_file is
2120 +  incremented.
2121 +- merging vma maintains the ref count too.
2122 +
2123 +This is not a good approach. It just fakes the printed path. But it
2124 +leaves all behaviour around f_mapping unchanged. This is surely an
2125 +advantage.
2126 +Actually aufs had adopted another complicated approach which calls
2127 +generic_file_mmap() and handles struct vm_operations_struct. In this
2128 +approach, aufs met a hard problem and I could not solve it without
2129 +switching the approach.
2130 +
2131 +There may be one more another approach which is
2132 +- bind-mount the branch-root onto the aufs-root internally
2133 +- grab the new vfsmount (ie. struct mount)
2134 +- lazy-umount the branch-root internally
2135 +- in open(2) the aufs-file, open the branch-file with the hidden
2136 +  vfsmount (instead of the original branch's vfsmount)
2137 +- ideally this "bind-mount and lazy-umount" should be done atomically,
2138 +  but it may be possible from userspace by the mount helper.
2139 +
2140 +Adding the internal hidden vfsmount and using it in opening a file, the
2141 +file path under /proc will be printed correctly. This approach looks
2142 +smarter, but is not possible I am afraid.
2143 +- aufs-root may be bind-mount later. when it happens, another hidden
2144 +  vfsmount will be required.
2145 +- it is hard to get the chance to bind-mount and lazy-umount
2146 +  + in kernel-space, FS can have vfsmount in open(2) via
2147 +    file->f_path, and aufs can know its vfsmount. But several locks are
2148 +    already acquired, and if aufs tries to bind-mount and lazy-umount
2149 +    here, then it may cause a deadlock.
2150 +  + in user-space, bind-mount doesn't invoke the mount helper.
2151 +- since /proc shows dev and ino, aufs has to give vma these info. it
2152 +  means a new member vm_prinode will be necessary. this is essentially
2153 +  equivalent to vm_prfile described above.
2154 +
2155 +I have to give up this "looks-smater" approach.
2156 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2157 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt  1970-01-01 01:00:00.000000000 +0100
2158 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt     2022-11-05 23:02:18.959222617 +0100
2159 @@ -0,0 +1,94 @@
2160 +
2161 +# Copyright (C) 2014-2022 Junjiro R. Okajima
2162 +#
2163 +# This program is free software; you can redistribute it and/or modify
2164 +# it under the terms of the GNU General Public License as published by
2165 +# the Free Software Foundation; either version 2 of the License, or
2166 +# (at your option) any later version.
2167 +#
2168 +# This program is distributed in the hope that it will be useful,
2169 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2170 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2171 +# GNU General Public License for more details.
2172 +#
2173 +# You should have received a copy of the GNU General Public License
2174 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2175 +
2176 +Listing XATTR/EA and getting the value
2177 +----------------------------------------------------------------------
2178 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2179 +shows the values from the topmost existing file. This behaviour is good
2180 +for the non-dir entries since the bahaviour exactly matches the shown
2181 +information. But for the directories, aufs considers all the same named
2182 +entries on the lower branches. Which means, if one of the lower entry
2183 +rejects readdir call, then aufs returns an error even if the topmost
2184 +entry allows it. This behaviour is necessary to respect the branch fs's
2185 +security, but can make users confused since the user-visible standard
2186 +attributes don't match the behaviour.
2187 +To address this issue, aufs has a mount option called dirperm1 which
2188 +checks the permission for the topmost entry only, and ignores the lower
2189 +entry's permission.
2190 +
2191 +A similar issue can happen around XATTR.
2192 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2193 +always set. Otherwise these very unpleasant situation would happen.
2194 +- listxattr(2) may return the duplicated entries.
2195 +- users may not be able to remove or reset the XATTR forever,
2196 +
2197 +
2198 +XATTR/EA support in the internal (copy,move)-(up,down)
2199 +----------------------------------------------------------------------
2200 +Generally the extended attributes of inode are categorized as these.
2201 +- "security" for LSM and capability.
2202 +- "system" for posix ACL, 'acl' mount option is required for the branch
2203 +  fs generally.
2204 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2205 +- "user" for userspace, 'user_xattr' mount option is required for the
2206 +  branch fs generally.
2207 +
2208 +Moreover there are some other categories. Aufs handles these rather
2209 +unpopular categories as the ordinary ones, ie. there is no special
2210 +condition nor exception.
2211 +
2212 +In copy-up, the support for XATTR on the dst branch may differ from the
2213 +src branch. In this case, the copy-up operation will get an error and
2214 +the original user operation which triggered the copy-up will fail. It
2215 +can happen that even all copy-up will fail.
2216 +When both of src and dst branches support XATTR and if an error occurs
2217 +during copying XATTR, then the copy-up should fail obviously. That is a
2218 +good reason and aufs should return an error to userspace. But when only
2219 +the src branch support that XATTR, aufs should not return an error.
2220 +For example, the src branch supports ACL but the dst branch doesn't
2221 +because the dst branch may natively un-support it or temporary
2222 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2223 +may NOT return an error even if the XATTR is not supported. It is
2224 +totally up to the branch fs.
2225 +
2226 +Anyway when the aufs internal copy-up gets an error from the dst branch
2227 +fs, then aufs tries removing the just copied entry and returns the error
2228 +to the userspace. The worst case of this situation will be all copy-up
2229 +will fail.
2230 +
2231 +For the copy-up operation, there two basic approaches.
2232 +- copy the specified XATTR only (by category above), and return the
2233 +  error unconditionally if it happens.
2234 +- copy all XATTR, and ignore the error on the specified category only.
2235 +
2236 +In order to support XATTR and to implement the correct behaviour, aufs
2237 +chooses the latter approach and introduces some new branch attributes,
2238 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2239 +They correspond to the XATTR namespaces (see above). Additionally, to be
2240 +convenient, "icex" is also provided which means all "icex*" attributes
2241 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2242 +
2243 +The meaning of these attributes is to ignore the error from setting
2244 +XATTR on that branch.
2245 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2246 +error from the dst branch according to the specified attributes.
2247 +
2248 +Some XATTR may have its default value. The default value may come from
2249 +the parent dir or the environment. If the default value is set at the
2250 +file creating-time, it will be overwritten by copy-up.
2251 +Some contradiction may happen I am afraid.
2252 +Do we need another attribute to stop copying XATTR? I am unsure. For
2253 +now, aufs implements the branch attributes to ignore the error.
2254 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2255 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2256 +++ linux/Documentation/filesystems/aufs/design/07export.txt    2022-11-05 23:02:18.959222617 +0100
2257 @@ -0,0 +1,58 @@
2258 +
2259 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2260 +#
2261 +# This program is free software; you can redistribute it and/or modify
2262 +# it under the terms of the GNU General Public License as published by
2263 +# the Free Software Foundation; either version 2 of the License, or
2264 +# (at your option) any later version.
2265 +#
2266 +# This program is distributed in the hope that it will be useful,
2267 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2268 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2269 +# GNU General Public License for more details.
2270 +#
2271 +# You should have received a copy of the GNU General Public License
2272 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2273 +
2274 +Export Aufs via NFS
2275 +----------------------------------------------------------------------
2276 +Here is an approach.
2277 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2278 +  generation.
2279 +- iget_locked(): initialize aufs inode generation for a new inode, and
2280 +  store it in xigen file.
2281 +- destroy_inode(): increment aufs inode generation and store it in xigen
2282 +  file. it is necessary even if it is not unlinked, because any data of
2283 +  inode may be changed by UDBA.
2284 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2285 +  build file handle by
2286 +  + branch id (4 bytes)
2287 +  + superblock generation (4 bytes)
2288 +  + inode number (4 or 8 bytes)
2289 +  + parent dir inode number (4 or 8 bytes)
2290 +  + inode generation (4 bytes))
2291 +  + return value of exportfs_encode_fh() for the parent on a branch (4
2292 +    bytes)
2293 +  + file handle for a branch (by exportfs_encode_fh())
2294 +- fh_to_dentry():
2295 +  + find the index of a branch from its id in handle, and check it is
2296 +    still exist in aufs.
2297 +  + 1st level: get the inode number from handle and search it in cache.
2298 +  + 2nd level: if not found in cache, get the parent inode number from
2299 +    the handle and search it in cache. and then open the found parent
2300 +    dir, find the matching inode number by vfs_readdir() and get its
2301 +    name, and call lookup_one_len() for the target dentry.
2302 +  + 3rd level: if the parent dir is not cached, call
2303 +    exportfs_decode_fh() for a branch and get the parent on a branch,
2304 +    build a pathname of it, convert it a pathname in aufs, call
2305 +    path_lookup(). now aufs gets a parent dir dentry, then handle it as
2306 +    the 2nd level.
2307 +  + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2308 +    for every branch, but not itself. to get this, (currently) aufs
2309 +    searches in current->nsproxy->mnt_ns list. it may not be a good
2310 +    idea, but I didn't get other approach.
2311 +  + test the generation of the gotten inode.
2312 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2313 +  convert it into ESTALE for NFSD.
2314 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2315 +  lookup_one_len(), vfs_getattr(), encode_fh() and others.
2316 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2317 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt   1970-01-01 01:00:00.000000000 +0100
2318 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt      2022-11-05 23:02:18.959222617 +0100
2319 @@ -0,0 +1,52 @@
2320 +
2321 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2322 +#
2323 +# This program is free software; you can redistribute it and/or modify
2324 +# it under the terms of the GNU General Public License as published by
2325 +# the Free Software Foundation; either version 2 of the License, or
2326 +# (at your option) any later version.
2327 +#
2328 +# This program is distributed in the hope that it will be useful,
2329 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2330 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2331 +# GNU General Public License for more details.
2332 +#
2333 +# You should have received a copy of the GNU General Public License
2334 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2335 +
2336 +Show Whiteout Mode (shwh)
2337 +----------------------------------------------------------------------
2338 +Generally aufs hides the name of whiteouts. But in some cases, to show
2339 +them is very useful for users. For instance, creating a new middle layer
2340 +(branch) by merging existing layers.
2341 +
2342 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2343 +When you have three branches,
2344 +- Bottom: 'system', squashfs (underlying base system), read-only
2345 +- Middle: 'mods', squashfs, read-only
2346 +- Top: 'overlay', ram (tmpfs), read-write
2347 +
2348 +The top layer is loaded at boot time and saved at shutdown, to preserve
2349 +the changes made to the system during the session.
2350 +When larger changes have been made, or smaller changes have accumulated,
2351 +the size of the saved top layer data grows. At this point, it would be
2352 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2353 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2354 +restoring save and load speed.
2355 +
2356 +This merging is simplified by the use of another aufs mount, of just the
2357 +two overlay branches using the 'shwh' option.
2358 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2359 +       aufs /livesys/merge_union
2360 +
2361 +A merged view of these two branches is then available at
2362 +/livesys/merge_union, and the new feature is that the whiteouts are
2363 +visible!
2364 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2365 +writing to all branches. Also the default mode for all branches is 'ro'.
2366 +It is now possible to save the combined contents of the two overlay
2367 +branches to a new squashfs, e.g.:
2368 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2369 +
2370 +This new squashfs archive can be stored on the boot device and the
2371 +initramfs will use it to replace the old one at the next boot.
2372 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2373 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt  1970-01-01 01:00:00.000000000 +0100
2374 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt     2022-11-05 23:02:18.959222617 +0100
2375 @@ -0,0 +1,47 @@
2376 +
2377 +# Copyright (C) 2010-2022 Junjiro R. Okajima
2378 +#
2379 +# This program is free software; you can redistribute it and/or modify
2380 +# it under the terms of the GNU General Public License as published by
2381 +# the Free Software Foundation; either version 2 of the License, or
2382 +# (at your option) any later version.
2383 +#
2384 +# This program is distributed in the hope that it will be useful,
2385 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2386 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2387 +# GNU General Public License for more details.
2388 +#
2389 +# You should have received a copy of the GNU General Public License
2390 +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
2391 +
2392 +Dynamically customizable FS operations
2393 +----------------------------------------------------------------------
2394 +Generally FS operations (struct inode_operations, struct
2395 +address_space_operations, struct file_operations, etc.) are defined as
2396 +"static const", but it never means that FS have only one set of
2397 +operation. Some FS have multiple sets of them. For instance, ext2 has
2398 +three sets, one for XIP, for NOBH, and for normal.
2399 +Since aufs overrides and redirects these operations, sometimes aufs has
2400 +to change its behaviour according to the branch FS type. More importantly
2401 +VFS acts differently if a function (member in the struct) is set or
2402 +not. It means aufs should have several sets of operations and select one
2403 +among them according to the branch FS definition.
2404 +
2405 +In order to solve this problem and not to affect the behaviour of VFS,
2406 +aufs defines these operations dynamically. For instance, aufs defines
2407 +dummy direct_IO function for struct address_space_operations, but it may
2408 +not be set to the address_space_operations actually. When the branch FS
2409 +doesn't have it, aufs doesn't set it to its address_space_operations
2410 +while the function definition itself is still alive. So the behaviour
2411 +itself will not change, and it will return an error when direct_IO is
2412 +not set.
2413 +
2414 +The lifetime of these dynamically generated operation object is
2415 +maintained by aufs branch object. When the branch is removed from aufs,
2416 +the reference counter of the object is decremented. When it reaches
2417 +zero, the dynamically generated operation object will be freed.
2418 +
2419 +This approach is designed to support AIO (io_submit), Direct I/O and
2420 +XIP (DAX) mainly.
2421 +Currently this approach is applied to address_space_operations for
2422 +regular files only.
2423 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2424 --- /usr/share/empty/Documentation/filesystems/aufs/README      1970-01-01 01:00:00.000000000 +0100
2425 +++ linux/Documentation/filesystems/aufs/README 2022-11-05 23:02:18.955889283 +0100
2426 @@ -0,0 +1,403 @@
2427 +
2428 +Aufs6 -- advanced multi layered unification filesystem version 6.x
2429 +http://aufs.sf.net
2430 +Junjiro R. Okajima
2431 +
2432 +
2433 +0. Introduction
2434 +----------------------------------------
2435 +In the early days, aufs was entirely re-designed and re-implemented
2436 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2437 +improvements and implementations, it became totally different from
2438 +Unionfs while keeping the basic features.
2439 +Later, Unionfs Version 2.x series began taking some of the same
2440 +approaches to aufs1's.
2441 +Unionfs was being developed by Professor Erez Zadok at Stony Brook
2442 +University and his team.
2443 +
2444 +Aufs6 supports linux-v6.0 and later, try aufs6.0 branch in
2445 +aufs-linux.git or aufs-standalone.git.
2446 +If you want older kernel version support,
2447 +- for linux-v5.x series, try aufs-linux.git or aufs-standalone.git
2448 +- for linux-v4.x series, try aufs4-linux.git or aufs4-standalone.git
2449 +- for linux-v3.x series, try aufs3-linux.git or aufs3-standalone.git
2450 +- for linux-v2.6.16 and later, try aufs2-2.6.git, aufs2-standalone.git
2451 +  or aufs1 from CVS on SourceForge.
2452 +
2453 +Note: the name of aufs5-linux.git and aufs5-standalone.git on github
2454 +      were changed. Now they are aufs-linux.git and
2455 +      aufs-standalone.git and they contain aufs5 and later branches.
2456 +
2457 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2458 +      According to Christoph Hellwig, linux rejects all union-type
2459 +      filesystems but UnionMount.
2460 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2461 +
2462 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2463 +    UnionMount, and he pointed out an issue around a directory mutex
2464 +    lock and aufs addressed it. But it is still unsure whether aufs will
2465 +    be merged (or any other union solution).
2466 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2467 +
2468 +
2469 +1. Features
2470 +----------------------------------------
2471 +- unite several directories into a single virtual filesystem. The member
2472 +  directory is called as a branch.
2473 +- you can specify the permission flags to the branch, which are 'readonly',
2474 +  'readwrite' and 'whiteout-able.'
2475 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2476 +  readonly branch are modifiable logically.
2477 +- dynamic branch manipulation, add, del.
2478 +- etc...
2479 +
2480 +Also there are many enhancements in aufs, such as:
2481 +- test only the highest one for the directory permission (dirperm1)
2482 +- copyup on open (coo=)
2483 +- 'move' policy for copy-up between two writable branches, after
2484 +  checking free space.
2485 +- xattr, acl
2486 +- readdir(3) in userspace.
2487 +- keep inode number by external inode number table
2488 +- keep the timestamps of file/dir in internal copyup operation
2489 +- seekable directory, supporting NFS readdir.
2490 +- whiteout is hardlinked in order to reduce the consumption of inodes
2491 +  on branch
2492 +- do not copyup, nor create a whiteout when it is unnecessary
2493 +- revert a single systemcall when an error occurs in aufs
2494 +- remount interface instead of ioctl
2495 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2496 +- loopback mounted filesystem as a branch
2497 +- kernel thread for removing the dir who has a plenty of whiteouts
2498 +- support copyup sparse file (a file which has a 'hole' in it)
2499 +- default permission flags for branches
2500 +- selectable permission flags for ro branch, whether whiteout can
2501 +  exist or not
2502 +- export via NFS.
2503 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2504 +- support multiple writable branches, some policies to select one
2505 +  among multiple writable branches.
2506 +- a new semantics for link(2) and rename(2) to support multiple
2507 +  writable branches.
2508 +- no glibc changes are required.
2509 +- pseudo hardlink (hardlink over branches)
2510 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2511 +  including NFS or remote filesystem branch.
2512 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2513 +- and more...
2514 +
2515 +Currently these features are dropped temporary from aufs6.
2516 +See design/08plan.txt in detail.
2517 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2518 +  (robr)
2519 +- statistics of aufs thread (/sys/fs/aufs/stat)
2520 +
2521 +Features or just an idea in the future (see also design/*.txt),
2522 +- reorder the branch index without del/re-add.
2523 +- permanent xino files for NFSD
2524 +- an option for refreshing the opened files after add/del branches
2525 +- light version, without branch manipulation. (unnecessary?)
2526 +- copyup in userspace
2527 +- inotify in userspace
2528 +- readv/writev
2529 +
2530 +
2531 +2. Download
2532 +----------------------------------------
2533 +There are three GIT trees for aufs6, aufs6-linux.git,
2534 +aufs6-standalone.git, and aufs-util.git. Note that there is no "6" in
2535 +"aufs-util.git."
2536 +While the aufs-util is always necessary, you need either of aufs6-linux
2537 +or aufs6-standalone.
2538 +
2539 +The aufs6-linux tree includes the whole linux mainline GIT tree,
2540 +git://git.kernel.org/.../torvalds/linux.git.
2541 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2542 +build aufs6 as an external kernel module.
2543 +Several extra patches are not included in this tree. Only
2544 +aufs6-standalone tree contains them. They are described in the later
2545 +section "Configuration and Compilation."
2546 +
2547 +On the other hand, the aufs6-standalone tree has only aufs source files
2548 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2549 +But you need to apply all aufs patches manually.
2550 +
2551 +You will find GIT branches whose name is in form of "aufs6.x" where "x"
2552 +represents the linux kernel version, "linux-6.x". For instance,
2553 +"aufs6.0" is for linux-6.0. For latest "linux-6.x-rcN", use
2554 +"aufs6.x-rcN" branch.
2555 +
2556 +o aufs6-linux tree
2557 +$ git clone --reference /your/linux/git/tree \
2558 +       git://github.com/sfjro/aufs6-linux.git aufs6-linux.git
2559 +- if you don't have linux GIT tree, then remove "--reference ..."
2560 +$ cd aufs6-linux.git
2561 +$ git checkout origin/aufs6.0
2562 +
2563 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2564 +leave the patch-work to GIT.
2565 +$ cd /your/linux/git/tree
2566 +$ git remote add aufs6 git://github.com/sfjro/aufs6-linux.git
2567 +$ git fetch aufs6
2568 +$ git checkout -b my6.0 v6.0
2569 +$ (add your local change...)
2570 +$ git pull aufs6 aufs6.0
2571 +- now you have v6.0 + your_changes + aufs6.0 in you my6.0 branch.
2572 +- you may need to solve some conflicts between your_changes and
2573 +  aufs6.0. in this case, git-rerere is recommended so that you can
2574 +  solve the similar conflicts automatically when you upgrade to 6.1 or
2575 +  later in the future.
2576 +
2577 +o aufs6-standalone tree
2578 +$ git clone git://github.com/sfjro/aufs6-standalone.git aufs6-standalone.git
2579 +$ cd aufs6-standalone.git
2580 +$ git checkout origin/aufs6.0
2581 +
2582 +o aufs-util tree
2583 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2584 +- note that the public aufs-util.git is on SourceForge instead of
2585 +  GitHUB.
2586 +$ cd aufs-util.git
2587 +$ git checkout origin/aufs6.0
2588 +
2589 +Note: The 6.x-rcN branch is to be used with `rc' kernel versions ONLY.
2590 +The minor version number, 'x' in '6.x', of aufs may not always
2591 +follow the minor version number of the kernel.
2592 +Because changes in the kernel that cause the use of a new
2593 +minor version number do not always require changes to aufs-util.
2594 +
2595 +Since aufs-util has its own minor version number, you may not be
2596 +able to find a GIT branch in aufs-util for your kernel's
2597 +exact minor version number.
2598 +In this case, you should git-checkout the branch for the
2599 +nearest lower number.
2600 +
2601 +For (an unreleased) example:
2602 +If you are using "linux-6.10" and the "aufs6.10" branch
2603 +does not exist in aufs-util repository, then "aufs6.9", "aufs6.8"
2604 +or something numerically smaller is the branch for your kernel.
2605 +
2606 +Also you can view all branches by
2607 +       $ git branch -a
2608 +
2609 +
2610 +3. Configuration and Compilation
2611 +----------------------------------------
2612 +Make sure you have git-checkout'ed the correct branch.
2613 +
2614 +For aufs6-linux tree,
2615 +- enable CONFIG_AUFS_FS.
2616 +- set other aufs configurations if necessary.
2617 +- for aufs5.13 and later
2618 +  Because aufs is not only an ordinary filesystem (callee of VFS), but
2619 +  also a caller of VFS functions for branch filesystems, subclassing of
2620 +  the internal locks for LOCKDEP is necessary. LOCKDEP is a debugging
2621 +  feature of linux kernel. If you enable CONFIG_LOCKDEP, then you will
2622 +  need to customize some LOCKDEP numbers. Here are what I use on my
2623 +  test environment.
2624 +       CONFIG_LOCKDEP_BITS=21
2625 +       CONFIG_LOCKDEP_CHAINS_BITS=21
2626 +       CONFIG_LOCKDEP_STACK_TRACE_BITS=24
2627 +
2628 +For aufs6-standalone tree,
2629 +There are several ways to build.
2630 +
2631 +1.
2632 +- apply ./aufs6-kbuild.patch to your kernel source files.
2633 +- apply ./aufs6-base.patch too.
2634 +- apply ./aufs6-mmap.patch too.
2635 +- apply ./aufs6-standalone.patch too, if you have a plan to set
2636 +  CONFIG_AUFS_FS=m. otherwise you don't need ./aufs6-standalone.patch.
2637 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2638 +  kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2639 +- enable CONFIG_AUFS_FS, you can select either
2640 +  =m or =y.
2641 +- and build your kernel as usual.
2642 +- install the built kernel.
2643 +- install the header files too by "make headers_install" to the
2644 +  directory where you specify. By default, it is $PWD/usr.
2645 +  "make help" shows a brief note for headers_install.
2646 +- and reboot your system.
2647 +
2648 +2.
2649 +- module only (CONFIG_AUFS_FS=m).
2650 +- apply ./aufs6-base.patch to your kernel source files.
2651 +- apply ./aufs6-mmap.patch too.
2652 +- apply ./aufs6-standalone.patch too.
2653 +- build your kernel, don't forget "make headers_install", and reboot.
2654 +- edit ./config.mk and set other aufs configurations if necessary.
2655 +  Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2656 +  every aufs configurations.
2657 +- build the module by simple "make".
2658 +- you can specify ${KDIR} make variable which points to your kernel
2659 +  source tree.
2660 +- install the files
2661 +  + run "make install" to install the aufs module, or copy the built
2662 +    $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2663 +  + run "make install_headers" (instead of headers_install) to install
2664 +    the modified aufs header file (you can specify DESTDIR which is
2665 +    available in aufs standalone version's Makefile only), or copy
2666 +    $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2667 +    you like manually. By default, the target directory is $PWD/usr.
2668 +- no need to apply aufs6-kbuild.patch, nor copying source files to your
2669 +  kernel source tree.
2670 +
2671 +Note: The header file aufs_type.h is necessary to build aufs-util
2672 +      as well as "make headers_install" in the kernel source tree.
2673 +      headers_install is subject to be forgotten, but it is essentially
2674 +      necessary, not only for building aufs-util.
2675 +      You may not meet problems without headers_install in some older
2676 +      version though.
2677 +
2678 +And then,
2679 +- read README in aufs-util, build and install it
2680 +- note that your distribution may contain an obsoleted version of
2681 +  aufs_type.h in /usr/include/linux or something. When you build aufs
2682 +  utilities, make sure that your compiler refers the correct aufs header
2683 +  file which is built by "make headers_install."
2684 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2685 +  then run "make install_ulib" too. And refer to the aufs manual in
2686 +  detail.
2687 +
2688 +There several other patches in aufs6-standalone.git. They are all
2689 +optional. When you meet some problems, they will help you.
2690 +- aufs6-loopback.patch
2691 +  Supports a nested loopback mount in a branch-fs. This patch is
2692 +  unnecessary until aufs produces a message like "you may want to try
2693 +  another patch for loopback file".
2694 +- vfs-ino.patch
2695 +  Modifies a system global kernel internal function get_next_ino() in
2696 +  order to stop assigning 0 for an inode-number. Not directly related to
2697 +  aufs, but recommended generally.
2698 +- tmpfs-idr.patch
2699 +  Keeps the tmpfs inode number as the lowest value. Effective to reduce
2700 +  the size of aufs XINO files for tmpfs branch. Also it prevents the
2701 +  duplication of inode number, which is important for backup tools and
2702 +  other utilities. When you find aufs XINO files for tmpfs branch
2703 +  growing too much, try this patch.
2704 +
2705 +
2706 +4. Usage
2707 +----------------------------------------
2708 +At first, make sure aufs-util are installed, and please read the aufs
2709 +manual, aufs.6 in aufs-util.git tree.
2710 +$ man -l aufs.6
2711 +
2712 +And then,
2713 +$ mkdir /tmp/rw /tmp/aufs
2714 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2715 +
2716 +Here is another example. The result is equivalent.
2717 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2718 +  Or
2719 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2720 +# mount -o remount,append:${HOME} /tmp/aufs
2721 +
2722 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2723 +you modify a file under /tmp/aufs, the one on your home directory is
2724 +not affected, instead the same named file will be newly created under
2725 +/tmp/rw. And all of your modification to a file will be applied to
2726 +the one under /tmp/rw. This is called the file based Copy on Write
2727 +(COW) method.
2728 +Aufs mount options are described in aufs.6.
2729 +If you run chroot or something and make your aufs as a root directory,
2730 +then you need to customize the shutdown script. See the aufs manual in
2731 +detail.
2732 +
2733 +Additionally, there are some sample usages of aufs which are a
2734 +diskless system with network booting, and LiveCD over NFS.
2735 +See sample dir in CVS tree on SourceForge.
2736 +
2737 +
2738 +5. Contact
2739 +----------------------------------------
2740 +When you have any problems or strange behaviour in aufs, please let me
2741 +know with:
2742 +- /proc/mounts (instead of the output of mount(8))
2743 +- /sys/module/aufs/*
2744 +- /sys/fs/aufs/* (if you have them)
2745 +- /debug/aufs/* (if you have them)
2746 +- linux kernel version
2747 +  if your kernel is not plain, for example modified by distributor,
2748 +  the url where i can download its source is necessary too.
2749 +- aufs version which was printed at loading the module or booting the
2750 +  system, instead of the date you downloaded.
2751 +- configuration (define/undefine CONFIG_AUFS_xxx)
2752 +- kernel configuration or /proc/config.gz (if you have it)
2753 +- LSM (linux security module, if you are using)
2754 +- behaviour which you think to be incorrect
2755 +- actual operation, reproducible one is better
2756 +- mailto: aufs-users at lists.sourceforge.net
2757 +
2758 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2759 +and Feature Requests) on SourceForge. Please join and write to
2760 +aufs-users ML.
2761 +
2762 +
2763 +6. Acknowledgements
2764 +----------------------------------------
2765 +Thanks to everyone who have tried and are using aufs, whoever
2766 +have reported a bug or any feedback.
2767 +
2768 +Especially donators:
2769 +Tomas Matejicek(slax.org) made a donation (much more than once).
2770 +       Since Apr 2010, Tomas M (the author of Slax and Linux Live
2771 +       scripts) is making "doubling" donations.
2772 +       Unfortunately I cannot list all of the donators, but I really
2773 +       appreciate.
2774 +       It ends Aug 2010, but the ordinary donation URL is still available.
2775 +       <http://sourceforge.net/donate/index.php?group_id=167503>
2776 +Dai Itasaka made a donation (2007/8).
2777 +Chuck Smith made a donation (2008/4, 10 and 12).
2778 +Henk Schoneveld made a donation (2008/9).
2779 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2780 +Francois Dupoux made a donation (2008/11).
2781 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2782 +       aufs2 GIT tree (2009/2).
2783 +William Grant made a donation (2009/3).
2784 +Patrick Lane made a donation (2009/4).
2785 +The Mail Archive (mail-archive.com) made donations (2009/5).
2786 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2787 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2788 +Pavel Pronskiy made a donation (2011/2).
2789 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2790 +       Networks (Ed Wildgoose) made a donation for hardware (2011/3).
2791 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2792 +11).
2793 +Sam Liddicott made a donation (2011/9).
2794 +Era Scarecrow made a donation (2013/4).
2795 +Bor Ratajc made a donation (2013/4).
2796 +Alessandro Gorreta made a donation (2013/4).
2797 +POIRETTE Marc made a donation (2013/4).
2798 +Alessandro Gorreta made a donation (2013/4).
2799 +lauri kasvandik made a donation (2013/5).
2800 +"pemasu from Finland" made a donation (2013/7).
2801 +The Parted Magic Project made a donation (2013/9 and 11).
2802 +Pavel Barta made a donation (2013/10).
2803 +Nikolay Pertsev made a donation (2014/5).
2804 +James B made a donation (2014/7, 2015/7, and 2021/12).
2805 +Stefano Di Biase made a donation (2014/8).
2806 +Daniel Epellei made a donation (2015/1).
2807 +OmegaPhil made a donation (2016/1, 2018/4).
2808 +Tomasz Szewczyk made a donation (2016/4).
2809 +James Burry made a donation (2016/12).
2810 +Carsten Rose made a donation (2018/9).
2811 +Porteus Kiosk made a donation (2018/10).
2812 +Enya Quetzalli Gomez Rodriguez made a donation (2022/5).
2813 +
2814 +Thank you very much.
2815 +Donations are always, including future donations, very important and
2816 +helpful for me to keep on developing aufs.
2817 +
2818 +
2819 +7.
2820 +----------------------------------------
2821 +If you are an experienced user, no explanation is needed. Aufs is
2822 +just a linux filesystem.
2823 +
2824 +
2825 +Enjoy!
2826 +
2827 +# Local variables: ;
2828 +# mode: text;
2829 +# End: ;
2830 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2831 --- /usr/share/empty/fs/aufs/aufs.h     1970-01-01 01:00:00.000000000 +0100
2832 +++ linux/fs/aufs/aufs.h        2022-11-05 23:02:18.959222617 +0100
2833 @@ -0,0 +1,62 @@
2834 +/* SPDX-License-Identifier: GPL-2.0 */
2835 +/*
2836 + * Copyright (C) 2005-2022 Junjiro R. Okajima
2837 + *
2838 + * This program is free software; you can redistribute it and/or modify
2839 + * it under the terms of the GNU General Public License as published by
2840 + * the Free Software Foundation; either version 2 of the License, or
2841 + * (at your option) any later version.
2842 + *
2843 + * This program is distributed in the hope that it will be useful,
2844 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2845 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2846 + * GNU General Public License for more details.
2847 + *
2848 + * You should have received a copy of the GNU General Public License
2849 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
2850 + */
2851 +
2852 +/*
2853 + * all header files
2854 + */
2855 +
2856 +#ifndef __AUFS_H__
2857 +#define __AUFS_H__
2858 +
2859 +#ifdef __KERNEL__
2860 +
2861 +#define AuStub(type, name, body, ...) \
2862 +       static inline type name(__VA_ARGS__) { body; }
2863 +
2864 +#define AuStubVoid(name, ...) \
2865 +       AuStub(void, name, , __VA_ARGS__)
2866 +#define AuStubInt0(name, ...) \
2867 +       AuStub(int, name, return 0, __VA_ARGS__)
2868 +
2869 +#include "debug.h"
2870 +
2871 +#include "branch.h"
2872 +#include "cpup.h"
2873 +#include "dcsub.h"
2874 +#include "dbgaufs.h"
2875 +#include "dentry.h"
2876 +#include "dir.h"
2877 +#include "dirren.h"
2878 +#include "dynop.h"
2879 +#include "file.h"
2880 +#include "fstype.h"
2881 +#include "hbl.h"
2882 +#include "inode.h"
2883 +#include "lcnt.h"
2884 +#include "loop.h"
2885 +#include "module.h"
2886 +#include "opts.h"
2887 +#include "rwsem.h"
2888 +#include "super.h"
2889 +#include "sysaufs.h"
2890 +#include "vfsub.h"
2891 +#include "whout.h"
2892 +#include "wkq.h"
2893 +
2894 +#endif /* __KERNEL__ */
2895 +#endif /* __AUFS_H__ */
2896 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2897 --- /usr/share/empty/fs/aufs/branch.c   1970-01-01 01:00:00.000000000 +0100
2898 +++ linux/fs/aufs/branch.c      2022-11-05 23:02:18.959222617 +0100
2899 @@ -0,0 +1,1427 @@
2900 +// SPDX-License-Identifier: GPL-2.0
2901 +/*
2902 + * Copyright (C) 2005-2022 Junjiro R. Okajima
2903 + *
2904 + * This program is free software; you can redistribute it and/or modify
2905 + * it under the terms of the GNU General Public License as published by
2906 + * the Free Software Foundation; either version 2 of the License, or
2907 + * (at your option) any later version.
2908 + *
2909 + * This program is distributed in the hope that it will be useful,
2910 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2911 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
2912 + * GNU General Public License for more details.
2913 + *
2914 + * You should have received a copy of the GNU General Public License
2915 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
2916 + */
2917 +
2918 +/*
2919 + * branch management
2920 + */
2921 +
2922 +#include <linux/compat.h>
2923 +#include <linux/statfs.h>
2924 +#include "aufs.h"
2925 +
2926 +/*
2927 + * free a single branch
2928 + */
2929 +static void au_br_do_free(struct au_branch *br)
2930 +{
2931 +       int i;
2932 +       struct au_wbr *wbr;
2933 +       struct au_dykey **key;
2934 +
2935 +       au_hnotify_fin_br(br);
2936 +       /* always, regardless the mount option */
2937 +       au_dr_hino_free(&br->br_dirren);
2938 +       au_xino_put(br);
2939 +
2940 +       AuLCntZero(au_lcnt_read(&br->br_nfiles, /*do_rev*/0));
2941 +       au_lcnt_fin(&br->br_nfiles, /*do_sync*/0);
2942 +       AuLCntZero(au_lcnt_read(&br->br_count, /*do_rev*/0));
2943 +       au_lcnt_fin(&br->br_count, /*do_sync*/0);
2944 +
2945 +       wbr = br->br_wbr;
2946 +       if (wbr) {
2947 +               for (i = 0; i < AuBrWh_Last; i++)
2948 +                       dput(wbr->wbr_wh[i]);
2949 +               AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2950 +               AuRwDestroy(&wbr->wbr_wh_rwsem);
2951 +       }
2952 +
2953 +       if (br->br_fhsm) {
2954 +               au_br_fhsm_fin(br->br_fhsm);
2955 +               au_kfree_try_rcu(br->br_fhsm);
2956 +       }
2957 +
2958 +       key = br->br_dykey;
2959 +       for (i = 0; i < AuBrDynOp; i++, key++)
2960 +               if (*key)
2961 +                       au_dy_put(*key);
2962 +               else
2963 +                       break;
2964 +
2965 +       /* recursive lock, s_umount of branch's */
2966 +       /* synchronize_rcu(); */ /* why? */
2967 +       lockdep_off();
2968 +       path_put(&br->br_path);
2969 +       lockdep_on();
2970 +       au_kfree_rcu(wbr);
2971 +       au_lcnt_wait_for_fin(&br->br_nfiles);
2972 +       au_lcnt_wait_for_fin(&br->br_count);
2973 +       /* I don't know why, but percpu_refcount requires this */
2974 +       /* synchronize_rcu(); */
2975 +       au_kfree_rcu(br);
2976 +}
2977 +
2978 +/*
2979 + * frees all branches
2980 + */
2981 +void au_br_free(struct au_sbinfo *sbinfo)
2982 +{
2983 +       aufs_bindex_t bmax;
2984 +       struct au_branch **br;
2985 +
2986 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
2987 +
2988 +       bmax = sbinfo->si_bbot + 1;
2989 +       br = sbinfo->si_branch;
2990 +       while (bmax--)
2991 +               au_br_do_free(*br++);
2992 +}
2993 +
2994 +/*
2995 + * find the index of a branch which is specified by @br_id.
2996 + */
2997 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2998 +{
2999 +       aufs_bindex_t bindex, bbot;
3000 +
3001 +       bbot = au_sbbot(sb);
3002 +       for (bindex = 0; bindex <= bbot; bindex++)
3003 +               if (au_sbr_id(sb, bindex) == br_id)
3004 +                       return bindex;
3005 +       return -1;
3006 +}
3007 +
3008 +/* ---------------------------------------------------------------------- */
3009 +
3010 +/*
3011 + * add a branch
3012 + */
3013 +
3014 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
3015 +                       struct dentry *h_root)
3016 +{
3017 +       if (unlikely(h_adding == h_root
3018 +                    || au_test_loopback_overlap(sb, h_adding)))
3019 +               return 1;
3020 +       if (h_adding->d_sb != h_root->d_sb)
3021 +               return 0;
3022 +       return au_test_subdir(h_adding, h_root)
3023 +               || au_test_subdir(h_root, h_adding);
3024 +}
3025 +
3026 +/*
3027 + * returns a newly allocated branch. @new_nbranch is a number of branches
3028 + * after adding a branch.
3029 + */
3030 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
3031 +                                    int perm)
3032 +{
3033 +       struct au_branch *add_branch;
3034 +       struct dentry *root;
3035 +       struct inode *inode;
3036 +       int err;
3037 +
3038 +       err = -ENOMEM;
3039 +       add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
3040 +       if (unlikely(!add_branch))
3041 +               goto out;
3042 +       add_branch->br_xino = au_xino_alloc(/*nfile*/1);
3043 +       if (unlikely(!add_branch->br_xino))
3044 +               goto out_br;
3045 +       err = au_hnotify_init_br(add_branch, perm);
3046 +       if (unlikely(err))
3047 +               goto out_xino;
3048 +
3049 +       if (au_br_writable(perm)) {
3050 +               /* may be freed separately at changing the branch permission */
3051 +               add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
3052 +                                            GFP_NOFS);
3053 +               if (unlikely(!add_branch->br_wbr))
3054 +                       goto out_hnotify;
3055 +       }
3056 +
3057 +       if (au_br_fhsm(perm)) {
3058 +               err = au_fhsm_br_alloc(add_branch);
3059 +               if (unlikely(err))
3060 +                       goto out_wbr;
3061 +       }
3062 +
3063 +       root = sb->s_root;
3064 +       err = au_sbr_realloc(au_sbi(sb), new_nbranch, /*may_shrink*/0);
3065 +       if (!err)
3066 +               err = au_di_realloc(au_di(root), new_nbranch, /*may_shrink*/0);
3067 +       if (!err) {
3068 +               inode = d_inode(root);
3069 +               err = au_hinode_realloc(au_ii(inode), new_nbranch,
3070 +                                       /*may_shrink*/0);
3071 +       }
3072 +       if (!err)
3073 +               return add_branch; /* success */
3074 +
3075 +out_wbr:
3076 +       au_kfree_rcu(add_branch->br_wbr);
3077 +out_hnotify:
3078 +       au_hnotify_fin_br(add_branch);
3079 +out_xino:
3080 +       au_xino_put(add_branch);
3081 +out_br:
3082 +       au_kfree_rcu(add_branch);
3083 +out:
3084 +       return ERR_PTR(err);
3085 +}
3086 +
3087 +/*
3088 + * test if the branch permission is legal or not.
3089 + */
3090 +static int test_br(struct inode *inode, int brperm, char *path)
3091 +{
3092 +       int err;
3093 +
3094 +       err = (au_br_writable(brperm) && IS_RDONLY(inode));
3095 +       if (!err)
3096 +               goto out;
3097 +
3098 +       err = -EINVAL;
3099 +       pr_err("write permission for readonly mount or inode, %s\n", path);
3100 +
3101 +out:
3102 +       return err;
3103 +}
3104 +
3105 +/*
3106 + * returns:
3107 + * 0: success, the caller will add it
3108 + * plus: success, it is already unified, the caller should ignore it
3109 + * minus: error
3110 + */
3111 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
3112 +{
3113 +       int err;
3114 +       aufs_bindex_t bbot, bindex;
3115 +       struct dentry *root, *h_dentry;
3116 +       struct inode *inode, *h_inode;
3117 +
3118 +       root = sb->s_root;
3119 +       bbot = au_sbbot(sb);
3120 +       if (unlikely(bbot >= 0
3121 +                    && au_find_dbindex(root, add->path.dentry) >= 0)) {
3122 +               err = 1;
3123 +               if (!remount) {
3124 +                       err = -EINVAL;
3125 +                       pr_err("%s duplicated\n", add->pathname);
3126 +               }
3127 +               goto out;
3128 +       }
3129 +
3130 +       err = -ENOSPC; /* -E2BIG; */
3131 +       if (unlikely(AUFS_BRANCH_MAX <= add->bindex
3132 +                    || AUFS_BRANCH_MAX - 1 <= bbot)) {
3133 +               pr_err("number of branches exceeded %s\n", add->pathname);
3134 +               goto out;
3135 +       }
3136 +
3137 +       err = -EDOM;
3138 +       if (unlikely(add->bindex < 0 || bbot + 1 < add->bindex)) {
3139 +               pr_err("bad index %d\n", add->bindex);
3140 +               goto out;
3141 +       }
3142 +
3143 +       inode = d_inode(add->path.dentry);
3144 +       err = -ENOENT;
3145 +       if (unlikely(!inode->i_nlink)) {
3146 +               pr_err("no existence %s\n", add->pathname);
3147 +               goto out;
3148 +       }
3149 +
3150 +       err = -EINVAL;
3151 +       if (unlikely(inode->i_sb == sb)) {
3152 +               pr_err("%s must be outside\n", add->pathname);
3153 +               goto out;
3154 +       }
3155 +
3156 +       if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
3157 +               pr_err("unsupported filesystem, %s (%s)\n",
3158 +                      add->pathname, au_sbtype(inode->i_sb));
3159 +               goto out;
3160 +       }
3161 +
3162 +       if (unlikely(inode->i_sb->s_stack_depth)) {
3163 +               pr_err("already stacked, %s (%s)\n",
3164 +                      add->pathname, au_sbtype(inode->i_sb));
3165 +               goto out;
3166 +       }
3167 +
3168 +       err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3169 +       if (unlikely(err))
3170 +               goto out;
3171 +
3172 +       if (bbot < 0)
3173 +               return 0; /* success */
3174 +
3175 +       err = -EINVAL;
3176 +       for (bindex = 0; bindex <= bbot; bindex++)
3177 +               if (unlikely(test_overlap(sb, add->path.dentry,
3178 +                                         au_h_dptr(root, bindex)))) {
3179 +                       pr_err("%s is overlapped\n", add->pathname);
3180 +                       goto out;
3181 +               }
3182 +
3183 +       err = 0;
3184 +       if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3185 +               h_dentry = au_h_dptr(root, 0);
3186 +               h_inode = d_inode(h_dentry);
3187 +               if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3188 +                   || !uid_eq(h_inode->i_uid, inode->i_uid)
3189 +                   || !gid_eq(h_inode->i_gid, inode->i_gid))
3190 +                       pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3191 +                               add->pathname,
3192 +                               i_uid_read(inode), i_gid_read(inode),
3193 +                               (inode->i_mode & S_IALLUGO),
3194 +                               i_uid_read(h_inode), i_gid_read(h_inode),
3195 +                               (h_inode->i_mode & S_IALLUGO));
3196 +       }
3197 +
3198 +out:
3199 +       return err;
3200 +}
3201 +
3202 +/*
3203 + * initialize or clean the whiteouts for an adding branch
3204 + */
3205 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3206 +                        int new_perm)
3207 +{
3208 +       int err, old_perm;
3209 +       aufs_bindex_t bindex;
3210 +       struct inode *h_inode;
3211 +       struct au_wbr *wbr;
3212 +       struct au_hinode *hdir;
3213 +       struct dentry *h_dentry;
3214 +
3215 +       err = vfsub_mnt_want_write(au_br_mnt(br));
3216 +       if (unlikely(err))
3217 +               goto out;
3218 +
3219 +       wbr = br->br_wbr;
3220 +       old_perm = br->br_perm;
3221 +       br->br_perm = new_perm;
3222 +       hdir = NULL;
3223 +       h_inode = NULL;
3224 +       bindex = au_br_index(sb, br->br_id);
3225 +       if (0 <= bindex) {
3226 +               hdir = au_hi(d_inode(sb->s_root), bindex);
3227 +               au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
3228 +       } else {
3229 +               h_dentry = au_br_dentry(br);
3230 +               h_inode = d_inode(h_dentry);
3231 +               inode_lock_nested(h_inode, AuLsc_I_PARENT);
3232 +       }
3233 +       if (!wbr)
3234 +               err = au_wh_init(br, sb);
3235 +       else {
3236 +               wbr_wh_write_lock(wbr);
3237 +               err = au_wh_init(br, sb);
3238 +               wbr_wh_write_unlock(wbr);
3239 +       }
3240 +       if (hdir)
3241 +               au_hn_inode_unlock(hdir);
3242 +       else
3243 +               inode_unlock(h_inode);
3244 +       vfsub_mnt_drop_write(au_br_mnt(br));
3245 +       br->br_perm = old_perm;
3246 +
3247 +       if (!err && wbr && !au_br_writable(new_perm)) {
3248 +               au_kfree_rcu(wbr);
3249 +               br->br_wbr = NULL;
3250 +       }
3251 +
3252 +out:
3253 +       return err;
3254 +}
3255 +
3256 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3257 +                      int perm)
3258 +{
3259 +       int err;
3260 +       struct kstatfs kst;
3261 +       struct au_wbr *wbr;
3262 +
3263 +       wbr = br->br_wbr;
3264 +       au_rw_init(&wbr->wbr_wh_rwsem);
3265 +       atomic_set(&wbr->wbr_wh_running, 0);
3266 +
3267 +       /*
3268 +        * a limit for rmdir/rename a dir
3269 +        * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3270 +        */
3271 +       err = vfs_statfs(&br->br_path, &kst);
3272 +       if (unlikely(err))
3273 +               goto out;
3274 +       err = -EINVAL;
3275 +       if (kst.f_namelen >= NAME_MAX)
3276 +               err = au_br_init_wh(sb, br, perm);
3277 +       else
3278 +               pr_err("%pd(%s), unsupported namelen %ld\n",
3279 +                      au_br_dentry(br),
3280 +                      au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3281 +
3282 +out:
3283 +       return err;
3284 +}
3285 +
3286 +/* initialize a new branch */
3287 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3288 +                     struct au_opt_add *add)
3289 +{
3290 +       int err;
3291 +       struct au_branch *brbase;
3292 +       struct file *xf;
3293 +       struct inode *h_inode;
3294 +
3295 +       err = 0;
3296 +       br->br_perm = add->perm;
3297 +       br->br_path = add->path; /* set first, path_get() later */
3298 +       spin_lock_init(&br->br_dykey_lock);
3299 +       au_lcnt_init(&br->br_nfiles, /*release*/NULL);
3300 +       au_lcnt_init(&br->br_count, /*release*/NULL);
3301 +       br->br_id = au_new_br_id(sb);
3302 +       AuDebugOn(br->br_id < 0);
3303 +
3304 +       /* always, regardless the given option */
3305 +       err = au_dr_br_init(sb, br, &add->path);
3306 +       if (unlikely(err))
3307 +               goto out_err;
3308 +
3309 +       if (au_br_writable(add->perm)) {
3310 +               err = au_wbr_init(br, sb, add->perm);
3311 +               if (unlikely(err))
3312 +                       goto out_err;
3313 +       }
3314 +
3315 +       if (au_opt_test(au_mntflags(sb), XINO)) {
3316 +               brbase = au_sbr(sb, 0);
3317 +               xf = au_xino_file(brbase->br_xino, /*idx*/-1);
3318 +               AuDebugOn(!xf);
3319 +               h_inode = d_inode(add->path.dentry);
3320 +               err = au_xino_init_br(sb, br, h_inode->i_ino, &xf->f_path);
3321 +               if (unlikely(err)) {
3322 +                       AuDebugOn(au_xino_file(br->br_xino, /*idx*/-1));
3323 +                       goto out_err;
3324 +               }
3325 +       }
3326 +
3327 +       sysaufs_br_init(br);
3328 +       path_get(&br->br_path);
3329 +       goto out; /* success */
3330 +
3331 +out_err:
3332 +       memset(&br->br_path, 0, sizeof(br->br_path));
3333 +out:
3334 +       return err;
3335 +}
3336 +
3337 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3338 +                            struct au_branch *br, aufs_bindex_t bbot,
3339 +                            aufs_bindex_t amount)
3340 +{
3341 +       struct au_branch **brp;
3342 +
3343 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
3344 +
3345 +       brp = sbinfo->si_branch + bindex;
3346 +       memmove(brp + 1, brp, sizeof(*brp) * amount);
3347 +       *brp = br;
3348 +       sbinfo->si_bbot++;
3349 +       if (unlikely(bbot < 0))
3350 +               sbinfo->si_bbot = 0;
3351 +}
3352 +
3353 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3354 +                            aufs_bindex_t bbot, aufs_bindex_t amount)
3355 +{
3356 +       struct au_hdentry *hdp;
3357 +
3358 +       AuRwMustWriteLock(&dinfo->di_rwsem);
3359 +
3360 +       hdp = au_hdentry(dinfo, bindex);
3361 +       memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3362 +       au_h_dentry_init(hdp);
3363 +       dinfo->di_bbot++;
3364 +       if (unlikely(bbot < 0))
3365 +               dinfo->di_btop = 0;
3366 +}
3367 +
3368 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3369 +                            aufs_bindex_t bbot, aufs_bindex_t amount)
3370 +{
3371 +       struct au_hinode *hip;
3372 +
3373 +       AuRwMustWriteLock(&iinfo->ii_rwsem);
3374 +
3375 +       hip = au_hinode(iinfo, bindex);
3376 +       memmove(hip + 1, hip, sizeof(*hip) * amount);
3377 +       au_hinode_init(hip);
3378 +       iinfo->ii_bbot++;
3379 +       if (unlikely(bbot < 0))
3380 +               iinfo->ii_btop = 0;
3381 +}
3382 +
3383 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3384 +                        aufs_bindex_t bindex)
3385 +{
3386 +       struct dentry *root, *h_dentry;
3387 +       struct inode *root_inode, *h_inode;
3388 +       aufs_bindex_t bbot, amount;
3389 +
3390 +       root = sb->s_root;
3391 +       root_inode = d_inode(root);
3392 +       bbot = au_sbbot(sb);
3393 +       amount = bbot + 1 - bindex;
3394 +       h_dentry = au_br_dentry(br);
3395 +       au_sbilist_lock();
3396 +       au_br_do_add_brp(au_sbi(sb), bindex, br, bbot, amount);
3397 +       au_br_do_add_hdp(au_di(root), bindex, bbot, amount);
3398 +       au_br_do_add_hip(au_ii(root_inode), bindex, bbot, amount);
3399 +       au_set_h_dptr(root, bindex, dget(h_dentry));
3400 +       h_inode = d_inode(h_dentry);
3401 +       au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3402 +       au_sbilist_unlock();
3403 +}
3404 +
3405 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3406 +{
3407 +       int err;
3408 +       aufs_bindex_t bbot, add_bindex;
3409 +       struct dentry *root, *h_dentry;
3410 +       struct inode *root_inode;
3411 +       struct au_branch *add_branch;
3412 +
3413 +       root = sb->s_root;
3414 +       root_inode = d_inode(root);
3415 +       IMustLock(root_inode);
3416 +       IiMustWriteLock(root_inode);
3417 +       err = test_add(sb, add, remount);
3418 +       if (unlikely(err < 0))
3419 +               goto out;
3420 +       if (err) {
3421 +               err = 0;
3422 +               goto out; /* success */
3423 +       }
3424 +
3425 +       bbot = au_sbbot(sb);
3426 +       add_branch = au_br_alloc(sb, bbot + 2, add->perm);
3427 +       err = PTR_ERR(add_branch);
3428 +       if (IS_ERR(add_branch))
3429 +               goto out;
3430 +
3431 +       err = au_br_init(add_branch, sb, add);
3432 +       if (unlikely(err)) {
3433 +               au_br_do_free(add_branch);
3434 +               goto out;
3435 +       }
3436 +
3437 +       add_bindex = add->bindex;
3438 +       sysaufs_brs_del(sb, add_bindex);        /* remove successors */
3439 +       au_br_do_add(sb, add_branch, add_bindex);
3440 +       sysaufs_brs_add(sb, add_bindex);        /* append successors */
3441 +       dbgaufs_brs_add(sb, add_bindex, /*topdown*/0);  /* rename successors */
3442 +
3443 +       h_dentry = add->path.dentry;
3444 +       if (!add_bindex) {
3445 +               au_cpup_attr_all(root_inode, /*force*/1);
3446 +               sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3447 +       } else
3448 +               au_add_nlink(root_inode, d_inode(h_dentry));
3449 +
3450 +out:
3451 +       return err;
3452 +}
3453 +
3454 +/* ---------------------------------------------------------------------- */
3455 +
3456 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3457 +                                      unsigned long long max __maybe_unused,
3458 +                                      void *arg)
3459 +{
3460 +       unsigned long long n;
3461 +       struct file **p, *f;
3462 +       struct hlist_bl_head *files;
3463 +       struct hlist_bl_node *pos;
3464 +       struct au_finfo *finfo;
3465 +
3466 +       n = 0;
3467 +       p = a;
3468 +       files = &au_sbi(sb)->si_files;
3469 +       hlist_bl_lock(files);
3470 +       hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
3471 +               f = finfo->fi_file;
3472 +               if (file_count(f)
3473 +                   && !special_file(file_inode(f)->i_mode)) {
3474 +                       get_file(f);
3475 +                       *p++ = f;
3476 +                       n++;
3477 +                       AuDebugOn(n > max);
3478 +               }
3479 +       }
3480 +       hlist_bl_unlock(files);
3481 +
3482 +       return n;
3483 +}
3484 +
3485 +static struct file **au_farray_alloc(struct super_block *sb,
3486 +                                    unsigned long long *max)
3487 +{
3488 +       struct au_sbinfo *sbi;
3489 +
3490 +       sbi = au_sbi(sb);
3491 +       *max = au_lcnt_read(&sbi->si_nfiles, /*do_rev*/1);
3492 +       return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3493 +}
3494 +
3495 +static void au_farray_free(struct file **a, unsigned long long max)
3496 +{
3497 +       unsigned long long ull;
3498 +
3499 +       for (ull = 0; ull < max; ull++)
3500 +               if (a[ull])
3501 +                       fput(a[ull]);
3502 +       kvfree(a);
3503 +}
3504 +
3505 +/* ---------------------------------------------------------------------- */
3506 +
3507 +/*
3508 + * delete a branch
3509 + */
3510 +
3511 +/* to show the line number, do not make it inlined function */
3512 +#define AuVerbose(do_info, fmt, ...) do { \
3513 +       if (do_info) \
3514 +               pr_info(fmt, ##__VA_ARGS__); \
3515 +} while (0)
3516 +
3517 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t btop,
3518 +                        aufs_bindex_t bbot)
3519 +{
3520 +       return (inode && !S_ISDIR(inode->i_mode)) || btop == bbot;
3521 +}
3522 +
3523 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t btop,
3524 +                        aufs_bindex_t bbot)
3525 +{
3526 +       return au_test_ibusy(d_inode(dentry), btop, bbot);
3527 +}
3528 +
3529 +/*
3530 + * test if the branch is deletable or not.
3531 + */
3532 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3533 +                           unsigned int sigen, const unsigned int verbose)
3534 +{
3535 +       int err, i, j, ndentry;
3536 +       aufs_bindex_t btop, bbot;
3537 +       struct au_dcsub_pages dpages;
3538 +       struct au_dpage *dpage;
3539 +       struct dentry *d;
3540 +
3541 +       err = au_dpages_init(&dpages, GFP_NOFS);
3542 +       if (unlikely(err))
3543 +               goto out;
3544 +       err = au_dcsub_pages(&dpages, root, NULL, NULL);
3545 +       if (unlikely(err))
3546 +               goto out_dpages;
3547 +
3548 +       for (i = 0; !err && i < dpages.ndpage; i++) {
3549 +               dpage = dpages.dpages + i;
3550 +               ndentry = dpage->ndentry;
3551 +               for (j = 0; !err && j < ndentry; j++) {
3552 +                       d = dpage->dentries[j];
3553 +                       AuDebugOn(au_dcount(d) <= 0);
3554 +                       if (!au_digen_test(d, sigen)) {
3555 +                               di_read_lock_child(d, AuLock_IR);
3556 +                               if (unlikely(au_dbrange_test(d))) {
3557 +                                       di_read_unlock(d, AuLock_IR);
3558 +                                       continue;
3559 +                               }
3560 +                       } else {
3561 +                               di_write_lock_child(d);
3562 +                               if (unlikely(au_dbrange_test(d))) {
3563 +                                       di_write_unlock(d);
3564 +                                       continue;
3565 +                               }
3566 +                               err = au_reval_dpath(d, sigen);
3567 +                               if (!err)
3568 +                                       di_downgrade_lock(d, AuLock_IR);
3569 +                               else {
3570 +                                       di_write_unlock(d);
3571 +                                       break;
3572 +                               }
3573 +                       }
3574 +
3575 +                       /* AuDbgDentry(d); */
3576 +                       btop = au_dbtop(d);
3577 +                       bbot = au_dbbot(d);
3578 +                       if (btop <= bindex
3579 +                           && bindex <= bbot
3580 +                           && au_h_dptr(d, bindex)
3581 +                           && au_test_dbusy(d, btop, bbot)) {
3582 +                               err = -EBUSY;
3583 +                               AuVerbose(verbose, "busy %pd\n", d);
3584 +                               AuDbgDentry(d);
3585 +                       }
3586 +                       di_read_unlock(d, AuLock_IR);
3587 +               }
3588 +       }
3589 +
3590 +out_dpages:
3591 +       au_dpages_free(&dpages);
3592 +out:
3593 +       return err;
3594 +}
3595 +
3596 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3597 +                          unsigned int sigen, const unsigned int verbose)
3598 +{
3599 +       int err;
3600 +       unsigned long long max, ull;
3601 +       struct inode *i, **array;
3602 +       aufs_bindex_t btop, bbot;
3603 +
3604 +       array = au_iarray_alloc(sb, &max);
3605 +       err = PTR_ERR(array);
3606 +       if (IS_ERR(array))
3607 +               goto out;
3608 +
3609 +       err = 0;
3610 +       AuDbg("b%d\n", bindex);
3611 +       for (ull = 0; !err && ull < max; ull++) {
3612 +               i = array[ull];
3613 +               if (unlikely(!i))
3614 +                       break;
3615 +               if (i->i_ino == AUFS_ROOT_INO)
3616 +                       continue;
3617 +
3618 +               /* AuDbgInode(i); */
3619 +               if (au_iigen(i, NULL) == sigen)
3620 +                       ii_read_lock_child(i);
3621 +               else {
3622 +                       ii_write_lock_child(i);
3623 +                       err = au_refresh_hinode_self(i);
3624 +                       au_iigen_dec(i);
3625 +                       if (!err)
3626 +                               ii_downgrade_lock(i);
3627 +                       else {
3628 +                               ii_write_unlock(i);
3629 +                               break;
3630 +                       }
3631 +               }
3632 +
3633 +               btop = au_ibtop(i);
3634 +               bbot = au_ibbot(i);
3635 +               if (btop <= bindex
3636 +                   && bindex <= bbot
3637 +                   && au_h_iptr(i, bindex)
3638 +                   && au_test_ibusy(i, btop, bbot)) {
3639 +                       err = -EBUSY;
3640 +                       AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3641 +                       AuDbgInode(i);
3642 +               }
3643 +               ii_read_unlock(i);
3644 +       }
3645 +       au_iarray_free(array, max);
3646 +
3647 +out:
3648 +       return err;
3649 +}
3650 +
3651 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3652 +                             const unsigned int verbose)
3653 +{
3654 +       int err;
3655 +       unsigned int sigen;
3656 +
3657 +       sigen = au_sigen(root->d_sb);
3658 +       DiMustNoWaiters(root);
3659 +       IiMustNoWaiters(d_inode(root));
3660 +       di_write_unlock(root);
3661 +       err = test_dentry_busy(root, bindex, sigen, verbose);
3662 +       if (!err)
3663 +               err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3664 +       di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3665 +
3666 +       return err;
3667 +}
3668 +
3669 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3670 +                        struct file **to_free, int *idx)
3671 +{
3672 +       int err;
3673 +       unsigned char matched, root;
3674 +       aufs_bindex_t bindex, bbot;
3675 +       struct au_fidir *fidir;
3676 +       struct au_hfile *hfile;
3677 +
3678 +       err = 0;
3679 +       root = IS_ROOT(file->f_path.dentry);
3680 +       if (root) {
3681 +               get_file(file);
3682 +               to_free[*idx] = file;
3683 +               (*idx)++;
3684 +               goto out;
3685 +       }
3686 +
3687 +       matched = 0;
3688 +       fidir = au_fi(file)->fi_hdir;
3689 +       AuDebugOn(!fidir);
3690 +       bbot = au_fbbot_dir(file);
3691 +       for (bindex = au_fbtop(file); bindex <= bbot; bindex++) {
3692 +               hfile = fidir->fd_hfile + bindex;
3693 +               if (!hfile->hf_file)
3694 +                       continue;
3695 +
3696 +               if (hfile->hf_br->br_id == br_id) {
3697 +                       matched = 1;
3698 +                       break;
3699 +               }
3700 +       }
3701 +       if (matched)
3702 +               err = -EBUSY;
3703 +
3704 +out:
3705 +       return err;
3706 +}
3707 +
3708 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3709 +                         struct file **to_free, int opened)
3710 +{
3711 +       int err, idx;
3712 +       unsigned long long ull, max;
3713 +       aufs_bindex_t btop;
3714 +       struct file *file, **array;
3715 +       struct dentry *root;
3716 +       struct au_hfile *hfile;
3717 +
3718 +       array = au_farray_alloc(sb, &max);
3719 +       err = PTR_ERR(array);
3720 +       if (IS_ERR(array))
3721 +               goto out;
3722 +
3723 +       err = 0;
3724 +       idx = 0;
3725 +       root = sb->s_root;
3726 +       di_write_unlock(root);
3727 +       for (ull = 0; ull < max; ull++) {
3728 +               file = array[ull];
3729 +               if (unlikely(!file))
3730 +                       break;
3731 +
3732 +               /* AuDbg("%pD\n", file); */
3733 +               fi_read_lock(file);
3734 +               btop = au_fbtop(file);
3735 +               if (!d_is_dir(file->f_path.dentry)) {
3736 +                       hfile = &au_fi(file)->fi_htop;
3737 +                       if (hfile->hf_br->br_id == br_id)
3738 +                               err = -EBUSY;
3739 +               } else
3740 +                       err = test_dir_busy(file, br_id, to_free, &idx);
3741 +               fi_read_unlock(file);
3742 +               if (unlikely(err))
3743 +                       break;
3744 +       }
3745 +       di_write_lock_child(root);
3746 +       au_farray_free(array, max);
3747 +       AuDebugOn(idx > opened);
3748 +
3749 +out:
3750 +       return err;
3751 +}
3752 +
3753 +static void br_del_file(struct file **to_free, unsigned long long opened,
3754 +                       aufs_bindex_t br_id)
3755 +{
3756 +       unsigned long long ull;
3757 +       aufs_bindex_t bindex, btop, bbot, bfound;
3758 +       struct file *file;
3759 +       struct au_fidir *fidir;
3760 +       struct au_hfile *hfile;
3761 +
3762 +       for (ull = 0; ull < opened; ull++) {
3763 +               file = to_free[ull];
3764 +               if (unlikely(!file))
3765 +                       break;
3766 +
3767 +               /* AuDbg("%pD\n", file); */
3768 +               AuDebugOn(!d_is_dir(file->f_path.dentry));
3769 +               bfound = -1;
3770 +               fidir = au_fi(file)->fi_hdir;
3771 +               AuDebugOn(!fidir);
3772 +               fi_write_lock(file);
3773 +               btop = au_fbtop(file);
3774 +               bbot = au_fbbot_dir(file);
3775 +               for (bindex = btop; bindex <= bbot; bindex++) {
3776 +                       hfile = fidir->fd_hfile + bindex;
3777 +                       if (!hfile->hf_file)
3778 +                               continue;
3779 +
3780 +                       if (hfile->hf_br->br_id == br_id) {
3781 +                               bfound = bindex;
3782 +                               break;
3783 +                       }
3784 +               }
3785 +               AuDebugOn(bfound < 0);
3786 +               au_set_h_fptr(file, bfound, NULL);
3787 +               if (bfound == btop) {
3788 +                       for (btop++; btop <= bbot; btop++)
3789 +                               if (au_hf_dir(file, btop)) {
3790 +                                       au_set_fbtop(file, btop);
3791 +                                       break;
3792 +                               }
3793 +               }
3794 +               fi_write_unlock(file);
3795 +       }
3796 +}
3797 +
3798 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3799 +                            const aufs_bindex_t bindex,
3800 +                            const aufs_bindex_t bbot)
3801 +{
3802 +       struct au_branch **brp, **p;
3803 +
3804 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
3805 +
3806 +       brp = sbinfo->si_branch + bindex;
3807 +       if (bindex < bbot)
3808 +               memmove(brp, brp + 1, sizeof(*brp) * (bbot - bindex));
3809 +       sbinfo->si_branch[0 + bbot] = NULL;
3810 +       sbinfo->si_bbot--;
3811 +
3812 +       p = au_krealloc(sbinfo->si_branch, sizeof(*p) * bbot, AuGFP_SBILIST,
3813 +                       /*may_shrink*/1);
3814 +       if (p)
3815 +               sbinfo->si_branch = p;
3816 +       /* harmless error */
3817 +}
3818 +
3819 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3820 +                            const aufs_bindex_t bbot)
3821 +{
3822 +       struct au_hdentry *hdp, *p;
3823 +
3824 +       AuRwMustWriteLock(&dinfo->di_rwsem);
3825 +
3826 +       hdp = au_hdentry(dinfo, bindex);
3827 +       if (bindex < bbot)
3828 +               memmove(hdp, hdp + 1, sizeof(*hdp) * (bbot - bindex));
3829 +       /* au_h_dentry_init(au_hdentry(dinfo, bbot); */
3830 +       dinfo->di_bbot--;
3831 +
3832 +       p = au_krealloc(dinfo->di_hdentry, sizeof(*p) * bbot, AuGFP_SBILIST,
3833 +                       /*may_shrink*/1);
3834 +       if (p)
3835 +               dinfo->di_hdentry = p;
3836 +       /* harmless error */
3837 +}
3838 +
3839 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3840 +                            const aufs_bindex_t bbot)
3841 +{
3842 +       struct au_hinode *hip, *p;
3843 +
3844 +       AuRwMustWriteLock(&iinfo->ii_rwsem);
3845 +
3846 +       hip = au_hinode(iinfo, bindex);
3847 +       if (bindex < bbot)
3848 +               memmove(hip, hip + 1, sizeof(*hip) * (bbot - bindex));
3849 +       /* au_hinode_init(au_hinode(iinfo, bbot)); */
3850 +       iinfo->ii_bbot--;
3851 +
3852 +       p = au_krealloc(iinfo->ii_hinode, sizeof(*p) * bbot, AuGFP_SBILIST,
3853 +                       /*may_shrink*/1);
3854 +       if (p)
3855 +               iinfo->ii_hinode = p;
3856 +       /* harmless error */
3857 +}
3858 +
3859 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3860 +                        struct au_branch *br)
3861 +{
3862 +       aufs_bindex_t bbot;
3863 +       struct au_sbinfo *sbinfo;
3864 +       struct dentry *root, *h_root;
3865 +       struct inode *inode, *h_inode;
3866 +       struct au_hinode *hinode;
3867 +
3868 +       SiMustWriteLock(sb);
3869 +
3870 +       root = sb->s_root;
3871 +       inode = d_inode(root);
3872 +       sbinfo = au_sbi(sb);
3873 +       bbot = sbinfo->si_bbot;
3874 +
3875 +       h_root = au_h_dptr(root, bindex);
3876 +       hinode = au_hi(inode, bindex);
3877 +       h_inode = au_igrab(hinode->hi_inode);
3878 +       au_hiput(hinode);
3879 +
3880 +       au_sbilist_lock();
3881 +       au_br_do_del_brp(sbinfo, bindex, bbot);
3882 +       au_br_do_del_hdp(au_di(root), bindex, bbot);
3883 +       au_br_do_del_hip(au_ii(inode), bindex, bbot);
3884 +       au_sbilist_unlock();
3885 +
3886 +       /* ignore an error */
3887 +       au_dr_br_fin(sb, br); /* always, regardless the mount option */
3888 +
3889 +       dput(h_root);
3890 +       iput(h_inode);
3891 +       au_br_do_free(br);
3892 +}
3893 +
3894 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3895 +                                  unsigned long long max, void *arg)
3896 +{
3897 +       return max;
3898 +}
3899 +
3900 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3901 +{
3902 +       int err, rerr, i;
3903 +       unsigned long long opened;
3904 +       unsigned int mnt_flags;
3905 +       aufs_bindex_t bindex, bbot, br_id;
3906 +       unsigned char do_wh, verbose;
3907 +       struct au_branch *br;
3908 +       struct au_wbr *wbr;
3909 +       struct dentry *root;
3910 +       struct file **to_free;
3911 +
3912 +       err = 0;
3913 +       opened = 0;
3914 +       to_free = NULL;
3915 +       root = sb->s_root;
3916 +       bindex = au_find_dbindex(root, del->h_path.dentry);
3917 +       if (bindex < 0) {
3918 +               if (remount)
3919 +                       goto out; /* success */
3920 +               err = -ENOENT;
3921 +               pr_err("%s no such branch\n", del->pathname);
3922 +               goto out;
3923 +       }
3924 +       AuDbg("bindex b%d\n", bindex);
3925 +
3926 +       err = -EBUSY;
3927 +       mnt_flags = au_mntflags(sb);
3928 +       verbose = !!au_opt_test(mnt_flags, VERBOSE);
3929 +       bbot = au_sbbot(sb);
3930 +       if (unlikely(!bbot)) {
3931 +               AuVerbose(verbose, "no more branches left\n");
3932 +               goto out;
3933 +       }
3934 +
3935 +       br = au_sbr(sb, bindex);
3936 +       AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3937 +       if (unlikely(au_lcnt_read(&br->br_count, /*do_rev*/1))) {
3938 +               AuVerbose(verbose, "br %pd2 is busy now\n", del->h_path.dentry);
3939 +               goto out;
3940 +       }
3941 +
3942 +       br_id = br->br_id;
3943 +       opened = au_lcnt_read(&br->br_nfiles, /*do_rev*/1);
3944 +       if (unlikely(opened)) {
3945 +               to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3946 +               err = PTR_ERR(to_free);
3947 +               if (IS_ERR(to_free))
3948 +                       goto out;
3949 +
3950 +               err = test_file_busy(sb, br_id, to_free, opened);
3951 +               if (unlikely(err)) {
3952 +                       AuVerbose(verbose, "%llu file(s) opened\n", opened);
3953 +                       goto out;
3954 +               }
3955 +       }
3956 +
3957 +       wbr = br->br_wbr;
3958 +       do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3959 +       if (do_wh) {
3960 +               /* instead of WbrWhMustWriteLock(wbr) */
3961 +               SiMustWriteLock(sb);
3962 +               for (i = 0; i < AuBrWh_Last; i++) {
3963 +                       dput(wbr->wbr_wh[i]);
3964 +                       wbr->wbr_wh[i] = NULL;
3965 +               }
3966 +       }
3967 +
3968 +       err = test_children_busy(root, bindex, verbose);
3969 +       if (unlikely(err)) {
3970 +               if (do_wh)
3971 +                       goto out_wh;
3972 +               goto out;
3973 +       }
3974 +
3975 +       err = 0;
3976 +       if (to_free) {
3977 +               /*
3978 +                * now we confirmed the branch is deletable.
3979 +                * let's free the remaining opened dirs on the branch.
3980 +                */
3981 +               di_write_unlock(root);
3982 +               br_del_file(to_free, opened, br_id);
3983 +               di_write_lock_child(root);
3984 +       }
3985 +
3986 +       sysaufs_brs_del(sb, bindex);    /* remove successors */
3987 +       dbgaufs_xino_del(br);           /* remove one */
3988 +       au_br_do_del(sb, bindex, br);
3989 +       sysaufs_brs_add(sb, bindex);    /* append successors */
3990 +       dbgaufs_brs_add(sb, bindex, /*topdown*/1);      /* rename successors */
3991 +
3992 +       if (!bindex) {
3993 +               au_cpup_attr_all(d_inode(root), /*force*/1);
3994 +               sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3995 +       } else
3996 +               au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
3997 +       if (au_opt_test(mnt_flags, PLINK))
3998 +               au_plink_half_refresh(sb, br_id);
3999 +
4000 +       goto out; /* success */
4001 +
4002 +out_wh:
4003 +       /* revert */
4004 +       rerr = au_br_init_wh(sb, br, br->br_perm);
4005 +       if (rerr)
4006 +               pr_warn("failed re-creating base whiteout, %s. (%d)\n",
4007 +                       del->pathname, rerr);
4008 +out:
4009 +       if (to_free)
4010 +               au_farray_free(to_free, opened);
4011 +       return err;
4012 +}
4013 +
4014 +/* ---------------------------------------------------------------------- */
4015 +
4016 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
4017 +{
4018 +       int err;
4019 +       aufs_bindex_t btop, bbot;
4020 +       struct aufs_ibusy ibusy;
4021 +       struct inode *inode, *h_inode;
4022 +
4023 +       err = -EPERM;
4024 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
4025 +               goto out;
4026 +
4027 +       err = copy_from_user(&ibusy, arg, sizeof(ibusy));
4028 +       if (!err)
4029 +               /* VERIFY_WRITE */
4030 +               err = !access_ok(&arg->h_ino, sizeof(arg->h_ino));
4031 +       if (unlikely(err)) {
4032 +               err = -EFAULT;
4033 +               AuTraceErr(err);
4034 +               goto out;
4035 +       }
4036 +
4037 +       err = -EINVAL;
4038 +       si_read_lock(sb, AuLock_FLUSH);
4039 +       if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbbot(sb)))
4040 +               goto out_unlock;
4041 +
4042 +       err = 0;
4043 +       ibusy.h_ino = 0; /* invalid */
4044 +       inode = ilookup(sb, ibusy.ino);
4045 +       if (!inode
4046 +           || inode->i_ino == AUFS_ROOT_INO
4047 +           || au_is_bad_inode(inode))
4048 +               goto out_unlock;
4049 +
4050 +       ii_read_lock_child(inode);
4051 +       btop = au_ibtop(inode);
4052 +       bbot = au_ibbot(inode);
4053 +       if (btop <= ibusy.bindex && ibusy.bindex <= bbot) {
4054 +               h_inode = au_h_iptr(inode, ibusy.bindex);
4055 +               if (h_inode && au_test_ibusy(inode, btop, bbot))
4056 +                       ibusy.h_ino = h_inode->i_ino;
4057 +       }
4058 +       ii_read_unlock(inode);
4059 +       iput(inode);
4060 +
4061 +out_unlock:
4062 +       si_read_unlock(sb);
4063 +       if (!err) {
4064 +               err = __put_user(ibusy.h_ino, &arg->h_ino);
4065 +               if (unlikely(err)) {
4066 +                       err = -EFAULT;
4067 +                       AuTraceErr(err);
4068 +               }
4069 +       }
4070 +out:
4071 +       return err;
4072 +}
4073 +
4074 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
4075 +{
4076 +       return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
4077 +}
4078 +
4079 +#ifdef CONFIG_COMPAT
4080 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
4081 +{
4082 +       return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
4083 +}
4084 +#endif
4085 +
4086 +/* ---------------------------------------------------------------------- */
4087 +
4088 +/*
4089 + * change a branch permission
4090 + */
4091 +
4092 +static void au_warn_ima(void)
4093 +{
4094 +#ifdef CONFIG_IMA
4095 +       /* since it doesn't support mark_files_ro() */
4096 +       AuWarn1("RW -> RO makes IMA to produce wrong message\n");
4097 +#endif
4098 +}
4099 +
4100 +static int do_need_sigen_inc(int a, int b)
4101 +{
4102 +       return au_br_whable(a) && !au_br_whable(b);
4103 +}
4104 +
4105 +static int need_sigen_inc(int old, int new)
4106 +{
4107 +       return do_need_sigen_inc(old, new)
4108 +               || do_need_sigen_inc(new, old);
4109 +}
4110 +
4111 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
4112 +{
4113 +       int err, do_warn;
4114 +       unsigned int mnt_flags;
4115 +       unsigned long long ull, max;
4116 +       aufs_bindex_t br_id;
4117 +       unsigned char verbose, writer;
4118 +       struct file *file, *hf, **array;
4119 +       struct au_hfile *hfile;
4120 +       struct inode *h_inode;
4121 +
4122 +       mnt_flags = au_mntflags(sb);
4123 +       verbose = !!au_opt_test(mnt_flags, VERBOSE);
4124 +
4125 +       array = au_farray_alloc(sb, &max);
4126 +       err = PTR_ERR(array);
4127 +       if (IS_ERR(array))
4128 +               goto out;
4129 +
4130 +       do_warn = 0;
4131 +       br_id = au_sbr_id(sb, bindex);
4132 +       for (ull = 0; ull < max; ull++) {
4133 +               file = array[ull];
4134 +               if (unlikely(!file))
4135 +                       break;
4136 +
4137 +               /* AuDbg("%pD\n", file); */
4138 +               fi_read_lock(file);
4139 +               if (unlikely(au_test_mmapped(file))) {
4140 +                       err = -EBUSY;
4141 +                       AuVerbose(verbose, "mmapped %pD\n", file);
4142 +                       AuDbgFile(file);
4143 +                       FiMustNoWaiters(file);
4144 +                       fi_read_unlock(file);
4145 +                       goto out_array;
4146 +               }
4147 +
4148 +               hfile = &au_fi(file)->fi_htop;
4149 +               hf = hfile->hf_file;
4150 +               if (!d_is_reg(file->f_path.dentry)
4151 +                   || !(file->f_mode & FMODE_WRITE)
4152 +                   || hfile->hf_br->br_id != br_id
4153 +                   || !(hf->f_mode & FMODE_WRITE))
4154 +                       array[ull] = NULL;
4155 +               else {
4156 +                       do_warn = 1;
4157 +                       get_file(file);
4158 +               }
4159 +
4160 +               FiMustNoWaiters(file);
4161 +               fi_read_unlock(file);
4162 +               fput(file);
4163 +       }
4164 +
4165 +       err = 0;
4166 +       if (do_warn)
4167 +               au_warn_ima();
4168 +
4169 +       for (ull = 0; ull < max; ull++) {
4170 +               file = array[ull];
4171 +               if (!file)
4172 +                       continue;
4173 +
4174 +               /* todo: already flushed? */
4175 +               /*
4176 +                * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4177 +                * approach which resets f_mode and calls mnt_drop_write() and
4178 +                * file_release_write() for each file, because the branch
4179 +                * attribute in aufs world is totally different from the native
4180 +                * fs rw/ro mode.
4181 +                */
4182 +               /* fi_read_lock(file); */
4183 +               hfile = &au_fi(file)->fi_htop;
4184 +               hf = hfile->hf_file;
4185 +               /* fi_read_unlock(file); */
4186 +               spin_lock(&hf->f_lock);
4187 +               writer = !!(hf->f_mode & FMODE_WRITER);
4188 +               hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4189 +               spin_unlock(&hf->f_lock);
4190 +               if (writer) {
4191 +                       h_inode = file_inode(hf);
4192 +                       if (hf->f_mode & FMODE_READ)
4193 +                               i_readcount_inc(h_inode);
4194 +                       put_write_access(h_inode);
4195 +                       __mnt_drop_write(hf->f_path.mnt);
4196 +               }
4197 +       }
4198 +
4199 +out_array:
4200 +       au_farray_free(array, max);
4201 +out:
4202 +       AuTraceErr(err);
4203 +       return err;
4204 +}
4205 +
4206 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4207 +             int *do_refresh)
4208 +{
4209 +       int err, rerr;
4210 +       aufs_bindex_t bindex;
4211 +       struct dentry *root;
4212 +       struct au_branch *br;
4213 +       struct au_br_fhsm *bf;
4214 +
4215 +       root = sb->s_root;
4216 +       bindex = au_find_dbindex(root, mod->h_root);
4217 +       if (bindex < 0) {
4218 +               if (remount)
4219 +                       return 0; /* success */
4220 +               err = -ENOENT;
4221 +               pr_err("%s no such branch\n", mod->path);
4222 +               goto out;
4223 +       }
4224 +       AuDbg("bindex b%d\n", bindex);
4225 +
4226 +       err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4227 +       if (unlikely(err))
4228 +               goto out;
4229 +
4230 +       br = au_sbr(sb, bindex);
4231 +       AuDebugOn(mod->h_root != au_br_dentry(br));
4232 +       if (br->br_perm == mod->perm)
4233 +               return 0; /* success */
4234 +
4235 +       /* pre-allocate for non-fhsm --> fhsm */
4236 +       bf = NULL;
4237 +       if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4238 +               err = au_fhsm_br_alloc(br);
4239 +               if (unlikely(err))
4240 +                       goto out;
4241 +               bf = br->br_fhsm;
4242 +               br->br_fhsm = NULL;
4243 +       }
4244 +
4245 +       if (au_br_writable(br->br_perm)) {
4246 +               /* remove whiteout base */
4247 +               err = au_br_init_wh(sb, br, mod->perm);
4248 +               if (unlikely(err))
4249 +                       goto out_bf;
4250 +
4251 +               if (!au_br_writable(mod->perm)) {
4252 +                       /* rw --> ro, file might be mmapped */
4253 +                       DiMustNoWaiters(root);
4254 +                       IiMustNoWaiters(d_inode(root));
4255 +                       di_write_unlock(root);
4256 +                       err = au_br_mod_files_ro(sb, bindex);
4257 +                       /* aufs_write_lock() calls ..._child() */
4258 +                       di_write_lock_child(root);
4259 +
4260 +                       if (unlikely(err)) {
4261 +                               rerr = -ENOMEM;
4262 +                               br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4263 +                                                    GFP_NOFS);
4264 +                               if (br->br_wbr)
4265 +                                       rerr = au_wbr_init(br, sb, br->br_perm);
4266 +                               if (unlikely(rerr)) {
4267 +                                       AuIOErr("nested error %d (%d)\n",
4268 +                                               rerr, err);
4269 +                                       br->br_perm = mod->perm;
4270 +                               }
4271 +                       }
4272 +               }
4273 +       } else if (au_br_writable(mod->perm)) {
4274 +               /* ro --> rw */
4275 +               err = -ENOMEM;
4276 +               br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4277 +               if (br->br_wbr) {
4278 +                       err = au_wbr_init(br, sb, mod->perm);
4279 +                       if (unlikely(err)) {
4280 +                               au_kfree_rcu(br->br_wbr);
4281 +                               br->br_wbr = NULL;
4282 +                       }
4283 +               }
4284 +       }
4285 +       if (unlikely(err))
4286 +               goto out_bf;
4287 +
4288 +       if (au_br_fhsm(br->br_perm)) {
4289 +               if (!au_br_fhsm(mod->perm)) {
4290 +                       /* fhsm --> non-fhsm */
4291 +                       au_br_fhsm_fin(br->br_fhsm);
4292 +                       au_kfree_rcu(br->br_fhsm);
4293 +                       br->br_fhsm = NULL;
4294 +               }
4295 +       } else if (au_br_fhsm(mod->perm))
4296 +               /* non-fhsm --> fhsm */
4297 +               br->br_fhsm = bf;
4298 +
4299 +       *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4300 +       br->br_perm = mod->perm;
4301 +       goto out; /* success */
4302 +
4303 +out_bf:
4304 +       au_kfree_try_rcu(bf);
4305 +out:
4306 +       AuTraceErr(err);
4307 +       return err;
4308 +}
4309 +
4310 +/* ---------------------------------------------------------------------- */
4311 +
4312 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4313 +{
4314 +       int err;
4315 +       struct kstatfs kstfs;
4316 +
4317 +       err = vfs_statfs(&br->br_path, &kstfs);
4318 +       if (!err) {
4319 +               stfs->f_blocks = kstfs.f_blocks;
4320 +               stfs->f_bavail = kstfs.f_bavail;
4321 +               stfs->f_files = kstfs.f_files;
4322 +               stfs->f_ffree = kstfs.f_ffree;
4323 +       }
4324 +
4325 +       return err;
4326 +}
4327 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4328 --- /usr/share/empty/fs/aufs/branch.h   1970-01-01 01:00:00.000000000 +0100
4329 +++ linux/fs/aufs/branch.h      2022-11-05 23:02:18.959222617 +0100
4330 @@ -0,0 +1,375 @@
4331 +/* SPDX-License-Identifier: GPL-2.0 */
4332 +/*
4333 + * Copyright (C) 2005-2022 Junjiro R. Okajima
4334 + *
4335 + * This program is free software; you can redistribute it and/or modify
4336 + * it under the terms of the GNU General Public License as published by
4337 + * the Free Software Foundation; either version 2 of the License, or
4338 + * (at your option) any later version.
4339 + *
4340 + * This program is distributed in the hope that it will be useful,
4341 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4342 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
4343 + * GNU General Public License for more details.
4344 + *
4345 + * You should have received a copy of the GNU General Public License
4346 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
4347 + */
4348 +
4349 +/*
4350 + * branch filesystems and xino for them
4351 + */
4352 +
4353 +#ifndef __AUFS_BRANCH_H__
4354 +#define __AUFS_BRANCH_H__
4355 +
4356 +#ifdef __KERNEL__
4357 +
4358 +#include <linux/mount.h>
4359 +#include "dirren.h"
4360 +#include "dynop.h"
4361 +#include "lcnt.h"
4362 +#include "rwsem.h"
4363 +#include "super.h"
4364 +
4365 +/* ---------------------------------------------------------------------- */
4366 +
4367 +/* a xino file */
4368 +struct au_xino {
4369 +       struct file             **xi_file;
4370 +       unsigned int            xi_nfile;
4371 +
4372 +       struct {
4373 +               spinlock_t              spin;
4374 +               ino_t                   *array;
4375 +               int                     total;
4376 +               /* reserved for future use */
4377 +               /* unsigned long        *bitmap; */
4378 +               wait_queue_head_t       wqh;
4379 +       } xi_nondir;
4380 +
4381 +       struct mutex            xi_mtx; /* protects xi_file array */
4382 +       struct hlist_bl_head    xi_writing;
4383 +
4384 +       atomic_t                xi_truncating;
4385 +
4386 +       struct kref             xi_kref;
4387 +};
4388 +
4389 +/* File-based Hierarchical Storage Management */
4390 +struct au_br_fhsm {
4391 +#ifdef CONFIG_AUFS_FHSM
4392 +       struct mutex            bf_lock;
4393 +       unsigned long           bf_jiffy;
4394 +       struct aufs_stfs        bf_stfs;
4395 +       int                     bf_readable;
4396 +#endif
4397 +};
4398 +
4399 +/* members for writable branch only */
4400 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4401 +struct au_wbr {
4402 +       struct au_rwsem         wbr_wh_rwsem;
4403 +       struct dentry           *wbr_wh[AuBrWh_Last];
4404 +       atomic_t                wbr_wh_running;
4405 +#define wbr_whbase             wbr_wh[AuBrWh_BASE]     /* whiteout base */
4406 +#define wbr_plink              wbr_wh[AuBrWh_PLINK]    /* pseudo-link dir */
4407 +#define wbr_orph               wbr_wh[AuBrWh_ORPH]     /* dir for orphans */
4408 +
4409 +       /* mfs mode */
4410 +       unsigned long long      wbr_bytes;
4411 +};
4412 +
4413 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4414 +#define AuBrDynOp (AuDyLast * 4)
4415 +
4416 +#ifdef CONFIG_AUFS_HFSNOTIFY
4417 +/* support for asynchronous destruction */
4418 +struct au_br_hfsnotify {
4419 +       struct fsnotify_group   *hfsn_group;
4420 +};
4421 +#endif
4422 +
4423 +/* sysfs entries */
4424 +struct au_brsysfs {
4425 +       char                    name[16];
4426 +       struct attribute        attr;
4427 +};
4428 +
4429 +enum {
4430 +       AuBrSysfs_BR,
4431 +       AuBrSysfs_BRID,
4432 +       AuBrSysfs_Last
4433 +};
4434 +
4435 +/* protected by superblock rwsem */
4436 +struct au_branch {
4437 +       struct au_xino          *br_xino;
4438 +
4439 +       aufs_bindex_t           br_id;
4440 +
4441 +       int                     br_perm;
4442 +       struct path             br_path;
4443 +       spinlock_t              br_dykey_lock;
4444 +       struct au_dykey         *br_dykey[AuBrDynOp];
4445 +       au_lcnt_t               br_nfiles;      /* opened files */
4446 +       au_lcnt_t               br_count;       /* in-use for other */
4447 +
4448 +       struct au_wbr           *br_wbr;
4449 +       struct au_br_fhsm       *br_fhsm;
4450 +
4451 +#ifdef CONFIG_AUFS_HFSNOTIFY
4452 +       struct au_br_hfsnotify  *br_hfsn;
4453 +#endif
4454 +
4455 +#ifdef CONFIG_SYSFS
4456 +       /* entries under sysfs per mount-point */
4457 +       struct au_brsysfs       br_sysfs[AuBrSysfs_Last];
4458 +#endif
4459 +
4460 +#ifdef CONFIG_DEBUG_FS
4461 +       struct dentry            *br_dbgaufs; /* xino */
4462 +#endif
4463 +
4464 +       struct au_dr_br         br_dirren;
4465 +};
4466 +
4467 +/* ---------------------------------------------------------------------- */
4468 +
4469 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4470 +{
4471 +       return br->br_path.mnt;
4472 +}
4473 +
4474 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4475 +{
4476 +       return br->br_path.dentry;
4477 +}
4478 +
4479 +static inline struct user_namespace *au_br_userns(struct au_branch *br)
4480 +{
4481 +       return mnt_user_ns(br->br_path.mnt);
4482 +}
4483 +
4484 +static inline struct super_block *au_br_sb(struct au_branch *br)
4485 +{
4486 +       return au_br_mnt(br)->mnt_sb;
4487 +}
4488 +
4489 +static inline int au_br_rdonly(struct au_branch *br)
4490 +{
4491 +       return (sb_rdonly(au_br_sb(br))
4492 +               || !au_br_writable(br->br_perm))
4493 +               ? -EROFS : 0;
4494 +}
4495 +
4496 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4497 +{
4498 +#ifdef CONFIG_AUFS_HNOTIFY
4499 +       return !(brperm & AuBrPerm_RR);
4500 +#else
4501 +       return 0;
4502 +#endif
4503 +}
4504 +
4505 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4506 +{
4507 +       int err, exec_flag;
4508 +
4509 +       err = 0;
4510 +       exec_flag = oflag & __FMODE_EXEC;
4511 +       if (unlikely(exec_flag && path_noexec(&br->br_path)))
4512 +               err = -EACCES;
4513 +
4514 +       return err;
4515 +}
4516 +
4517 +static inline void au_xino_get(struct au_branch *br)
4518 +{
4519 +       struct au_xino *xi;
4520 +
4521 +       xi = br->br_xino;
4522 +       if (xi)
4523 +               kref_get(&xi->xi_kref);
4524 +}
4525 +
4526 +static inline int au_xino_count(struct au_branch *br)
4527 +{
4528 +       int v;
4529 +       struct au_xino *xi;
4530 +
4531 +       v = 0;
4532 +       xi = br->br_xino;
4533 +       if (xi)
4534 +               v = kref_read(&xi->xi_kref);
4535 +
4536 +       return v;
4537 +}
4538 +
4539 +/* ---------------------------------------------------------------------- */
4540 +
4541 +/* branch.c */
4542 +struct au_sbinfo;
4543 +void au_br_free(struct au_sbinfo *sinfo);
4544 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4545 +struct au_opt_add;
4546 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4547 +struct au_opt_del;
4548 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4549 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4550 +#ifdef CONFIG_COMPAT
4551 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4552 +#endif
4553 +struct au_opt_mod;
4554 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4555 +             int *do_refresh);
4556 +struct aufs_stfs;
4557 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4558 +
4559 +/* xino.c */
4560 +static const loff_t au_loff_max = LLONG_MAX;
4561 +
4562 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry);
4563 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
4564 +                           int wbrtop);
4565 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
4566 +                            struct file *copy_src);
4567 +struct au_xi_new {
4568 +       struct au_xino *xi;     /* switch between xino and xigen */
4569 +       int idx;
4570 +       struct path *base;
4571 +       struct file *copy_src;
4572 +};
4573 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew);
4574 +
4575 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4576 +                ino_t *ino);
4577 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4578 +                 ino_t ino);
4579 +ssize_t xino_fread(struct file *file, void *buf, size_t size, loff_t *pos);
4580 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos);
4581 +
4582 +int au_xib_trunc(struct super_block *sb);
4583 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin);
4584 +
4585 +struct au_xino *au_xino_alloc(unsigned int nfile);
4586 +int au_xino_put(struct au_branch *br);
4587 +struct file *au_xino_file1(struct au_xino *xi);
4588 +
4589 +struct au_opt_xino;
4590 +void au_xino_clr(struct super_block *sb);
4591 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount);
4592 +struct file *au_xino_def(struct super_block *sb);
4593 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4594 +                   struct path *base);
4595 +
4596 +ino_t au_xino_new_ino(struct super_block *sb);
4597 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4598 +
4599 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
4600 +                      ino_t h_ino, int idx);
4601 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4602 +                     int *idx);
4603 +
4604 +int au_xino_path(struct seq_file *seq, struct file *file);
4605 +
4606 +/* ---------------------------------------------------------------------- */
4607 +
4608 +/* @idx is signed to accept -1 meaning the first file */
4609 +static inline struct file *au_xino_file(struct au_xino *xi, int idx)
4610 +{
4611 +       struct file *file;
4612 +
4613 +       file = NULL;
4614 +       if (!xi)
4615 +               goto out;
4616 +
4617 +       if (idx >= 0) {
4618 +               if (idx < xi->xi_nfile)
4619 +                       file = xi->xi_file[idx];
4620 +       } else
4621 +               file = au_xino_file1(xi);
4622 +
4623 +out:
4624 +       return file;
4625 +}
4626 +
4627 +/* ---------------------------------------------------------------------- */
4628 +
4629 +/* Superblock to branch */
4630 +static inline
4631 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4632 +{
4633 +       return au_sbr(sb, bindex)->br_id;
4634 +}
4635 +
4636 +static inline
4637 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4638 +{
4639 +       return au_br_mnt(au_sbr(sb, bindex));
4640 +}
4641 +
4642 +static inline
4643 +struct user_namespace *au_sbr_userns(struct super_block *sb, aufs_bindex_t bindex)
4644 +{
4645 +       return au_br_userns(au_sbr(sb, bindex));
4646 +}
4647 +
4648 +static inline
4649 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4650 +{
4651 +       return au_br_sb(au_sbr(sb, bindex));
4652 +}
4653 +
4654 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4655 +{
4656 +       return au_sbr(sb, bindex)->br_perm;
4657 +}
4658 +
4659 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4660 +{
4661 +       return au_br_whable(au_sbr_perm(sb, bindex));
4662 +}
4663 +
4664 +/* ---------------------------------------------------------------------- */
4665 +
4666 +#define wbr_wh_read_lock(wbr)  au_rw_read_lock(&(wbr)->wbr_wh_rwsem)
4667 +#define wbr_wh_write_lock(wbr) au_rw_write_lock(&(wbr)->wbr_wh_rwsem)
4668 +#define wbr_wh_read_trylock(wbr)       au_rw_read_trylock(&(wbr)->wbr_wh_rwsem)
4669 +#define wbr_wh_write_trylock(wbr) au_rw_write_trylock(&(wbr)->wbr_wh_rwsem)
4670 +/*
4671 +#define wbr_wh_read_trylock_nested(wbr) \
4672 +       au_rw_read_trylock_nested(&(wbr)->wbr_wh_rwsem)
4673 +#define wbr_wh_write_trylock_nested(wbr) \
4674 +       au_rw_write_trylock_nested(&(wbr)->wbr_wh_rwsem)
4675 +*/
4676 +
4677 +#define wbr_wh_read_unlock(wbr)        au_rw_read_unlock(&(wbr)->wbr_wh_rwsem)
4678 +#define wbr_wh_write_unlock(wbr)       au_rw_write_unlock(&(wbr)->wbr_wh_rwsem)
4679 +#define wbr_wh_downgrade_lock(wbr)     au_rw_dgrade_lock(&(wbr)->wbr_wh_rwsem)
4680 +
4681 +#define WbrWhMustNoWaiters(wbr)        AuRwMustNoWaiters(&(wbr)->wbr_wh_rwsem)
4682 +#define WbrWhMustAnyLock(wbr)  AuRwMustAnyLock(&(wbr)->wbr_wh_rwsem)
4683 +#define WbrWhMustWriteLock(wbr)        AuRwMustWriteLock(&(wbr)->wbr_wh_rwsem)
4684 +
4685 +/* ---------------------------------------------------------------------- */
4686 +
4687 +#ifdef CONFIG_AUFS_FHSM
4688 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4689 +{
4690 +       mutex_init(&brfhsm->bf_lock);
4691 +       brfhsm->bf_jiffy = 0;
4692 +       brfhsm->bf_readable = 0;
4693 +}
4694 +
4695 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4696 +{
4697 +       mutex_destroy(&brfhsm->bf_lock);
4698 +}
4699 +#else
4700 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4701 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4702 +#endif
4703 +
4704 +#endif /* __KERNEL__ */
4705 +#endif /* __AUFS_BRANCH_H__ */
4706 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4707 --- /usr/share/empty/fs/aufs/conf.mk    1970-01-01 01:00:00.000000000 +0100
4708 +++ linux/fs/aufs/conf.mk       2022-11-05 23:02:18.959222617 +0100
4709 @@ -0,0 +1,40 @@
4710 +# SPDX-License-Identifier: GPL-2.0
4711 +
4712 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4713 +
4714 +define AuConf
4715 +ifdef ${1}
4716 +AuConfStr += ${1}=${${1}}
4717 +endif
4718 +endef
4719 +
4720 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4721 +       SBILIST \
4722 +       HNOTIFY HFSNOTIFY \
4723 +       EXPORT INO_T_64 \
4724 +       XATTR \
4725 +       FHSM \
4726 +       RDU \
4727 +       DIRREN \
4728 +       SHWH \
4729 +       BR_RAMFS \
4730 +       BR_FUSE POLL \
4731 +       BR_HFSPLUS \
4732 +       BDEV_LOOP \
4733 +       DEBUG MAGIC_SYSRQ
4734 +$(foreach i, ${AuConfAll}, \
4735 +       $(eval $(call AuConf,CONFIG_AUFS_${i})))
4736 +
4737 +AuConfName = ${obj}/conf.str
4738 +${AuConfName}.tmp: FORCE
4739 +       @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4740 +${AuConfName}: ${AuConfName}.tmp
4741 +       @diff -q $< $@ > /dev/null 2>&1 || { \
4742 +       echo '  GEN    ' $@; \
4743 +       cp -p $< $@; \
4744 +       }
4745 +FORCE:
4746 +clean-files += ${AuConfName} ${AuConfName}.tmp
4747 +${obj}/sysfs.o: ${AuConfName}
4748 +
4749 +-include ${srctree}/${src}/conf_priv.mk
4750 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4751 --- /usr/share/empty/fs/aufs/cpup.c     1970-01-01 01:00:00.000000000 +0100
4752 +++ linux/fs/aufs/cpup.c        2022-11-05 23:02:18.962555950 +0100
4753 @@ -0,0 +1,1459 @@
4754 +// SPDX-License-Identifier: GPL-2.0
4755 +/*
4756 + * Copyright (C) 2005-2022 Junjiro R. Okajima
4757 + *
4758 + * This program is free software; you can redistribute it and/or modify
4759 + * it under the terms of the GNU General Public License as published by
4760 + * the Free Software Foundation; either version 2 of the License, or
4761 + * (at your option) any later version.
4762 + *
4763 + * This program is distributed in the hope that it will be useful,
4764 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4765 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
4766 + * GNU General Public License for more details.
4767 + *
4768 + * You should have received a copy of the GNU General Public License
4769 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
4770 + */
4771 +
4772 +/*
4773 + * copy-up functions, see wbr_policy.c for copy-down
4774 + */
4775 +
4776 +#include <linux/fs_stack.h>
4777 +#include <linux/mm.h>
4778 +#include <linux/task_work.h>
4779 +#include "aufs.h"
4780 +
4781 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4782 +{
4783 +       const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4784 +               | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4785 +
4786 +       BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4787 +
4788 +       dst->i_flags |= iflags & ~mask;
4789 +       if (au_test_fs_notime(dst->i_sb))
4790 +               dst->i_flags |= S_NOATIME | S_NOCMTIME;
4791 +}
4792 +
4793 +void au_cpup_attr_timesizes(struct inode *inode)
4794 +{
4795 +       struct inode *h_inode;
4796 +
4797 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
4798 +       fsstack_copy_attr_times(inode, h_inode);
4799 +       fsstack_copy_inode_size(inode, h_inode);
4800 +}
4801 +
4802 +void au_cpup_attr_nlink(struct inode *inode, int force)
4803 +{
4804 +       struct inode *h_inode;
4805 +       struct super_block *sb;
4806 +       aufs_bindex_t bindex, bbot;
4807 +
4808 +       sb = inode->i_sb;
4809 +       bindex = au_ibtop(inode);
4810 +       h_inode = au_h_iptr(inode, bindex);
4811 +       if (!force
4812 +           && !S_ISDIR(h_inode->i_mode)
4813 +           && au_opt_test(au_mntflags(sb), PLINK)
4814 +           && au_plink_test(inode))
4815 +               return;
4816 +
4817 +       /*
4818 +        * 0 can happen in revalidating.
4819 +        * h_inode->i_mutex may not be held here, but it is harmless since once
4820 +        * i_nlink reaches 0, it will never become positive except O_TMPFILE
4821 +        * case.
4822 +        * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4823 +        *       the incorrect link count.
4824 +        */
4825 +       set_nlink(inode, h_inode->i_nlink);
4826 +
4827 +       /*
4828 +        * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4829 +        * it may includes whplink directory.
4830 +        */
4831 +       if (S_ISDIR(h_inode->i_mode)) {
4832 +               bbot = au_ibbot(inode);
4833 +               for (bindex++; bindex <= bbot; bindex++) {
4834 +                       h_inode = au_h_iptr(inode, bindex);
4835 +                       if (h_inode)
4836 +                               au_add_nlink(inode, h_inode);
4837 +               }
4838 +       }
4839 +}
4840 +
4841 +void au_cpup_attr_changeable(struct inode *inode)
4842 +{
4843 +       struct inode *h_inode;
4844 +
4845 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
4846 +       inode->i_mode = h_inode->i_mode;
4847 +       inode->i_uid = h_inode->i_uid;
4848 +       inode->i_gid = h_inode->i_gid;
4849 +       au_cpup_attr_timesizes(inode);
4850 +       au_cpup_attr_flags(inode, h_inode->i_flags);
4851 +}
4852 +
4853 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4854 +{
4855 +       struct au_iinfo *iinfo = au_ii(inode);
4856 +
4857 +       IiMustWriteLock(inode);
4858 +
4859 +       iinfo->ii_higen = h_inode->i_generation;
4860 +       iinfo->ii_hsb1 = h_inode->i_sb;
4861 +}
4862 +
4863 +void au_cpup_attr_all(struct inode *inode, int force)
4864 +{
4865 +       struct inode *h_inode;
4866 +
4867 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
4868 +       au_cpup_attr_changeable(inode);
4869 +       if (inode->i_nlink > 0)
4870 +               au_cpup_attr_nlink(inode, force);
4871 +       inode->i_rdev = h_inode->i_rdev;
4872 +       inode->i_blkbits = h_inode->i_blkbits;
4873 +       au_cpup_igen(inode, h_inode);
4874 +}
4875 +
4876 +/* ---------------------------------------------------------------------- */
4877 +
4878 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4879 +
4880 +/* keep the timestamps of the parent dir when cpup */
4881 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4882 +                   struct path *h_path)
4883 +{
4884 +       struct inode *h_inode;
4885 +
4886 +       dt->dt_dentry = dentry;
4887 +       dt->dt_h_path = *h_path;
4888 +       h_inode = d_inode(h_path->dentry);
4889 +       dt->dt_atime = h_inode->i_atime;
4890 +       dt->dt_mtime = h_inode->i_mtime;
4891 +       /* smp_mb(); */
4892 +}
4893 +
4894 +void au_dtime_revert(struct au_dtime *dt)
4895 +{
4896 +       struct iattr attr;
4897 +       int err;
4898 +
4899 +       attr.ia_atime = dt->dt_atime;
4900 +       attr.ia_mtime = dt->dt_mtime;
4901 +       attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4902 +               | ATTR_ATIME | ATTR_ATIME_SET;
4903 +
4904 +       /* no delegation since this is a directory */
4905 +       err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4906 +       if (unlikely(err))
4907 +               pr_warn("restoring timestamps failed(%d). ignored\n", err);
4908 +}
4909 +
4910 +/* ---------------------------------------------------------------------- */
4911 +
4912 +/* internal use only */
4913 +struct au_cpup_reg_attr {
4914 +       int             valid;
4915 +       struct kstat    st;
4916 +       unsigned int    iflags; /* inode->i_flags */
4917 +};
4918 +
4919 +static noinline_for_stack
4920 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct path *h_src,
4921 +              struct au_cpup_reg_attr *h_src_attr)
4922 +{
4923 +       int err, sbits, icex;
4924 +       unsigned int mnt_flags;
4925 +       unsigned char verbose;
4926 +       struct iattr ia;
4927 +       struct path h_path;
4928 +       struct inode *h_isrc, *h_idst;
4929 +       struct kstat *h_st;
4930 +       struct au_branch *br;
4931 +
4932 +       br = au_sbr(dst->d_sb, bindex);
4933 +       h_path.mnt = au_br_mnt(br);
4934 +       h_path.dentry = au_h_dptr(dst, bindex);
4935 +       h_idst = d_inode(h_path.dentry);
4936 +       h_isrc = d_inode(h_src->dentry);
4937 +       ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4938 +               | ATTR_ATIME | ATTR_MTIME
4939 +               | ATTR_ATIME_SET | ATTR_MTIME_SET;
4940 +       if (h_src_attr && h_src_attr->valid) {
4941 +               h_st = &h_src_attr->st;
4942 +               ia.ia_uid = h_st->uid;
4943 +               ia.ia_gid = h_st->gid;
4944 +               ia.ia_atime = h_st->atime;
4945 +               ia.ia_mtime = h_st->mtime;
4946 +               if (h_idst->i_mode != h_st->mode
4947 +                   && !S_ISLNK(h_idst->i_mode)) {
4948 +                       ia.ia_valid |= ATTR_MODE;
4949 +                       ia.ia_mode = h_st->mode;
4950 +               }
4951 +               sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4952 +               au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4953 +       } else {
4954 +               ia.ia_uid = h_isrc->i_uid;
4955 +               ia.ia_gid = h_isrc->i_gid;
4956 +               ia.ia_atime = h_isrc->i_atime;
4957 +               ia.ia_mtime = h_isrc->i_mtime;
4958 +               if (h_idst->i_mode != h_isrc->i_mode
4959 +                   && !S_ISLNK(h_idst->i_mode)) {
4960 +                       ia.ia_valid |= ATTR_MODE;
4961 +                       ia.ia_mode = h_isrc->i_mode;
4962 +               }
4963 +               sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4964 +               au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4965 +       }
4966 +       /* no delegation since it is just created */
4967 +       err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4968 +
4969 +       /* is this nfs only? */
4970 +       if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4971 +               ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4972 +               ia.ia_mode = h_isrc->i_mode;
4973 +               err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4974 +       }
4975 +
4976 +       icex = br->br_perm & AuBrAttr_ICEX;
4977 +       if (!err) {
4978 +               mnt_flags = au_mntflags(dst->d_sb);
4979 +               verbose = !!au_opt_test(mnt_flags, VERBOSE);
4980 +               err = au_cpup_xattr(&h_path, h_src, icex, verbose);
4981 +       }
4982 +
4983 +       return err;
4984 +}
4985 +
4986 +/* ---------------------------------------------------------------------- */
4987 +
4988 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4989 +                          char *buf, unsigned long blksize)
4990 +{
4991 +       int err;
4992 +       size_t sz, rbytes, wbytes;
4993 +       unsigned char all_zero;
4994 +       char *p, *zp;
4995 +       struct inode *h_inode;
4996 +       /* reduce stack usage */
4997 +       struct iattr *ia;
4998 +
4999 +       zp = page_address(ZERO_PAGE(0));
5000 +       if (unlikely(!zp))
5001 +               return -ENOMEM; /* possible? */
5002 +
5003 +       err = 0;
5004 +       all_zero = 0;
5005 +       while (len) {
5006 +               AuDbg("len %lld\n", len);
5007 +               sz = blksize;
5008 +               if (len < blksize)
5009 +                       sz = len;
5010 +
5011 +               rbytes = 0;
5012 +               /* todo: signal_pending? */
5013 +               while (!rbytes || err == -EAGAIN || err == -EINTR) {
5014 +                       rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
5015 +                       err = rbytes;
5016 +               }
5017 +               if (unlikely(err < 0))
5018 +                       break;
5019 +
5020 +               all_zero = 0;
5021 +               if (len >= rbytes && rbytes == blksize)
5022 +                       all_zero = !memcmp(buf, zp, rbytes);
5023 +               if (!all_zero) {
5024 +                       wbytes = rbytes;
5025 +                       p = buf;
5026 +                       while (wbytes) {
5027 +                               size_t b;
5028 +
5029 +                               b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
5030 +                               err = b;
5031 +                               /* todo: signal_pending? */
5032 +                               if (unlikely(err == -EAGAIN || err == -EINTR))
5033 +                                       continue;
5034 +                               if (unlikely(err < 0))
5035 +                                       break;
5036 +                               wbytes -= b;
5037 +                               p += b;
5038 +                       }
5039 +                       if (unlikely(err < 0))
5040 +                               break;
5041 +               } else {
5042 +                       loff_t res;
5043 +
5044 +                       AuLabel(hole);
5045 +                       res = vfsub_llseek(dst, rbytes, SEEK_CUR);
5046 +                       err = res;
5047 +                       if (unlikely(res < 0))
5048 +                               break;
5049 +               }
5050 +               len -= rbytes;
5051 +               err = 0;
5052 +       }
5053 +
5054 +       /* the last block may be a hole */
5055 +       if (!err && all_zero) {
5056 +               AuLabel(last hole);
5057 +
5058 +               err = 1;
5059 +               if (au_test_nfs(dst->f_path.dentry->d_sb)) {
5060 +                       /* nfs requires this step to make last hole */
5061 +                       /* is this only nfs? */
5062 +                       do {
5063 +                               /* todo: signal_pending? */
5064 +                               err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
5065 +                       } while (err == -EAGAIN || err == -EINTR);
5066 +                       if (err == 1)
5067 +                               dst->f_pos--;
5068 +               }
5069 +
5070 +               if (err == 1) {
5071 +                       ia = (void *)buf;
5072 +                       ia->ia_size = dst->f_pos;
5073 +                       ia->ia_valid = ATTR_SIZE | ATTR_FILE;
5074 +                       ia->ia_file = dst;
5075 +                       h_inode = file_inode(dst);
5076 +                       inode_lock_nested(h_inode, AuLsc_I_CHILD2);
5077 +                       /* no delegation since it is just created */
5078 +                       err = vfsub_notify_change(&dst->f_path, ia,
5079 +                                                 /*delegated*/NULL);
5080 +                       inode_unlock(h_inode);
5081 +               }
5082 +       }
5083 +
5084 +       return err;
5085 +}
5086 +
5087 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
5088 +{
5089 +       int err;
5090 +       unsigned long blksize;
5091 +       unsigned char do_kfree;
5092 +       char *buf;
5093 +       struct super_block *h_sb;
5094 +
5095 +       err = -ENOMEM;
5096 +       h_sb = file_inode(dst)->i_sb;
5097 +       blksize = h_sb->s_blocksize;
5098 +       if (!blksize || PAGE_SIZE < blksize)
5099 +               blksize = PAGE_SIZE;
5100 +       AuDbg("blksize %lu\n", blksize);
5101 +       do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
5102 +       if (do_kfree)
5103 +               buf = kmalloc(blksize, GFP_NOFS);
5104 +       else
5105 +               buf = (void *)__get_free_page(GFP_NOFS);
5106 +       if (unlikely(!buf))
5107 +               goto out;
5108 +
5109 +       if (len > (1 << 22))
5110 +               AuDbg("copying a large file %lld\n", (long long)len);
5111 +
5112 +       src->f_pos = 0;
5113 +       dst->f_pos = 0;
5114 +       err = au_do_copy_file(dst, src, len, buf, blksize);
5115 +       if (do_kfree) {
5116 +               AuDebugOn(!au_kfree_do_sz_test(blksize));
5117 +               au_kfree_do_rcu(buf);
5118 +       } else
5119 +               free_page((unsigned long)buf);
5120 +
5121 +out:
5122 +       return err;
5123 +}
5124 +
5125 +static int au_do_copy(struct file *dst, struct file *src, loff_t len)
5126 +{
5127 +       int err;
5128 +       struct super_block *h_src_sb;
5129 +       struct inode *h_src_inode;
5130 +
5131 +       h_src_inode = file_inode(src);
5132 +       h_src_sb = h_src_inode->i_sb;
5133 +
5134 +       /* XFS acquires inode_lock */
5135 +       if (!au_test_xfs(h_src_sb))
5136 +               err = au_copy_file(dst, src, len);
5137 +       else {
5138 +               inode_unlock_shared(h_src_inode);
5139 +               err = au_copy_file(dst, src, len);
5140 +               inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5141 +       }
5142 +
5143 +       return err;
5144 +}
5145 +
5146 +static int au_clone_or_copy(struct file *dst, struct file *src, loff_t len)
5147 +{
5148 +       int err;
5149 +       loff_t lo;
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 +       if (h_src_sb != file_inode(dst)->i_sb
5156 +           || !dst->f_op->remap_file_range) {
5157 +               err = au_do_copy(dst, src, len);
5158 +               goto out;
5159 +       }
5160 +
5161 +       if (!au_test_nfs(h_src_sb)) {
5162 +               inode_unlock_shared(h_src_inode);
5163 +               lo = vfsub_clone_file_range(src, dst, len);
5164 +               inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5165 +       } else
5166 +               lo = vfsub_clone_file_range(src, dst, len);
5167 +       if (lo == len) {
5168 +               err = 0;
5169 +               goto out; /* success */
5170 +       } else if (lo >= 0)
5171 +               /* todo: possible? */
5172 +               /* paritially succeeded */
5173 +               AuDbg("lo %lld, len %lld. Retrying.\n", lo, len);
5174 +       else if (lo != -EOPNOTSUPP) {
5175 +               /* older XFS has a condition in cloning */
5176 +               err = lo;
5177 +               goto out;
5178 +       }
5179 +
5180 +       /* the backend fs on NFS may not support cloning */
5181 +       err = au_do_copy(dst, src, len);
5182 +
5183 +out:
5184 +       AuTraceErr(err);
5185 +       return err;
5186 +}
5187 +
5188 +/*
5189 + * to support a sparse file which is opened with O_APPEND,
5190 + * we need to close the file.
5191 + */
5192 +static int au_cp_regular(struct au_cp_generic *cpg)
5193 +{
5194 +       int err, i;
5195 +       enum { SRC, DST };
5196 +       struct {
5197 +               aufs_bindex_t bindex;
5198 +               unsigned int flags;
5199 +               struct dentry *dentry;
5200 +               int force_wr;
5201 +               struct file *file;
5202 +       } *f, file[] = {
5203 +               {
5204 +                       .bindex = cpg->bsrc,
5205 +                       .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
5206 +               },
5207 +               {
5208 +                       .bindex = cpg->bdst,
5209 +                       .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
5210 +                       .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
5211 +               }
5212 +       };
5213 +       struct au_branch *br;
5214 +       struct super_block *sb, *h_src_sb;
5215 +       struct inode *h_src_inode;
5216 +       struct task_struct *tsk = current;
5217 +
5218 +       /* bsrc branch can be ro/rw. */
5219 +       sb = cpg->dentry->d_sb;
5220 +       f = file;
5221 +       for (i = 0; i < 2; i++, f++) {
5222 +               f->dentry = au_h_dptr(cpg->dentry, f->bindex);
5223 +               f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
5224 +                                   /*file*/NULL, f->force_wr);
5225 +               if (IS_ERR(f->file)) {
5226 +                       err = PTR_ERR(f->file);
5227 +                       if (i == SRC)
5228 +                               goto out;
5229 +                       else
5230 +                               goto out_src;
5231 +               }
5232 +       }
5233 +
5234 +       /* try stopping to update while we copyup */
5235 +       h_src_inode = d_inode(file[SRC].dentry);
5236 +       h_src_sb = h_src_inode->i_sb;
5237 +       if (!au_test_nfs(h_src_sb))
5238 +               IMustLock(h_src_inode);
5239 +       err = au_clone_or_copy(file[DST].file, file[SRC].file, cpg->len);
5240 +
5241 +       /* i wonder if we had O_NO_DELAY_FPUT flag */
5242 +       if (tsk->flags & PF_KTHREAD)
5243 +               __fput_sync(file[DST].file);
5244 +       else {
5245 +               /* it happened actually */
5246 +               fput(file[DST].file);
5247 +               /*
5248 +                * too bad.
5249 +                * we have to call both since we don't know which place the file
5250 +                * was added to.
5251 +                */
5252 +               task_work_run();
5253 +               flush_delayed_fput();
5254 +       }
5255 +       br = au_sbr(sb, file[DST].bindex);
5256 +       au_lcnt_dec(&br->br_nfiles);
5257 +
5258 +out_src:
5259 +       fput(file[SRC].file);
5260 +       br = au_sbr(sb, file[SRC].bindex);
5261 +       au_lcnt_dec(&br->br_nfiles);
5262 +out:
5263 +       return err;
5264 +}
5265 +
5266 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
5267 +                             struct au_cpup_reg_attr *h_src_attr)
5268 +{
5269 +       int err, rerr;
5270 +       loff_t l;
5271 +       struct path h_path;
5272 +       struct inode *h_src_inode, *h_dst_inode;
5273 +
5274 +       err = 0;
5275 +       h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
5276 +       l = i_size_read(h_src_inode);
5277 +       if (cpg->len == -1 || l < cpg->len)
5278 +               cpg->len = l;
5279 +       if (cpg->len) {
5280 +               /* try stopping to update while we are referencing */
5281 +               inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5282 +               au_pin_hdir_unlock(cpg->pin);
5283 +
5284 +               h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5285 +               h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
5286 +               h_src_attr->iflags = h_src_inode->i_flags;
5287 +               if (!au_test_nfs(h_src_inode->i_sb))
5288 +                       err = vfsub_getattr(&h_path, &h_src_attr->st);
5289 +               else {
5290 +                       inode_unlock_shared(h_src_inode);
5291 +                       err = vfsub_getattr(&h_path, &h_src_attr->st);
5292 +                       inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5293 +               }
5294 +               if (unlikely(err)) {
5295 +                       inode_unlock_shared(h_src_inode);
5296 +                       goto out;
5297 +               }
5298 +               h_src_attr->valid = 1;
5299 +               if (!au_test_nfs(h_src_inode->i_sb)) {
5300 +                       err = au_cp_regular(cpg);
5301 +                       inode_unlock_shared(h_src_inode);
5302 +               } else {
5303 +                       inode_unlock_shared(h_src_inode);
5304 +                       err = au_cp_regular(cpg);
5305 +               }
5306 +               rerr = au_pin_hdir_relock(cpg->pin);
5307 +               if (!err && rerr)
5308 +                       err = rerr;
5309 +       }
5310 +       if (!err && (h_src_inode->i_state & I_LINKABLE)) {
5311 +               h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5312 +               h_dst_inode = d_inode(h_path.dentry);
5313 +               spin_lock(&h_dst_inode->i_lock);
5314 +               h_dst_inode->i_state |= I_LINKABLE;
5315 +               spin_unlock(&h_dst_inode->i_lock);
5316 +       }
5317 +
5318 +out:
5319 +       return err;
5320 +}
5321 +
5322 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
5323 +                             struct inode *h_dir)
5324 +{
5325 +       int err;
5326 +       DEFINE_DELAYED_CALL(done);
5327 +       const char *sym;
5328 +
5329 +       sym = vfs_get_link(h_src, &done);
5330 +       err = PTR_ERR(sym);
5331 +       if (IS_ERR(sym))
5332 +               goto out;
5333 +
5334 +       err = vfsub_symlink(h_dir, h_path, sym);
5335 +
5336 +out:
5337 +       do_delayed_call(&done);
5338 +       return err;
5339 +}
5340 +
5341 +/*
5342 + * regardless 'acl' option, reset all ACL.
5343 + * All ACL will be copied up later from the original entry on the lower branch.
5344 + */
5345 +static int au_reset_acl(struct inode *h_dir, struct path *h_path, umode_t mode)
5346 +{
5347 +       int err;
5348 +       struct dentry *h_dentry;
5349 +       struct inode *h_inode;
5350 +       struct user_namespace *h_userns;
5351 +
5352 +       h_userns = mnt_user_ns(h_path->mnt);
5353 +       h_dentry = h_path->dentry;
5354 +       h_inode = d_inode(h_dentry);
5355 +       /* forget_all_cached_acls(h_inode)); */
5356 +       err = vfsub_removexattr(h_userns, h_dentry, XATTR_NAME_POSIX_ACL_ACCESS);
5357 +       AuTraceErr(err);
5358 +       if (err == -EOPNOTSUPP)
5359 +               err = 0;
5360 +       if (!err)
5361 +               err = vfsub_acl_chmod(h_userns, h_inode, mode);
5362 +
5363 +       AuTraceErr(err);
5364 +       return err;
5365 +}
5366 +
5367 +static int au_do_cpup_dir(struct au_cp_generic *cpg, struct dentry *dst_parent,
5368 +                         struct inode *h_dir, struct path *h_path)
5369 +{
5370 +       int err;
5371 +       struct inode *dir, *inode;
5372 +       struct user_namespace *h_userns;
5373 +
5374 +       h_userns = mnt_user_ns(h_path->mnt);
5375 +       err = vfsub_removexattr(h_userns, h_path->dentry,
5376 +                               XATTR_NAME_POSIX_ACL_DEFAULT);
5377 +       AuTraceErr(err);
5378 +       if (err == -EOPNOTSUPP)
5379 +               err = 0;
5380 +       if (unlikely(err))
5381 +               goto out;
5382 +
5383 +       /*
5384 +        * strange behaviour from the users view,
5385 +        * particularly setattr case
5386 +        */
5387 +       dir = d_inode(dst_parent);
5388 +       if (au_ibtop(dir) == cpg->bdst)
5389 +               au_cpup_attr_nlink(dir, /*force*/1);
5390 +       inode = d_inode(cpg->dentry);
5391 +       au_cpup_attr_nlink(inode, /*force*/1);
5392 +
5393 +out:
5394 +       return err;
5395 +}
5396 +
5397 +static noinline_for_stack
5398 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5399 +              struct au_cpup_reg_attr *h_src_attr)
5400 +{
5401 +       int err;
5402 +       umode_t mode;
5403 +       unsigned int mnt_flags;
5404 +       unsigned char isdir, isreg, force;
5405 +       const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5406 +       struct au_dtime dt;
5407 +       struct path h_path;
5408 +       struct dentry *h_src, *h_dst, *h_parent;
5409 +       struct inode *h_inode, *h_dir;
5410 +       struct super_block *sb;
5411 +
5412 +       /* bsrc branch can be ro/rw. */
5413 +       h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5414 +       h_inode = d_inode(h_src);
5415 +       AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5416 +
5417 +       /* try stopping to be referenced while we are creating */
5418 +       h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5419 +       if (au_ftest_cpup(cpg->flags, RENAME))
5420 +               AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5421 +                                 AUFS_WH_PFX_LEN));
5422 +       h_parent = h_dst->d_parent; /* dir inode is locked */
5423 +       h_dir = d_inode(h_parent);
5424 +       IMustLock(h_dir);
5425 +       AuDebugOn(h_parent != h_dst->d_parent);
5426 +
5427 +       sb = cpg->dentry->d_sb;
5428 +       h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5429 +       if (do_dt) {
5430 +               h_path.dentry = h_parent;
5431 +               au_dtime_store(&dt, dst_parent, &h_path);
5432 +       }
5433 +       h_path.dentry = h_dst;
5434 +
5435 +       isreg = 0;
5436 +       isdir = 0;
5437 +       mode = h_inode->i_mode;
5438 +       switch (mode & S_IFMT) {
5439 +       case S_IFREG:
5440 +               isreg = 1;
5441 +               err = vfsub_create(h_dir, &h_path, 0600, /*want_excl*/true);
5442 +               if (!err)
5443 +                       err = au_do_cpup_regular(cpg, h_src_attr);
5444 +               break;
5445 +       case S_IFDIR:
5446 +               isdir = 1;
5447 +               err = vfsub_mkdir(h_dir, &h_path, mode);
5448 +               if (!err)
5449 +                       err = au_do_cpup_dir(cpg, dst_parent, h_dir, &h_path);
5450 +               break;
5451 +       case S_IFLNK:
5452 +               err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5453 +               break;
5454 +       case S_IFCHR:
5455 +       case S_IFBLK:
5456 +               AuDebugOn(!capable(CAP_MKNOD));
5457 +               fallthrough;
5458 +       case S_IFIFO:
5459 +       case S_IFSOCK:
5460 +               err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5461 +               break;
5462 +       default:
5463 +               AuIOErr("Unknown inode type 0%o\n", mode);
5464 +               err = -EIO;
5465 +       }
5466 +       if (!err)
5467 +               err = au_reset_acl(h_dir, &h_path, mode);
5468 +
5469 +       mnt_flags = au_mntflags(sb);
5470 +       if (!au_opt_test(mnt_flags, UDBA_NONE)
5471 +           && !isdir
5472 +           && au_opt_test(mnt_flags, XINO)
5473 +           && (h_inode->i_nlink == 1
5474 +               || (h_inode->i_state & I_LINKABLE))
5475 +           /* todo: unnecessary? */
5476 +           /* && d_inode(cpg->dentry)->i_nlink == 1 */
5477 +           && cpg->bdst < cpg->bsrc
5478 +           && !au_ftest_cpup(cpg->flags, KEEPLINO))
5479 +               au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5480 +               /* ignore this error */
5481 +
5482 +       if (!err) {
5483 +               force = 0;
5484 +               if (isreg) {
5485 +                       force = !!cpg->len;
5486 +                       if (cpg->len == -1)
5487 +                               force = !!i_size_read(h_inode);
5488 +               }
5489 +               au_fhsm_wrote(sb, cpg->bdst, force);
5490 +       }
5491 +
5492 +       if (do_dt)
5493 +               au_dtime_revert(&dt);
5494 +       return err;
5495 +}
5496 +
5497 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5498 +{
5499 +       int err;
5500 +       struct dentry *dentry, *h_dentry, *h_parent, *parent;
5501 +       struct path h_ppath;
5502 +       struct inode *h_dir;
5503 +       aufs_bindex_t bdst;
5504 +
5505 +       dentry = cpg->dentry;
5506 +       bdst = cpg->bdst;
5507 +       h_ppath.mnt = au_sbr_mnt(dentry->d_sb, bdst);
5508 +       h_dentry = au_h_dptr(dentry, bdst);
5509 +       if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5510 +               dget(h_dentry);
5511 +               au_set_h_dptr(dentry, bdst, NULL);
5512 +               err = au_lkup_neg(dentry, bdst, /*wh*/0);
5513 +               if (!err)
5514 +                       h_path->dentry = dget(au_h_dptr(dentry, bdst));
5515 +               au_set_h_dptr(dentry, bdst, h_dentry);
5516 +       } else {
5517 +               err = 0;
5518 +               parent = dget_parent(dentry);
5519 +               h_ppath.dentry = au_h_dptr(parent, bdst);
5520 +               dput(parent);
5521 +               h_path->dentry = vfsub_lkup_one(&dentry->d_name, &h_ppath);
5522 +               if (IS_ERR(h_path->dentry))
5523 +                       err = PTR_ERR(h_path->dentry);
5524 +       }
5525 +       if (unlikely(err))
5526 +               goto out;
5527 +
5528 +       h_parent = h_dentry->d_parent; /* dir inode is locked */
5529 +       h_dir = d_inode(h_parent);
5530 +       IMustLock(h_dir);
5531 +       AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5532 +       /* no delegation since it is just created */
5533 +       err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL,
5534 +                          /*flags*/0);
5535 +       dput(h_path->dentry);
5536 +
5537 +out:
5538 +       return err;
5539 +}
5540 +
5541 +/*
5542 + * copyup the @dentry from @bsrc to @bdst.
5543 + * the caller must set the both of lower dentries.
5544 + * @len is for truncating when it is -1 copyup the entire file.
5545 + * in link/rename cases, @dst_parent may be different from the real one.
5546 + * basic->bsrc can be larger than basic->bdst.
5547 + * aufs doesn't touch the credential so
5548 + * security_inode_copy_up{,_xattr}() are unnecessary.
5549 + */
5550 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5551 +{
5552 +       int err, rerr;
5553 +       aufs_bindex_t old_ibtop;
5554 +       unsigned char isdir, plink;
5555 +       struct dentry *h_src, *h_dst, *h_parent;
5556 +       struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5557 +       struct super_block *sb;
5558 +       struct au_branch *br;
5559 +       struct path h_src_path;
5560 +       /* to reduce stack size */
5561 +       struct {
5562 +               struct au_dtime dt;
5563 +               struct path h_path;
5564 +               struct au_cpup_reg_attr h_src_attr;
5565 +       } *a;
5566 +
5567 +       err = -ENOMEM;
5568 +       a = kmalloc(sizeof(*a), GFP_NOFS);
5569 +       if (unlikely(!a))
5570 +               goto out;
5571 +       a->h_src_attr.valid = 0;
5572 +
5573 +       sb = cpg->dentry->d_sb;
5574 +       br = au_sbr(sb, cpg->bdst);
5575 +       a->h_path.mnt = au_br_mnt(br);
5576 +       h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5577 +       h_parent = h_dst->d_parent; /* dir inode is locked */
5578 +       h_dir = d_inode(h_parent);
5579 +       IMustLock(h_dir);
5580 +
5581 +       h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5582 +       inode = d_inode(cpg->dentry);
5583 +
5584 +       if (!dst_parent)
5585 +               dst_parent = dget_parent(cpg->dentry);
5586 +       else
5587 +               dget(dst_parent);
5588 +
5589 +       plink = !!au_opt_test(au_mntflags(sb), PLINK);
5590 +       dst_inode = au_h_iptr(inode, cpg->bdst);
5591 +       if (dst_inode) {
5592 +               if (unlikely(!plink)) {
5593 +                       err = -EIO;
5594 +                       AuIOErr("hi%lu(i%lu) exists on b%d "
5595 +                               "but plink is disabled\n",
5596 +                               dst_inode->i_ino, inode->i_ino, cpg->bdst);
5597 +                       goto out_parent;
5598 +               }
5599 +
5600 +               if (dst_inode->i_nlink) {
5601 +                       const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5602 +
5603 +                       h_src = au_plink_lkup(inode, cpg->bdst);
5604 +                       err = PTR_ERR(h_src);
5605 +                       if (IS_ERR(h_src))
5606 +                               goto out_parent;
5607 +                       if (unlikely(d_is_negative(h_src))) {
5608 +                               err = -EIO;
5609 +                               AuIOErr("i%lu exists on b%d "
5610 +                                       "but not pseudo-linked\n",
5611 +                                       inode->i_ino, cpg->bdst);
5612 +                               dput(h_src);
5613 +                               goto out_parent;
5614 +                       }
5615 +
5616 +                       if (do_dt) {
5617 +                               a->h_path.dentry = h_parent;
5618 +                               au_dtime_store(&a->dt, dst_parent, &a->h_path);
5619 +                       }
5620 +
5621 +                       a->h_path.dentry = h_dst;
5622 +                       delegated = NULL;
5623 +                       err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5624 +                       if (!err && au_ftest_cpup(cpg->flags, RENAME))
5625 +                               err = au_do_ren_after_cpup(cpg, &a->h_path);
5626 +                       if (do_dt)
5627 +                               au_dtime_revert(&a->dt);
5628 +                       if (unlikely(err == -EWOULDBLOCK)) {
5629 +                               pr_warn("cannot retry for NFSv4 delegation"
5630 +                                       " for an internal link\n");
5631 +                               iput(delegated);
5632 +                       }
5633 +                       dput(h_src);
5634 +                       goto out_parent;
5635 +               } else
5636 +                       /* todo: cpup_wh_file? */
5637 +                       /* udba work */
5638 +                       au_update_ibrange(inode, /*do_put_zero*/1);
5639 +       }
5640 +
5641 +       isdir = S_ISDIR(inode->i_mode);
5642 +       old_ibtop = au_ibtop(inode);
5643 +       err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5644 +       if (unlikely(err))
5645 +               goto out_rev;
5646 +       dst_inode = d_inode(h_dst);
5647 +       inode_lock_nested(dst_inode, AuLsc_I_CHILD2);
5648 +       /* todo: necessary? */
5649 +       /* au_pin_hdir_unlock(cpg->pin); */
5650 +
5651 +       h_src_path.dentry = h_src;
5652 +       h_src_path.mnt = au_sbr_mnt(sb, cpg->bsrc);
5653 +       err = cpup_iattr(cpg->dentry, cpg->bdst, &h_src_path, &a->h_src_attr);
5654 +       if (unlikely(err)) {
5655 +               /* todo: necessary? */
5656 +               /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5657 +               inode_unlock(dst_inode);
5658 +               goto out_rev;
5659 +       }
5660 +
5661 +       if (cpg->bdst < old_ibtop) {
5662 +               if (S_ISREG(inode->i_mode)) {
5663 +                       err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5664 +                       if (unlikely(err)) {
5665 +                               /* ignore an error */
5666 +                               /* au_pin_hdir_relock(cpg->pin); */
5667 +                               inode_unlock(dst_inode);
5668 +                               goto out_rev;
5669 +                       }
5670 +               }
5671 +               au_set_ibtop(inode, cpg->bdst);
5672 +       } else
5673 +               au_set_ibbot(inode, cpg->bdst);
5674 +       au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5675 +                     au_hi_flags(inode, isdir));
5676 +
5677 +       /* todo: necessary? */
5678 +       /* err = au_pin_hdir_relock(cpg->pin); */
5679 +       inode_unlock(dst_inode);
5680 +       if (unlikely(err))
5681 +               goto out_rev;
5682 +
5683 +       src_inode = d_inode(h_src);
5684 +       if (!isdir
5685 +           && (src_inode->i_nlink > 1
5686 +               || src_inode->i_state & I_LINKABLE)
5687 +           && plink)
5688 +               au_plink_append(inode, cpg->bdst, h_dst);
5689 +
5690 +       if (au_ftest_cpup(cpg->flags, RENAME)) {
5691 +               a->h_path.dentry = h_dst;
5692 +               err = au_do_ren_after_cpup(cpg, &a->h_path);
5693 +       }
5694 +       if (!err)
5695 +               goto out_parent; /* success */
5696 +
5697 +       /* revert */
5698 +out_rev:
5699 +       a->h_path.dentry = h_parent;
5700 +       au_dtime_store(&a->dt, dst_parent, &a->h_path);
5701 +       a->h_path.dentry = h_dst;
5702 +       rerr = 0;
5703 +       if (d_is_positive(h_dst)) {
5704 +               if (!isdir) {
5705 +                       /* no delegation since it is just created */
5706 +                       rerr = vfsub_unlink(h_dir, &a->h_path,
5707 +                                           /*delegated*/NULL, /*force*/0);
5708 +               } else
5709 +                       rerr = vfsub_rmdir(h_dir, &a->h_path);
5710 +       }
5711 +       au_dtime_revert(&a->dt);
5712 +       if (rerr) {
5713 +               AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5714 +               err = -EIO;
5715 +       }
5716 +out_parent:
5717 +       dput(dst_parent);
5718 +       au_kfree_rcu(a);
5719 +out:
5720 +       return err;
5721 +}
5722 +
5723 +#if 0 /* reserved */
5724 +struct au_cpup_single_args {
5725 +       int *errp;
5726 +       struct au_cp_generic *cpg;
5727 +       struct dentry *dst_parent;
5728 +};
5729 +
5730 +static void au_call_cpup_single(void *args)
5731 +{
5732 +       struct au_cpup_single_args *a = args;
5733 +
5734 +       au_pin_hdir_acquire_nest(a->cpg->pin);
5735 +       *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5736 +       au_pin_hdir_release(a->cpg->pin);
5737 +}
5738 +#endif
5739 +
5740 +/*
5741 + * prevent SIGXFSZ in copy-up.
5742 + * testing CAP_MKNOD is for generic fs,
5743 + * but CAP_FSETID is for xfs only, currently.
5744 + */
5745 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5746 +{
5747 +       int do_sio;
5748 +       struct super_block *sb;
5749 +       struct inode *h_dir;
5750 +
5751 +       do_sio = 0;
5752 +       sb = au_pinned_parent(pin)->d_sb;
5753 +       if (!au_wkq_test()
5754 +           && (!au_sbi(sb)->si_plink_maint_pid
5755 +               || au_plink_maint(sb, AuLock_NOPLM))) {
5756 +               switch (mode & S_IFMT) {
5757 +               case S_IFREG:
5758 +                       /* no condition about RLIMIT_FSIZE and the file size */
5759 +                       do_sio = 1;
5760 +                       break;
5761 +               case S_IFCHR:
5762 +               case S_IFBLK:
5763 +                       do_sio = !capable(CAP_MKNOD);
5764 +                       break;
5765 +               }
5766 +               if (!do_sio)
5767 +                       do_sio = ((mode & (S_ISUID | S_ISGID))
5768 +                                 && !capable(CAP_FSETID));
5769 +               /* this workaround may be removed in the future */
5770 +               if (!do_sio) {
5771 +                       h_dir = au_pinned_h_dir(pin);
5772 +                       do_sio = h_dir->i_mode & S_ISVTX;
5773 +               }
5774 +       }
5775 +
5776 +       return do_sio;
5777 +}
5778 +
5779 +#if 0 /* reserved */
5780 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5781 +{
5782 +       int err, wkq_err;
5783 +       struct dentry *h_dentry;
5784 +
5785 +       h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5786 +       if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5787 +               err = au_cpup_single(cpg, dst_parent);
5788 +       else {
5789 +               struct au_cpup_single_args args = {
5790 +                       .errp           = &err,
5791 +                       .cpg            = cpg,
5792 +                       .dst_parent     = dst_parent
5793 +               };
5794 +               wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5795 +               if (unlikely(wkq_err))
5796 +                       err = wkq_err;
5797 +       }
5798 +
5799 +       return err;
5800 +}
5801 +#endif
5802 +
5803 +/*
5804 + * copyup the @dentry from the first active lower branch to @bdst,
5805 + * using au_cpup_single().
5806 + */
5807 +static int au_cpup_simple(struct au_cp_generic *cpg)
5808 +{
5809 +       int err;
5810 +       unsigned int flags_orig;
5811 +       struct dentry *dentry;
5812 +
5813 +       AuDebugOn(cpg->bsrc < 0);
5814 +
5815 +       dentry = cpg->dentry;
5816 +       DiMustWriteLock(dentry);
5817 +
5818 +       err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5819 +       if (!err) {
5820 +               flags_orig = cpg->flags;
5821 +               au_fset_cpup(cpg->flags, RENAME);
5822 +               err = au_cpup_single(cpg, NULL);
5823 +               cpg->flags = flags_orig;
5824 +               if (!err)
5825 +                       return 0; /* success */
5826 +
5827 +               /* revert */
5828 +               au_set_h_dptr(dentry, cpg->bdst, NULL);
5829 +               au_set_dbtop(dentry, cpg->bsrc);
5830 +       }
5831 +
5832 +       return err;
5833 +}
5834 +
5835 +struct au_cpup_simple_args {
5836 +       int *errp;
5837 +       struct au_cp_generic *cpg;
5838 +};
5839 +
5840 +static void au_call_cpup_simple(void *args)
5841 +{
5842 +       struct au_cpup_simple_args *a = args;
5843 +
5844 +       au_pin_hdir_acquire_nest(a->cpg->pin);
5845 +       *a->errp = au_cpup_simple(a->cpg);
5846 +       au_pin_hdir_release(a->cpg->pin);
5847 +}
5848 +
5849 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5850 +{
5851 +       int err, wkq_err;
5852 +       struct dentry *dentry, *parent;
5853 +       struct file *h_file;
5854 +       struct inode *h_dir;
5855 +       struct user_namespace *h_userns;
5856 +
5857 +       dentry = cpg->dentry;
5858 +       h_file = NULL;
5859 +       if (au_ftest_cpup(cpg->flags, HOPEN)) {
5860 +               AuDebugOn(cpg->bsrc < 0);
5861 +               h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5862 +               err = PTR_ERR(h_file);
5863 +               if (IS_ERR(h_file))
5864 +                       goto out;
5865 +       }
5866 +
5867 +       parent = dget_parent(dentry);
5868 +       h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5869 +       h_userns = au_sbr_userns(dentry->d_sb, cpg->bdst);
5870 +       if (!au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC | MAY_WRITE)
5871 +           && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5872 +               err = au_cpup_simple(cpg);
5873 +       else {
5874 +               struct au_cpup_simple_args args = {
5875 +                       .errp           = &err,
5876 +                       .cpg            = cpg
5877 +               };
5878 +               wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5879 +               if (unlikely(wkq_err))
5880 +                       err = wkq_err;
5881 +       }
5882 +
5883 +       dput(parent);
5884 +       if (h_file)
5885 +               au_h_open_post(dentry, cpg->bsrc, h_file);
5886 +
5887 +out:
5888 +       return err;
5889 +}
5890 +
5891 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5892 +{
5893 +       aufs_bindex_t bsrc, bbot;
5894 +       struct dentry *dentry, *h_dentry;
5895 +
5896 +       if (cpg->bsrc < 0) {
5897 +               dentry = cpg->dentry;
5898 +               bbot = au_dbbot(dentry);
5899 +               for (bsrc = cpg->bdst + 1; bsrc <= bbot; bsrc++) {
5900 +                       h_dentry = au_h_dptr(dentry, bsrc);
5901 +                       if (h_dentry) {
5902 +                               AuDebugOn(d_is_negative(h_dentry));
5903 +                               break;
5904 +                       }
5905 +               }
5906 +               AuDebugOn(bsrc > bbot);
5907 +               cpg->bsrc = bsrc;
5908 +       }
5909 +       AuDebugOn(cpg->bsrc <= cpg->bdst);
5910 +       return au_do_sio_cpup_simple(cpg);
5911 +}
5912 +
5913 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5914 +{
5915 +       AuDebugOn(cpg->bdst <= cpg->bsrc);
5916 +       return au_do_sio_cpup_simple(cpg);
5917 +}
5918 +
5919 +/* ---------------------------------------------------------------------- */
5920 +
5921 +/*
5922 + * copyup the deleted file for writing.
5923 + */
5924 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5925 +                        struct file *file)
5926 +{
5927 +       int err;
5928 +       unsigned int flags_orig;
5929 +       aufs_bindex_t bsrc_orig;
5930 +       struct au_dinfo *dinfo;
5931 +       struct {
5932 +               struct au_hdentry *hd;
5933 +               struct dentry *h_dentry;
5934 +       } hdst, hsrc;
5935 +
5936 +       dinfo = au_di(cpg->dentry);
5937 +       AuRwMustWriteLock(&dinfo->di_rwsem);
5938 +
5939 +       bsrc_orig = cpg->bsrc;
5940 +       cpg->bsrc = dinfo->di_btop;
5941 +       hdst.hd = au_hdentry(dinfo, cpg->bdst);
5942 +       hdst.h_dentry = hdst.hd->hd_dentry;
5943 +       hdst.hd->hd_dentry = wh_dentry;
5944 +       dinfo->di_btop = cpg->bdst;
5945 +
5946 +       hsrc.h_dentry = NULL;
5947 +       if (file) {
5948 +               hsrc.hd = au_hdentry(dinfo, cpg->bsrc);
5949 +               hsrc.h_dentry = hsrc.hd->hd_dentry;
5950 +               hsrc.hd->hd_dentry = au_hf_top(file)->f_path.dentry;
5951 +       }
5952 +       flags_orig = cpg->flags;
5953 +       cpg->flags = !AuCpup_DTIME;
5954 +       err = au_cpup_single(cpg, /*h_parent*/NULL);
5955 +       cpg->flags = flags_orig;
5956 +       if (file) {
5957 +               if (!err)
5958 +                       err = au_reopen_nondir(file);
5959 +               hsrc.hd->hd_dentry = hsrc.h_dentry;
5960 +       }
5961 +       hdst.hd->hd_dentry = hdst.h_dentry;
5962 +       dinfo->di_btop = cpg->bsrc;
5963 +       cpg->bsrc = bsrc_orig;
5964 +
5965 +       return err;
5966 +}
5967 +
5968 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5969 +{
5970 +       int err;
5971 +       aufs_bindex_t bdst;
5972 +       struct au_dtime dt;
5973 +       struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5974 +       struct au_branch *br;
5975 +       struct path h_path;
5976 +
5977 +       dentry = cpg->dentry;
5978 +       bdst = cpg->bdst;
5979 +       br = au_sbr(dentry->d_sb, bdst);
5980 +       parent = dget_parent(dentry);
5981 +       h_parent = au_h_dptr(parent, bdst);
5982 +       wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5983 +       err = PTR_ERR(wh_dentry);
5984 +       if (IS_ERR(wh_dentry))
5985 +               goto out;
5986 +
5987 +       h_path.dentry = h_parent;
5988 +       h_path.mnt = au_br_mnt(br);
5989 +       au_dtime_store(&dt, parent, &h_path);
5990 +       err = au_do_cpup_wh(cpg, wh_dentry, file);
5991 +       if (unlikely(err))
5992 +               goto out_wh;
5993 +
5994 +       dget(wh_dentry);
5995 +       h_path.dentry = wh_dentry;
5996 +       if (!d_is_dir(wh_dentry)) {
5997 +               /* no delegation since it is just created */
5998 +               err = vfsub_unlink(d_inode(h_parent), &h_path,
5999 +                                  /*delegated*/NULL, /*force*/0);
6000 +       } else
6001 +               err = vfsub_rmdir(d_inode(h_parent), &h_path);
6002 +       if (unlikely(err)) {
6003 +               AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
6004 +                       wh_dentry, err);
6005 +               err = -EIO;
6006 +       }
6007 +       au_dtime_revert(&dt);
6008 +       au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
6009 +
6010 +out_wh:
6011 +       dput(wh_dentry);
6012 +out:
6013 +       dput(parent);
6014 +       return err;
6015 +}
6016 +
6017 +struct au_cpup_wh_args {
6018 +       int *errp;
6019 +       struct au_cp_generic *cpg;
6020 +       struct file *file;
6021 +};
6022 +
6023 +static void au_call_cpup_wh(void *args)
6024 +{
6025 +       struct au_cpup_wh_args *a = args;
6026 +
6027 +       au_pin_hdir_acquire_nest(a->cpg->pin);
6028 +       *a->errp = au_cpup_wh(a->cpg, a->file);
6029 +       au_pin_hdir_release(a->cpg->pin);
6030 +}
6031 +
6032 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6033 +{
6034 +       int err, wkq_err;
6035 +       aufs_bindex_t bdst;
6036 +       struct dentry *dentry, *parent, *h_orph, *h_parent;
6037 +       struct inode *dir, *h_dir, *h_tmpdir;
6038 +       struct au_wbr *wbr;
6039 +       struct au_pin wh_pin, *pin_orig;
6040 +       struct user_namespace *h_userns;
6041 +
6042 +       dentry = cpg->dentry;
6043 +       bdst = cpg->bdst;
6044 +       parent = dget_parent(dentry);
6045 +       dir = d_inode(parent);
6046 +       h_orph = NULL;
6047 +       h_parent = NULL;
6048 +       h_dir = au_igrab(au_h_iptr(dir, bdst));
6049 +       h_tmpdir = h_dir;
6050 +       pin_orig = NULL;
6051 +       if (!h_dir->i_nlink) {
6052 +               wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
6053 +               h_orph = wbr->wbr_orph;
6054 +
6055 +               h_parent = dget(au_h_dptr(parent, bdst));
6056 +               au_set_h_dptr(parent, bdst, dget(h_orph));
6057 +               h_tmpdir = d_inode(h_orph);
6058 +               au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
6059 +
6060 +               inode_lock_nested(h_tmpdir, AuLsc_I_PARENT3);
6061 +               /* todo: au_h_open_pre()? */
6062 +
6063 +               pin_orig = cpg->pin;
6064 +               au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
6065 +                           AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
6066 +               cpg->pin = &wh_pin;
6067 +       }
6068 +
6069 +       h_userns = au_sbr_userns(dentry->d_sb, bdst);
6070 +       if (!au_test_h_perm_sio(h_userns, h_tmpdir, MAY_EXEC | MAY_WRITE)
6071 +           && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6072 +               err = au_cpup_wh(cpg, file);
6073 +       else {
6074 +               struct au_cpup_wh_args args = {
6075 +                       .errp   = &err,
6076 +                       .cpg    = cpg,
6077 +                       .file   = file
6078 +               };
6079 +               wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
6080 +               if (unlikely(wkq_err))
6081 +                       err = wkq_err;
6082 +       }
6083 +
6084 +       if (h_orph) {
6085 +               inode_unlock(h_tmpdir);
6086 +               /* todo: au_h_open_post()? */
6087 +               au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
6088 +               au_set_h_dptr(parent, bdst, h_parent);
6089 +               AuDebugOn(!pin_orig);
6090 +               cpg->pin = pin_orig;
6091 +       }
6092 +       iput(h_dir);
6093 +       dput(parent);
6094 +
6095 +       return err;
6096 +}
6097 +
6098 +/* ---------------------------------------------------------------------- */
6099 +
6100 +/*
6101 + * generic routine for both of copy-up and copy-down.
6102 + */
6103 +/* cf. revalidate function in file.c */
6104 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6105 +              int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6106 +                        struct au_pin *pin,
6107 +                        struct dentry *h_parent, void *arg),
6108 +              void *arg)
6109 +{
6110 +       int err;
6111 +       struct au_pin pin;
6112 +       struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
6113 +
6114 +       err = 0;
6115 +       parent = dget_parent(dentry);
6116 +       if (IS_ROOT(parent))
6117 +               goto out;
6118 +
6119 +       au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
6120 +                   au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
6121 +
6122 +       /* do not use au_dpage */
6123 +       real_parent = parent;
6124 +       while (1) {
6125 +               dput(parent);
6126 +               parent = dget_parent(dentry);
6127 +               h_parent = au_h_dptr(parent, bdst);
6128 +               if (h_parent)
6129 +                       goto out; /* success */
6130 +
6131 +               /* find top dir which is necessary to cpup */
6132 +               do {
6133 +                       d = parent;
6134 +                       dput(parent);
6135 +                       parent = dget_parent(d);
6136 +                       di_read_lock_parent3(parent, !AuLock_IR);
6137 +                       h_parent = au_h_dptr(parent, bdst);
6138 +                       di_read_unlock(parent, !AuLock_IR);
6139 +               } while (!h_parent);
6140 +
6141 +               if (d != real_parent)
6142 +                       di_write_lock_child3(d);
6143 +
6144 +               /* somebody else might create while we were sleeping */
6145 +               h_dentry = au_h_dptr(d, bdst);
6146 +               if (!h_dentry || d_is_negative(h_dentry)) {
6147 +                       if (h_dentry)
6148 +                               au_update_dbtop(d);
6149 +
6150 +                       au_pin_set_dentry(&pin, d);
6151 +                       err = au_do_pin(&pin);
6152 +                       if (!err) {
6153 +                               err = cp(d, bdst, &pin, h_parent, arg);
6154 +                               au_unpin(&pin);
6155 +                       }
6156 +               }
6157 +
6158 +               if (d != real_parent)
6159 +                       di_write_unlock(d);
6160 +               if (unlikely(err))
6161 +                       break;
6162 +       }
6163 +
6164 +out:
6165 +       dput(parent);
6166 +       return err;
6167 +}
6168 +
6169 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
6170 +                      struct au_pin *pin,
6171 +                      struct dentry *h_parent __maybe_unused,
6172 +                      void *arg __maybe_unused)
6173 +{
6174 +       struct au_cp_generic cpg = {
6175 +               .dentry = dentry,
6176 +               .bdst   = bdst,
6177 +               .bsrc   = -1,
6178 +               .len    = 0,
6179 +               .pin    = pin,
6180 +               .flags  = AuCpup_DTIME
6181 +       };
6182 +       return au_sio_cpup_simple(&cpg);
6183 +}
6184 +
6185 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6186 +{
6187 +       return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
6188 +}
6189 +
6190 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6191 +{
6192 +       int err;
6193 +       struct dentry *parent;
6194 +       struct inode *dir;
6195 +
6196 +       parent = dget_parent(dentry);
6197 +       dir = d_inode(parent);
6198 +       err = 0;
6199 +       if (au_h_iptr(dir, bdst))
6200 +               goto out;
6201 +
6202 +       di_read_unlock(parent, AuLock_IR);
6203 +       di_write_lock_parent(parent);
6204 +       /* someone else might change our inode while we were sleeping */
6205 +       if (!au_h_iptr(dir, bdst))
6206 +               err = au_cpup_dirs(dentry, bdst);
6207 +       di_downgrade_lock(parent, AuLock_IR);
6208 +
6209 +out:
6210 +       dput(parent);
6211 +       return err;
6212 +}
6213 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
6214 --- /usr/share/empty/fs/aufs/cpup.h     1970-01-01 01:00:00.000000000 +0100
6215 +++ linux/fs/aufs/cpup.h        2022-11-05 23:02:18.962555950 +0100
6216 @@ -0,0 +1,100 @@
6217 +/* SPDX-License-Identifier: GPL-2.0 */
6218 +/*
6219 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6220 + *
6221 + * This program is free software; you can redistribute it and/or modify
6222 + * it under the terms of the GNU General Public License as published by
6223 + * the Free Software Foundation; either version 2 of the License, or
6224 + * (at your option) any later version.
6225 + *
6226 + * This program is distributed in the hope that it will be useful,
6227 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6228 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6229 + * GNU General Public License for more details.
6230 + *
6231 + * You should have received a copy of the GNU General Public License
6232 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6233 + */
6234 +
6235 +/*
6236 + * copy-up/down functions
6237 + */
6238 +
6239 +#ifndef __AUFS_CPUP_H__
6240 +#define __AUFS_CPUP_H__
6241 +
6242 +#ifdef __KERNEL__
6243 +
6244 +#include <linux/path.h>
6245 +
6246 +struct inode;
6247 +struct file;
6248 +struct au_pin;
6249 +
6250 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
6251 +void au_cpup_attr_timesizes(struct inode *inode);
6252 +void au_cpup_attr_nlink(struct inode *inode, int force);
6253 +void au_cpup_attr_changeable(struct inode *inode);
6254 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
6255 +void au_cpup_attr_all(struct inode *inode, int force);
6256 +
6257 +/* ---------------------------------------------------------------------- */
6258 +
6259 +struct au_cp_generic {
6260 +       struct dentry   *dentry;
6261 +       aufs_bindex_t   bdst, bsrc;
6262 +       loff_t          len;
6263 +       struct au_pin   *pin;
6264 +       unsigned int    flags;
6265 +};
6266 +
6267 +/* cpup flags */
6268 +#define AuCpup_DTIME           1               /* do dtime_store/revert */
6269 +#define AuCpup_KEEPLINO                (1 << 1)        /* do not clear the lower xino,
6270 +                                                  for link(2) */
6271 +#define AuCpup_RENAME          (1 << 2)        /* rename after cpup */
6272 +#define AuCpup_HOPEN           (1 << 3)        /* call h_open_pre/post() in
6273 +                                                  cpup */
6274 +#define AuCpup_OVERWRITE       (1 << 4)        /* allow overwriting the
6275 +                                                  existing entry */
6276 +#define AuCpup_RWDST           (1 << 5)        /* force write target even if
6277 +                                                  the branch is marked as RO */
6278 +
6279 +#ifndef CONFIG_AUFS_BR_HFSPLUS
6280 +#undef AuCpup_HOPEN
6281 +#define AuCpup_HOPEN           0
6282 +#endif
6283 +
6284 +#define au_ftest_cpup(flags, name)     ((flags) & AuCpup_##name)
6285 +#define au_fset_cpup(flags, name) \
6286 +       do { (flags) |= AuCpup_##name; } while (0)
6287 +#define au_fclr_cpup(flags, name) \
6288 +       do { (flags) &= ~AuCpup_##name; } while (0)
6289 +
6290 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
6291 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
6292 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
6293 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
6294 +
6295 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6296 +              int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6297 +                        struct au_pin *pin,
6298 +                        struct dentry *h_parent, void *arg),
6299 +              void *arg);
6300 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6301 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6302 +
6303 +/* ---------------------------------------------------------------------- */
6304 +
6305 +/* keep timestamps when copyup */
6306 +struct au_dtime {
6307 +       struct dentry *dt_dentry;
6308 +       struct path dt_h_path;
6309 +       struct timespec64 dt_atime, dt_mtime;
6310 +};
6311 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
6312 +                   struct path *h_path);
6313 +void au_dtime_revert(struct au_dtime *dt);
6314 +
6315 +#endif /* __KERNEL__ */
6316 +#endif /* __AUFS_CPUP_H__ */
6317 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
6318 --- /usr/share/empty/fs/aufs/dbgaufs.c  1970-01-01 01:00:00.000000000 +0100
6319 +++ linux/fs/aufs/dbgaufs.c     2022-11-05 23:02:18.962555950 +0100
6320 @@ -0,0 +1,526 @@
6321 +// SPDX-License-Identifier: GPL-2.0
6322 +/*
6323 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6324 + *
6325 + * This program is free software; you can redistribute it and/or modify
6326 + * it under the terms of the GNU General Public License as published by
6327 + * the Free Software Foundation; either version 2 of the License, or
6328 + * (at your option) any later version.
6329 + *
6330 + * This program is distributed in the hope that it will be useful,
6331 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6332 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6333 + * GNU General Public License for more details.
6334 + *
6335 + * You should have received a copy of the GNU General Public License
6336 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6337 + */
6338 +
6339 +/*
6340 + * debugfs interface
6341 + */
6342 +
6343 +#include <linux/debugfs.h>
6344 +#include "aufs.h"
6345 +
6346 +#ifndef CONFIG_SYSFS
6347 +#error DEBUG_FS depends upon SYSFS
6348 +#endif
6349 +
6350 +static struct dentry *dbgaufs;
6351 +static const mode_t dbgaufs_mode = 0444;
6352 +
6353 +/* 20 is max digits length of ulong 64 */
6354 +struct dbgaufs_arg {
6355 +       int n;
6356 +       char a[20 * 4];
6357 +};
6358 +
6359 +/*
6360 + * common function for all XINO files
6361 + */
6362 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
6363 +                             struct file *file)
6364 +{
6365 +       void *p;
6366 +
6367 +       p = file->private_data;
6368 +       if (p) {
6369 +               /* this is struct dbgaufs_arg */
6370 +               AuDebugOn(!au_kfree_sz_test(p));
6371 +               au_kfree_do_rcu(p);
6372 +       }
6373 +       return 0;
6374 +}
6375 +
6376 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt,
6377 +                          int cnt)
6378 +{
6379 +       int err;
6380 +       struct kstat st;
6381 +       struct dbgaufs_arg *p;
6382 +
6383 +       err = -ENOMEM;
6384 +       p = kmalloc(sizeof(*p), GFP_NOFS);
6385 +       if (unlikely(!p))
6386 +               goto out;
6387 +
6388 +       err = 0;
6389 +       p->n = 0;
6390 +       file->private_data = p;
6391 +       if (!xf)
6392 +               goto out;
6393 +
6394 +       err = vfsub_getattr(&xf->f_path, &st);
6395 +       if (!err) {
6396 +               if (do_fcnt)
6397 +                       p->n = snprintf
6398 +                               (p->a, sizeof(p->a), "%d, %llux%u %lld\n",
6399 +                                cnt, st.blocks, st.blksize,
6400 +                                (long long)st.size);
6401 +               else
6402 +                       p->n = snprintf(p->a, sizeof(p->a), "%llux%u %lld\n",
6403 +                                       st.blocks, st.blksize,
6404 +                                       (long long)st.size);
6405 +               AuDebugOn(p->n >= sizeof(p->a));
6406 +       } else {
6407 +               p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
6408 +               err = 0;
6409 +       }
6410 +
6411 +out:
6412 +       return err;
6413 +}
6414 +
6415 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
6416 +                              size_t count, loff_t *ppos)
6417 +{
6418 +       struct dbgaufs_arg *p;
6419 +
6420 +       p = file->private_data;
6421 +       return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6422 +}
6423 +
6424 +/* ---------------------------------------------------------------------- */
6425 +
6426 +struct dbgaufs_plink_arg {
6427 +       int n;
6428 +       char a[];
6429 +};
6430 +
6431 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6432 +                                struct file *file)
6433 +{
6434 +       free_page((unsigned long)file->private_data);
6435 +       return 0;
6436 +}
6437 +
6438 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6439 +{
6440 +       int err, i, limit;
6441 +       unsigned long n, sum;
6442 +       struct dbgaufs_plink_arg *p;
6443 +       struct au_sbinfo *sbinfo;
6444 +       struct super_block *sb;
6445 +       struct hlist_bl_head *hbl;
6446 +
6447 +       err = -ENOMEM;
6448 +       p = (void *)get_zeroed_page(GFP_NOFS);
6449 +       if (unlikely(!p))
6450 +               goto out;
6451 +
6452 +       err = -EFBIG;
6453 +       sbinfo = inode->i_private;
6454 +       sb = sbinfo->si_sb;
6455 +       si_noflush_read_lock(sb);
6456 +       if (au_opt_test(au_mntflags(sb), PLINK)) {
6457 +               limit = PAGE_SIZE - sizeof(p->n);
6458 +
6459 +               /* the number of buckets */
6460 +               n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6461 +               p->n += n;
6462 +               limit -= n;
6463 +
6464 +               sum = 0;
6465 +               for (i = 0, hbl = sbinfo->si_plink; i < AuPlink_NHASH;
6466 +                    i++, hbl++) {
6467 +                       n = au_hbl_count(hbl);
6468 +                       sum += n;
6469 +
6470 +                       n = snprintf(p->a + p->n, limit, "%lu ", n);
6471 +                       p->n += n;
6472 +                       limit -= n;
6473 +                       if (unlikely(limit <= 0))
6474 +                               goto out_free;
6475 +               }
6476 +               p->a[p->n - 1] = '\n';
6477 +
6478 +               /* the sum of plinks */
6479 +               n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6480 +               p->n += n;
6481 +               limit -= n;
6482 +               if (unlikely(limit <= 0))
6483 +                       goto out_free;
6484 +       } else {
6485 +#define str "1\n0\n0\n"
6486 +               p->n = sizeof(str) - 1;
6487 +               strcpy(p->a, str);
6488 +#undef str
6489 +       }
6490 +       si_read_unlock(sb);
6491 +
6492 +       err = 0;
6493 +       file->private_data = p;
6494 +       goto out; /* success */
6495 +
6496 +out_free:
6497 +       free_page((unsigned long)p);
6498 +out:
6499 +       return err;
6500 +}
6501 +
6502 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6503 +                                 size_t count, loff_t *ppos)
6504 +{
6505 +       struct dbgaufs_plink_arg *p;
6506 +
6507 +       p = file->private_data;
6508 +       return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6509 +}
6510 +
6511 +static const struct file_operations dbgaufs_plink_fop = {
6512 +       .owner          = THIS_MODULE,
6513 +       .open           = dbgaufs_plink_open,
6514 +       .release        = dbgaufs_plink_release,
6515 +       .read           = dbgaufs_plink_read
6516 +};
6517 +
6518 +/* ---------------------------------------------------------------------- */
6519 +
6520 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6521 +{
6522 +       int err;
6523 +       struct au_sbinfo *sbinfo;
6524 +       struct super_block *sb;
6525 +
6526 +       sbinfo = inode->i_private;
6527 +       sb = sbinfo->si_sb;
6528 +       si_noflush_read_lock(sb);
6529 +       err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0, /*cnt*/0);
6530 +       si_read_unlock(sb);
6531 +       return err;
6532 +}
6533 +
6534 +static const struct file_operations dbgaufs_xib_fop = {
6535 +       .owner          = THIS_MODULE,
6536 +       .open           = dbgaufs_xib_open,
6537 +       .release        = dbgaufs_xi_release,
6538 +       .read           = dbgaufs_xi_read
6539 +};
6540 +
6541 +/* ---------------------------------------------------------------------- */
6542 +
6543 +#define DbgaufsXi_PREFIX "xi"
6544 +
6545 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6546 +{
6547 +       int err, idx;
6548 +       long l;
6549 +       aufs_bindex_t bindex;
6550 +       char *p, a[sizeof(DbgaufsXi_PREFIX) + 8];
6551 +       struct au_sbinfo *sbinfo;
6552 +       struct super_block *sb;
6553 +       struct au_xino *xi;
6554 +       struct file *xf;
6555 +       struct qstr *name;
6556 +       struct au_branch *br;
6557 +
6558 +       err = -ENOENT;
6559 +       name = &file->f_path.dentry->d_name;
6560 +       if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6561 +                    || memcmp(name->name, DbgaufsXi_PREFIX,
6562 +                              sizeof(DbgaufsXi_PREFIX) - 1)))
6563 +               goto out;
6564 +
6565 +       AuDebugOn(name->len >= sizeof(a));
6566 +       memcpy(a, name->name, name->len);
6567 +       a[name->len] = '\0';
6568 +       p = strchr(a, '-');
6569 +       if (p)
6570 +               *p = '\0';
6571 +       err = kstrtol(a + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6572 +       if (unlikely(err))
6573 +               goto out;
6574 +       bindex = l;
6575 +       idx = 0;
6576 +       if (p) {
6577 +               err = kstrtol(p + 1, 10, &l);
6578 +               if (unlikely(err))
6579 +                       goto out;
6580 +               idx = l;
6581 +       }
6582 +
6583 +       err = -ENOENT;
6584 +       sbinfo = inode->i_private;
6585 +       sb = sbinfo->si_sb;
6586 +       si_noflush_read_lock(sb);
6587 +       if (unlikely(bindex < 0 || bindex > au_sbbot(sb)))
6588 +               goto out_si;
6589 +       br = au_sbr(sb, bindex);
6590 +       xi = br->br_xino;
6591 +       if (unlikely(idx >= xi->xi_nfile))
6592 +               goto out_si;
6593 +       xf = au_xino_file(xi, idx);
6594 +       if (xf)
6595 +               err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1,
6596 +                                     au_xino_count(br));
6597 +
6598 +out_si:
6599 +       si_read_unlock(sb);
6600 +out:
6601 +       AuTraceErr(err);
6602 +       return err;
6603 +}
6604 +
6605 +static const struct file_operations dbgaufs_xino_fop = {
6606 +       .owner          = THIS_MODULE,
6607 +       .open           = dbgaufs_xino_open,
6608 +       .release        = dbgaufs_xi_release,
6609 +       .read           = dbgaufs_xi_read
6610 +};
6611 +
6612 +void dbgaufs_xino_del(struct au_branch *br)
6613 +{
6614 +       struct dentry *dbgaufs;
6615 +
6616 +       dbgaufs = br->br_dbgaufs;
6617 +       if (!dbgaufs)
6618 +               return;
6619 +
6620 +       br->br_dbgaufs = NULL;
6621 +       /* debugfs acquires the parent i_mutex */
6622 +       lockdep_off();
6623 +       debugfs_remove(dbgaufs);
6624 +       lockdep_on();
6625 +}
6626 +
6627 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6628 +{
6629 +       aufs_bindex_t bbot;
6630 +       struct au_branch *br;
6631 +
6632 +       if (!au_sbi(sb)->si_dbgaufs)
6633 +               return;
6634 +
6635 +       bbot = au_sbbot(sb);
6636 +       for (; bindex <= bbot; bindex++) {
6637 +               br = au_sbr(sb, bindex);
6638 +               dbgaufs_xino_del(br);
6639 +       }
6640 +}
6641 +
6642 +static void dbgaufs_br_do_add(struct super_block *sb, aufs_bindex_t bindex,
6643 +                             unsigned int idx, struct dentry *parent,
6644 +                             struct au_sbinfo *sbinfo)
6645 +{
6646 +       struct au_branch *br;
6647 +       struct dentry *d;
6648 +       /* "xi" bindex(5) "-" idx(2) NULL */
6649 +       char name[sizeof(DbgaufsXi_PREFIX) + 8];
6650 +
6651 +       if (!idx)
6652 +               snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6653 +       else
6654 +               snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d-%u",
6655 +                        bindex, idx);
6656 +       br = au_sbr(sb, bindex);
6657 +       if (br->br_dbgaufs) {
6658 +               struct qstr qstr = QSTR_INIT(name, strlen(name));
6659 +
6660 +               if (!au_qstreq(&br->br_dbgaufs->d_name, &qstr)) {
6661 +                       /* debugfs acquires the parent i_mutex */
6662 +                       lockdep_off();
6663 +                       d = debugfs_rename(parent, br->br_dbgaufs, parent,
6664 +                                          name);
6665 +                       lockdep_on();
6666 +                       if (unlikely(!d))
6667 +                               pr_warn("failed renaming %pd/%s, ignored.\n",
6668 +                                       parent, name);
6669 +               }
6670 +       } else {
6671 +               lockdep_off();
6672 +               br->br_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6673 +                                                    sbinfo, &dbgaufs_xino_fop);
6674 +               lockdep_on();
6675 +               if (unlikely(!br->br_dbgaufs))
6676 +                       pr_warn("failed creating %pd/%s, ignored.\n",
6677 +                               parent, name);
6678 +       }
6679 +}
6680 +
6681 +static void dbgaufs_br_add(struct super_block *sb, aufs_bindex_t bindex,
6682 +                          struct dentry *parent, struct au_sbinfo *sbinfo)
6683 +{
6684 +       struct au_branch *br;
6685 +       struct au_xino *xi;
6686 +       unsigned int u;
6687 +
6688 +       br = au_sbr(sb, bindex);
6689 +       xi = br->br_xino;
6690 +       for (u = 0; u < xi->xi_nfile; u++)
6691 +               dbgaufs_br_do_add(sb, bindex, u, parent, sbinfo);
6692 +}
6693 +
6694 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown)
6695 +{
6696 +       struct au_sbinfo *sbinfo;
6697 +       struct dentry *parent;
6698 +       aufs_bindex_t bbot;
6699 +
6700 +       if (!au_opt_test(au_mntflags(sb), XINO))
6701 +               return;
6702 +
6703 +       sbinfo = au_sbi(sb);
6704 +       parent = sbinfo->si_dbgaufs;
6705 +       if (!parent)
6706 +               return;
6707 +
6708 +       bbot = au_sbbot(sb);
6709 +       if (topdown)
6710 +               for (; bindex <= bbot; bindex++)
6711 +                       dbgaufs_br_add(sb, bindex, parent, sbinfo);
6712 +       else
6713 +               for (; bbot >= bindex; bbot--)
6714 +                       dbgaufs_br_add(sb, bbot, parent, sbinfo);
6715 +}
6716 +
6717 +/* ---------------------------------------------------------------------- */
6718 +
6719 +#ifdef CONFIG_AUFS_EXPORT
6720 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6721 +{
6722 +       int err;
6723 +       struct au_sbinfo *sbinfo;
6724 +       struct super_block *sb;
6725 +
6726 +       sbinfo = inode->i_private;
6727 +       sb = sbinfo->si_sb;
6728 +       si_noflush_read_lock(sb);
6729 +       err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0, /*cnt*/0);
6730 +       si_read_unlock(sb);
6731 +       return err;
6732 +}
6733 +
6734 +static const struct file_operations dbgaufs_xigen_fop = {
6735 +       .owner          = THIS_MODULE,
6736 +       .open           = dbgaufs_xigen_open,
6737 +       .release        = dbgaufs_xi_release,
6738 +       .read           = dbgaufs_xi_read
6739 +};
6740 +
6741 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6742 +{
6743 +       int err;
6744 +
6745 +       /*
6746 +        * This function is a dynamic '__init' function actually,
6747 +        * so the tiny check for si_rwsem is unnecessary.
6748 +        */
6749 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6750 +
6751 +       err = -EIO;
6752 +       sbinfo->si_dbgaufs_xigen = debugfs_create_file
6753 +               ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6754 +                &dbgaufs_xigen_fop);
6755 +       if (sbinfo->si_dbgaufs_xigen)
6756 +               err = 0;
6757 +
6758 +       return err;
6759 +}
6760 +#else
6761 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6762 +{
6763 +       return 0;
6764 +}
6765 +#endif /* CONFIG_AUFS_EXPORT */
6766 +
6767 +/* ---------------------------------------------------------------------- */
6768 +
6769 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6770 +{
6771 +       /*
6772 +        * This function is a dynamic '__fin' function actually,
6773 +        * so the tiny check for si_rwsem is unnecessary.
6774 +        */
6775 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6776 +
6777 +       debugfs_remove_recursive(sbinfo->si_dbgaufs);
6778 +       sbinfo->si_dbgaufs = NULL;
6779 +}
6780 +
6781 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6782 +{
6783 +       int err;
6784 +       char name[SysaufsSiNameLen];
6785 +
6786 +       /*
6787 +        * This function is a dynamic '__init' function actually,
6788 +        * so the tiny check for si_rwsem is unnecessary.
6789 +        */
6790 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6791 +
6792 +       err = -ENOENT;
6793 +       if (!dbgaufs) {
6794 +               AuErr1("/debug/aufs is uninitialized\n");
6795 +               goto out;
6796 +       }
6797 +
6798 +       err = -EIO;
6799 +       sysaufs_name(sbinfo, name);
6800 +       sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6801 +       if (unlikely(!sbinfo->si_dbgaufs))
6802 +               goto out;
6803 +
6804 +       /* regardless plink/noplink option */
6805 +       sbinfo->si_dbgaufs_plink = debugfs_create_file
6806 +               ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6807 +                &dbgaufs_plink_fop);
6808 +       if (unlikely(!sbinfo->si_dbgaufs_plink))
6809 +               goto out_dir;
6810 +
6811 +       /* regardless xino/noxino option */
6812 +       sbinfo->si_dbgaufs_xib = debugfs_create_file
6813 +               ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6814 +                &dbgaufs_xib_fop);
6815 +       if (unlikely(!sbinfo->si_dbgaufs_xib))
6816 +               goto out_dir;
6817 +
6818 +       err = dbgaufs_xigen_init(sbinfo);
6819 +       if (!err)
6820 +               goto out; /* success */
6821 +
6822 +out_dir:
6823 +       dbgaufs_si_fin(sbinfo);
6824 +out:
6825 +       if (unlikely(err))
6826 +               pr_err("debugfs/aufs failed\n");
6827 +       return err;
6828 +}
6829 +
6830 +/* ---------------------------------------------------------------------- */
6831 +
6832 +void dbgaufs_fin(void)
6833 +{
6834 +       debugfs_remove(dbgaufs);
6835 +}
6836 +
6837 +int __init dbgaufs_init(void)
6838 +{
6839 +       int err;
6840 +
6841 +       err = -EIO;
6842 +       dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6843 +       if (dbgaufs)
6844 +               err = 0;
6845 +       return err;
6846 +}
6847 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6848 --- /usr/share/empty/fs/aufs/dbgaufs.h  1970-01-01 01:00:00.000000000 +0100
6849 +++ linux/fs/aufs/dbgaufs.h     2022-11-05 23:02:18.962555950 +0100
6850 @@ -0,0 +1,53 @@
6851 +/* SPDX-License-Identifier: GPL-2.0 */
6852 +/*
6853 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6854 + *
6855 + * This program is free software; you can redistribute it and/or modify
6856 + * it under the terms of the GNU General Public License as published by
6857 + * the Free Software Foundation; either version 2 of the License, or
6858 + * (at your option) any later version.
6859 + *
6860 + * This program is distributed in the hope that it will be useful,
6861 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6862 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6863 + * GNU General Public License for more details.
6864 + *
6865 + * You should have received a copy of the GNU General Public License
6866 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6867 + */
6868 +
6869 +/*
6870 + * debugfs interface
6871 + */
6872 +
6873 +#ifndef __DBGAUFS_H__
6874 +#define __DBGAUFS_H__
6875 +
6876 +#ifdef __KERNEL__
6877 +
6878 +struct super_block;
6879 +struct au_sbinfo;
6880 +struct au_branch;
6881 +
6882 +#ifdef CONFIG_DEBUG_FS
6883 +/* dbgaufs.c */
6884 +void dbgaufs_xino_del(struct au_branch *br);
6885 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6886 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown);
6887 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6888 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6889 +void dbgaufs_fin(void);
6890 +int __init dbgaufs_init(void);
6891 +#else
6892 +AuStubVoid(dbgaufs_xino_del, struct au_branch *br)
6893 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6894 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex,
6895 +          int topdown)
6896 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6897 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6898 +AuStubVoid(dbgaufs_fin, void)
6899 +AuStubInt0(__init dbgaufs_init, void)
6900 +#endif /* CONFIG_DEBUG_FS */
6901 +
6902 +#endif /* __KERNEL__ */
6903 +#endif /* __DBGAUFS_H__ */
6904 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6905 --- /usr/share/empty/fs/aufs/dcsub.c    1970-01-01 01:00:00.000000000 +0100
6906 +++ linux/fs/aufs/dcsub.c       2022-11-05 23:02:18.962555950 +0100
6907 @@ -0,0 +1,225 @@
6908 +// SPDX-License-Identifier: GPL-2.0
6909 +/*
6910 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6911 + *
6912 + * This program is free software; you can redistribute it and/or modify
6913 + * it under the terms of the GNU General Public License as published by
6914 + * the Free Software Foundation; either version 2 of the License, or
6915 + * (at your option) any later version.
6916 + *
6917 + * This program is distributed in the hope that it will be useful,
6918 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6919 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
6920 + * GNU General Public License for more details.
6921 + *
6922 + * You should have received a copy of the GNU General Public License
6923 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
6924 + */
6925 +
6926 +/*
6927 + * sub-routines for dentry cache
6928 + */
6929 +
6930 +#include "aufs.h"
6931 +
6932 +static void au_dpage_free(struct au_dpage *dpage)
6933 +{
6934 +       int i;
6935 +       struct dentry **p;
6936 +
6937 +       p = dpage->dentries;
6938 +       for (i = 0; i < dpage->ndentry; i++)
6939 +               dput(*p++);
6940 +       free_page((unsigned long)dpage->dentries);
6941 +}
6942 +
6943 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6944 +{
6945 +       int err;
6946 +       void *p;
6947 +
6948 +       err = -ENOMEM;
6949 +       dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6950 +       if (unlikely(!dpages->dpages))
6951 +               goto out;
6952 +
6953 +       p = (void *)__get_free_page(gfp);
6954 +       if (unlikely(!p))
6955 +               goto out_dpages;
6956 +
6957 +       dpages->dpages[0].ndentry = 0;
6958 +       dpages->dpages[0].dentries = p;
6959 +       dpages->ndpage = 1;
6960 +       return 0; /* success */
6961 +
6962 +out_dpages:
6963 +       au_kfree_try_rcu(dpages->dpages);
6964 +out:
6965 +       return err;
6966 +}
6967 +
6968 +void au_dpages_free(struct au_dcsub_pages *dpages)
6969 +{
6970 +       int i;
6971 +       struct au_dpage *p;
6972 +
6973 +       p = dpages->dpages;
6974 +       for (i = 0; i < dpages->ndpage; i++)
6975 +               au_dpage_free(p++);
6976 +       au_kfree_try_rcu(dpages->dpages);
6977 +}
6978 +
6979 +static int au_dpages_append(struct au_dcsub_pages *dpages,
6980 +                           struct dentry *dentry, gfp_t gfp)
6981 +{
6982 +       int err, sz;
6983 +       struct au_dpage *dpage;
6984 +       void *p;
6985 +
6986 +       dpage = dpages->dpages + dpages->ndpage - 1;
6987 +       sz = PAGE_SIZE / sizeof(dentry);
6988 +       if (unlikely(dpage->ndentry >= sz)) {
6989 +               AuLabel(new dpage);
6990 +               err = -ENOMEM;
6991 +               sz = dpages->ndpage * sizeof(*dpages->dpages);
6992 +               p = au_kzrealloc(dpages->dpages, sz,
6993 +                                sz + sizeof(*dpages->dpages), gfp,
6994 +                                /*may_shrink*/0);
6995 +               if (unlikely(!p))
6996 +                       goto out;
6997 +
6998 +               dpages->dpages = p;
6999 +               dpage = dpages->dpages + dpages->ndpage;
7000 +               p = (void *)__get_free_page(gfp);
7001 +               if (unlikely(!p))
7002 +                       goto out;
7003 +
7004 +               dpage->ndentry = 0;
7005 +               dpage->dentries = p;
7006 +               dpages->ndpage++;
7007 +       }
7008 +
7009 +       AuDebugOn(au_dcount(dentry) <= 0);
7010 +       dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
7011 +       return 0; /* success */
7012 +
7013 +out:
7014 +       return err;
7015 +}
7016 +
7017 +/* todo: BAD approach */
7018 +/* copied from linux/fs/dcache.c */
7019 +enum d_walk_ret {
7020 +       D_WALK_CONTINUE,
7021 +       D_WALK_QUIT,
7022 +       D_WALK_NORETRY,
7023 +       D_WALK_SKIP,
7024 +};
7025 +
7026 +extern void d_walk(struct dentry *parent, void *data,
7027 +                  enum d_walk_ret (*enter)(void *, struct dentry *));
7028 +
7029 +struct ac_dpages_arg {
7030 +       int err;
7031 +       struct au_dcsub_pages *dpages;
7032 +       struct super_block *sb;
7033 +       au_dpages_test test;
7034 +       void *arg;
7035 +};
7036 +
7037 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
7038 +{
7039 +       enum d_walk_ret ret;
7040 +       struct ac_dpages_arg *arg = _arg;
7041 +
7042 +       ret = D_WALK_CONTINUE;
7043 +       if (dentry->d_sb == arg->sb
7044 +           && !IS_ROOT(dentry)
7045 +           && au_dcount(dentry) > 0
7046 +           && au_di(dentry)
7047 +           && (!arg->test || arg->test(dentry, arg->arg))) {
7048 +               arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
7049 +               if (unlikely(arg->err))
7050 +                       ret = D_WALK_QUIT;
7051 +       }
7052 +
7053 +       return ret;
7054 +}
7055 +
7056 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7057 +                  au_dpages_test test, void *arg)
7058 +{
7059 +       struct ac_dpages_arg args = {
7060 +               .err    = 0,
7061 +               .dpages = dpages,
7062 +               .sb     = root->d_sb,
7063 +               .test   = test,
7064 +               .arg    = arg
7065 +       };
7066 +
7067 +       d_walk(root, &args, au_call_dpages_append);
7068 +
7069 +       return args.err;
7070 +}
7071 +
7072 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7073 +                      int do_include, au_dpages_test test, void *arg)
7074 +{
7075 +       int err;
7076 +
7077 +       err = 0;
7078 +       write_seqlock(&rename_lock);
7079 +       spin_lock(&dentry->d_lock);
7080 +       if (do_include
7081 +           && au_dcount(dentry) > 0
7082 +           && (!test || test(dentry, arg)))
7083 +               err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7084 +       spin_unlock(&dentry->d_lock);
7085 +       if (unlikely(err))
7086 +               goto out;
7087 +
7088 +       /*
7089 +        * RCU for vfsmount is unnecessary since this is a traverse in a single
7090 +        * mount
7091 +        */
7092 +       while (!IS_ROOT(dentry)) {
7093 +               dentry = dentry->d_parent; /* rename_lock is locked */
7094 +               spin_lock(&dentry->d_lock);
7095 +               if (au_dcount(dentry) > 0
7096 +                   && (!test || test(dentry, arg)))
7097 +                       err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7098 +               spin_unlock(&dentry->d_lock);
7099 +               if (unlikely(err))
7100 +                       break;
7101 +       }
7102 +
7103 +out:
7104 +       write_sequnlock(&rename_lock);
7105 +       return err;
7106 +}
7107 +
7108 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
7109 +{
7110 +       return au_di(dentry) && dentry->d_sb == arg;
7111 +}
7112 +
7113 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7114 +                           struct dentry *dentry, int do_include)
7115 +{
7116 +       return au_dcsub_pages_rev(dpages, dentry, do_include,
7117 +                                 au_dcsub_dpages_aufs, dentry->d_sb);
7118 +}
7119 +
7120 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
7121 +{
7122 +       struct path path[2] = {
7123 +               {
7124 +                       .dentry = d1
7125 +               },
7126 +               {
7127 +                       .dentry = d2
7128 +               }
7129 +       };
7130 +
7131 +       return path_is_under(path + 0, path + 1);
7132 +}
7133 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
7134 --- /usr/share/empty/fs/aufs/dcsub.h    1970-01-01 01:00:00.000000000 +0100
7135 +++ linux/fs/aufs/dcsub.h       2022-11-05 23:02:18.962555950 +0100
7136 @@ -0,0 +1,137 @@
7137 +/* SPDX-License-Identifier: GPL-2.0 */
7138 +/*
7139 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7140 + *
7141 + * This program is free software; you can redistribute it and/or modify
7142 + * it under the terms of the GNU General Public License as published by
7143 + * the Free Software Foundation; either version 2 of the License, or
7144 + * (at your option) any later version.
7145 + *
7146 + * This program is distributed in the hope that it will be useful,
7147 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7148 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7149 + * GNU General Public License for more details.
7150 + *
7151 + * You should have received a copy of the GNU General Public License
7152 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7153 + */
7154 +
7155 +/*
7156 + * sub-routines for dentry cache
7157 + */
7158 +
7159 +#ifndef __AUFS_DCSUB_H__
7160 +#define __AUFS_DCSUB_H__
7161 +
7162 +#ifdef __KERNEL__
7163 +
7164 +#include <linux/dcache.h>
7165 +#include <linux/fs.h>
7166 +
7167 +struct au_dpage {
7168 +       int ndentry;
7169 +       struct dentry **dentries;
7170 +};
7171 +
7172 +struct au_dcsub_pages {
7173 +       int ndpage;
7174 +       struct au_dpage *dpages;
7175 +};
7176 +
7177 +/* ---------------------------------------------------------------------- */
7178 +
7179 +/* dcsub.c */
7180 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
7181 +void au_dpages_free(struct au_dcsub_pages *dpages);
7182 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
7183 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7184 +                  au_dpages_test test, void *arg);
7185 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7186 +                      int do_include, au_dpages_test test, void *arg);
7187 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7188 +                           struct dentry *dentry, int do_include);
7189 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
7190 +
7191 +/* ---------------------------------------------------------------------- */
7192 +
7193 +/*
7194 + * todo: in linux-3.13, several similar (but faster) helpers are added to
7195 + * include/linux/dcache.h. Try them (in the future).
7196 + */
7197 +
7198 +static inline int au_d_hashed_positive(struct dentry *d)
7199 +{
7200 +       int err;
7201 +       struct inode *inode = d_inode(d);
7202 +
7203 +       err = 0;
7204 +       if (unlikely(d_unhashed(d)
7205 +                    || d_is_negative(d)
7206 +                    || !inode->i_nlink))
7207 +               err = -ENOENT;
7208 +       return err;
7209 +}
7210 +
7211 +static inline int au_d_linkable(struct dentry *d)
7212 +{
7213 +       int err;
7214 +       struct inode *inode = d_inode(d);
7215 +
7216 +       err = au_d_hashed_positive(d);
7217 +       if (err
7218 +           && d_is_positive(d)
7219 +           && (inode->i_state & I_LINKABLE))
7220 +               err = 0;
7221 +       return err;
7222 +}
7223 +
7224 +static inline int au_d_alive(struct dentry *d)
7225 +{
7226 +       int err;
7227 +       struct inode *inode;
7228 +
7229 +       err = 0;
7230 +       if (!IS_ROOT(d))
7231 +               err = au_d_hashed_positive(d);
7232 +       else {
7233 +               inode = d_inode(d);
7234 +               if (unlikely(d_unlinked(d)
7235 +                            || d_is_negative(d)
7236 +                            || !inode->i_nlink))
7237 +                       err = -ENOENT;
7238 +       }
7239 +       return err;
7240 +}
7241 +
7242 +static inline int au_alive_dir(struct dentry *d)
7243 +{
7244 +       int err;
7245 +
7246 +       err = au_d_alive(d);
7247 +       if (unlikely(err || IS_DEADDIR(d_inode(d))))
7248 +               err = -ENOENT;
7249 +       return err;
7250 +}
7251 +
7252 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
7253 +{
7254 +       return a->len == b->len
7255 +               && !memcmp(a->name, b->name, a->len);
7256 +}
7257 +
7258 +/*
7259 + * by the commit
7260 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
7261 + *                     taking d_lock
7262 + * the type of d_lockref.count became int, but the inlined function d_count()
7263 + * still returns unsigned int.
7264 + * I don't know why. Maybe it is for every d_count() users?
7265 + * Anyway au_dcount() lives on.
7266 + */
7267 +static inline int au_dcount(struct dentry *d)
7268 +{
7269 +       return (int)d_count(d);
7270 +}
7271 +
7272 +#endif /* __KERNEL__ */
7273 +#endif /* __AUFS_DCSUB_H__ */
7274 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
7275 --- /usr/share/empty/fs/aufs/debug.c    1970-01-01 01:00:00.000000000 +0100
7276 +++ linux/fs/aufs/debug.c       2022-11-05 23:02:18.962555950 +0100
7277 @@ -0,0 +1,444 @@
7278 +// SPDX-License-Identifier: GPL-2.0
7279 +/*
7280 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7281 + *
7282 + * This program is free software; you can redistribute it and/or modify
7283 + * it under the terms of the GNU General Public License as published by
7284 + * the Free Software Foundation; either version 2 of the License, or
7285 + * (at your option) any later version.
7286 + *
7287 + * This program is distributed in the hope that it will be useful,
7288 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7289 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7290 + * GNU General Public License for more details.
7291 + *
7292 + * You should have received a copy of the GNU General Public License
7293 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7294 + */
7295 +
7296 +/*
7297 + * debug print functions
7298 + */
7299 +
7300 +#include <linux/iversion.h>
7301 +#include "aufs.h"
7302 +
7303 +/* Returns 0, or -errno.  arg is in kp->arg. */
7304 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
7305 +{
7306 +       int err, n;
7307 +
7308 +       err = kstrtoint(val, 0, &n);
7309 +       if (!err) {
7310 +               if (n > 0)
7311 +                       au_debug_on();
7312 +               else
7313 +                       au_debug_off();
7314 +       }
7315 +       return err;
7316 +}
7317 +
7318 +/* Returns length written or -errno.  Buffer is 4k (ie. be short!) */
7319 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
7320 +{
7321 +       atomic_t *a;
7322 +
7323 +       a = kp->arg;
7324 +       return sprintf(buffer, "%d", atomic_read(a));
7325 +}
7326 +
7327 +static const struct kernel_param_ops param_ops_atomic_t = {
7328 +       .set = param_atomic_t_set,
7329 +       .get = param_atomic_t_get
7330 +       /* void (*free)(void *arg) */
7331 +};
7332 +
7333 +atomic_t aufs_debug = ATOMIC_INIT(0);
7334 +MODULE_PARM_DESC(debug, "debug print");
7335 +module_param_named(debug, aufs_debug, atomic_t, 0664);
7336 +
7337 +DEFINE_MUTEX(au_dbg_mtx);      /* just to serialize the dbg msgs */
7338 +char *au_plevel = KERN_DEBUG;
7339 +#define dpri(fmt, ...) do {                                    \
7340 +       if ((au_plevel                                          \
7341 +            && strcmp(au_plevel, KERN_DEBUG))                  \
7342 +           || au_debug_test())                                 \
7343 +               printk("%s" fmt, au_plevel, ##__VA_ARGS__);     \
7344 +} while (0)
7345 +
7346 +/* ---------------------------------------------------------------------- */
7347 +
7348 +void au_dpri_whlist(struct au_nhash *whlist)
7349 +{
7350 +       unsigned long ul, n;
7351 +       struct hlist_head *head;
7352 +       struct au_vdir_wh *pos;
7353 +
7354 +       n = whlist->nh_num;
7355 +       head = whlist->nh_head;
7356 +       for (ul = 0; ul < n; ul++) {
7357 +               hlist_for_each_entry(pos, head, wh_hash)
7358 +                       dpri("b%d, %.*s, %d\n",
7359 +                            pos->wh_bindex,
7360 +                            pos->wh_str.len, pos->wh_str.name,
7361 +                            pos->wh_str.len);
7362 +               head++;
7363 +       }
7364 +}
7365 +
7366 +void au_dpri_vdir(struct au_vdir *vdir)
7367 +{
7368 +       unsigned long ul;
7369 +       union au_vdir_deblk_p p;
7370 +       unsigned char *o;
7371 +
7372 +       if (!vdir || IS_ERR(vdir)) {
7373 +               dpri("err %ld\n", PTR_ERR(vdir));
7374 +               return;
7375 +       }
7376 +
7377 +       dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %llu\n",
7378 +            vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
7379 +            vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
7380 +       for (ul = 0; ul < vdir->vd_nblk; ul++) {
7381 +               p.deblk = vdir->vd_deblk[ul];
7382 +               o = p.deblk;
7383 +               dpri("[%lu]: %p\n", ul, o);
7384 +       }
7385 +}
7386 +
7387 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
7388 +                       struct dentry *wh)
7389 +{
7390 +       char *n = NULL;
7391 +       int l = 0;
7392 +
7393 +       if (!inode || IS_ERR(inode)) {
7394 +               dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
7395 +               return -1;
7396 +       }
7397 +
7398 +       /* the type of i_blocks depends upon CONFIG_LBDAF */
7399 +       BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
7400 +                    && sizeof(inode->i_blocks) != sizeof(u64));
7401 +       if (wh) {
7402 +               n = (void *)wh->d_name.name;
7403 +               l = wh->d_name.len;
7404 +       }
7405 +
7406 +       dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
7407 +            " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
7408 +            bindex, inode,
7409 +            inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
7410 +            atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
7411 +            i_size_read(inode), (unsigned long long)inode->i_blocks,
7412 +            hn, (long long)timespec64_to_ns(&inode->i_ctime) & 0x0ffff,
7413 +            inode->i_mapping ? inode->i_mapping->nrpages : 0,
7414 +            inode->i_state, inode->i_flags, inode_peek_iversion(inode),
7415 +            inode->i_generation,
7416 +            l ? ", wh " : "", l, n);
7417 +       return 0;
7418 +}
7419 +
7420 +void au_dpri_inode(struct inode *inode)
7421 +{
7422 +       struct au_iinfo *iinfo;
7423 +       struct au_hinode *hi;
7424 +       aufs_bindex_t bindex;
7425 +       int err, hn;
7426 +
7427 +       err = do_pri_inode(-1, inode, -1, NULL);
7428 +       if (err || !au_test_aufs(inode->i_sb) || au_is_bad_inode(inode))
7429 +               return;
7430 +
7431 +       iinfo = au_ii(inode);
7432 +       dpri("i-1: btop %d, bbot %d, gen %d\n",
7433 +            iinfo->ii_btop, iinfo->ii_bbot, au_iigen(inode, NULL));
7434 +       if (iinfo->ii_btop < 0)
7435 +               return;
7436 +       hn = 0;
7437 +       for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot; bindex++) {
7438 +               hi = au_hinode(iinfo, bindex);
7439 +               hn = !!au_hn(hi);
7440 +               do_pri_inode(bindex, hi->hi_inode, hn, hi->hi_whdentry);
7441 +       }
7442 +}
7443 +
7444 +void au_dpri_dalias(struct inode *inode)
7445 +{
7446 +       struct dentry *d;
7447 +
7448 +       spin_lock(&inode->i_lock);
7449 +       hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
7450 +               au_dpri_dentry(d);
7451 +       spin_unlock(&inode->i_lock);
7452 +}
7453 +
7454 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
7455 +{
7456 +       struct dentry *wh = NULL;
7457 +       int hn;
7458 +       struct inode *inode;
7459 +       struct au_iinfo *iinfo;
7460 +       struct au_hinode *hi;
7461 +
7462 +       if (!dentry || IS_ERR(dentry)) {
7463 +               dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
7464 +               return -1;
7465 +       }
7466 +       /* do not call dget_parent() here */
7467 +       /* note: access d_xxx without d_lock */
7468 +       dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
7469 +            bindex, dentry, dentry,
7470 +            dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
7471 +            au_dcount(dentry), dentry->d_flags,
7472 +            d_unhashed(dentry) ? "un" : "");
7473 +       hn = -1;
7474 +       inode = NULL;
7475 +       if (d_is_positive(dentry))
7476 +               inode = d_inode(dentry);
7477 +       if (inode
7478 +           && au_test_aufs(dentry->d_sb)
7479 +           && bindex >= 0
7480 +           && !au_is_bad_inode(inode)) {
7481 +               iinfo = au_ii(inode);
7482 +               hi = au_hinode(iinfo, bindex);
7483 +               hn = !!au_hn(hi);
7484 +               wh = hi->hi_whdentry;
7485 +       }
7486 +       do_pri_inode(bindex, inode, hn, wh);
7487 +       return 0;
7488 +}
7489 +
7490 +void au_dpri_dentry(struct dentry *dentry)
7491 +{
7492 +       struct au_dinfo *dinfo;
7493 +       aufs_bindex_t bindex;
7494 +       int err;
7495 +
7496 +       err = do_pri_dentry(-1, dentry);
7497 +       if (err || !au_test_aufs(dentry->d_sb))
7498 +               return;
7499 +
7500 +       dinfo = au_di(dentry);
7501 +       if (!dinfo)
7502 +               return;
7503 +       dpri("d-1: btop %d, bbot %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
7504 +            dinfo->di_btop, dinfo->di_bbot,
7505 +            dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
7506 +            dinfo->di_tmpfile);
7507 +       if (dinfo->di_btop < 0)
7508 +               return;
7509 +       for (bindex = dinfo->di_btop; bindex <= dinfo->di_bbot; bindex++)
7510 +               do_pri_dentry(bindex, au_hdentry(dinfo, bindex)->hd_dentry);
7511 +}
7512 +
7513 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7514 +{
7515 +       char a[32];
7516 +
7517 +       if (!file || IS_ERR(file)) {
7518 +               dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7519 +               return -1;
7520 +       }
7521 +       a[0] = 0;
7522 +       if (bindex < 0
7523 +           && !IS_ERR_OR_NULL(file->f_path.dentry)
7524 +           && au_test_aufs(file->f_path.dentry->d_sb)
7525 +           && au_fi(file))
7526 +               snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7527 +                        au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7528 +       dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7529 +            bindex, file->f_mode, file->f_flags, (long)file_count(file),
7530 +            file->f_version, file->f_pos, a);
7531 +       if (!IS_ERR_OR_NULL(file->f_path.dentry))
7532 +               do_pri_dentry(bindex, file->f_path.dentry);
7533 +       return 0;
7534 +}
7535 +
7536 +void au_dpri_file(struct file *file)
7537 +{
7538 +       struct au_finfo *finfo;
7539 +       struct au_fidir *fidir;
7540 +       struct au_hfile *hfile;
7541 +       aufs_bindex_t bindex;
7542 +       int err;
7543 +
7544 +       err = do_pri_file(-1, file);
7545 +       if (err
7546 +           || IS_ERR_OR_NULL(file->f_path.dentry)
7547 +           || !au_test_aufs(file->f_path.dentry->d_sb))
7548 +               return;
7549 +
7550 +       finfo = au_fi(file);
7551 +       if (!finfo)
7552 +               return;
7553 +       if (finfo->fi_btop < 0)
7554 +               return;
7555 +       fidir = finfo->fi_hdir;
7556 +       if (!fidir)
7557 +               do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7558 +       else
7559 +               for (bindex = finfo->fi_btop;
7560 +                    bindex >= 0 && bindex <= fidir->fd_bbot;
7561 +                    bindex++) {
7562 +                       hfile = fidir->fd_hfile + bindex;
7563 +                       do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7564 +               }
7565 +}
7566 +
7567 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7568 +{
7569 +       struct vfsmount *mnt;
7570 +       struct super_block *sb;
7571 +
7572 +       if (!br || IS_ERR(br))
7573 +               goto out;
7574 +       mnt = au_br_mnt(br);
7575 +       if (!mnt || IS_ERR(mnt))
7576 +               goto out;
7577 +       sb = mnt->mnt_sb;
7578 +       if (!sb || IS_ERR(sb))
7579 +               goto out;
7580 +
7581 +       dpri("s%d: {perm 0x%x, id %d, wbr %p}, "
7582 +            "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7583 +            "xino %d\n",
7584 +            bindex, br->br_perm, br->br_id, br->br_wbr,
7585 +            au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7586 +            sb->s_flags, sb->s_count,
7587 +            atomic_read(&sb->s_active),
7588 +            !!au_xino_file(br->br_xino, /*idx*/-1));
7589 +       return 0;
7590 +
7591 +out:
7592 +       dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7593 +       return -1;
7594 +}
7595 +
7596 +void au_dpri_sb(struct super_block *sb)
7597 +{
7598 +       struct au_sbinfo *sbinfo;
7599 +       aufs_bindex_t bindex;
7600 +       int err;
7601 +       /* to reduce stack size */
7602 +       struct {
7603 +               struct vfsmount mnt;
7604 +               struct au_branch fake;
7605 +       } *a;
7606 +
7607 +       /* this function can be called from magic sysrq */
7608 +       a = kzalloc(sizeof(*a), GFP_ATOMIC);
7609 +       if (unlikely(!a)) {
7610 +               dpri("no memory\n");
7611 +               return;
7612 +       }
7613 +
7614 +       a->mnt.mnt_sb = sb;
7615 +       a->fake.br_path.mnt = &a->mnt;
7616 +       err = do_pri_br(-1, &a->fake);
7617 +       au_kfree_rcu(a);
7618 +       dpri("dev 0x%x\n", sb->s_dev);
7619 +       if (err || !au_test_aufs(sb))
7620 +               return;
7621 +
7622 +       sbinfo = au_sbi(sb);
7623 +       if (!sbinfo)
7624 +               return;
7625 +       dpri("nw %d, gen %u, kobj %d\n",
7626 +            atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7627 +            kref_read(&sbinfo->si_kobj.kref));
7628 +       for (bindex = 0; bindex <= sbinfo->si_bbot; bindex++)
7629 +               do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7630 +}
7631 +
7632 +/* ---------------------------------------------------------------------- */
7633 +
7634 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7635 +{
7636 +       struct inode *h_inode, *inode = d_inode(dentry);
7637 +       struct dentry *h_dentry;
7638 +       aufs_bindex_t bindex, bbot, bi;
7639 +
7640 +       if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7641 +               return;
7642 +
7643 +       bbot = au_dbbot(dentry);
7644 +       bi = au_ibbot(inode);
7645 +       if (bi < bbot)
7646 +               bbot = bi;
7647 +       bindex = au_dbtop(dentry);
7648 +       bi = au_ibtop(inode);
7649 +       if (bi > bindex)
7650 +               bindex = bi;
7651 +
7652 +       for (; bindex <= bbot; bindex++) {
7653 +               h_dentry = au_h_dptr(dentry, bindex);
7654 +               if (!h_dentry)
7655 +                       continue;
7656 +               h_inode = au_h_iptr(inode, bindex);
7657 +               if (unlikely(h_inode != d_inode(h_dentry))) {
7658 +                       au_debug_on();
7659 +                       AuDbg("b%d, %s:%d\n", bindex, func, line);
7660 +                       AuDbgDentry(dentry);
7661 +                       AuDbgInode(inode);
7662 +                       au_debug_off();
7663 +                       if (au_test_fuse(h_inode->i_sb))
7664 +                               WARN_ON_ONCE(1);
7665 +                       else
7666 +                               BUG();
7667 +               }
7668 +       }
7669 +}
7670 +
7671 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7672 +{
7673 +       int err, i, j;
7674 +       struct au_dcsub_pages dpages;
7675 +       struct au_dpage *dpage;
7676 +       struct dentry **dentries;
7677 +
7678 +       err = au_dpages_init(&dpages, GFP_NOFS);
7679 +       AuDebugOn(err);
7680 +       err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7681 +       AuDebugOn(err);
7682 +       for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7683 +               dpage = dpages.dpages + i;
7684 +               dentries = dpage->dentries;
7685 +               for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7686 +                       AuDebugOn(au_digen_test(dentries[j], sigen));
7687 +       }
7688 +       au_dpages_free(&dpages);
7689 +}
7690 +
7691 +void au_dbg_verify_kthread(void)
7692 +{
7693 +       if (au_wkq_test()) {
7694 +               au_dbg_blocked();
7695 +               /*
7696 +                * It may be recursive, but udba=notify between two aufs mounts,
7697 +                * where a single ro branch is shared, is not a problem.
7698 +                */
7699 +               /* WARN_ON(1); */
7700 +       }
7701 +}
7702 +
7703 +/* ---------------------------------------------------------------------- */
7704 +
7705 +int __init au_debug_init(void)
7706 +{
7707 +       aufs_bindex_t bindex;
7708 +       struct au_vdir_destr destr;
7709 +
7710 +       bindex = -1;
7711 +       AuDebugOn(bindex >= 0);
7712 +
7713 +       destr.len = -1;
7714 +       AuDebugOn(destr.len < NAME_MAX);
7715 +
7716 +#ifdef CONFIG_4KSTACKS
7717 +       pr_warn("CONFIG_4KSTACKS is defined.\n");
7718 +#endif
7719 +
7720 +       return 0;
7721 +}
7722 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7723 --- /usr/share/empty/fs/aufs/debug.h    1970-01-01 01:00:00.000000000 +0100
7724 +++ linux/fs/aufs/debug.h       2022-11-05 23:02:18.962555950 +0100
7725 @@ -0,0 +1,226 @@
7726 +/* SPDX-License-Identifier: GPL-2.0 */
7727 +/*
7728 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7729 + *
7730 + * This program is free software; you can redistribute it and/or modify
7731 + * it under the terms of the GNU General Public License as published by
7732 + * the Free Software Foundation; either version 2 of the License, or
7733 + * (at your option) any later version.
7734 + *
7735 + * This program is distributed in the hope that it will be useful,
7736 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7737 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7738 + * GNU General Public License for more details.
7739 + *
7740 + * You should have received a copy of the GNU General Public License
7741 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7742 + */
7743 +
7744 +/*
7745 + * debug print functions
7746 + */
7747 +
7748 +#ifndef __AUFS_DEBUG_H__
7749 +#define __AUFS_DEBUG_H__
7750 +
7751 +#ifdef __KERNEL__
7752 +
7753 +#include <linux/atomic.h>
7754 +#include <linux/module.h>
7755 +#include <linux/kallsyms.h>
7756 +#include <linux/sysrq.h>
7757 +
7758 +#ifdef CONFIG_AUFS_DEBUG
7759 +#define AuDebugOn(a)           BUG_ON(a)
7760 +
7761 +/* module parameter */
7762 +extern atomic_t aufs_debug;
7763 +static inline void au_debug_on(void)
7764 +{
7765 +       atomic_inc(&aufs_debug);
7766 +}
7767 +static inline void au_debug_off(void)
7768 +{
7769 +       atomic_dec_if_positive(&aufs_debug);
7770 +}
7771 +
7772 +static inline int au_debug_test(void)
7773 +{
7774 +       return atomic_read(&aufs_debug) > 0;
7775 +}
7776 +#else
7777 +#define AuDebugOn(a)           do {} while (0)
7778 +AuStubVoid(au_debug_on, void)
7779 +AuStubVoid(au_debug_off, void)
7780 +AuStubInt0(au_debug_test, void)
7781 +#endif /* CONFIG_AUFS_DEBUG */
7782 +
7783 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7784 +
7785 +/* ---------------------------------------------------------------------- */
7786 +
7787 +/* debug print */
7788 +
7789 +#define AuDbg(fmt, ...) do { \
7790 +       if (au_debug_test()) \
7791 +               pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7792 +} while (0)
7793 +#define AuLabel(l)             AuDbg(#l "\n")
7794 +#define AuIOErr(fmt, ...)      pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7795 +#define AuWarn1(fmt, ...) do { \
7796 +       static unsigned char _c; \
7797 +       if (!_c++) \
7798 +               pr_warn(fmt, ##__VA_ARGS__); \
7799 +} while (0)
7800 +
7801 +#define AuErr1(fmt, ...) do { \
7802 +       static unsigned char _c; \
7803 +       if (!_c++) \
7804 +               pr_err(fmt, ##__VA_ARGS__); \
7805 +} while (0)
7806 +
7807 +#define AuIOErr1(fmt, ...) do { \
7808 +       static unsigned char _c; \
7809 +       if (!_c++) \
7810 +               AuIOErr(fmt, ##__VA_ARGS__); \
7811 +} while (0)
7812 +
7813 +#define AuUnsupportMsg "This operation is not supported." \
7814 +                       " Please report this application to aufs-users ML."
7815 +#define AuUnsupport(fmt, ...) do { \
7816 +       pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7817 +       dump_stack(); \
7818 +} while (0)
7819 +
7820 +#define AuTraceErr(e) do { \
7821 +       if (unlikely((e) < 0)) \
7822 +               AuDbg("err %d\n", (int)(e)); \
7823 +} while (0)
7824 +
7825 +#define AuTraceErrPtr(p) do { \
7826 +       if (IS_ERR(p)) \
7827 +               AuDbg("err %ld\n", PTR_ERR(p)); \
7828 +} while (0)
7829 +
7830 +/* dirty macros for debug print, use with "%.*s" and caution */
7831 +#define AuLNPair(qstr)         (qstr)->len, (qstr)->name
7832 +
7833 +/* ---------------------------------------------------------------------- */
7834 +
7835 +struct dentry;
7836 +#ifdef CONFIG_AUFS_DEBUG
7837 +extern struct mutex au_dbg_mtx;
7838 +extern char *au_plevel;
7839 +struct au_nhash;
7840 +void au_dpri_whlist(struct au_nhash *whlist);
7841 +struct au_vdir;
7842 +void au_dpri_vdir(struct au_vdir *vdir);
7843 +struct inode;
7844 +void au_dpri_inode(struct inode *inode);
7845 +void au_dpri_dalias(struct inode *inode);
7846 +void au_dpri_dentry(struct dentry *dentry);
7847 +struct file;
7848 +void au_dpri_file(struct file *filp);
7849 +struct super_block;
7850 +void au_dpri_sb(struct super_block *sb);
7851 +
7852 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7853 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7854 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7855 +void au_dbg_verify_kthread(void);
7856 +
7857 +int __init au_debug_init(void);
7858 +
7859 +#define AuDbgWhlist(w) do { \
7860 +       mutex_lock(&au_dbg_mtx); \
7861 +       AuDbg(#w "\n"); \
7862 +       au_dpri_whlist(w); \
7863 +       mutex_unlock(&au_dbg_mtx); \
7864 +} while (0)
7865 +
7866 +#define AuDbgVdir(v) do { \
7867 +       mutex_lock(&au_dbg_mtx); \
7868 +       AuDbg(#v "\n"); \
7869 +       au_dpri_vdir(v); \
7870 +       mutex_unlock(&au_dbg_mtx); \
7871 +} while (0)
7872 +
7873 +#define AuDbgInode(i) do { \
7874 +       mutex_lock(&au_dbg_mtx); \
7875 +       AuDbg(#i "\n"); \
7876 +       au_dpri_inode(i); \
7877 +       mutex_unlock(&au_dbg_mtx); \
7878 +} while (0)
7879 +
7880 +#define AuDbgDAlias(i) do { \
7881 +       mutex_lock(&au_dbg_mtx); \
7882 +       AuDbg(#i "\n"); \
7883 +       au_dpri_dalias(i); \
7884 +       mutex_unlock(&au_dbg_mtx); \
7885 +} while (0)
7886 +
7887 +#define AuDbgDentry(d) do { \
7888 +       mutex_lock(&au_dbg_mtx); \
7889 +       AuDbg(#d "\n"); \
7890 +       au_dpri_dentry(d); \
7891 +       mutex_unlock(&au_dbg_mtx); \
7892 +} while (0)
7893 +
7894 +#define AuDbgFile(f) do { \
7895 +       mutex_lock(&au_dbg_mtx); \
7896 +       AuDbg(#f "\n"); \
7897 +       au_dpri_file(f); \
7898 +       mutex_unlock(&au_dbg_mtx); \
7899 +} while (0)
7900 +
7901 +#define AuDbgSb(sb) do { \
7902 +       mutex_lock(&au_dbg_mtx); \
7903 +       AuDbg(#sb "\n"); \
7904 +       au_dpri_sb(sb); \
7905 +       mutex_unlock(&au_dbg_mtx); \
7906 +} while (0)
7907 +
7908 +#define AuDbgSym(addr) do {                            \
7909 +       char sym[KSYM_SYMBOL_LEN];                      \
7910 +       sprint_symbol(sym, (unsigned long)addr);        \
7911 +       AuDbg("%s\n", sym);                             \
7912 +} while (0)
7913 +#else
7914 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7915 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7916 +AuStubVoid(au_dbg_verify_kthread, void)
7917 +AuStubInt0(__init au_debug_init, void)
7918 +
7919 +#define AuDbgWhlist(w)         do {} while (0)
7920 +#define AuDbgVdir(v)           do {} while (0)
7921 +#define AuDbgInode(i)          do {} while (0)
7922 +#define AuDbgDAlias(i)         do {} while (0)
7923 +#define AuDbgDentry(d)         do {} while (0)
7924 +#define AuDbgFile(f)           do {} while (0)
7925 +#define AuDbgSb(sb)            do {} while (0)
7926 +#define AuDbgSym(addr)         do {} while (0)
7927 +#endif /* CONFIG_AUFS_DEBUG */
7928 +
7929 +/* ---------------------------------------------------------------------- */
7930 +
7931 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7932 +int __init au_sysrq_init(void);
7933 +void au_sysrq_fin(void);
7934 +
7935 +#ifdef CONFIG_HW_CONSOLE
7936 +#define au_dbg_blocked() do { \
7937 +       WARN_ON(1); \
7938 +       handle_sysrq('w'); \
7939 +} while (0)
7940 +#else
7941 +AuStubVoid(au_dbg_blocked, void)
7942 +#endif
7943 +
7944 +#else
7945 +AuStubInt0(__init au_sysrq_init, void)
7946 +AuStubVoid(au_sysrq_fin, void)
7947 +AuStubVoid(au_dbg_blocked, void)
7948 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7949 +
7950 +#endif /* __KERNEL__ */
7951 +#endif /* __AUFS_DEBUG_H__ */
7952 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7953 --- /usr/share/empty/fs/aufs/dentry.c   1970-01-01 01:00:00.000000000 +0100
7954 +++ linux/fs/aufs/dentry.c      2022-11-05 23:02:18.962555950 +0100
7955 @@ -0,0 +1,1168 @@
7956 +// SPDX-License-Identifier: GPL-2.0
7957 +/*
7958 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7959 + *
7960 + * This program is free software; you can redistribute it and/or modify
7961 + * it under the terms of the GNU General Public License as published by
7962 + * the Free Software Foundation; either version 2 of the License, or
7963 + * (at your option) any later version.
7964 + *
7965 + * This program is distributed in the hope that it will be useful,
7966 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7967 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
7968 + * GNU General Public License for more details.
7969 + *
7970 + * You should have received a copy of the GNU General Public License
7971 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
7972 + */
7973 +
7974 +/*
7975 + * lookup and dentry operations
7976 + */
7977 +
7978 +#include <linux/iversion.h>
7979 +#include "aufs.h"
7980 +
7981 +/*
7982 + * returns positive/negative dentry, NULL or an error.
7983 + * NULL means whiteout-ed or not-found.
7984 + */
7985 +static struct dentry*
7986 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7987 +            aufs_bindex_t bindex, struct au_do_lookup_args *args)
7988 +{
7989 +       struct dentry *h_dentry;
7990 +       struct inode *h_inode;
7991 +       struct au_branch *br;
7992 +       struct user_namespace *h_userns;
7993 +       struct path h_path;
7994 +       int wh_found, opq;
7995 +       unsigned char wh_able;
7996 +       const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
7997 +       const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
7998 +                                                         IGNORE_PERM);
7999 +
8000 +       wh_found = 0;
8001 +       br = au_sbr(dentry->d_sb, bindex);
8002 +       h_path.dentry = h_parent;
8003 +       h_path.mnt = au_br_mnt(br);
8004 +       h_userns = au_br_userns(br);
8005 +       wh_able = !!au_br_whable(br->br_perm);
8006 +       if (wh_able)
8007 +               wh_found = au_wh_test(h_userns, &h_path, &args->whname,
8008 +                                     ignore_perm);
8009 +       h_dentry = ERR_PTR(wh_found);
8010 +       if (!wh_found)
8011 +               goto real_lookup;
8012 +       if (unlikely(wh_found < 0))
8013 +               goto out;
8014 +
8015 +       /* We found a whiteout */
8016 +       /* au_set_dbbot(dentry, bindex); */
8017 +       au_set_dbwh(dentry, bindex);
8018 +       if (!allow_neg)
8019 +               return NULL; /* success */
8020 +
8021 +real_lookup:
8022 +       if (!ignore_perm)
8023 +               h_dentry = vfsub_lkup_one(args->name, &h_path);
8024 +       else
8025 +               h_dentry = au_sio_lkup_one(h_userns, args->name, &h_path);
8026 +       if (IS_ERR(h_dentry)) {
8027 +               if (PTR_ERR(h_dentry) == -ENAMETOOLONG
8028 +                   && !allow_neg)
8029 +                       h_dentry = NULL;
8030 +               goto out;
8031 +       }
8032 +
8033 +       h_inode = d_inode(h_dentry);
8034 +       if (d_is_negative(h_dentry)) {
8035 +               if (!allow_neg)
8036 +                       goto out_neg;
8037 +       } else if (wh_found
8038 +                  || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
8039 +               goto out_neg;
8040 +       else if (au_ftest_lkup(args->flags, DIRREN)
8041 +                /* && h_inode */
8042 +                && !au_dr_lkup_h_ino(args, bindex, h_inode->i_ino)) {
8043 +               AuDbg("b%d %pd ignored hi%llu\n", bindex, h_dentry,
8044 +                     (unsigned long long)h_inode->i_ino);
8045 +               goto out_neg;
8046 +       }
8047 +
8048 +       if (au_dbbot(dentry) <= bindex)
8049 +               au_set_dbbot(dentry, bindex);
8050 +       if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
8051 +               au_set_dbtop(dentry, bindex);
8052 +       au_set_h_dptr(dentry, bindex, h_dentry);
8053 +
8054 +       if (!d_is_dir(h_dentry)
8055 +           || !wh_able
8056 +           || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
8057 +               goto out; /* success */
8058 +
8059 +       h_path.dentry = h_dentry;
8060 +       inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
8061 +       opq = au_diropq_test(h_userns, &h_path);
8062 +       inode_unlock_shared(h_inode);
8063 +       if (opq > 0)
8064 +               au_set_dbdiropq(dentry, bindex);
8065 +       else if (unlikely(opq < 0)) {
8066 +               au_set_h_dptr(dentry, bindex, NULL);
8067 +               h_dentry = ERR_PTR(opq);
8068 +       }
8069 +       goto out;
8070 +
8071 +out_neg:
8072 +       dput(h_dentry);
8073 +       h_dentry = NULL;
8074 +out:
8075 +       return h_dentry;
8076 +}
8077 +
8078 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
8079 +{
8080 +       if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
8081 +                    && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
8082 +               return -EPERM;
8083 +       return 0;
8084 +}
8085 +
8086 +/*
8087 + * returns the number of lower positive dentries,
8088 + * otherwise an error.
8089 + * can be called at unlinking with @type is zero.
8090 + */
8091 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
8092 +                  unsigned int flags)
8093 +{
8094 +       int npositive, err;
8095 +       aufs_bindex_t bindex, btail, bdiropq;
8096 +       unsigned char isdir, dirperm1, dirren;
8097 +       struct au_do_lookup_args args = {
8098 +               .flags          = flags,
8099 +               .name           = &dentry->d_name
8100 +       };
8101 +       struct dentry *parent;
8102 +       struct super_block *sb;
8103 +
8104 +       sb = dentry->d_sb;
8105 +       err = au_test_shwh(sb, args.name);
8106 +       if (unlikely(err))
8107 +               goto out;
8108 +
8109 +       err = au_wh_name_alloc(&args.whname, args.name);
8110 +       if (unlikely(err))
8111 +               goto out;
8112 +
8113 +       isdir = !!d_is_dir(dentry);
8114 +       dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
8115 +       dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
8116 +       if (dirren)
8117 +               au_fset_lkup(args.flags, DIRREN);
8118 +
8119 +       npositive = 0;
8120 +       parent = dget_parent(dentry);
8121 +       btail = au_dbtaildir(parent);
8122 +       for (bindex = btop; bindex <= btail; bindex++) {
8123 +               struct dentry *h_parent, *h_dentry;
8124 +               struct inode *h_inode, *h_dir;
8125 +               struct au_branch *br;
8126 +
8127 +               h_dentry = au_h_dptr(dentry, bindex);
8128 +               if (h_dentry) {
8129 +                       if (d_is_positive(h_dentry))
8130 +                               npositive++;
8131 +                       break;
8132 +               }
8133 +               h_parent = au_h_dptr(parent, bindex);
8134 +               if (!h_parent || !d_is_dir(h_parent))
8135 +                       continue;
8136 +
8137 +               if (dirren) {
8138 +                       /* if the inum matches, then use the prepared name */
8139 +                       err = au_dr_lkup_name(&args, bindex);
8140 +                       if (unlikely(err))
8141 +                               goto out_parent;
8142 +               }
8143 +
8144 +               h_dir = d_inode(h_parent);
8145 +               inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
8146 +               h_dentry = au_do_lookup(h_parent, dentry, bindex, &args);
8147 +               inode_unlock_shared(h_dir);
8148 +               err = PTR_ERR(h_dentry);
8149 +               if (IS_ERR(h_dentry))
8150 +                       goto out_parent;
8151 +               if (h_dentry)
8152 +                       au_fclr_lkup(args.flags, ALLOW_NEG);
8153 +               if (dirperm1)
8154 +                       au_fset_lkup(args.flags, IGNORE_PERM);
8155 +
8156 +               if (au_dbwh(dentry) == bindex)
8157 +                       break;
8158 +               if (!h_dentry)
8159 +                       continue;
8160 +               if (d_is_negative(h_dentry))
8161 +                       continue;
8162 +               h_inode = d_inode(h_dentry);
8163 +               npositive++;
8164 +               if (!args.type)
8165 +                       args.type = h_inode->i_mode & S_IFMT;
8166 +               if (args.type != S_IFDIR)
8167 +                       break;
8168 +               else if (isdir) {
8169 +                       /* the type of lower may be different */
8170 +                       bdiropq = au_dbdiropq(dentry);
8171 +                       if (bdiropq >= 0 && bdiropq <= bindex)
8172 +                               break;
8173 +               }
8174 +               br = au_sbr(sb, bindex);
8175 +               if (dirren
8176 +                   && au_dr_hino_test_add(&br->br_dirren, h_inode->i_ino,
8177 +                                          /*add_ent*/NULL)) {
8178 +                       /* prepare next name to lookup */
8179 +                       err = au_dr_lkup(&args, dentry, bindex);
8180 +                       if (unlikely(err))
8181 +                               goto out_parent;
8182 +               }
8183 +       }
8184 +
8185 +       if (npositive) {
8186 +               AuLabel(positive);
8187 +               au_update_dbtop(dentry);
8188 +       }
8189 +       err = npositive;
8190 +       if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
8191 +                    && au_dbtop(dentry) < 0)) {
8192 +               err = -EIO;
8193 +               AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
8194 +                       dentry, err);
8195 +       }
8196 +
8197 +out_parent:
8198 +       dput(parent);
8199 +       au_kfree_try_rcu(args.whname.name);
8200 +       if (dirren)
8201 +               au_dr_lkup_fin(&args);
8202 +out:
8203 +       return err;
8204 +}
8205 +
8206 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
8207 +                              struct path *ppath)
8208 +{
8209 +       struct dentry *dentry;
8210 +       int wkq_err;
8211 +
8212 +       if (!au_test_h_perm_sio(userns, d_inode(ppath->dentry), MAY_EXEC))
8213 +               dentry = vfsub_lkup_one(name, ppath);
8214 +       else {
8215 +               struct vfsub_lkup_one_args args = {
8216 +                       .errp   = &dentry,
8217 +                       .name   = name,
8218 +                       .ppath  = ppath
8219 +               };
8220 +
8221 +               wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
8222 +               if (unlikely(wkq_err))
8223 +                       dentry = ERR_PTR(wkq_err);
8224 +       }
8225 +
8226 +       return dentry;
8227 +}
8228 +
8229 +/*
8230 + * lookup @dentry on @bindex which should be negative.
8231 + */
8232 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
8233 +{
8234 +       int err;
8235 +       struct dentry *parent, *h_dentry;
8236 +       struct au_branch *br;
8237 +       struct user_namespace *h_userns;
8238 +       struct path h_ppath;
8239 +
8240 +       parent = dget_parent(dentry);
8241 +       br = au_sbr(dentry->d_sb, bindex);
8242 +       h_ppath.dentry = au_h_dptr(parent, bindex);
8243 +       h_ppath.mnt = au_br_mnt(br);
8244 +       h_userns = au_br_userns(br);
8245 +       if (wh)
8246 +               h_dentry = au_whtmp_lkup(h_ppath.dentry, br, &dentry->d_name);
8247 +       else
8248 +               h_dentry = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
8249 +       err = PTR_ERR(h_dentry);
8250 +       if (IS_ERR(h_dentry))
8251 +               goto out;
8252 +       if (unlikely(d_is_positive(h_dentry))) {
8253 +               err = -EIO;
8254 +               AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
8255 +               dput(h_dentry);
8256 +               goto out;
8257 +       }
8258 +
8259 +       err = 0;
8260 +       if (bindex < au_dbtop(dentry))
8261 +               au_set_dbtop(dentry, bindex);
8262 +       if (au_dbbot(dentry) < bindex)
8263 +               au_set_dbbot(dentry, bindex);
8264 +       au_set_h_dptr(dentry, bindex, h_dentry);
8265 +
8266 +out:
8267 +       dput(parent);
8268 +       return err;
8269 +}
8270 +
8271 +/* ---------------------------------------------------------------------- */
8272 +
8273 +/* subset of struct inode */
8274 +struct au_iattr {
8275 +       unsigned long           i_ino;
8276 +       /* unsigned int         i_nlink; */
8277 +       kuid_t                  i_uid;
8278 +       kgid_t                  i_gid;
8279 +       u64                     i_version;
8280 +/*
8281 +       loff_t                  i_size;
8282 +       blkcnt_t                i_blocks;
8283 +*/
8284 +       umode_t                 i_mode;
8285 +};
8286 +
8287 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
8288 +{
8289 +       ia->i_ino = h_inode->i_ino;
8290 +       /* ia->i_nlink = h_inode->i_nlink; */
8291 +       ia->i_uid = h_inode->i_uid;
8292 +       ia->i_gid = h_inode->i_gid;
8293 +       ia->i_version = inode_query_iversion(h_inode);
8294 +/*
8295 +       ia->i_size = h_inode->i_size;
8296 +       ia->i_blocks = h_inode->i_blocks;
8297 +*/
8298 +       ia->i_mode = (h_inode->i_mode & S_IFMT);
8299 +}
8300 +
8301 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
8302 +{
8303 +       return ia->i_ino != h_inode->i_ino
8304 +               /* || ia->i_nlink != h_inode->i_nlink */
8305 +               || !uid_eq(ia->i_uid, h_inode->i_uid)
8306 +               || !gid_eq(ia->i_gid, h_inode->i_gid)
8307 +               || !inode_eq_iversion(h_inode, ia->i_version)
8308 +/*
8309 +               || ia->i_size != h_inode->i_size
8310 +               || ia->i_blocks != h_inode->i_blocks
8311 +*/
8312 +               || ia->i_mode != (h_inode->i_mode & S_IFMT);
8313 +}
8314 +
8315 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
8316 +                             struct au_branch *br)
8317 +{
8318 +       int err;
8319 +       struct au_iattr ia;
8320 +       struct inode *h_inode;
8321 +       struct dentry *h_d;
8322 +       struct super_block *h_sb;
8323 +       struct path h_ppath;
8324 +
8325 +       err = 0;
8326 +       memset(&ia, -1, sizeof(ia));
8327 +       h_sb = h_dentry->d_sb;
8328 +       h_inode = NULL;
8329 +       if (d_is_positive(h_dentry)) {
8330 +               h_inode = d_inode(h_dentry);
8331 +               au_iattr_save(&ia, h_inode);
8332 +       } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
8333 +               /* nfs d_revalidate may return 0 for negative dentry */
8334 +               /* fuse d_revalidate always return 0 for negative dentry */
8335 +               goto out;
8336 +
8337 +       /* main purpose is namei.c:cached_lookup() and d_revalidate */
8338 +       h_ppath.dentry = h_parent;
8339 +       h_ppath.mnt = au_br_mnt(br);
8340 +       h_d = vfsub_lkup_one(&h_dentry->d_name, &h_ppath);
8341 +       err = PTR_ERR(h_d);
8342 +       if (IS_ERR(h_d))
8343 +               goto out;
8344 +
8345 +       err = 0;
8346 +       if (unlikely(h_d != h_dentry
8347 +                    || d_inode(h_d) != h_inode
8348 +                    || (h_inode && au_iattr_test(&ia, h_inode))))
8349 +               err = au_busy_or_stale();
8350 +       dput(h_d);
8351 +
8352 +out:
8353 +       AuTraceErr(err);
8354 +       return err;
8355 +}
8356 +
8357 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8358 +               struct dentry *h_parent, struct au_branch *br)
8359 +{
8360 +       int err;
8361 +
8362 +       err = 0;
8363 +       if (udba == AuOpt_UDBA_REVAL
8364 +           && !au_test_fs_remote(h_dentry->d_sb)) {
8365 +               IMustLock(h_dir);
8366 +               err = (d_inode(h_dentry->d_parent) != h_dir);
8367 +       } else if (udba != AuOpt_UDBA_NONE)
8368 +               err = au_h_verify_dentry(h_dentry, h_parent, br);
8369 +
8370 +       return err;
8371 +}
8372 +
8373 +/* ---------------------------------------------------------------------- */
8374 +
8375 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
8376 +{
8377 +       int err;
8378 +       aufs_bindex_t new_bindex, bindex, bbot, bwh, bdiropq;
8379 +       struct au_hdentry tmp, *p, *q;
8380 +       struct au_dinfo *dinfo;
8381 +       struct super_block *sb;
8382 +
8383 +       DiMustWriteLock(dentry);
8384 +
8385 +       sb = dentry->d_sb;
8386 +       dinfo = au_di(dentry);
8387 +       bbot = dinfo->di_bbot;
8388 +       bwh = dinfo->di_bwh;
8389 +       bdiropq = dinfo->di_bdiropq;
8390 +       bindex = dinfo->di_btop;
8391 +       p = au_hdentry(dinfo, bindex);
8392 +       for (; bindex <= bbot; bindex++, p++) {
8393 +               if (!p->hd_dentry)
8394 +                       continue;
8395 +
8396 +               new_bindex = au_br_index(sb, p->hd_id);
8397 +               if (new_bindex == bindex)
8398 +                       continue;
8399 +
8400 +               if (dinfo->di_bwh == bindex)
8401 +                       bwh = new_bindex;
8402 +               if (dinfo->di_bdiropq == bindex)
8403 +                       bdiropq = new_bindex;
8404 +               if (new_bindex < 0) {
8405 +                       au_hdput(p);
8406 +                       p->hd_dentry = NULL;
8407 +                       continue;
8408 +               }
8409 +
8410 +               /* swap two lower dentries, and loop again */
8411 +               q = au_hdentry(dinfo, new_bindex);
8412 +               tmp = *q;
8413 +               *q = *p;
8414 +               *p = tmp;
8415 +               if (tmp.hd_dentry) {
8416 +                       bindex--;
8417 +                       p--;
8418 +               }
8419 +       }
8420 +
8421 +       dinfo->di_bwh = -1;
8422 +       if (bwh >= 0 && bwh <= au_sbbot(sb) && au_sbr_whable(sb, bwh))
8423 +               dinfo->di_bwh = bwh;
8424 +
8425 +       dinfo->di_bdiropq = -1;
8426 +       if (bdiropq >= 0
8427 +           && bdiropq <= au_sbbot(sb)
8428 +           && au_sbr_whable(sb, bdiropq))
8429 +               dinfo->di_bdiropq = bdiropq;
8430 +
8431 +       err = -EIO;
8432 +       dinfo->di_btop = -1;
8433 +       dinfo->di_bbot = -1;
8434 +       bbot = au_dbbot(parent);
8435 +       bindex = 0;
8436 +       p = au_hdentry(dinfo, bindex);
8437 +       for (; bindex <= bbot; bindex++, p++)
8438 +               if (p->hd_dentry) {
8439 +                       dinfo->di_btop = bindex;
8440 +                       break;
8441 +               }
8442 +
8443 +       if (dinfo->di_btop >= 0) {
8444 +               bindex = bbot;
8445 +               p = au_hdentry(dinfo, bindex);
8446 +               for (; bindex >= 0; bindex--, p--)
8447 +                       if (p->hd_dentry) {
8448 +                               dinfo->di_bbot = bindex;
8449 +                               err = 0;
8450 +                               break;
8451 +                       }
8452 +       }
8453 +
8454 +       return err;
8455 +}
8456 +
8457 +static void au_do_hide(struct dentry *dentry)
8458 +{
8459 +       struct inode *inode;
8460 +
8461 +       if (d_really_is_positive(dentry)) {
8462 +               inode = d_inode(dentry);
8463 +               if (!d_is_dir(dentry)) {
8464 +                       if (inode->i_nlink && !d_unhashed(dentry))
8465 +                               drop_nlink(inode);
8466 +               } else {
8467 +                       clear_nlink(inode);
8468 +                       /* stop next lookup */
8469 +                       inode->i_flags |= S_DEAD;
8470 +               }
8471 +               smp_mb(); /* necessary? */
8472 +       }
8473 +       d_drop(dentry);
8474 +}
8475 +
8476 +static int au_hide_children(struct dentry *parent)
8477 +{
8478 +       int err, i, j, ndentry;
8479 +       struct au_dcsub_pages dpages;
8480 +       struct au_dpage *dpage;
8481 +       struct dentry *dentry;
8482 +
8483 +       err = au_dpages_init(&dpages, GFP_NOFS);
8484 +       if (unlikely(err))
8485 +               goto out;
8486 +       err = au_dcsub_pages(&dpages, parent, NULL, NULL);
8487 +       if (unlikely(err))
8488 +               goto out_dpages;
8489 +
8490 +       /* in reverse order */
8491 +       for (i = dpages.ndpage - 1; i >= 0; i--) {
8492 +               dpage = dpages.dpages + i;
8493 +               ndentry = dpage->ndentry;
8494 +               for (j = ndentry - 1; j >= 0; j--) {
8495 +                       dentry = dpage->dentries[j];
8496 +                       if (dentry != parent)
8497 +                               au_do_hide(dentry);
8498 +               }
8499 +       }
8500 +
8501 +out_dpages:
8502 +       au_dpages_free(&dpages);
8503 +out:
8504 +       return err;
8505 +}
8506 +
8507 +static void au_hide(struct dentry *dentry)
8508 +{
8509 +       int err;
8510 +
8511 +       AuDbgDentry(dentry);
8512 +       if (d_is_dir(dentry)) {
8513 +               /* shrink_dcache_parent(dentry); */
8514 +               err = au_hide_children(dentry);
8515 +               if (unlikely(err))
8516 +                       AuIOErr("%pd, failed hiding children, ignored %d\n",
8517 +                               dentry, err);
8518 +       }
8519 +       au_do_hide(dentry);
8520 +}
8521 +
8522 +/*
8523 + * By adding a dirty branch, a cached dentry may be affected in various ways.
8524 + *
8525 + * a dirty branch is added
8526 + * - on the top of layers
8527 + * - in the middle of layers
8528 + * - to the bottom of layers
8529 + *
8530 + * on the added branch there exists
8531 + * - a whiteout
8532 + * - a diropq
8533 + * - a same named entry
8534 + *   + exist
8535 + *     * negative --> positive
8536 + *     * positive --> positive
8537 + *      - type is unchanged
8538 + *      - type is changed
8539 + *   + doesn't exist
8540 + *     * negative --> negative
8541 + *     * positive --> negative (rejected by au_br_del() for non-dir case)
8542 + * - none
8543 + */
8544 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8545 +                              struct au_dinfo *tmp)
8546 +{
8547 +       int err;
8548 +       aufs_bindex_t bindex, bbot;
8549 +       struct {
8550 +               struct dentry *dentry;
8551 +               struct inode *inode;
8552 +               mode_t mode;
8553 +       } orig_h, tmp_h = {
8554 +               .dentry = NULL
8555 +       };
8556 +       struct au_hdentry *hd;
8557 +       struct inode *inode, *h_inode;
8558 +       struct dentry *h_dentry;
8559 +
8560 +       err = 0;
8561 +       AuDebugOn(dinfo->di_btop < 0);
8562 +       orig_h.mode = 0;
8563 +       orig_h.dentry = au_hdentry(dinfo, dinfo->di_btop)->hd_dentry;
8564 +       orig_h.inode = NULL;
8565 +       if (d_is_positive(orig_h.dentry)) {
8566 +               orig_h.inode = d_inode(orig_h.dentry);
8567 +               orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8568 +       }
8569 +       if (tmp->di_btop >= 0) {
8570 +               tmp_h.dentry = au_hdentry(tmp, tmp->di_btop)->hd_dentry;
8571 +               if (d_is_positive(tmp_h.dentry)) {
8572 +                       tmp_h.inode = d_inode(tmp_h.dentry);
8573 +                       tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8574 +               }
8575 +       }
8576 +
8577 +       inode = NULL;
8578 +       if (d_really_is_positive(dentry))
8579 +               inode = d_inode(dentry);
8580 +       if (!orig_h.inode) {
8581 +               AuDbg("negative originally\n");
8582 +               if (inode) {
8583 +                       au_hide(dentry);
8584 +                       goto out;
8585 +               }
8586 +               AuDebugOn(inode);
8587 +               AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8588 +               AuDebugOn(dinfo->di_bdiropq != -1);
8589 +
8590 +               if (!tmp_h.inode) {
8591 +                       AuDbg("negative --> negative\n");
8592 +                       /* should have only one negative lower */
8593 +                       if (tmp->di_btop >= 0
8594 +                           && tmp->di_btop < dinfo->di_btop) {
8595 +                               AuDebugOn(tmp->di_btop != tmp->di_bbot);
8596 +                               AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8597 +                               au_set_h_dptr(dentry, dinfo->di_btop, NULL);
8598 +                               au_di_cp(dinfo, tmp);
8599 +                               hd = au_hdentry(tmp, tmp->di_btop);
8600 +                               au_set_h_dptr(dentry, tmp->di_btop,
8601 +                                             dget(hd->hd_dentry));
8602 +                       }
8603 +                       au_dbg_verify_dinode(dentry);
8604 +               } else {
8605 +                       AuDbg("negative --> positive\n");
8606 +                       /*
8607 +                        * similar to the behaviour of creating with bypassing
8608 +                        * aufs.
8609 +                        * unhash it in order to force an error in the
8610 +                        * succeeding create operation.
8611 +                        * we should not set S_DEAD here.
8612 +                        */
8613 +                       d_drop(dentry);
8614 +                       /* au_di_swap(tmp, dinfo); */
8615 +                       au_dbg_verify_dinode(dentry);
8616 +               }
8617 +       } else {
8618 +               AuDbg("positive originally\n");
8619 +               /* inode may be NULL */
8620 +               AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8621 +               if (!tmp_h.inode) {
8622 +                       AuDbg("positive --> negative\n");
8623 +                       /* or bypassing aufs */
8624 +                       au_hide(dentry);
8625 +                       if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_btop)
8626 +                               dinfo->di_bwh = tmp->di_bwh;
8627 +                       if (inode)
8628 +                               err = au_refresh_hinode_self(inode);
8629 +                       au_dbg_verify_dinode(dentry);
8630 +               } else if (orig_h.mode == tmp_h.mode) {
8631 +                       AuDbg("positive --> positive, same type\n");
8632 +                       if (!S_ISDIR(orig_h.mode)
8633 +                           && dinfo->di_btop > tmp->di_btop) {
8634 +                               /*
8635 +                                * similar to the behaviour of removing and
8636 +                                * creating.
8637 +                                */
8638 +                               au_hide(dentry);
8639 +                               if (inode)
8640 +                                       err = au_refresh_hinode_self(inode);
8641 +                               au_dbg_verify_dinode(dentry);
8642 +                       } else {
8643 +                               /* fill empty slots */
8644 +                               if (dinfo->di_btop > tmp->di_btop)
8645 +                                       dinfo->di_btop = tmp->di_btop;
8646 +                               if (dinfo->di_bbot < tmp->di_bbot)
8647 +                                       dinfo->di_bbot = tmp->di_bbot;
8648 +                               dinfo->di_bwh = tmp->di_bwh;
8649 +                               dinfo->di_bdiropq = tmp->di_bdiropq;
8650 +                               bbot = dinfo->di_bbot;
8651 +                               bindex = tmp->di_btop;
8652 +                               hd = au_hdentry(tmp, bindex);
8653 +                               for (; bindex <= bbot; bindex++, hd++) {
8654 +                                       if (au_h_dptr(dentry, bindex))
8655 +                                               continue;
8656 +                                       h_dentry = hd->hd_dentry;
8657 +                                       if (!h_dentry)
8658 +                                               continue;
8659 +                                       AuDebugOn(d_is_negative(h_dentry));
8660 +                                       h_inode = d_inode(h_dentry);
8661 +                                       AuDebugOn(orig_h.mode
8662 +                                                 != (h_inode->i_mode
8663 +                                                     & S_IFMT));
8664 +                                       au_set_h_dptr(dentry, bindex,
8665 +                                                     dget(h_dentry));
8666 +                               }
8667 +                               if (inode)
8668 +                                       err = au_refresh_hinode(inode, dentry);
8669 +                               au_dbg_verify_dinode(dentry);
8670 +                       }
8671 +               } else {
8672 +                       AuDbg("positive --> positive, different type\n");
8673 +                       /* similar to the behaviour of removing and creating */
8674 +                       au_hide(dentry);
8675 +                       if (inode)
8676 +                               err = au_refresh_hinode_self(inode);
8677 +                       au_dbg_verify_dinode(dentry);
8678 +               }
8679 +       }
8680 +
8681 +out:
8682 +       return err;
8683 +}
8684 +
8685 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8686 +{
8687 +       const struct dentry_operations *dop
8688 +               = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8689 +       static const unsigned int mask
8690 +               = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8691 +
8692 +       BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8693 +
8694 +       if (dentry->d_op == dop)
8695 +               return;
8696 +
8697 +       AuDbg("%pd\n", dentry);
8698 +       spin_lock(&dentry->d_lock);
8699 +       if (dop == &aufs_dop)
8700 +               dentry->d_flags |= mask;
8701 +       else
8702 +               dentry->d_flags &= ~mask;
8703 +       dentry->d_op = dop;
8704 +       spin_unlock(&dentry->d_lock);
8705 +}
8706 +
8707 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8708 +{
8709 +       int err, ebrange, nbr;
8710 +       unsigned int sigen;
8711 +       struct au_dinfo *dinfo, *tmp;
8712 +       struct super_block *sb;
8713 +       struct inode *inode;
8714 +
8715 +       DiMustWriteLock(dentry);
8716 +       AuDebugOn(IS_ROOT(dentry));
8717 +       AuDebugOn(d_really_is_negative(parent));
8718 +
8719 +       sb = dentry->d_sb;
8720 +       sigen = au_sigen(sb);
8721 +       err = au_digen_test(parent, sigen);
8722 +       if (unlikely(err))
8723 +               goto out;
8724 +
8725 +       nbr = au_sbbot(sb) + 1;
8726 +       dinfo = au_di(dentry);
8727 +       err = au_di_realloc(dinfo, nbr, /*may_shrink*/0);
8728 +       if (unlikely(err))
8729 +               goto out;
8730 +       ebrange = au_dbrange_test(dentry);
8731 +       if (!ebrange)
8732 +               ebrange = au_do_refresh_hdentry(dentry, parent);
8733 +
8734 +       if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8735 +               AuDebugOn(au_dbtop(dentry) < 0 && au_dbbot(dentry) >= 0);
8736 +               if (d_really_is_positive(dentry)) {
8737 +                       inode = d_inode(dentry);
8738 +                       err = au_refresh_hinode_self(inode);
8739 +               }
8740 +               au_dbg_verify_dinode(dentry);
8741 +               if (!err)
8742 +                       goto out_dgen; /* success */
8743 +               goto out;
8744 +       }
8745 +
8746 +       /* temporary dinfo */
8747 +       AuDbgDentry(dentry);
8748 +       err = -ENOMEM;
8749 +       tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8750 +       if (unlikely(!tmp))
8751 +               goto out;
8752 +       au_di_swap(tmp, dinfo);
8753 +       /* returns the number of positive dentries */
8754 +       /*
8755 +        * if current working dir is removed, it returns an error.
8756 +        * but the dentry is legal.
8757 +        */
8758 +       err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
8759 +       AuDbgDentry(dentry);
8760 +       au_di_swap(tmp, dinfo);
8761 +       if (err == -ENOENT)
8762 +               err = 0;
8763 +       if (err >= 0) {
8764 +               /* compare/refresh by dinfo */
8765 +               AuDbgDentry(dentry);
8766 +               err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8767 +               au_dbg_verify_dinode(dentry);
8768 +               AuTraceErr(err);
8769 +       }
8770 +       au_di_realloc(dinfo, nbr, /*may_shrink*/1); /* harmless if err */
8771 +       au_rw_write_unlock(&tmp->di_rwsem);
8772 +       au_di_free(tmp);
8773 +       if (unlikely(err))
8774 +               goto out;
8775 +
8776 +out_dgen:
8777 +       au_update_digen(dentry);
8778 +out:
8779 +       if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8780 +               AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8781 +               AuDbgDentry(dentry);
8782 +       }
8783 +       AuTraceErr(err);
8784 +       return err;
8785 +}
8786 +
8787 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8788 +                          struct dentry *dentry, aufs_bindex_t bindex)
8789 +{
8790 +       int err, valid;
8791 +
8792 +       err = 0;
8793 +       if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8794 +               goto out;
8795 +
8796 +       AuDbg("b%d\n", bindex);
8797 +       /*
8798 +        * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8799 +        * due to whiteout and branch permission.
8800 +        */
8801 +       flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8802 +                  | LOOKUP_FOLLOW | LOOKUP_EXCL);
8803 +       /* it may return tri-state */
8804 +       valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8805 +
8806 +       if (unlikely(valid < 0))
8807 +               err = valid;
8808 +       else if (!valid)
8809 +               err = -EINVAL;
8810 +
8811 +out:
8812 +       AuTraceErr(err);
8813 +       return err;
8814 +}
8815 +
8816 +/* todo: remove this */
8817 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8818 +                         unsigned int flags, int do_udba, int dirren)
8819 +{
8820 +       int err;
8821 +       umode_t mode, h_mode;
8822 +       aufs_bindex_t bindex, btail, btop, ibs, ibe;
8823 +       unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8824 +       struct inode *h_inode, *h_cached_inode;
8825 +       struct dentry *h_dentry;
8826 +       struct qstr *name, *h_name;
8827 +
8828 +       err = 0;
8829 +       plus = 0;
8830 +       mode = 0;
8831 +       ibs = -1;
8832 +       ibe = -1;
8833 +       unhashed = !!d_unhashed(dentry);
8834 +       is_root = !!IS_ROOT(dentry);
8835 +       name = &dentry->d_name;
8836 +       tmpfile = au_di(dentry)->di_tmpfile;
8837 +
8838 +       /*
8839 +        * Theoretically, REVAL test should be unnecessary in case of
8840 +        * {FS,I}NOTIFY.
8841 +        * But {fs,i}notify doesn't fire some necessary events,
8842 +        *      IN_ATTRIB for atime/nlink/pageio
8843 +        * Let's do REVAL test too.
8844 +        */
8845 +       if (do_udba && inode) {
8846 +               mode = (inode->i_mode & S_IFMT);
8847 +               plus = (inode->i_nlink > 0);
8848 +               ibs = au_ibtop(inode);
8849 +               ibe = au_ibbot(inode);
8850 +       }
8851 +
8852 +       btop = au_dbtop(dentry);
8853 +       btail = btop;
8854 +       if (inode && S_ISDIR(inode->i_mode))
8855 +               btail = au_dbtaildir(dentry);
8856 +       for (bindex = btop; bindex <= btail; bindex++) {
8857 +               h_dentry = au_h_dptr(dentry, bindex);
8858 +               if (!h_dentry)
8859 +                       continue;
8860 +
8861 +               AuDbg("b%d, %pd\n", bindex, h_dentry);
8862 +               h_nfs = !!au_test_nfs(h_dentry->d_sb);
8863 +               spin_lock(&h_dentry->d_lock);
8864 +               h_name = &h_dentry->d_name;
8865 +               if (unlikely(do_udba
8866 +                            && !is_root
8867 +                            && ((!h_nfs
8868 +                                 && (unhashed != !!d_unhashed(h_dentry)
8869 +                                     || (!tmpfile && !dirren
8870 +                                         && !au_qstreq(name, h_name))
8871 +                                         ))
8872 +                                || (h_nfs
8873 +                                    && !(flags & LOOKUP_OPEN)
8874 +                                    && (h_dentry->d_flags
8875 +                                        & DCACHE_NFSFS_RENAMED)))
8876 +                           )) {
8877 +                       int h_unhashed;
8878 +
8879 +                       h_unhashed = d_unhashed(h_dentry);
8880 +                       spin_unlock(&h_dentry->d_lock);
8881 +                       AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8882 +                             unhashed, h_unhashed, dentry, h_dentry);
8883 +                       goto err;
8884 +               }
8885 +               spin_unlock(&h_dentry->d_lock);
8886 +
8887 +               err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8888 +               if (unlikely(err))
8889 +                       /* do not goto err, to keep the errno */
8890 +                       break;
8891 +
8892 +               /* todo: plink too? */
8893 +               if (!do_udba)
8894 +                       continue;
8895 +
8896 +               /* UDBA tests */
8897 +               if (unlikely(!!inode != d_is_positive(h_dentry)))
8898 +                       goto err;
8899 +
8900 +               h_inode = NULL;
8901 +               if (d_is_positive(h_dentry))
8902 +                       h_inode = d_inode(h_dentry);
8903 +               h_plus = plus;
8904 +               h_mode = mode;
8905 +               h_cached_inode = h_inode;
8906 +               if (h_inode) {
8907 +                       h_mode = (h_inode->i_mode & S_IFMT);
8908 +                       h_plus = (h_inode->i_nlink > 0);
8909 +               }
8910 +               if (inode && ibs <= bindex && bindex <= ibe)
8911 +                       h_cached_inode = au_h_iptr(inode, bindex);
8912 +
8913 +               if (!h_nfs) {
8914 +                       if (unlikely(plus != h_plus && !tmpfile))
8915 +                               goto err;
8916 +               } else {
8917 +                       if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8918 +                                    && !is_root
8919 +                                    && !IS_ROOT(h_dentry)
8920 +                                    && unhashed != d_unhashed(h_dentry)))
8921 +                               goto err;
8922 +               }
8923 +               if (unlikely(mode != h_mode
8924 +                            || h_cached_inode != h_inode))
8925 +                       goto err;
8926 +               continue;
8927 +
8928 +err:
8929 +               err = -EINVAL;
8930 +               break;
8931 +       }
8932 +
8933 +       AuTraceErr(err);
8934 +       return err;
8935 +}
8936 +
8937 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8938 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8939 +{
8940 +       int err;
8941 +       struct dentry *parent;
8942 +
8943 +       if (!au_digen_test(dentry, sigen))
8944 +               return 0;
8945 +
8946 +       parent = dget_parent(dentry);
8947 +       di_read_lock_parent(parent, AuLock_IR);
8948 +       AuDebugOn(au_digen_test(parent, sigen));
8949 +       au_dbg_verify_gen(parent, sigen);
8950 +       err = au_refresh_dentry(dentry, parent);
8951 +       di_read_unlock(parent, AuLock_IR);
8952 +       dput(parent);
8953 +       AuTraceErr(err);
8954 +       return err;
8955 +}
8956 +
8957 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8958 +{
8959 +       int err;
8960 +       struct dentry *d, *parent;
8961 +
8962 +       if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8963 +               return simple_reval_dpath(dentry, sigen);
8964 +
8965 +       /* slow loop, keep it simple and stupid */
8966 +       /* cf: au_cpup_dirs() */
8967 +       err = 0;
8968 +       parent = NULL;
8969 +       while (au_digen_test(dentry, sigen)) {
8970 +               d = dentry;
8971 +               while (1) {
8972 +                       dput(parent);
8973 +                       parent = dget_parent(d);
8974 +                       if (!au_digen_test(parent, sigen))
8975 +                               break;
8976 +                       d = parent;
8977 +               }
8978 +
8979 +               if (d != dentry)
8980 +                       di_write_lock_child2(d);
8981 +
8982 +               /* someone might update our dentry while we were sleeping */
8983 +               if (au_digen_test(d, sigen)) {
8984 +                       /*
8985 +                        * todo: consolidate with simple_reval_dpath(),
8986 +                        * do_refresh() and au_reval_for_attr().
8987 +                        */
8988 +                       di_read_lock_parent(parent, AuLock_IR);
8989 +                       err = au_refresh_dentry(d, parent);
8990 +                       di_read_unlock(parent, AuLock_IR);
8991 +               }
8992 +
8993 +               if (d != dentry)
8994 +                       di_write_unlock(d);
8995 +               dput(parent);
8996 +               if (unlikely(err))
8997 +                       break;
8998 +       }
8999 +
9000 +       return err;
9001 +}
9002 +
9003 +/*
9004 + * if valid returns 1, otherwise 0.
9005 + */
9006 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
9007 +{
9008 +       int valid, err;
9009 +       unsigned int sigen;
9010 +       unsigned char do_udba, dirren;
9011 +       struct super_block *sb;
9012 +       struct inode *inode;
9013 +
9014 +       /* todo: support rcu-walk? */
9015 +       if (flags & LOOKUP_RCU)
9016 +               return -ECHILD;
9017 +
9018 +       valid = 0;
9019 +       if (unlikely(!au_di(dentry)))
9020 +               goto out;
9021 +
9022 +       valid = 1;
9023 +       sb = dentry->d_sb;
9024 +       /*
9025 +        * todo: very ugly
9026 +        * i_mutex of parent dir may be held,
9027 +        * but we should not return 'invalid' due to busy.
9028 +        */
9029 +       err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
9030 +       if (unlikely(err)) {
9031 +               valid = err;
9032 +               AuTraceErr(err);
9033 +               goto out;
9034 +       }
9035 +       inode = NULL;
9036 +       if (d_really_is_positive(dentry))
9037 +               inode = d_inode(dentry);
9038 +       if (unlikely(inode && au_is_bad_inode(inode))) {
9039 +               err = -EINVAL;
9040 +               AuTraceErr(err);
9041 +               goto out_dgrade;
9042 +       }
9043 +       if (unlikely(au_dbrange_test(dentry))) {
9044 +               err = -EINVAL;
9045 +               AuTraceErr(err);
9046 +               goto out_dgrade;
9047 +       }
9048 +
9049 +       sigen = au_sigen(sb);
9050 +       if (au_digen_test(dentry, sigen)) {
9051 +               AuDebugOn(IS_ROOT(dentry));
9052 +               err = au_reval_dpath(dentry, sigen);
9053 +               if (unlikely(err)) {
9054 +                       AuTraceErr(err);
9055 +                       goto out_dgrade;
9056 +               }
9057 +       }
9058 +       di_downgrade_lock(dentry, AuLock_IR);
9059 +
9060 +       err = -EINVAL;
9061 +       if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
9062 +           && inode
9063 +           && !(inode->i_state && I_LINKABLE)
9064 +           && (IS_DEADDIR(inode) || !inode->i_nlink)) {
9065 +               AuTraceErr(err);
9066 +               goto out_inval;
9067 +       }
9068 +
9069 +       do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
9070 +       if (do_udba && inode) {
9071 +               aufs_bindex_t btop = au_ibtop(inode);
9072 +               struct inode *h_inode;
9073 +
9074 +               if (btop >= 0) {
9075 +                       h_inode = au_h_iptr(inode, btop);
9076 +                       if (h_inode && au_test_higen(inode, h_inode)) {
9077 +                               AuTraceErr(err);
9078 +                               goto out_inval;
9079 +                       }
9080 +               }
9081 +       }
9082 +
9083 +       dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
9084 +       err = h_d_revalidate(dentry, inode, flags, do_udba, dirren);
9085 +       if (unlikely(!err && do_udba && au_dbtop(dentry) < 0)) {
9086 +               err = -EIO;
9087 +               AuDbg("both of real entry and whiteout found, %p, err %d\n",
9088 +                     dentry, err);
9089 +       }
9090 +       goto out_inval;
9091 +
9092 +out_dgrade:
9093 +       di_downgrade_lock(dentry, AuLock_IR);
9094 +out_inval:
9095 +       aufs_read_unlock(dentry, AuLock_IR);
9096 +       AuTraceErr(err);
9097 +       valid = !err;
9098 +out:
9099 +       if (!valid) {
9100 +               AuDbg("%pd invalid, %d\n", dentry, valid);
9101 +               d_drop(dentry);
9102 +       }
9103 +       return valid;
9104 +}
9105 +
9106 +static void aufs_d_release(struct dentry *dentry)
9107 +{
9108 +       if (au_di(dentry)) {
9109 +               au_di_fin(dentry);
9110 +               au_hn_di_reinit(dentry);
9111 +       }
9112 +}
9113 +
9114 +const struct dentry_operations aufs_dop = {
9115 +       .d_revalidate           = aufs_d_revalidate,
9116 +       .d_weak_revalidate      = aufs_d_revalidate,
9117 +       .d_release              = aufs_d_release
9118 +};
9119 +
9120 +/* aufs_dop without d_revalidate */
9121 +const struct dentry_operations aufs_dop_noreval = {
9122 +       .d_release              = aufs_d_release
9123 +};
9124 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
9125 --- /usr/share/empty/fs/aufs/dentry.h   1970-01-01 01:00:00.000000000 +0100
9126 +++ linux/fs/aufs/dentry.h      2022-11-05 23:02:18.962555950 +0100
9127 @@ -0,0 +1,269 @@
9128 +/* SPDX-License-Identifier: GPL-2.0 */
9129 +/*
9130 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9131 + *
9132 + * This program is free software; you can redistribute it and/or modify
9133 + * it under the terms of the GNU General Public License as published by
9134 + * the Free Software Foundation; either version 2 of the License, or
9135 + * (at your option) any later version.
9136 + *
9137 + * This program is distributed in the hope that it will be useful,
9138 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9139 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9140 + * GNU General Public License for more details.
9141 + *
9142 + * You should have received a copy of the GNU General Public License
9143 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
9144 + */
9145 +
9146 +/*
9147 + * lookup and dentry operations
9148 + */
9149 +
9150 +#ifndef __AUFS_DENTRY_H__
9151 +#define __AUFS_DENTRY_H__
9152 +
9153 +#ifdef __KERNEL__
9154 +
9155 +#include <linux/dcache.h>
9156 +#include "dirren.h"
9157 +#include "rwsem.h"
9158 +
9159 +struct au_hdentry {
9160 +       struct dentry           *hd_dentry;
9161 +       aufs_bindex_t           hd_id;
9162 +};
9163 +
9164 +struct au_dinfo {
9165 +       atomic_t                di_generation;
9166 +
9167 +       struct au_rwsem         di_rwsem;
9168 +       aufs_bindex_t           di_btop, di_bbot, di_bwh, di_bdiropq;
9169 +       unsigned char           di_tmpfile; /* to allow the different name */
9170 +       struct au_hdentry       *di_hdentry;
9171 +       struct rcu_head         rcu;
9172 +} ____cacheline_aligned_in_smp;
9173 +
9174 +/* ---------------------------------------------------------------------- */
9175 +
9176 +/* flags for au_lkup_dentry() */
9177 +#define AuLkup_ALLOW_NEG       1
9178 +#define AuLkup_IGNORE_PERM     (1 << 1)
9179 +#define AuLkup_DIRREN          (1 << 2)
9180 +#define au_ftest_lkup(flags, name)     ((flags) & AuLkup_##name)
9181 +#define au_fset_lkup(flags, name) \
9182 +       do { (flags) |= AuLkup_##name; } while (0)
9183 +#define au_fclr_lkup(flags, name) \
9184 +       do { (flags) &= ~AuLkup_##name; } while (0)
9185 +
9186 +#ifndef CONFIG_AUFS_DIRREN
9187 +#undef AuLkup_DIRREN
9188 +#define AuLkup_DIRREN 0
9189 +#endif
9190 +
9191 +struct au_do_lookup_args {
9192 +       unsigned int            flags;
9193 +       mode_t                  type;
9194 +       struct qstr             whname, *name;
9195 +       struct au_dr_lookup     dirren;
9196 +};
9197 +
9198 +/* ---------------------------------------------------------------------- */
9199 +
9200 +/* dentry.c */
9201 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
9202 +struct au_branch;
9203 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
9204 +                              struct path *ppath);
9205 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
9206 +               struct dentry *h_parent, struct au_branch *br);
9207 +
9208 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
9209 +                  unsigned int flags);
9210 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
9211 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
9212 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
9213 +void au_refresh_dop(struct dentry *dentry, int force_reval);
9214 +
9215 +/* dinfo.c */
9216 +void au_di_init_once(void *_di);
9217 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
9218 +void au_di_free(struct au_dinfo *dinfo);
9219 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
9220 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
9221 +int au_di_init(struct dentry *dentry);
9222 +void au_di_fin(struct dentry *dentry);
9223 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink);
9224 +
9225 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
9226 +void di_read_unlock(struct dentry *d, int flags);
9227 +void di_downgrade_lock(struct dentry *d, int flags);
9228 +void di_write_lock(struct dentry *d, unsigned int lsc);
9229 +void di_write_unlock(struct dentry *d);
9230 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
9231 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
9232 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
9233 +
9234 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
9235 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
9236 +aufs_bindex_t au_dbtail(struct dentry *dentry);
9237 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
9238 +
9239 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9240 +                  struct dentry *h_dentry);
9241 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
9242 +int au_dbrange_test(struct dentry *dentry);
9243 +void au_update_digen(struct dentry *dentry);
9244 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
9245 +void au_update_dbtop(struct dentry *dentry);
9246 +void au_update_dbbot(struct dentry *dentry);
9247 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
9248 +
9249 +/* ---------------------------------------------------------------------- */
9250 +
9251 +static inline struct au_dinfo *au_di(struct dentry *dentry)
9252 +{
9253 +       return dentry->d_fsdata;
9254 +}
9255 +
9256 +/* ---------------------------------------------------------------------- */
9257 +
9258 +/* lock subclass for dinfo */
9259 +enum {
9260 +       AuLsc_DI_CHILD,         /* child first */
9261 +       AuLsc_DI_CHILD2,        /* rename(2), link(2), and cpup at hnotify */
9262 +       AuLsc_DI_CHILD3,        /* copyup dirs */
9263 +       AuLsc_DI_PARENT,
9264 +       AuLsc_DI_PARENT2,
9265 +       AuLsc_DI_PARENT3,
9266 +       AuLsc_DI_TMP            /* temp for replacing dinfo */
9267 +};
9268 +
9269 +/*
9270 + * di_read_lock_child, di_write_lock_child,
9271 + * di_read_lock_child2, di_write_lock_child2,
9272 + * di_read_lock_child3, di_write_lock_child3,
9273 + * di_read_lock_parent, di_write_lock_parent,
9274 + * di_read_lock_parent2, di_write_lock_parent2,
9275 + * di_read_lock_parent3, di_write_lock_parent3,
9276 + */
9277 +#define AuReadLockFunc(name, lsc) \
9278 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
9279 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
9280 +
9281 +#define AuWriteLockFunc(name, lsc) \
9282 +static inline void di_write_lock_##name(struct dentry *d) \
9283 +{ di_write_lock(d, AuLsc_DI_##lsc); }
9284 +
9285 +#define AuRWLockFuncs(name, lsc) \
9286 +       AuReadLockFunc(name, lsc) \
9287 +       AuWriteLockFunc(name, lsc)
9288 +
9289 +AuRWLockFuncs(child, CHILD);
9290 +AuRWLockFuncs(child2, CHILD2);
9291 +AuRWLockFuncs(child3, CHILD3);
9292 +AuRWLockFuncs(parent, PARENT);
9293 +AuRWLockFuncs(parent2, PARENT2);
9294 +AuRWLockFuncs(parent3, PARENT3);
9295 +
9296 +#undef AuReadLockFunc
9297 +#undef AuWriteLockFunc
9298 +#undef AuRWLockFuncs
9299 +
9300 +#define DiMustNoWaiters(d)     AuRwMustNoWaiters(&au_di(d)->di_rwsem)
9301 +#define DiMustAnyLock(d)       AuRwMustAnyLock(&au_di(d)->di_rwsem)
9302 +#define DiMustWriteLock(d)     AuRwMustWriteLock(&au_di(d)->di_rwsem)
9303 +
9304 +/* ---------------------------------------------------------------------- */
9305 +
9306 +/* todo: memory barrier? */
9307 +static inline unsigned int au_digen(struct dentry *d)
9308 +{
9309 +       return atomic_read(&au_di(d)->di_generation);
9310 +}
9311 +
9312 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
9313 +{
9314 +       hdentry->hd_dentry = NULL;
9315 +}
9316 +
9317 +static inline struct au_hdentry *au_hdentry(struct au_dinfo *di,
9318 +                                           aufs_bindex_t bindex)
9319 +{
9320 +       return di->di_hdentry + bindex;
9321 +}
9322 +
9323 +static inline void au_hdput(struct au_hdentry *hd)
9324 +{
9325 +       if (hd)
9326 +               dput(hd->hd_dentry);
9327 +}
9328 +
9329 +static inline aufs_bindex_t au_dbtop(struct dentry *dentry)
9330 +{
9331 +       DiMustAnyLock(dentry);
9332 +       return au_di(dentry)->di_btop;
9333 +}
9334 +
9335 +static inline aufs_bindex_t au_dbbot(struct dentry *dentry)
9336 +{
9337 +       DiMustAnyLock(dentry);
9338 +       return au_di(dentry)->di_bbot;
9339 +}
9340 +
9341 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
9342 +{
9343 +       DiMustAnyLock(dentry);
9344 +       return au_di(dentry)->di_bwh;
9345 +}
9346 +
9347 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
9348 +{
9349 +       DiMustAnyLock(dentry);
9350 +       return au_di(dentry)->di_bdiropq;
9351 +}
9352 +
9353 +/* todo: hard/soft set? */
9354 +static inline void au_set_dbtop(struct dentry *dentry, aufs_bindex_t bindex)
9355 +{
9356 +       DiMustWriteLock(dentry);
9357 +       au_di(dentry)->di_btop = bindex;
9358 +}
9359 +
9360 +static inline void au_set_dbbot(struct dentry *dentry, aufs_bindex_t bindex)
9361 +{
9362 +       DiMustWriteLock(dentry);
9363 +       au_di(dentry)->di_bbot = bindex;
9364 +}
9365 +
9366 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
9367 +{
9368 +       DiMustWriteLock(dentry);
9369 +       /* dbwh can be outside of btop - bbot range */
9370 +       au_di(dentry)->di_bwh = bindex;
9371 +}
9372 +
9373 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
9374 +{
9375 +       DiMustWriteLock(dentry);
9376 +       au_di(dentry)->di_bdiropq = bindex;
9377 +}
9378 +
9379 +/* ---------------------------------------------------------------------- */
9380 +
9381 +#ifdef CONFIG_AUFS_HNOTIFY
9382 +static inline void au_digen_dec(struct dentry *d)
9383 +{
9384 +       atomic_dec(&au_di(d)->di_generation);
9385 +}
9386 +
9387 +static inline void au_hn_di_reinit(struct dentry *dentry)
9388 +{
9389 +       dentry->d_fsdata = NULL;
9390 +}
9391 +#else
9392 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
9393 +#endif /* CONFIG_AUFS_HNOTIFY */
9394 +
9395 +#endif /* __KERNEL__ */
9396 +#endif /* __AUFS_DENTRY_H__ */
9397 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
9398 --- /usr/share/empty/fs/aufs/dinfo.c    1970-01-01 01:00:00.000000000 +0100
9399 +++ linux/fs/aufs/dinfo.c       2022-11-05 23:02:18.962555950 +0100
9400 @@ -0,0 +1,554 @@
9401 +// SPDX-License-Identifier: GPL-2.0
9402 +/*
9403 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9404 + *
9405 + * This program is free software; you can redistribute it and/or modify
9406 + * it under the terms of the GNU General Public License as published by
9407 + * the Free Software Foundation; either version 2 of the License, or
9408 + * (at your option) any later version.
9409 + *
9410 + * This program is distributed in the hope that it will be useful,
9411 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9412 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9413 + * GNU General Public License for more details.
9414 + *
9415 + * You should have received a copy of the GNU General Public License
9416 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
9417 + */
9418 +
9419 +/*
9420 + * dentry private data
9421 + */
9422 +
9423 +#include "aufs.h"
9424 +
9425 +void au_di_init_once(void *_dinfo)
9426 +{
9427 +       struct au_dinfo *dinfo = _dinfo;
9428 +
9429 +       au_rw_init(&dinfo->di_rwsem);
9430 +}
9431 +
9432 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
9433 +{
9434 +       struct au_dinfo *dinfo;
9435 +       int nbr, i;
9436 +
9437 +       dinfo = au_cache_alloc_dinfo();
9438 +       if (unlikely(!dinfo))
9439 +               goto out;
9440 +
9441 +       nbr = au_sbbot(sb) + 1;
9442 +       if (nbr <= 0)
9443 +               nbr = 1;
9444 +       dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
9445 +       if (dinfo->di_hdentry) {
9446 +               au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
9447 +               dinfo->di_btop = -1;
9448 +               dinfo->di_bbot = -1;
9449 +               dinfo->di_bwh = -1;
9450 +               dinfo->di_bdiropq = -1;
9451 +               dinfo->di_tmpfile = 0;
9452 +               for (i = 0; i < nbr; i++)
9453 +                       dinfo->di_hdentry[i].hd_id = -1;
9454 +               goto out;
9455 +       }
9456 +
9457 +       au_cache_free_dinfo(dinfo);
9458 +       dinfo = NULL;
9459 +
9460 +out:
9461 +       return dinfo;
9462 +}
9463 +
9464 +void au_di_free(struct au_dinfo *dinfo)
9465 +{
9466 +       struct au_hdentry *p;
9467 +       aufs_bindex_t bbot, bindex;
9468 +
9469 +       /* dentry may not be revalidated */
9470 +       bindex = dinfo->di_btop;
9471 +       if (bindex >= 0) {
9472 +               bbot = dinfo->di_bbot;
9473 +               p = au_hdentry(dinfo, bindex);
9474 +               while (bindex++ <= bbot)
9475 +                       au_hdput(p++);
9476 +       }
9477 +       au_kfree_try_rcu(dinfo->di_hdentry);
9478 +       au_cache_free_dinfo(dinfo);
9479 +}
9480 +
9481 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
9482 +{
9483 +       struct au_hdentry *p;
9484 +       aufs_bindex_t bi;
9485 +
9486 +       AuRwMustWriteLock(&a->di_rwsem);
9487 +       AuRwMustWriteLock(&b->di_rwsem);
9488 +
9489 +#define DiSwap(v, name)                                \
9490 +       do {                                    \
9491 +               v = a->di_##name;               \
9492 +               a->di_##name = b->di_##name;    \
9493 +               b->di_##name = v;               \
9494 +       } while (0)
9495 +
9496 +       DiSwap(p, hdentry);
9497 +       DiSwap(bi, btop);
9498 +       DiSwap(bi, bbot);
9499 +       DiSwap(bi, bwh);
9500 +       DiSwap(bi, bdiropq);
9501 +       /* smp_mb(); */
9502 +
9503 +#undef DiSwap
9504 +}
9505 +
9506 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
9507 +{
9508 +       AuRwMustWriteLock(&dst->di_rwsem);
9509 +       AuRwMustWriteLock(&src->di_rwsem);
9510 +
9511 +       dst->di_btop = src->di_btop;
9512 +       dst->di_bbot = src->di_bbot;
9513 +       dst->di_bwh = src->di_bwh;
9514 +       dst->di_bdiropq = src->di_bdiropq;
9515 +       /* smp_mb(); */
9516 +}
9517 +
9518 +int au_di_init(struct dentry *dentry)
9519 +{
9520 +       int err;
9521 +       struct super_block *sb;
9522 +       struct au_dinfo *dinfo;
9523 +
9524 +       err = 0;
9525 +       sb = dentry->d_sb;
9526 +       dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
9527 +       if (dinfo) {
9528 +               atomic_set(&dinfo->di_generation, au_sigen(sb));
9529 +               /* smp_mb(); */ /* atomic_set */
9530 +               dentry->d_fsdata = dinfo;
9531 +       } else
9532 +               err = -ENOMEM;
9533 +
9534 +       return err;
9535 +}
9536 +
9537 +void au_di_fin(struct dentry *dentry)
9538 +{
9539 +       struct au_dinfo *dinfo;
9540 +
9541 +       dinfo = au_di(dentry);
9542 +       AuRwDestroy(&dinfo->di_rwsem);
9543 +       au_di_free(dinfo);
9544 +}
9545 +
9546 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink)
9547 +{
9548 +       int err, sz;
9549 +       struct au_hdentry *hdp;
9550 +
9551 +       AuRwMustWriteLock(&dinfo->di_rwsem);
9552 +
9553 +       err = -ENOMEM;
9554 +       sz = sizeof(*hdp) * (dinfo->di_bbot + 1);
9555 +       if (!sz)
9556 +               sz = sizeof(*hdp);
9557 +       hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS,
9558 +                          may_shrink);
9559 +       if (hdp) {
9560 +               dinfo->di_hdentry = hdp;
9561 +               err = 0;
9562 +       }
9563 +
9564 +       return err;
9565 +}
9566 +
9567 +/* ---------------------------------------------------------------------- */
9568 +
9569 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
9570 +{
9571 +       switch (lsc) {
9572 +       case AuLsc_DI_CHILD:
9573 +               ii_write_lock_child(inode);
9574 +               break;
9575 +       case AuLsc_DI_CHILD2:
9576 +               ii_write_lock_child2(inode);
9577 +               break;
9578 +       case AuLsc_DI_CHILD3:
9579 +               ii_write_lock_child3(inode);
9580 +               break;
9581 +       case AuLsc_DI_PARENT:
9582 +               ii_write_lock_parent(inode);
9583 +               break;
9584 +       case AuLsc_DI_PARENT2:
9585 +               ii_write_lock_parent2(inode);
9586 +               break;
9587 +       case AuLsc_DI_PARENT3:
9588 +               ii_write_lock_parent3(inode);
9589 +               break;
9590 +       default:
9591 +               BUG();
9592 +       }
9593 +}
9594 +
9595 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9596 +{
9597 +       switch (lsc) {
9598 +       case AuLsc_DI_CHILD:
9599 +               ii_read_lock_child(inode);
9600 +               break;
9601 +       case AuLsc_DI_CHILD2:
9602 +               ii_read_lock_child2(inode);
9603 +               break;
9604 +       case AuLsc_DI_CHILD3:
9605 +               ii_read_lock_child3(inode);
9606 +               break;
9607 +       case AuLsc_DI_PARENT:
9608 +               ii_read_lock_parent(inode);
9609 +               break;
9610 +       case AuLsc_DI_PARENT2:
9611 +               ii_read_lock_parent2(inode);
9612 +               break;
9613 +       case AuLsc_DI_PARENT3:
9614 +               ii_read_lock_parent3(inode);
9615 +               break;
9616 +       default:
9617 +               BUG();
9618 +       }
9619 +}
9620 +
9621 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9622 +{
9623 +       struct inode *inode;
9624 +
9625 +       au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9626 +       if (d_really_is_positive(d)) {
9627 +               inode = d_inode(d);
9628 +               if (au_ftest_lock(flags, IW))
9629 +                       do_ii_write_lock(inode, lsc);
9630 +               else if (au_ftest_lock(flags, IR))
9631 +                       do_ii_read_lock(inode, lsc);
9632 +       }
9633 +}
9634 +
9635 +void di_read_unlock(struct dentry *d, int flags)
9636 +{
9637 +       struct inode *inode;
9638 +
9639 +       if (d_really_is_positive(d)) {
9640 +               inode = d_inode(d);
9641 +               if (au_ftest_lock(flags, IW)) {
9642 +                       au_dbg_verify_dinode(d);
9643 +                       ii_write_unlock(inode);
9644 +               } else if (au_ftest_lock(flags, IR)) {
9645 +                       au_dbg_verify_dinode(d);
9646 +                       ii_read_unlock(inode);
9647 +               }
9648 +       }
9649 +       au_rw_read_unlock(&au_di(d)->di_rwsem);
9650 +}
9651 +
9652 +void di_downgrade_lock(struct dentry *d, int flags)
9653 +{
9654 +       if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9655 +               ii_downgrade_lock(d_inode(d));
9656 +       au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9657 +}
9658 +
9659 +void di_write_lock(struct dentry *d, unsigned int lsc)
9660 +{
9661 +       au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9662 +       if (d_really_is_positive(d))
9663 +               do_ii_write_lock(d_inode(d), lsc);
9664 +}
9665 +
9666 +void di_write_unlock(struct dentry *d)
9667 +{
9668 +       au_dbg_verify_dinode(d);
9669 +       if (d_really_is_positive(d))
9670 +               ii_write_unlock(d_inode(d));
9671 +       au_rw_write_unlock(&au_di(d)->di_rwsem);
9672 +}
9673 +
9674 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9675 +{
9676 +       AuDebugOn(d1 == d2
9677 +                 || d_inode(d1) == d_inode(d2)
9678 +                 || d1->d_sb != d2->d_sb);
9679 +
9680 +       if ((isdir && au_test_subdir(d1, d2))
9681 +           || d1 < d2) {
9682 +               di_write_lock_child(d1);
9683 +               di_write_lock_child2(d2);
9684 +       } else {
9685 +               di_write_lock_child(d2);
9686 +               di_write_lock_child2(d1);
9687 +       }
9688 +}
9689 +
9690 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9691 +{
9692 +       AuDebugOn(d1 == d2
9693 +                 || d_inode(d1) == d_inode(d2)
9694 +                 || d1->d_sb != d2->d_sb);
9695 +
9696 +       if ((isdir && au_test_subdir(d1, d2))
9697 +           || d1 < d2) {
9698 +               di_write_lock_parent(d1);
9699 +               di_write_lock_parent2(d2);
9700 +       } else {
9701 +               di_write_lock_parent(d2);
9702 +               di_write_lock_parent2(d1);
9703 +       }
9704 +}
9705 +
9706 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9707 +{
9708 +       di_write_unlock(d1);
9709 +       if (d_inode(d1) == d_inode(d2))
9710 +               au_rw_write_unlock(&au_di(d2)->di_rwsem);
9711 +       else
9712 +               di_write_unlock(d2);
9713 +}
9714 +
9715 +/* ---------------------------------------------------------------------- */
9716 +
9717 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9718 +{
9719 +       struct dentry *d;
9720 +
9721 +       DiMustAnyLock(dentry);
9722 +
9723 +       if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
9724 +               return NULL;
9725 +       AuDebugOn(bindex < 0);
9726 +       d = au_hdentry(au_di(dentry), bindex)->hd_dentry;
9727 +       AuDebugOn(d && au_dcount(d) <= 0);
9728 +       return d;
9729 +}
9730 +
9731 +/*
9732 + * extended version of au_h_dptr().
9733 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9734 + * error.
9735 + */
9736 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9737 +{
9738 +       struct dentry *h_dentry;
9739 +       struct inode *inode, *h_inode;
9740 +
9741 +       AuDebugOn(d_really_is_negative(dentry));
9742 +
9743 +       h_dentry = NULL;
9744 +       if (au_dbtop(dentry) <= bindex
9745 +           && bindex <= au_dbbot(dentry))
9746 +               h_dentry = au_h_dptr(dentry, bindex);
9747 +       if (h_dentry && !au_d_linkable(h_dentry)) {
9748 +               dget(h_dentry);
9749 +               goto out; /* success */
9750 +       }
9751 +
9752 +       inode = d_inode(dentry);
9753 +       AuDebugOn(bindex < au_ibtop(inode));
9754 +       AuDebugOn(au_ibbot(inode) < bindex);
9755 +       h_inode = au_h_iptr(inode, bindex);
9756 +       h_dentry = d_find_alias(h_inode);
9757 +       if (h_dentry) {
9758 +               if (!IS_ERR(h_dentry)) {
9759 +                       if (!au_d_linkable(h_dentry))
9760 +                               goto out; /* success */
9761 +                       dput(h_dentry);
9762 +               } else
9763 +                       goto out;
9764 +       }
9765 +
9766 +       if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9767 +               h_dentry = au_plink_lkup(inode, bindex);
9768 +               AuDebugOn(!h_dentry);
9769 +               if (!IS_ERR(h_dentry)) {
9770 +                       if (!au_d_hashed_positive(h_dentry))
9771 +                               goto out; /* success */
9772 +                       dput(h_dentry);
9773 +                       h_dentry = NULL;
9774 +               }
9775 +       }
9776 +
9777 +out:
9778 +       AuDbgDentry(h_dentry);
9779 +       return h_dentry;
9780 +}
9781 +
9782 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9783 +{
9784 +       aufs_bindex_t bbot, bwh;
9785 +
9786 +       bbot = au_dbbot(dentry);
9787 +       if (0 <= bbot) {
9788 +               bwh = au_dbwh(dentry);
9789 +               if (!bwh)
9790 +                       return bwh;
9791 +               if (0 < bwh && bwh < bbot)
9792 +                       return bwh - 1;
9793 +       }
9794 +       return bbot;
9795 +}
9796 +
9797 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9798 +{
9799 +       aufs_bindex_t bbot, bopq;
9800 +
9801 +       bbot = au_dbtail(dentry);
9802 +       if (0 <= bbot) {
9803 +               bopq = au_dbdiropq(dentry);
9804 +               if (0 <= bopq && bopq < bbot)
9805 +                       bbot = bopq;
9806 +       }
9807 +       return bbot;
9808 +}
9809 +
9810 +/* ---------------------------------------------------------------------- */
9811 +
9812 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9813 +                  struct dentry *h_dentry)
9814 +{
9815 +       struct au_dinfo *dinfo;
9816 +       struct au_hdentry *hd;
9817 +       struct au_branch *br;
9818 +
9819 +       DiMustWriteLock(dentry);
9820 +
9821 +       dinfo = au_di(dentry);
9822 +       hd = au_hdentry(dinfo, bindex);
9823 +       au_hdput(hd);
9824 +       hd->hd_dentry = h_dentry;
9825 +       if (h_dentry) {
9826 +               br = au_sbr(dentry->d_sb, bindex);
9827 +               hd->hd_id = br->br_id;
9828 +       }
9829 +}
9830 +
9831 +int au_dbrange_test(struct dentry *dentry)
9832 +{
9833 +       int err;
9834 +       aufs_bindex_t btop, bbot;
9835 +
9836 +       err = 0;
9837 +       btop = au_dbtop(dentry);
9838 +       bbot = au_dbbot(dentry);
9839 +       if (btop >= 0)
9840 +               AuDebugOn(bbot < 0 && btop > bbot);
9841 +       else {
9842 +               err = -EIO;
9843 +               AuDebugOn(bbot >= 0);
9844 +       }
9845 +
9846 +       return err;
9847 +}
9848 +
9849 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9850 +{
9851 +       int err;
9852 +
9853 +       err = 0;
9854 +       if (unlikely(au_digen(dentry) != sigen
9855 +                    || au_iigen_test(d_inode(dentry), sigen)))
9856 +               err = -EIO;
9857 +
9858 +       return err;
9859 +}
9860 +
9861 +void au_update_digen(struct dentry *dentry)
9862 +{
9863 +       atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9864 +       /* smp_mb(); */ /* atomic_set */
9865 +}
9866 +
9867 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9868 +{
9869 +       struct au_dinfo *dinfo;
9870 +       struct dentry *h_d;
9871 +       struct au_hdentry *hdp;
9872 +       aufs_bindex_t bindex, bbot;
9873 +
9874 +       DiMustWriteLock(dentry);
9875 +
9876 +       dinfo = au_di(dentry);
9877 +       if (!dinfo || dinfo->di_btop < 0)
9878 +               return;
9879 +
9880 +       if (do_put_zero) {
9881 +               bbot = dinfo->di_bbot;
9882 +               bindex = dinfo->di_btop;
9883 +               hdp = au_hdentry(dinfo, bindex);
9884 +               for (; bindex <= bbot; bindex++, hdp++) {
9885 +                       h_d = hdp->hd_dentry;
9886 +                       if (h_d && d_is_negative(h_d))
9887 +                               au_set_h_dptr(dentry, bindex, NULL);
9888 +               }
9889 +       }
9890 +
9891 +       dinfo->di_btop = 0;
9892 +       hdp = au_hdentry(dinfo, dinfo->di_btop);
9893 +       for (; dinfo->di_btop <= dinfo->di_bbot; dinfo->di_btop++, hdp++)
9894 +               if (hdp->hd_dentry)
9895 +                       break;
9896 +       if (dinfo->di_btop > dinfo->di_bbot) {
9897 +               dinfo->di_btop = -1;
9898 +               dinfo->di_bbot = -1;
9899 +               return;
9900 +       }
9901 +
9902 +       hdp = au_hdentry(dinfo, dinfo->di_bbot);
9903 +       for (; dinfo->di_bbot >= 0; dinfo->di_bbot--, hdp--)
9904 +               if (hdp->hd_dentry)
9905 +                       break;
9906 +       AuDebugOn(dinfo->di_btop > dinfo->di_bbot || dinfo->di_bbot < 0);
9907 +}
9908 +
9909 +void au_update_dbtop(struct dentry *dentry)
9910 +{
9911 +       aufs_bindex_t bindex, bbot;
9912 +       struct dentry *h_dentry;
9913 +
9914 +       bbot = au_dbbot(dentry);
9915 +       for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
9916 +               h_dentry = au_h_dptr(dentry, bindex);
9917 +               if (!h_dentry)
9918 +                       continue;
9919 +               if (d_is_positive(h_dentry)) {
9920 +                       au_set_dbtop(dentry, bindex);
9921 +                       return;
9922 +               }
9923 +               au_set_h_dptr(dentry, bindex, NULL);
9924 +       }
9925 +}
9926 +
9927 +void au_update_dbbot(struct dentry *dentry)
9928 +{
9929 +       aufs_bindex_t bindex, btop;
9930 +       struct dentry *h_dentry;
9931 +
9932 +       btop = au_dbtop(dentry);
9933 +       for (bindex = au_dbbot(dentry); bindex >= btop; bindex--) {
9934 +               h_dentry = au_h_dptr(dentry, bindex);
9935 +               if (!h_dentry)
9936 +                       continue;
9937 +               if (d_is_positive(h_dentry)) {
9938 +                       au_set_dbbot(dentry, bindex);
9939 +                       return;
9940 +               }
9941 +               au_set_h_dptr(dentry, bindex, NULL);
9942 +       }
9943 +}
9944 +
9945 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9946 +{
9947 +       aufs_bindex_t bindex, bbot;
9948 +
9949 +       bbot = au_dbbot(dentry);
9950 +       for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++)
9951 +               if (au_h_dptr(dentry, bindex) == h_dentry)
9952 +                       return bindex;
9953 +       return -1;
9954 +}
9955 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9956 --- /usr/share/empty/fs/aufs/dir.c      1970-01-01 01:00:00.000000000 +0100
9957 +++ linux/fs/aufs/dir.c 2022-11-05 23:02:18.962555950 +0100
9958 @@ -0,0 +1,765 @@
9959 +// SPDX-License-Identifier: GPL-2.0
9960 +/*
9961 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9962 + *
9963 + * This program is free software; you can redistribute it and/or modify
9964 + * it under the terms of the GNU General Public License as published by
9965 + * the Free Software Foundation; either version 2 of the License, or
9966 + * (at your option) any later version.
9967 + *
9968 + * This program is distributed in the hope that it will be useful,
9969 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9970 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9971 + * GNU General Public License for more details.
9972 + *
9973 + * You should have received a copy of the GNU General Public License
9974 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
9975 + */
9976 +
9977 +/*
9978 + * directory operations
9979 + */
9980 +
9981 +#include <linux/fs_stack.h>
9982 +#include <linux/iversion.h>
9983 +#include "aufs.h"
9984 +
9985 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
9986 +{
9987 +       unsigned int nlink;
9988 +
9989 +       AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9990 +
9991 +       nlink = dir->i_nlink;
9992 +       nlink += h_dir->i_nlink - 2;
9993 +       if (h_dir->i_nlink < 2)
9994 +               nlink += 2;
9995 +       smp_mb(); /* for i_nlink */
9996 +       /* 0 can happen in revaliding */
9997 +       set_nlink(dir, nlink);
9998 +}
9999 +
10000 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
10001 +{
10002 +       unsigned int nlink;
10003 +
10004 +       AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10005 +
10006 +       nlink = dir->i_nlink;
10007 +       nlink -= h_dir->i_nlink - 2;
10008 +       if (h_dir->i_nlink < 2)
10009 +               nlink -= 2;
10010 +       smp_mb(); /* for i_nlink */
10011 +       /* nlink == 0 means the branch-fs is broken */
10012 +       set_nlink(dir, nlink);
10013 +}
10014 +
10015 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
10016 +{
10017 +       loff_t sz;
10018 +       aufs_bindex_t bindex, bbot;
10019 +       struct file *h_file;
10020 +       struct dentry *h_dentry;
10021 +
10022 +       sz = 0;
10023 +       if (file) {
10024 +               AuDebugOn(!d_is_dir(file->f_path.dentry));
10025 +
10026 +               bbot = au_fbbot_dir(file);
10027 +               for (bindex = au_fbtop(file);
10028 +                    bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10029 +                    bindex++) {
10030 +                       h_file = au_hf_dir(file, bindex);
10031 +                       if (h_file && file_inode(h_file))
10032 +                               sz += vfsub_f_size_read(h_file);
10033 +               }
10034 +       } else {
10035 +               AuDebugOn(!dentry);
10036 +               AuDebugOn(!d_is_dir(dentry));
10037 +
10038 +               bbot = au_dbtaildir(dentry);
10039 +               for (bindex = au_dbtop(dentry);
10040 +                    bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10041 +                    bindex++) {
10042 +                       h_dentry = au_h_dptr(dentry, bindex);
10043 +                       if (h_dentry && d_is_positive(h_dentry))
10044 +                               sz += i_size_read(d_inode(h_dentry));
10045 +               }
10046 +       }
10047 +       if (sz < KMALLOC_MAX_SIZE)
10048 +               sz = roundup_pow_of_two(sz);
10049 +       if (sz > KMALLOC_MAX_SIZE)
10050 +               sz = KMALLOC_MAX_SIZE;
10051 +       else if (sz < NAME_MAX) {
10052 +               BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
10053 +               sz = AUFS_RDBLK_DEF;
10054 +       }
10055 +       return sz;
10056 +}
10057 +
10058 +struct au_dir_ts_arg {
10059 +       struct dentry *dentry;
10060 +       aufs_bindex_t brid;
10061 +};
10062 +
10063 +static void au_do_dir_ts(void *arg)
10064 +{
10065 +       struct au_dir_ts_arg *a = arg;
10066 +       struct au_dtime dt;
10067 +       struct path h_path;
10068 +       struct inode *dir, *h_dir;
10069 +       struct super_block *sb;
10070 +       struct au_branch *br;
10071 +       struct au_hinode *hdir;
10072 +       int err;
10073 +       aufs_bindex_t btop, bindex;
10074 +
10075 +       sb = a->dentry->d_sb;
10076 +       if (d_really_is_negative(a->dentry))
10077 +               goto out;
10078 +       /* no dir->i_mutex lock */
10079 +       aufs_read_lock(a->dentry, AuLock_DW); /* noflush */
10080 +
10081 +       dir = d_inode(a->dentry);
10082 +       btop = au_ibtop(dir);
10083 +       bindex = au_br_index(sb, a->brid);
10084 +       if (bindex < btop)
10085 +               goto out_unlock;
10086 +
10087 +       br = au_sbr(sb, bindex);
10088 +       h_path.dentry = au_h_dptr(a->dentry, bindex);
10089 +       if (!h_path.dentry)
10090 +               goto out_unlock;
10091 +       h_path.mnt = au_br_mnt(br);
10092 +       au_dtime_store(&dt, a->dentry, &h_path);
10093 +
10094 +       br = au_sbr(sb, btop);
10095 +       if (!au_br_writable(br->br_perm))
10096 +               goto out_unlock;
10097 +       h_path.dentry = au_h_dptr(a->dentry, btop);
10098 +       h_path.mnt = au_br_mnt(br);
10099 +       err = vfsub_mnt_want_write(h_path.mnt);
10100 +       if (err)
10101 +               goto out_unlock;
10102 +       hdir = au_hi(dir, btop);
10103 +       au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
10104 +       h_dir = au_h_iptr(dir, btop);
10105 +       if (h_dir->i_nlink
10106 +           && timespec64_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
10107 +               dt.dt_h_path = h_path;
10108 +               au_dtime_revert(&dt);
10109 +       }
10110 +       au_hn_inode_unlock(hdir);
10111 +       vfsub_mnt_drop_write(h_path.mnt);
10112 +       au_cpup_attr_timesizes(dir);
10113 +
10114 +out_unlock:
10115 +       aufs_read_unlock(a->dentry, AuLock_DW);
10116 +out:
10117 +       dput(a->dentry);
10118 +       au_nwt_done(&au_sbi(sb)->si_nowait);
10119 +       au_kfree_try_rcu(arg);
10120 +}
10121 +
10122 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
10123 +{
10124 +       int perm, wkq_err;
10125 +       aufs_bindex_t btop;
10126 +       struct au_dir_ts_arg *arg;
10127 +       struct dentry *dentry;
10128 +       struct super_block *sb;
10129 +
10130 +       IMustLock(dir);
10131 +
10132 +       dentry = d_find_any_alias(dir);
10133 +       AuDebugOn(!dentry);
10134 +       sb = dentry->d_sb;
10135 +       btop = au_ibtop(dir);
10136 +       if (btop == bindex) {
10137 +               au_cpup_attr_timesizes(dir);
10138 +               goto out;
10139 +       }
10140 +
10141 +       perm = au_sbr_perm(sb, btop);
10142 +       if (!au_br_writable(perm))
10143 +               goto out;
10144 +
10145 +       arg = kmalloc(sizeof(*arg), GFP_NOFS);
10146 +       if (!arg)
10147 +               goto out;
10148 +
10149 +       arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
10150 +       arg->brid = au_sbr_id(sb, bindex);
10151 +       wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
10152 +       if (unlikely(wkq_err)) {
10153 +               pr_err("wkq %d\n", wkq_err);
10154 +               dput(dentry);
10155 +               au_kfree_try_rcu(arg);
10156 +       }
10157 +
10158 +out:
10159 +       dput(dentry);
10160 +}
10161 +
10162 +/* ---------------------------------------------------------------------- */
10163 +
10164 +static int reopen_dir(struct file *file)
10165 +{
10166 +       int err;
10167 +       unsigned int flags;
10168 +       aufs_bindex_t bindex, btail, btop;
10169 +       struct dentry *dentry, *h_dentry;
10170 +       struct file *h_file;
10171 +
10172 +       /* open all lower dirs */
10173 +       dentry = file->f_path.dentry;
10174 +       btop = au_dbtop(dentry);
10175 +       for (bindex = au_fbtop(file); bindex < btop; bindex++)
10176 +               au_set_h_fptr(file, bindex, NULL);
10177 +       au_set_fbtop(file, btop);
10178 +
10179 +       btail = au_dbtaildir(dentry);
10180 +       for (bindex = au_fbbot_dir(file); btail < bindex; bindex--)
10181 +               au_set_h_fptr(file, bindex, NULL);
10182 +       au_set_fbbot_dir(file, btail);
10183 +
10184 +       flags = vfsub_file_flags(file);
10185 +       for (bindex = btop; bindex <= btail; bindex++) {
10186 +               h_dentry = au_h_dptr(dentry, bindex);
10187 +               if (!h_dentry)
10188 +                       continue;
10189 +               h_file = au_hf_dir(file, bindex);
10190 +               if (h_file)
10191 +                       continue;
10192 +
10193 +               h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10194 +               err = PTR_ERR(h_file);
10195 +               if (IS_ERR(h_file))
10196 +                       goto out; /* close all? */
10197 +               au_set_h_fptr(file, bindex, h_file);
10198 +       }
10199 +       au_update_figen(file);
10200 +       /* todo: necessary? */
10201 +       /* file->f_ra = h_file->f_ra; */
10202 +       err = 0;
10203 +
10204 +out:
10205 +       return err;
10206 +}
10207 +
10208 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
10209 +{
10210 +       int err;
10211 +       aufs_bindex_t bindex, btail;
10212 +       struct dentry *dentry, *h_dentry;
10213 +       struct vfsmount *mnt;
10214 +
10215 +       FiMustWriteLock(file);
10216 +       AuDebugOn(h_file);
10217 +
10218 +       err = 0;
10219 +       mnt = file->f_path.mnt;
10220 +       dentry = file->f_path.dentry;
10221 +       file->f_version = inode_query_iversion(d_inode(dentry));
10222 +       bindex = au_dbtop(dentry);
10223 +       au_set_fbtop(file, bindex);
10224 +       btail = au_dbtaildir(dentry);
10225 +       au_set_fbbot_dir(file, btail);
10226 +       for (; !err && bindex <= btail; bindex++) {
10227 +               h_dentry = au_h_dptr(dentry, bindex);
10228 +               if (!h_dentry)
10229 +                       continue;
10230 +
10231 +               err = vfsub_test_mntns(mnt, h_dentry->d_sb);
10232 +               if (unlikely(err))
10233 +                       break;
10234 +               h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10235 +               if (IS_ERR(h_file)) {
10236 +                       err = PTR_ERR(h_file);
10237 +                       break;
10238 +               }
10239 +               au_set_h_fptr(file, bindex, h_file);
10240 +       }
10241 +       au_update_figen(file);
10242 +       /* todo: necessary? */
10243 +       /* file->f_ra = h_file->f_ra; */
10244 +       if (!err)
10245 +               return 0; /* success */
10246 +
10247 +       /* close all */
10248 +       for (bindex = au_fbtop(file); bindex <= btail; bindex++)
10249 +               au_set_h_fptr(file, bindex, NULL);
10250 +       au_set_fbtop(file, -1);
10251 +       au_set_fbbot_dir(file, -1);
10252 +
10253 +       return err;
10254 +}
10255 +
10256 +static int aufs_open_dir(struct inode *inode __maybe_unused,
10257 +                        struct file *file)
10258 +{
10259 +       int err;
10260 +       struct super_block *sb;
10261 +       struct au_fidir *fidir;
10262 +
10263 +       err = -ENOMEM;
10264 +       sb = file->f_path.dentry->d_sb;
10265 +       si_read_lock(sb, AuLock_FLUSH);
10266 +       fidir = au_fidir_alloc(sb);
10267 +       if (fidir) {
10268 +               struct au_do_open_args args = {
10269 +                       .open   = do_open_dir,
10270 +                       .fidir  = fidir
10271 +               };
10272 +               err = au_do_open(file, &args);
10273 +               if (unlikely(err))
10274 +                       au_kfree_rcu(fidir);
10275 +       }
10276 +       si_read_unlock(sb);
10277 +       return err;
10278 +}
10279 +
10280 +static int aufs_release_dir(struct inode *inode __maybe_unused,
10281 +                           struct file *file)
10282 +{
10283 +       struct au_vdir *vdir_cache;
10284 +       struct au_finfo *finfo;
10285 +       struct au_fidir *fidir;
10286 +       struct au_hfile *hf;
10287 +       aufs_bindex_t bindex, bbot;
10288 +
10289 +       finfo = au_fi(file);
10290 +       fidir = finfo->fi_hdir;
10291 +       if (fidir) {
10292 +               au_hbl_del(&finfo->fi_hlist,
10293 +                          &au_sbi(file->f_path.dentry->d_sb)->si_files);
10294 +               vdir_cache = fidir->fd_vdir_cache; /* lock-free */
10295 +               if (vdir_cache)
10296 +                       au_vdir_free(vdir_cache);
10297 +
10298 +               bindex = finfo->fi_btop;
10299 +               if (bindex >= 0) {
10300 +                       hf = fidir->fd_hfile + bindex;
10301 +                       /*
10302 +                        * calls fput() instead of filp_close(),
10303 +                        * since no dnotify or lock for the lower file.
10304 +                        */
10305 +                       bbot = fidir->fd_bbot;
10306 +                       for (; bindex <= bbot; bindex++, hf++)
10307 +                               if (hf->hf_file)
10308 +                                       au_hfput(hf, /*execed*/0);
10309 +               }
10310 +               au_kfree_rcu(fidir);
10311 +               finfo->fi_hdir = NULL;
10312 +       }
10313 +       au_finfo_fin(file);
10314 +       return 0;
10315 +}
10316 +
10317 +/* ---------------------------------------------------------------------- */
10318 +
10319 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
10320 +{
10321 +       int err;
10322 +       aufs_bindex_t bindex, bbot;
10323 +       struct file *h_file;
10324 +
10325 +       err = 0;
10326 +       bbot = au_fbbot_dir(file);
10327 +       for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10328 +               h_file = au_hf_dir(file, bindex);
10329 +               if (h_file)
10330 +                       err = vfsub_flush(h_file, id);
10331 +       }
10332 +       return err;
10333 +}
10334 +
10335 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
10336 +{
10337 +       return au_do_flush(file, id, au_do_flush_dir);
10338 +}
10339 +
10340 +/* ---------------------------------------------------------------------- */
10341 +
10342 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
10343 +{
10344 +       int err;
10345 +       aufs_bindex_t bbot, bindex;
10346 +       struct inode *inode;
10347 +       struct super_block *sb;
10348 +
10349 +       err = 0;
10350 +       sb = dentry->d_sb;
10351 +       inode = d_inode(dentry);
10352 +       IMustLock(inode);
10353 +       bbot = au_dbbot(dentry);
10354 +       for (bindex = au_dbtop(dentry); !err && bindex <= bbot; bindex++) {
10355 +               struct path h_path;
10356 +
10357 +               if (au_test_ro(sb, bindex, inode))
10358 +                       continue;
10359 +               h_path.dentry = au_h_dptr(dentry, bindex);
10360 +               if (!h_path.dentry)
10361 +                       continue;
10362 +
10363 +               h_path.mnt = au_sbr_mnt(sb, bindex);
10364 +               err = vfsub_fsync(NULL, &h_path, datasync);
10365 +       }
10366 +
10367 +       return err;
10368 +}
10369 +
10370 +static int au_do_fsync_dir(struct file *file, int datasync)
10371 +{
10372 +       int err;
10373 +       aufs_bindex_t bbot, bindex;
10374 +       struct file *h_file;
10375 +       struct super_block *sb;
10376 +       struct inode *inode;
10377 +
10378 +       err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10379 +       if (unlikely(err))
10380 +               goto out;
10381 +
10382 +       inode = file_inode(file);
10383 +       sb = inode->i_sb;
10384 +       bbot = au_fbbot_dir(file);
10385 +       for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10386 +               h_file = au_hf_dir(file, bindex);
10387 +               if (!h_file || au_test_ro(sb, bindex, inode))
10388 +                       continue;
10389 +
10390 +               err = vfsub_fsync(h_file, &h_file->f_path, datasync);
10391 +       }
10392 +
10393 +out:
10394 +       return err;
10395 +}
10396 +
10397 +/*
10398 + * @file may be NULL
10399 + */
10400 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
10401 +                         int datasync)
10402 +{
10403 +       int err;
10404 +       struct dentry *dentry;
10405 +       struct inode *inode;
10406 +       struct super_block *sb;
10407 +
10408 +       err = 0;
10409 +       dentry = file->f_path.dentry;
10410 +       inode = d_inode(dentry);
10411 +       inode_lock(inode);
10412 +       sb = dentry->d_sb;
10413 +       si_noflush_read_lock(sb);
10414 +       if (file)
10415 +               err = au_do_fsync_dir(file, datasync);
10416 +       else {
10417 +               di_write_lock_child(dentry);
10418 +               err = au_do_fsync_dir_no_file(dentry, datasync);
10419 +       }
10420 +       au_cpup_attr_timesizes(inode);
10421 +       di_write_unlock(dentry);
10422 +       if (file)
10423 +               fi_write_unlock(file);
10424 +
10425 +       si_read_unlock(sb);
10426 +       inode_unlock(inode);
10427 +       return err;
10428 +}
10429 +
10430 +/* ---------------------------------------------------------------------- */
10431 +
10432 +static int aufs_iterate_shared(struct file *file, struct dir_context *ctx)
10433 +{
10434 +       int err;
10435 +       struct dentry *dentry;
10436 +       struct inode *inode, *h_inode;
10437 +       struct super_block *sb;
10438 +
10439 +       AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
10440 +
10441 +       dentry = file->f_path.dentry;
10442 +       inode = d_inode(dentry);
10443 +       IMustLock(inode);
10444 +
10445 +       sb = dentry->d_sb;
10446 +       si_read_lock(sb, AuLock_FLUSH);
10447 +       err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10448 +       if (unlikely(err))
10449 +               goto out;
10450 +       err = au_alive_dir(dentry);
10451 +       if (!err)
10452 +               err = au_vdir_init(file);
10453 +       di_downgrade_lock(dentry, AuLock_IR);
10454 +       if (unlikely(err))
10455 +               goto out_unlock;
10456 +
10457 +       h_inode = au_h_iptr(inode, au_ibtop(inode));
10458 +       if (!au_test_nfsd()) {
10459 +               err = au_vdir_fill_de(file, ctx);
10460 +               fsstack_copy_attr_atime(inode, h_inode);
10461 +       } else {
10462 +               /*
10463 +                * nfsd filldir may call lookup_one_len(), vfs_getattr(),
10464 +                * encode_fh() and others.
10465 +                */
10466 +               atomic_inc(&h_inode->i_count);
10467 +               di_read_unlock(dentry, AuLock_IR);
10468 +               si_read_unlock(sb);
10469 +               err = au_vdir_fill_de(file, ctx);
10470 +               fsstack_copy_attr_atime(inode, h_inode);
10471 +               fi_write_unlock(file);
10472 +               iput(h_inode);
10473 +
10474 +               AuTraceErr(err);
10475 +               return err;
10476 +       }
10477 +
10478 +out_unlock:
10479 +       di_read_unlock(dentry, AuLock_IR);
10480 +       fi_write_unlock(file);
10481 +out:
10482 +       si_read_unlock(sb);
10483 +       return err;
10484 +}
10485 +
10486 +/* ---------------------------------------------------------------------- */
10487 +
10488 +#define AuTestEmpty_WHONLY     1
10489 +#define AuTestEmpty_CALLED     (1 << 1)
10490 +#define AuTestEmpty_SHWH       (1 << 2)
10491 +#define au_ftest_testempty(flags, name)        ((flags) & AuTestEmpty_##name)
10492 +#define au_fset_testempty(flags, name) \
10493 +       do { (flags) |= AuTestEmpty_##name; } while (0)
10494 +#define au_fclr_testempty(flags, name) \
10495 +       do { (flags) &= ~AuTestEmpty_##name; } while (0)
10496 +
10497 +#ifndef CONFIG_AUFS_SHWH
10498 +#undef AuTestEmpty_SHWH
10499 +#define AuTestEmpty_SHWH       0
10500 +#endif
10501 +
10502 +struct test_empty_arg {
10503 +       struct dir_context ctx;
10504 +       struct au_nhash *whlist;
10505 +       unsigned int flags;
10506 +       int err;
10507 +       aufs_bindex_t bindex;
10508 +};
10509 +
10510 +static int test_empty_cb(struct dir_context *ctx, const char *__name,
10511 +                        int namelen, loff_t offset __maybe_unused, u64 ino,
10512 +                        unsigned int d_type)
10513 +{
10514 +       struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
10515 +                                                 ctx);
10516 +       char *name = (void *)__name;
10517 +
10518 +       arg->err = 0;
10519 +       au_fset_testempty(arg->flags, CALLED);
10520 +       /* smp_mb(); */
10521 +       if (name[0] == '.'
10522 +           && (namelen == 1 || (name[1] == '.' && namelen == 2)))
10523 +               goto out; /* success */
10524 +
10525 +       if (namelen <= AUFS_WH_PFX_LEN
10526 +           || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
10527 +               if (au_ftest_testempty(arg->flags, WHONLY)
10528 +                   && !au_nhash_test_known_wh(arg->whlist, name, namelen))
10529 +                       arg->err = -ENOTEMPTY;
10530 +               goto out;
10531 +       }
10532 +
10533 +       name += AUFS_WH_PFX_LEN;
10534 +       namelen -= AUFS_WH_PFX_LEN;
10535 +       if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
10536 +               arg->err = au_nhash_append_wh
10537 +                       (arg->whlist, name, namelen, ino, d_type, arg->bindex,
10538 +                        au_ftest_testempty(arg->flags, SHWH));
10539 +
10540 +out:
10541 +       /* smp_mb(); */
10542 +       AuTraceErr(arg->err);
10543 +       return arg->err;
10544 +}
10545 +
10546 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10547 +{
10548 +       int err;
10549 +       struct file *h_file;
10550 +       struct au_branch *br;
10551 +
10552 +       h_file = au_h_open(dentry, arg->bindex,
10553 +                          O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
10554 +                          /*file*/NULL, /*force_wr*/0);
10555 +       err = PTR_ERR(h_file);
10556 +       if (IS_ERR(h_file))
10557 +               goto out;
10558 +
10559 +       err = 0;
10560 +       if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
10561 +           && !file_inode(h_file)->i_nlink)
10562 +               goto out_put;
10563 +
10564 +       do {
10565 +               arg->err = 0;
10566 +               au_fclr_testempty(arg->flags, CALLED);
10567 +               /* smp_mb(); */
10568 +               err = vfsub_iterate_dir(h_file, &arg->ctx);
10569 +               if (err >= 0)
10570 +                       err = arg->err;
10571 +       } while (!err && au_ftest_testempty(arg->flags, CALLED));
10572 +
10573 +out_put:
10574 +       fput(h_file);
10575 +       br = au_sbr(dentry->d_sb, arg->bindex);
10576 +       au_lcnt_dec(&br->br_nfiles);
10577 +out:
10578 +       return err;
10579 +}
10580 +
10581 +struct do_test_empty_args {
10582 +       int *errp;
10583 +       struct dentry *dentry;
10584 +       struct test_empty_arg *arg;
10585 +};
10586 +
10587 +static void call_do_test_empty(void *args)
10588 +{
10589 +       struct do_test_empty_args *a = args;
10590 +       *a->errp = do_test_empty(a->dentry, a->arg);
10591 +}
10592 +
10593 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10594 +{
10595 +       int err, wkq_err;
10596 +       struct dentry *h_dentry;
10597 +       struct inode *h_inode;
10598 +       struct user_namespace *h_userns;
10599 +
10600 +       h_userns = au_sbr_userns(dentry->d_sb, arg->bindex);
10601 +       h_dentry = au_h_dptr(dentry, arg->bindex);
10602 +       h_inode = d_inode(h_dentry);
10603 +       /* todo: i_mode changes anytime? */
10604 +       inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
10605 +       err = au_test_h_perm_sio(h_userns, h_inode, MAY_EXEC | MAY_READ);
10606 +       inode_unlock_shared(h_inode);
10607 +       if (!err)
10608 +               err = do_test_empty(dentry, arg);
10609 +       else {
10610 +               struct do_test_empty_args args = {
10611 +                       .errp   = &err,
10612 +                       .dentry = dentry,
10613 +                       .arg    = arg
10614 +               };
10615 +               unsigned int flags = arg->flags;
10616 +
10617 +               wkq_err = au_wkq_wait(call_do_test_empty, &args);
10618 +               if (unlikely(wkq_err))
10619 +                       err = wkq_err;
10620 +               arg->flags = flags;
10621 +       }
10622 +
10623 +       return err;
10624 +}
10625 +
10626 +int au_test_empty_lower(struct dentry *dentry)
10627 +{
10628 +       int err;
10629 +       unsigned int rdhash;
10630 +       aufs_bindex_t bindex, btop, btail;
10631 +       struct au_nhash whlist;
10632 +       struct test_empty_arg arg = {
10633 +               .ctx = {
10634 +                       .actor = test_empty_cb
10635 +               }
10636 +       };
10637 +       int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10638 +
10639 +       SiMustAnyLock(dentry->d_sb);
10640 +
10641 +       rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10642 +       if (!rdhash)
10643 +               rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10644 +       err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10645 +       if (unlikely(err))
10646 +               goto out;
10647 +
10648 +       arg.flags = 0;
10649 +       arg.whlist = &whlist;
10650 +       btop = au_dbtop(dentry);
10651 +       if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10652 +               au_fset_testempty(arg.flags, SHWH);
10653 +       test_empty = do_test_empty;
10654 +       if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10655 +               test_empty = sio_test_empty;
10656 +       arg.bindex = btop;
10657 +       err = test_empty(dentry, &arg);
10658 +       if (unlikely(err))
10659 +               goto out_whlist;
10660 +
10661 +       au_fset_testempty(arg.flags, WHONLY);
10662 +       btail = au_dbtaildir(dentry);
10663 +       for (bindex = btop + 1; !err && bindex <= btail; bindex++) {
10664 +               struct dentry *h_dentry;
10665 +
10666 +               h_dentry = au_h_dptr(dentry, bindex);
10667 +               if (h_dentry && d_is_positive(h_dentry)) {
10668 +                       arg.bindex = bindex;
10669 +                       err = test_empty(dentry, &arg);
10670 +               }
10671 +       }
10672 +
10673 +out_whlist:
10674 +       au_nhash_wh_free(&whlist);
10675 +out:
10676 +       return err;
10677 +}
10678 +
10679 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10680 +{
10681 +       int err;
10682 +       struct test_empty_arg arg = {
10683 +               .ctx = {
10684 +                       .actor = test_empty_cb
10685 +               }
10686 +       };
10687 +       aufs_bindex_t bindex, btail;
10688 +
10689 +       err = 0;
10690 +       arg.whlist = whlist;
10691 +       arg.flags = AuTestEmpty_WHONLY;
10692 +       if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10693 +               au_fset_testempty(arg.flags, SHWH);
10694 +       btail = au_dbtaildir(dentry);
10695 +       for (bindex = au_dbtop(dentry); !err && bindex <= btail; bindex++) {
10696 +               struct dentry *h_dentry;
10697 +
10698 +               h_dentry = au_h_dptr(dentry, bindex);
10699 +               if (h_dentry && d_is_positive(h_dentry)) {
10700 +                       arg.bindex = bindex;
10701 +                       err = sio_test_empty(dentry, &arg);
10702 +               }
10703 +       }
10704 +
10705 +       return err;
10706 +}
10707 +
10708 +/* ---------------------------------------------------------------------- */
10709 +
10710 +const struct file_operations aufs_dir_fop = {
10711 +       .owner          = THIS_MODULE,
10712 +       .llseek         = default_llseek,
10713 +       .read           = generic_read_dir,
10714 +       .iterate_shared = aufs_iterate_shared,
10715 +       .unlocked_ioctl = aufs_ioctl_dir,
10716 +#ifdef CONFIG_COMPAT
10717 +       .compat_ioctl   = aufs_compat_ioctl_dir,
10718 +#endif
10719 +       .open           = aufs_open_dir,
10720 +       .release        = aufs_release_dir,
10721 +       .flush          = aufs_flush_dir,
10722 +       .fsync          = aufs_fsync_dir
10723 +};
10724 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10725 --- /usr/share/empty/fs/aufs/dir.h      1970-01-01 01:00:00.000000000 +0100
10726 +++ linux/fs/aufs/dir.h 2022-11-05 23:02:18.962555950 +0100
10727 @@ -0,0 +1,134 @@
10728 +/* SPDX-License-Identifier: GPL-2.0 */
10729 +/*
10730 + * Copyright (C) 2005-2022 Junjiro R. Okajima
10731 + *
10732 + * This program is free software; you can redistribute it and/or modify
10733 + * it under the terms of the GNU General Public License as published by
10734 + * the Free Software Foundation; either version 2 of the License, or
10735 + * (at your option) any later version.
10736 + *
10737 + * This program is distributed in the hope that it will be useful,
10738 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10739 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
10740 + * GNU General Public License for more details.
10741 + *
10742 + * You should have received a copy of the GNU General Public License
10743 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
10744 + */
10745 +
10746 +/*
10747 + * directory operations
10748 + */
10749 +
10750 +#ifndef __AUFS_DIR_H__
10751 +#define __AUFS_DIR_H__
10752 +
10753 +#ifdef __KERNEL__
10754 +
10755 +#include <linux/fs.h>
10756 +
10757 +/* ---------------------------------------------------------------------- */
10758 +
10759 +/* need to be faster and smaller */
10760 +
10761 +struct au_nhash {
10762 +       unsigned int            nh_num;
10763 +       struct hlist_head       *nh_head;
10764 +};
10765 +
10766 +struct au_vdir_destr {
10767 +       unsigned char   len;
10768 +       unsigned char   name[];
10769 +} __packed;
10770 +
10771 +struct au_vdir_dehstr {
10772 +       struct hlist_node       hash;
10773 +       struct au_vdir_destr    *str;
10774 +       struct rcu_head         rcu;
10775 +} ____cacheline_aligned_in_smp;
10776 +
10777 +struct au_vdir_de {
10778 +       ino_t                   de_ino;
10779 +       unsigned char           de_type;
10780 +       /* caution: packed */
10781 +       struct au_vdir_destr    de_str;
10782 +} __packed;
10783 +
10784 +struct au_vdir_wh {
10785 +       struct hlist_node       wh_hash;
10786 +#ifdef CONFIG_AUFS_SHWH
10787 +       ino_t                   wh_ino;
10788 +       aufs_bindex_t           wh_bindex;
10789 +       unsigned char           wh_type;
10790 +#else
10791 +       aufs_bindex_t           wh_bindex;
10792 +#endif
10793 +       /* caution: packed */
10794 +       struct au_vdir_destr    wh_str;
10795 +} __packed;
10796 +
10797 +union au_vdir_deblk_p {
10798 +       unsigned char           *deblk;
10799 +       struct au_vdir_de       *de;
10800 +};
10801 +
10802 +struct au_vdir {
10803 +       unsigned char   **vd_deblk;
10804 +       unsigned long   vd_nblk;
10805 +       struct {
10806 +               unsigned long           ul;
10807 +               union au_vdir_deblk_p   p;
10808 +       } vd_last;
10809 +
10810 +       u64             vd_version;
10811 +       unsigned int    vd_deblk_sz;
10812 +       unsigned long   vd_jiffy;
10813 +       struct rcu_head rcu;
10814 +} ____cacheline_aligned_in_smp;
10815 +
10816 +/* ---------------------------------------------------------------------- */
10817 +
10818 +/* dir.c */
10819 +extern const struct file_operations aufs_dir_fop;
10820 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10821 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10822 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10823 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10824 +int au_test_empty_lower(struct dentry *dentry);
10825 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10826 +
10827 +/* vdir.c */
10828 +unsigned int au_rdhash_est(loff_t sz);
10829 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10830 +void au_nhash_wh_free(struct au_nhash *whlist);
10831 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10832 +                           int limit);
10833 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10834 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10835 +                      unsigned int d_type, aufs_bindex_t bindex,
10836 +                      unsigned char shwh);
10837 +void au_vdir_free(struct au_vdir *vdir);
10838 +int au_vdir_init(struct file *file);
10839 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10840 +
10841 +/* ioctl.c */
10842 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10843 +
10844 +#ifdef CONFIG_AUFS_RDU
10845 +/* rdu.c */
10846 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10847 +#ifdef CONFIG_COMPAT
10848 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10849 +                        unsigned long arg);
10850 +#endif
10851 +#else
10852 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10853 +       unsigned int cmd, unsigned long arg)
10854 +#ifdef CONFIG_COMPAT
10855 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10856 +       unsigned int cmd, unsigned long arg)
10857 +#endif
10858 +#endif
10859 +
10860 +#endif /* __KERNEL__ */
10861 +#endif /* __AUFS_DIR_H__ */
10862 diff -urN /usr/share/empty/fs/aufs/dirren.c linux/fs/aufs/dirren.c
10863 --- /usr/share/empty/fs/aufs/dirren.c   1970-01-01 01:00:00.000000000 +0100
10864 +++ linux/fs/aufs/dirren.c      2022-11-05 23:02:18.962555950 +0100
10865 @@ -0,0 +1,1315 @@
10866 +// SPDX-License-Identifier: GPL-2.0
10867 +/*
10868 + * Copyright (C) 2017-2022 Junjiro R. Okajima
10869 + *
10870 + * This program is free software; you can redistribute it and/or modify
10871 + * it under the terms of the GNU General Public License as published by
10872 + * the Free Software Foundation; either version 2 of the License, or
10873 + * (at your option) any later version.
10874 + *
10875 + * This program is distributed in the hope that it will be useful,
10876 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10877 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
10878 + * GNU General Public License for more details.
10879 + *
10880 + * You should have received a copy of the GNU General Public License
10881 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
10882 + */
10883 +
10884 +/*
10885 + * special handling in renaming a directory
10886 + * in order to support looking-up the before-renamed name on the lower readonly
10887 + * branches
10888 + */
10889 +
10890 +#include <linux/byteorder/generic.h>
10891 +#include "aufs.h"
10892 +
10893 +static void au_dr_hino_del(struct au_dr_br *dr, struct au_dr_hino *ent)
10894 +{
10895 +       int idx;
10896 +
10897 +       idx = au_dr_ihash(ent->dr_h_ino);
10898 +       au_hbl_del(&ent->dr_hnode, dr->dr_h_ino + idx);
10899 +}
10900 +
10901 +static int au_dr_hino_test_empty(struct au_dr_br *dr)
10902 +{
10903 +       int ret, i;
10904 +       struct hlist_bl_head *hbl;
10905 +
10906 +       ret = 1;
10907 +       for (i = 0; ret && i < AuDirren_NHASH; i++) {
10908 +               hbl = dr->dr_h_ino + i;
10909 +               hlist_bl_lock(hbl);
10910 +               ret &= hlist_bl_empty(hbl);
10911 +               hlist_bl_unlock(hbl);
10912 +       }
10913 +
10914 +       return ret;
10915 +}
10916 +
10917 +static struct au_dr_hino *au_dr_hino_find(struct au_dr_br *dr, ino_t ino)
10918 +{
10919 +       struct au_dr_hino *found, *ent;
10920 +       struct hlist_bl_head *hbl;
10921 +       struct hlist_bl_node *pos;
10922 +       int idx;
10923 +
10924 +       found = NULL;
10925 +       idx = au_dr_ihash(ino);
10926 +       hbl = dr->dr_h_ino + idx;
10927 +       hlist_bl_lock(hbl);
10928 +       hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10929 +               if (ent->dr_h_ino == ino) {
10930 +                       found = ent;
10931 +                       break;
10932 +               }
10933 +       hlist_bl_unlock(hbl);
10934 +
10935 +       return found;
10936 +}
10937 +
10938 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t ino,
10939 +                       struct au_dr_hino *add_ent)
10940 +{
10941 +       int found, idx;
10942 +       struct hlist_bl_head *hbl;
10943 +       struct hlist_bl_node *pos;
10944 +       struct au_dr_hino *ent;
10945 +
10946 +       found = 0;
10947 +       idx = au_dr_ihash(ino);
10948 +       hbl = dr->dr_h_ino + idx;
10949 +#if 0 /* debug print */
10950 +       {
10951 +               struct hlist_bl_node *tmp;
10952 +
10953 +               hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10954 +                       AuDbg("hi%llu\n", (unsigned long long)ent->dr_h_ino);
10955 +       }
10956 +#endif
10957 +       hlist_bl_lock(hbl);
10958 +       hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10959 +               if (ent->dr_h_ino == ino) {
10960 +                       found = 1;
10961 +                       break;
10962 +               }
10963 +       if (!found && add_ent)
10964 +               hlist_bl_add_head(&add_ent->dr_hnode, hbl);
10965 +       hlist_bl_unlock(hbl);
10966 +
10967 +       if (!found && add_ent)
10968 +               AuDbg("i%llu added\n", (unsigned long long)add_ent->dr_h_ino);
10969 +
10970 +       return found;
10971 +}
10972 +
10973 +void au_dr_hino_free(struct au_dr_br *dr)
10974 +{
10975 +       int i;
10976 +       struct hlist_bl_head *hbl;
10977 +       struct hlist_bl_node *pos, *tmp;
10978 +       struct au_dr_hino *ent;
10979 +
10980 +       /* SiMustWriteLock(sb); */
10981 +
10982 +       for (i = 0; i < AuDirren_NHASH; i++) {
10983 +               hbl = dr->dr_h_ino + i;
10984 +               /* no spinlock since sbinfo must be write-locked */
10985 +               hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10986 +                       au_kfree_rcu(ent);
10987 +               INIT_HLIST_BL_HEAD(hbl);
10988 +       }
10989 +}
10990 +
10991 +/* returns the number of inodes or an error */
10992 +static int au_dr_hino_store(struct super_block *sb, struct au_branch *br,
10993 +                           struct file *hinofile)
10994 +{
10995 +       int err, i;
10996 +       ssize_t ssz;
10997 +       loff_t pos, oldsize;
10998 +       __be64 u64;
10999 +       struct inode *hinoinode;
11000 +       struct hlist_bl_head *hbl;
11001 +       struct hlist_bl_node *n1, *n2;
11002 +       struct au_dr_hino *ent;
11003 +
11004 +       SiMustWriteLock(sb);
11005 +       AuDebugOn(!au_br_writable(br->br_perm));
11006 +
11007 +       hinoinode = file_inode(hinofile);
11008 +       oldsize = i_size_read(hinoinode);
11009 +
11010 +       err = 0;
11011 +       pos = 0;
11012 +       hbl = br->br_dirren.dr_h_ino;
11013 +       for (i = 0; !err && i < AuDirren_NHASH; i++, hbl++) {
11014 +               /* no bit-lock since sbinfo must be write-locked */
11015 +               hlist_bl_for_each_entry_safe(ent, n1, n2, hbl, dr_hnode) {
11016 +                       AuDbg("hi%llu, %pD2\n",
11017 +                             (unsigned long long)ent->dr_h_ino, hinofile);
11018 +                       u64 = cpu_to_be64(ent->dr_h_ino);
11019 +                       ssz = vfsub_write_k(hinofile, &u64, sizeof(u64), &pos);
11020 +                       if (ssz == sizeof(u64))
11021 +                               continue;
11022 +
11023 +                       /* write error */
11024 +                       pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11025 +                       err = -ENOSPC;
11026 +                       if (ssz < 0)
11027 +                               err = ssz;
11028 +                       break;
11029 +               }
11030 +       }
11031 +       /* regardless the error */
11032 +       if (pos < oldsize) {
11033 +               err = vfsub_trunc(&hinofile->f_path, pos, /*attr*/0, hinofile);
11034 +               AuTraceErr(err);
11035 +       }
11036 +
11037 +       AuTraceErr(err);
11038 +       return err;
11039 +}
11040 +
11041 +static int au_dr_hino_load(struct au_dr_br *dr, struct file *hinofile)
11042 +{
11043 +       int err, hidx;
11044 +       ssize_t ssz;
11045 +       size_t sz, n;
11046 +       loff_t pos;
11047 +       uint64_t u64;
11048 +       struct au_dr_hino *ent;
11049 +       struct inode *hinoinode;
11050 +       struct hlist_bl_head *hbl;
11051 +
11052 +       err = 0;
11053 +       pos = 0;
11054 +       hbl = dr->dr_h_ino;
11055 +       hinoinode = file_inode(hinofile);
11056 +       sz = i_size_read(hinoinode);
11057 +       AuDebugOn(sz % sizeof(u64));
11058 +       n = sz / sizeof(u64);
11059 +       while (n--) {
11060 +               ssz = vfsub_read_k(hinofile, &u64, sizeof(u64), &pos);
11061 +               if (unlikely(ssz != sizeof(u64))) {
11062 +                       pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11063 +                       err = -EINVAL;
11064 +                       if (ssz < 0)
11065 +                               err = ssz;
11066 +                       goto out_free;
11067 +               }
11068 +
11069 +               ent = kmalloc(sizeof(*ent), GFP_NOFS);
11070 +               if (!ent) {
11071 +                       err = -ENOMEM;
11072 +                       AuTraceErr(err);
11073 +                       goto out_free;
11074 +               }
11075 +               ent->dr_h_ino = be64_to_cpu((__force __be64)u64);
11076 +               AuDbg("hi%llu, %pD2\n",
11077 +                     (unsigned long long)ent->dr_h_ino, hinofile);
11078 +               hidx = au_dr_ihash(ent->dr_h_ino);
11079 +               au_hbl_add(&ent->dr_hnode, hbl + hidx);
11080 +       }
11081 +       goto out; /* success */
11082 +
11083 +out_free:
11084 +       au_dr_hino_free(dr);
11085 +out:
11086 +       AuTraceErr(err);
11087 +       return err;
11088 +}
11089 +
11090 +/*
11091 + * @bindex/@br is a switch to distinguish whether suspending hnotify or not.
11092 + * @path is a switch to distinguish load and store.
11093 + */
11094 +static int au_dr_hino(struct super_block *sb, aufs_bindex_t bindex,
11095 +                     struct au_branch *br, const struct path *path)
11096 +{
11097 +       int err, flags;
11098 +       unsigned char load, suspend;
11099 +       struct file *hinofile;
11100 +       struct au_hinode *hdir;
11101 +       struct inode *dir, *delegated;
11102 +       struct path hinopath;
11103 +       struct qstr hinoname = QSTR_INIT(AUFS_WH_DR_BRHINO,
11104 +                                        sizeof(AUFS_WH_DR_BRHINO) - 1);
11105 +
11106 +       AuDebugOn(bindex < 0 && !br);
11107 +       AuDebugOn(bindex >= 0 && br);
11108 +
11109 +       err = -EINVAL;
11110 +       suspend = !br;
11111 +       if (suspend)
11112 +               br = au_sbr(sb, bindex);
11113 +       load = !!path;
11114 +       if (!load) {
11115 +               path = &br->br_path;
11116 +               AuDebugOn(!au_br_writable(br->br_perm));
11117 +               if (unlikely(!au_br_writable(br->br_perm)))
11118 +                       goto out;
11119 +       }
11120 +
11121 +       hdir = NULL;
11122 +       if (suspend) {
11123 +               dir = d_inode(sb->s_root);
11124 +               hdir = au_hinode(au_ii(dir), bindex);
11125 +               dir = hdir->hi_inode;
11126 +               au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11127 +       } else {
11128 +               dir = d_inode(path->dentry);
11129 +               inode_lock_nested(dir, AuLsc_I_CHILD);
11130 +       }
11131 +       hinopath.mnt = path->mnt;
11132 +       hinopath.dentry = vfsub_lkup_one(&hinoname, (struct path *)path);
11133 +       err = PTR_ERR(hinopath.dentry);
11134 +       if (IS_ERR(hinopath.dentry))
11135 +               goto out_unlock;
11136 +
11137 +       err = 0;
11138 +       flags = O_RDONLY;
11139 +       if (load) {
11140 +               if (d_is_negative(hinopath.dentry))
11141 +                       goto out_dput; /* success */
11142 +       } else {
11143 +               if (au_dr_hino_test_empty(&br->br_dirren)) {
11144 +                       if (d_is_positive(hinopath.dentry)) {
11145 +                               delegated = NULL;
11146 +                               err = vfsub_unlink(dir, &hinopath, &delegated,
11147 +                                                  /*force*/0);
11148 +                               AuTraceErr(err);
11149 +                               if (unlikely(err))
11150 +                                       pr_err("ignored err %d, %pd2\n",
11151 +                                              err, hinopath.dentry);
11152 +                               if (unlikely(err == -EWOULDBLOCK))
11153 +                                       iput(delegated);
11154 +                               err = 0;
11155 +                       }
11156 +                       goto out_dput;
11157 +               } else if (!d_is_positive(hinopath.dentry)) {
11158 +                       err = vfsub_create(dir, &hinopath, 0600,
11159 +                                          /*want_excl*/false);
11160 +                       AuTraceErr(err);
11161 +                       if (unlikely(err))
11162 +                               goto out_dput;
11163 +               }
11164 +               flags = O_WRONLY;
11165 +       }
11166 +       hinofile = vfsub_dentry_open(&hinopath, flags);
11167 +       if (suspend)
11168 +               au_hn_inode_unlock(hdir);
11169 +       else
11170 +               inode_unlock(dir);
11171 +       dput(hinopath.dentry);
11172 +       AuTraceErrPtr(hinofile);
11173 +       if (IS_ERR(hinofile)) {
11174 +               err = PTR_ERR(hinofile);
11175 +               goto out;
11176 +       }
11177 +
11178 +       if (load)
11179 +               err = au_dr_hino_load(&br->br_dirren, hinofile);
11180 +       else
11181 +               err = au_dr_hino_store(sb, br, hinofile);
11182 +       fput(hinofile);
11183 +       goto out;
11184 +
11185 +out_dput:
11186 +       dput(hinopath.dentry);
11187 +out_unlock:
11188 +       if (suspend)
11189 +               au_hn_inode_unlock(hdir);
11190 +       else
11191 +               inode_unlock(dir);
11192 +out:
11193 +       AuTraceErr(err);
11194 +       return err;
11195 +}
11196 +
11197 +/* ---------------------------------------------------------------------- */
11198 +
11199 +static int au_dr_brid_init(struct au_dr_brid *brid, const struct path *path)
11200 +{
11201 +       int err;
11202 +       struct kstatfs kstfs;
11203 +       dev_t dev;
11204 +       struct dentry *dentry;
11205 +       struct super_block *sb;
11206 +
11207 +       err = vfs_statfs((void *)path, &kstfs);
11208 +       AuTraceErr(err);
11209 +       if (unlikely(err))
11210 +               goto out;
11211 +
11212 +       /* todo: support for UUID */
11213 +
11214 +       if (kstfs.f_fsid.val[0] || kstfs.f_fsid.val[1]) {
11215 +               brid->type = AuBrid_FSID;
11216 +               brid->fsid = kstfs.f_fsid;
11217 +       } else {
11218 +               dentry = path->dentry;
11219 +               sb = dentry->d_sb;
11220 +               dev = sb->s_dev;
11221 +               if (dev) {
11222 +                       brid->type = AuBrid_DEV;
11223 +                       brid->dev = dev;
11224 +               }
11225 +       }
11226 +
11227 +out:
11228 +       return err;
11229 +}
11230 +
11231 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
11232 +                 const struct path *path)
11233 +{
11234 +       int err, i;
11235 +       struct au_dr_br *dr;
11236 +       struct hlist_bl_head *hbl;
11237 +
11238 +       dr = &br->br_dirren;
11239 +       hbl = dr->dr_h_ino;
11240 +       for (i = 0; i < AuDirren_NHASH; i++, hbl++)
11241 +               INIT_HLIST_BL_HEAD(hbl);
11242 +
11243 +       err = au_dr_brid_init(&dr->dr_brid, path);
11244 +       if (unlikely(err))
11245 +               goto out;
11246 +
11247 +       if (au_opt_test(au_mntflags(sb), DIRREN))
11248 +               err = au_dr_hino(sb, /*bindex*/-1, br, path);
11249 +
11250 +out:
11251 +       AuTraceErr(err);
11252 +       return err;
11253 +}
11254 +
11255 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br)
11256 +{
11257 +       int err;
11258 +
11259 +       err = 0;
11260 +       if (au_br_writable(br->br_perm))
11261 +               err = au_dr_hino(sb, /*bindex*/-1, br, /*path*/NULL);
11262 +       if (!err)
11263 +               au_dr_hino_free(&br->br_dirren);
11264 +
11265 +       return err;
11266 +}
11267 +
11268 +/* ---------------------------------------------------------------------- */
11269 +
11270 +static int au_brid_str(struct au_dr_brid *brid, struct inode *h_inode,
11271 +                      char *buf, size_t sz)
11272 +{
11273 +       int err;
11274 +       unsigned int major, minor;
11275 +       char *p;
11276 +
11277 +       p = buf;
11278 +       err = snprintf(p, sz, "%d_", brid->type);
11279 +       AuDebugOn(err > sz);
11280 +       p += err;
11281 +       sz -= err;
11282 +       switch (brid->type) {
11283 +       case AuBrid_Unset:
11284 +               return -EINVAL;
11285 +       case AuBrid_UUID:
11286 +               err = snprintf(p, sz, "%pU", brid->uuid.b);
11287 +               break;
11288 +       case AuBrid_FSID:
11289 +               err = snprintf(p, sz, "%08x-%08x",
11290 +                              brid->fsid.val[0], brid->fsid.val[1]);
11291 +               break;
11292 +       case AuBrid_DEV:
11293 +               major = MAJOR(brid->dev);
11294 +               minor = MINOR(brid->dev);
11295 +               if (major <= 0xff && minor <= 0xff)
11296 +                       err = snprintf(p, sz, "%02x%02x", major, minor);
11297 +               else
11298 +                       err = snprintf(p, sz, "%03x:%05x", major, minor);
11299 +               break;
11300 +       }
11301 +       AuDebugOn(err > sz);
11302 +       p += err;
11303 +       sz -= err;
11304 +       err = snprintf(p, sz, "_%llu", (unsigned long long)h_inode->i_ino);
11305 +       AuDebugOn(err > sz);
11306 +       p += err;
11307 +       sz -= err;
11308 +
11309 +       return p - buf;
11310 +}
11311 +
11312 +static int au_drinfo_name(struct au_branch *br, char *name, int len)
11313 +{
11314 +       int rlen;
11315 +       struct dentry *br_dentry;
11316 +       struct inode *br_inode;
11317 +
11318 +       br_dentry = au_br_dentry(br);
11319 +       br_inode = d_inode(br_dentry);
11320 +       rlen = au_brid_str(&br->br_dirren.dr_brid, br_inode, name, len);
11321 +       AuDebugOn(rlen >= AUFS_DIRREN_ENV_VAL_SZ);
11322 +       AuDebugOn(rlen > len);
11323 +
11324 +       return rlen;
11325 +}
11326 +
11327 +/* ---------------------------------------------------------------------- */
11328 +
11329 +/*
11330 + * from the given @h_dentry, construct drinfo at @*fdata.
11331 + * when the size of @*fdata is not enough, reallocate and return new @fdata and
11332 + * @allocated.
11333 + */
11334 +static int au_drinfo_construct(struct au_drinfo_fdata **fdata,
11335 +                              struct dentry *h_dentry,
11336 +                              unsigned char *allocated)
11337 +{
11338 +       int err, v;
11339 +       struct au_drinfo_fdata *f, *p;
11340 +       struct au_drinfo *drinfo;
11341 +       struct inode *h_inode;
11342 +       struct qstr *qname;
11343 +
11344 +       err = 0;
11345 +       f = *fdata;
11346 +       h_inode = d_inode(h_dentry);
11347 +       qname = &h_dentry->d_name;
11348 +       drinfo = &f->drinfo;
11349 +       drinfo->ino = (__force uint64_t)cpu_to_be64(h_inode->i_ino);
11350 +       drinfo->oldnamelen = qname->len;
11351 +       if (*allocated < sizeof(*f) + qname->len) {
11352 +               v = roundup_pow_of_two(*allocated + qname->len);
11353 +               p = au_krealloc(f, v, GFP_NOFS, /*may_shrink*/0);
11354 +               if (unlikely(!p)) {
11355 +                       err = -ENOMEM;
11356 +                       AuTraceErr(err);
11357 +                       goto out;
11358 +               }
11359 +               f = p;
11360 +               *fdata = f;
11361 +               *allocated = v;
11362 +               drinfo = &f->drinfo;
11363 +       }
11364 +       memcpy(drinfo->oldname, qname->name, qname->len);
11365 +       AuDbg("i%llu, %.*s\n",
11366 +             be64_to_cpu((__force __be64)drinfo->ino), drinfo->oldnamelen,
11367 +             drinfo->oldname);
11368 +
11369 +out:
11370 +       AuTraceErr(err);
11371 +       return err;
11372 +}
11373 +
11374 +/* callers have to free the return value */
11375 +static struct au_drinfo *au_drinfo_read_k(struct file *file, ino_t h_ino)
11376 +{
11377 +       struct au_drinfo *ret, *drinfo;
11378 +       struct au_drinfo_fdata fdata;
11379 +       int len;
11380 +       loff_t pos;
11381 +       ssize_t ssz;
11382 +
11383 +       ret = ERR_PTR(-EIO);
11384 +       pos = 0;
11385 +       ssz = vfsub_read_k(file, &fdata, sizeof(fdata), &pos);
11386 +       if (unlikely(ssz != sizeof(fdata))) {
11387 +               AuIOErr("ssz %zd, %u, %pD2\n",
11388 +                       ssz, (unsigned int)sizeof(fdata), file);
11389 +               goto out;
11390 +       }
11391 +
11392 +       fdata.magic = ntohl((__force __be32)fdata.magic);
11393 +       switch (fdata.magic) {
11394 +       case AUFS_DRINFO_MAGIC_V1:
11395 +               break;
11396 +       default:
11397 +               AuIOErr("magic-num 0x%x, 0x%x, %pD2\n",
11398 +                       fdata.magic, AUFS_DRINFO_MAGIC_V1, file);
11399 +               goto out;
11400 +       }
11401 +
11402 +       drinfo = &fdata.drinfo;
11403 +       len = drinfo->oldnamelen;
11404 +       if (!len) {
11405 +               AuIOErr("broken drinfo %pD2\n", file);
11406 +               goto out;
11407 +       }
11408 +
11409 +       ret = NULL;
11410 +       drinfo->ino = be64_to_cpu((__force __be64)drinfo->ino);
11411 +       if (unlikely(h_ino && drinfo->ino != h_ino)) {
11412 +               AuDbg("ignored i%llu, i%llu, %pD2\n",
11413 +                     (unsigned long long)drinfo->ino,
11414 +                     (unsigned long long)h_ino, file);
11415 +               goto out; /* success */
11416 +       }
11417 +
11418 +       ret = kmalloc(sizeof(*ret) + len, GFP_NOFS);
11419 +       if (unlikely(!ret)) {
11420 +               ret = ERR_PTR(-ENOMEM);
11421 +               AuTraceErrPtr(ret);
11422 +               goto out;
11423 +       }
11424 +
11425 +       *ret = *drinfo;
11426 +       ssz = vfsub_read_k(file, (void *)ret->oldname, len, &pos);
11427 +       if (unlikely(ssz != len)) {
11428 +               au_kfree_rcu(ret);
11429 +               ret = ERR_PTR(-EIO);
11430 +               AuIOErr("ssz %zd, %u, %pD2\n", ssz, len, file);
11431 +               goto out;
11432 +       }
11433 +
11434 +       AuDbg("oldname %.*s\n", ret->oldnamelen, ret->oldname);
11435 +
11436 +out:
11437 +       return ret;
11438 +}
11439 +
11440 +/* ---------------------------------------------------------------------- */
11441 +
11442 +/* in order to be revertible */
11443 +struct au_drinfo_rev_elm {
11444 +       int                     created;
11445 +       struct dentry           *info_dentry;
11446 +       struct au_drinfo        *info_last;
11447 +};
11448 +
11449 +struct au_drinfo_rev {
11450 +       unsigned char                   already;
11451 +       aufs_bindex_t                   nelm;
11452 +       struct au_drinfo_rev_elm        elm[];
11453 +};
11454 +
11455 +/* todo: isn't it too large? */
11456 +struct au_drinfo_store {
11457 +       struct path h_ppath;
11458 +       struct dentry *h_dentry;
11459 +       struct au_drinfo_fdata *fdata;
11460 +       char *infoname;                 /* inside of whname, just after PFX */
11461 +       char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ];
11462 +       aufs_bindex_t btgt, btail;
11463 +       unsigned char no_sio,
11464 +               allocated,              /* current size of *fdata */
11465 +               infonamelen,            /* room size for p */
11466 +               whnamelen,              /* length of the generated name */
11467 +               renameback;             /* renamed back */
11468 +};
11469 +
11470 +/* on rename(2) error, the caller should revert it using @elm */
11471 +static int au_drinfo_do_store(struct au_drinfo_store *w,
11472 +                             struct au_drinfo_rev_elm *elm)
11473 +{
11474 +       int err, len;
11475 +       ssize_t ssz;
11476 +       loff_t pos;
11477 +       struct path infopath = {
11478 +               .mnt = w->h_ppath.mnt
11479 +       };
11480 +       struct inode *h_dir, *h_inode, *delegated;
11481 +       struct file *infofile;
11482 +       struct qstr *qname;
11483 +
11484 +       AuDebugOn(elm
11485 +                 && memcmp(elm, page_address(ZERO_PAGE(0)), sizeof(*elm)));
11486 +
11487 +       infopath.dentry = vfsub_lookup_one_len(w->whname, &w->h_ppath,
11488 +                                              w->whnamelen);
11489 +       AuTraceErrPtr(infopath.dentry);
11490 +       if (IS_ERR(infopath.dentry)) {
11491 +               err = PTR_ERR(infopath.dentry);
11492 +               goto out;
11493 +       }
11494 +
11495 +       err = 0;
11496 +       h_dir = d_inode(w->h_ppath.dentry);
11497 +       if (elm && d_is_negative(infopath.dentry)) {
11498 +               err = vfsub_create(h_dir, &infopath, 0600, /*want_excl*/true);
11499 +               AuTraceErr(err);
11500 +               if (unlikely(err))
11501 +                       goto out_dput;
11502 +               elm->created = 1;
11503 +               elm->info_dentry = dget(infopath.dentry);
11504 +       }
11505 +
11506 +       infofile = vfsub_dentry_open(&infopath, O_RDWR);
11507 +       AuTraceErrPtr(infofile);
11508 +       if (IS_ERR(infofile)) {
11509 +               err = PTR_ERR(infofile);
11510 +               goto out_dput;
11511 +       }
11512 +
11513 +       h_inode = d_inode(infopath.dentry);
11514 +       if (elm && i_size_read(h_inode)) {
11515 +               h_inode = d_inode(w->h_dentry);
11516 +               elm->info_last = au_drinfo_read_k(infofile, h_inode->i_ino);
11517 +               AuTraceErrPtr(elm->info_last);
11518 +               if (IS_ERR(elm->info_last)) {
11519 +                       err = PTR_ERR(elm->info_last);
11520 +                       elm->info_last = NULL;
11521 +                       AuDebugOn(elm->info_dentry);
11522 +                       goto out_fput;
11523 +               }
11524 +       }
11525 +
11526 +       if (elm && w->renameback) {
11527 +               delegated = NULL;
11528 +               err = vfsub_unlink(h_dir, &infopath, &delegated, /*force*/0);
11529 +               AuTraceErr(err);
11530 +               if (unlikely(err == -EWOULDBLOCK))
11531 +                       iput(delegated);
11532 +               goto out_fput;
11533 +       }
11534 +
11535 +       pos = 0;
11536 +       qname = &w->h_dentry->d_name;
11537 +       len = sizeof(*w->fdata) + qname->len;
11538 +       if (!elm)
11539 +               len = sizeof(*w->fdata) + w->fdata->drinfo.oldnamelen;
11540 +       ssz = vfsub_write_k(infofile, w->fdata, len, &pos);
11541 +       if (ssz == len) {
11542 +               AuDbg("hi%llu, %.*s\n", w->fdata->drinfo.ino,
11543 +                     w->fdata->drinfo.oldnamelen, w->fdata->drinfo.oldname);
11544 +               goto out_fput; /* success */
11545 +       } else {
11546 +               err = -EIO;
11547 +               if (ssz < 0)
11548 +                       err = ssz;
11549 +               /* the caller should revert it using @elm */
11550 +       }
11551 +
11552 +out_fput:
11553 +       fput(infofile);
11554 +out_dput:
11555 +       dput(infopath.dentry);
11556 +out:
11557 +       AuTraceErr(err);
11558 +       return err;
11559 +}
11560 +
11561 +struct au_call_drinfo_do_store_args {
11562 +       int *errp;
11563 +       struct au_drinfo_store *w;
11564 +       struct au_drinfo_rev_elm *elm;
11565 +};
11566 +
11567 +static void au_call_drinfo_do_store(void *args)
11568 +{
11569 +       struct au_call_drinfo_do_store_args *a = args;
11570 +
11571 +       *a->errp = au_drinfo_do_store(a->w, a->elm);
11572 +}
11573 +
11574 +static int au_drinfo_store_sio(struct au_drinfo_store *w,
11575 +                              struct au_drinfo_rev_elm *elm)
11576 +{
11577 +       int err, wkq_err;
11578 +
11579 +       if (w->no_sio)
11580 +               err = au_drinfo_do_store(w, elm);
11581 +       else {
11582 +               struct au_call_drinfo_do_store_args a = {
11583 +                       .errp   = &err,
11584 +                       .w      = w,
11585 +                       .elm    = elm
11586 +               };
11587 +               wkq_err = au_wkq_wait(au_call_drinfo_do_store, &a);
11588 +               if (unlikely(wkq_err))
11589 +                       err = wkq_err;
11590 +       }
11591 +       AuTraceErr(err);
11592 +
11593 +       return err;
11594 +}
11595 +
11596 +static int au_drinfo_store_work_init(struct au_drinfo_store *w,
11597 +                                    aufs_bindex_t btgt)
11598 +{
11599 +       int err;
11600 +
11601 +       memset(w, 0, sizeof(*w));
11602 +       w->allocated = roundup_pow_of_two(sizeof(*w->fdata) + 40);
11603 +       strcpy(w->whname, AUFS_WH_DR_INFO_PFX);
11604 +       w->infoname = w->whname + sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11605 +       w->infonamelen = sizeof(w->whname) - sizeof(AUFS_WH_DR_INFO_PFX);
11606 +       w->btgt = btgt;
11607 +       w->no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
11608 +
11609 +       err = -ENOMEM;
11610 +       w->fdata = kcalloc(1, w->allocated, GFP_NOFS);
11611 +       if (unlikely(!w->fdata)) {
11612 +               AuTraceErr(err);
11613 +               goto out;
11614 +       }
11615 +       w->fdata->magic = (__force uint32_t)htonl(AUFS_DRINFO_MAGIC_V1);
11616 +       err = 0;
11617 +
11618 +out:
11619 +       return err;
11620 +}
11621 +
11622 +static void au_drinfo_store_work_fin(struct au_drinfo_store *w)
11623 +{
11624 +       au_kfree_rcu(w->fdata);
11625 +}
11626 +
11627 +static void au_drinfo_store_rev(struct au_drinfo_rev *rev,
11628 +                               struct au_drinfo_store *w)
11629 +{
11630 +       struct au_drinfo_rev_elm *elm;
11631 +       struct inode *h_dir, *delegated;
11632 +       int err, nelm;
11633 +       struct path infopath = {
11634 +               .mnt = w->h_ppath.mnt
11635 +       };
11636 +
11637 +       h_dir = d_inode(w->h_ppath.dentry);
11638 +       IMustLock(h_dir);
11639 +
11640 +       err = 0;
11641 +       elm = rev->elm;
11642 +       for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11643 +               AuDebugOn(elm->created && elm->info_last);
11644 +               if (elm->created) {
11645 +                       AuDbg("here\n");
11646 +                       delegated = NULL;
11647 +                       infopath.dentry = elm->info_dentry;
11648 +                       err = vfsub_unlink(h_dir, &infopath, &delegated,
11649 +                                          !w->no_sio);
11650 +                       AuTraceErr(err);
11651 +                       if (unlikely(err == -EWOULDBLOCK))
11652 +                               iput(delegated);
11653 +                       dput(elm->info_dentry);
11654 +               } else if (elm->info_last) {
11655 +                       AuDbg("here\n");
11656 +                       w->fdata->drinfo = *elm->info_last;
11657 +                       memcpy(w->fdata->drinfo.oldname,
11658 +                              elm->info_last->oldname,
11659 +                              elm->info_last->oldnamelen);
11660 +                       err = au_drinfo_store_sio(w, /*elm*/NULL);
11661 +                       au_kfree_rcu(elm->info_last);
11662 +               }
11663 +               if (unlikely(err))
11664 +                       AuIOErr("%d, %s\n", err, w->whname);
11665 +               /* go on even if err */
11666 +       }
11667 +}
11668 +
11669 +/* caller has to call au_dr_rename_fin() later */
11670 +static int au_drinfo_store(struct dentry *dentry, aufs_bindex_t btgt,
11671 +                          struct qstr *dst_name, void *_rev)
11672 +{
11673 +       int err, sz, nelm;
11674 +       aufs_bindex_t bindex, btail;
11675 +       struct au_drinfo_store work;
11676 +       struct au_drinfo_rev *rev, **p;
11677 +       struct au_drinfo_rev_elm *elm;
11678 +       struct super_block *sb;
11679 +       struct au_branch *br;
11680 +       struct au_hinode *hdir;
11681 +
11682 +       err = au_drinfo_store_work_init(&work, btgt);
11683 +       AuTraceErr(err);
11684 +       if (unlikely(err))
11685 +               goto out;
11686 +
11687 +       err = -ENOMEM;
11688 +       btail = au_dbtaildir(dentry);
11689 +       nelm = btail - btgt;
11690 +       sz = sizeof(*rev) + sizeof(*elm) * nelm;
11691 +       rev = kcalloc(1, sz, GFP_NOFS);
11692 +       if (unlikely(!rev)) {
11693 +               AuTraceErr(err);
11694 +               goto out_args;
11695 +       }
11696 +       rev->nelm = nelm;
11697 +       elm = rev->elm;
11698 +       p = _rev;
11699 +       *p = rev;
11700 +
11701 +       err = 0;
11702 +       sb = dentry->d_sb;
11703 +       work.h_ppath.dentry = au_h_dptr(dentry, btgt);
11704 +       work.h_ppath.mnt = au_sbr_mnt(sb, btgt);
11705 +       hdir = au_hi(d_inode(dentry), btgt);
11706 +       au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11707 +       for (bindex = btgt + 1; bindex <= btail; bindex++, elm++) {
11708 +               work.h_dentry = au_h_dptr(dentry, bindex);
11709 +               if (!work.h_dentry)
11710 +                       continue;
11711 +
11712 +               err = au_drinfo_construct(&work.fdata, work.h_dentry,
11713 +                                         &work.allocated);
11714 +               AuTraceErr(err);
11715 +               if (unlikely(err))
11716 +                       break;
11717 +
11718 +               work.renameback = au_qstreq(&work.h_dentry->d_name, dst_name);
11719 +               br = au_sbr(sb, bindex);
11720 +               work.whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11721 +               work.whnamelen += au_drinfo_name(br, work.infoname,
11722 +                                                work.infonamelen);
11723 +               AuDbg("whname %.*s, i%llu, %.*s\n",
11724 +                     work.whnamelen, work.whname,
11725 +                     be64_to_cpu((__force __be64)work.fdata->drinfo.ino),
11726 +                     work.fdata->drinfo.oldnamelen,
11727 +                     work.fdata->drinfo.oldname);
11728 +
11729 +               err = au_drinfo_store_sio(&work, elm);
11730 +               AuTraceErr(err);
11731 +               if (unlikely(err))
11732 +                       break;
11733 +       }
11734 +       if (unlikely(err)) {
11735 +               /* revert all drinfo */
11736 +               au_drinfo_store_rev(rev, &work);
11737 +               au_kfree_try_rcu(rev);
11738 +               *p = NULL;
11739 +       }
11740 +       au_hn_inode_unlock(hdir);
11741 +
11742 +out_args:
11743 +       au_drinfo_store_work_fin(&work);
11744 +out:
11745 +       return err;
11746 +}
11747 +
11748 +/* ---------------------------------------------------------------------- */
11749 +
11750 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
11751 +                struct qstr *dst_name, void *_rev)
11752 +{
11753 +       int err, already;
11754 +       ino_t ino;
11755 +       struct super_block *sb;
11756 +       struct au_branch *br;
11757 +       struct au_dr_br *dr;
11758 +       struct dentry *h_dentry;
11759 +       struct inode *h_inode;
11760 +       struct au_dr_hino *ent;
11761 +       struct au_drinfo_rev *rev, **p;
11762 +
11763 +       AuDbg("bindex %d\n", bindex);
11764 +
11765 +       err = -ENOMEM;
11766 +       ent = kmalloc(sizeof(*ent), GFP_NOFS);
11767 +       if (unlikely(!ent))
11768 +               goto out;
11769 +
11770 +       sb = src->d_sb;
11771 +       br = au_sbr(sb, bindex);
11772 +       dr = &br->br_dirren;
11773 +       h_dentry = au_h_dptr(src, bindex);
11774 +       h_inode = d_inode(h_dentry);
11775 +       ino = h_inode->i_ino;
11776 +       ent->dr_h_ino = ino;
11777 +       already = au_dr_hino_test_add(dr, ino, ent);
11778 +       AuDbg("b%d, hi%llu, already %d\n",
11779 +             bindex, (unsigned long long)ino, already);
11780 +
11781 +       err = au_drinfo_store(src, bindex, dst_name, _rev);
11782 +       AuTraceErr(err);
11783 +       if (!err) {
11784 +               p = _rev;
11785 +               rev = *p;
11786 +               rev->already = already;
11787 +               goto out; /* success */
11788 +       }
11789 +
11790 +       /* revert */
11791 +       if (!already)
11792 +               au_dr_hino_del(dr, ent);
11793 +       au_kfree_rcu(ent);
11794 +
11795 +out:
11796 +       AuTraceErr(err);
11797 +       return err;
11798 +}
11799 +
11800 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11801 +{
11802 +       struct au_drinfo_rev *rev;
11803 +       struct au_drinfo_rev_elm *elm;
11804 +       int nelm;
11805 +
11806 +       rev = _rev;
11807 +       elm = rev->elm;
11808 +       for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11809 +               dput(elm->info_dentry);
11810 +               au_kfree_rcu(elm->info_last);
11811 +       }
11812 +       au_kfree_try_rcu(rev);
11813 +}
11814 +
11815 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11816 +{
11817 +       int err;
11818 +       struct au_drinfo_store work;
11819 +       struct au_drinfo_rev *rev = _rev;
11820 +       struct super_block *sb;
11821 +       struct au_branch *br;
11822 +       struct inode *h_inode;
11823 +       struct au_dr_br *dr;
11824 +       struct au_dr_hino *ent;
11825 +
11826 +       err = au_drinfo_store_work_init(&work, btgt);
11827 +       if (unlikely(err))
11828 +               goto out;
11829 +
11830 +       sb = src->d_sb;
11831 +       br = au_sbr(sb, btgt);
11832 +       work.h_ppath.dentry = au_h_dptr(src, btgt);
11833 +       work.h_ppath.mnt = au_br_mnt(br);
11834 +       au_drinfo_store_rev(rev, &work);
11835 +       au_drinfo_store_work_fin(&work);
11836 +       if (rev->already)
11837 +               goto out;
11838 +
11839 +       dr = &br->br_dirren;
11840 +       h_inode = d_inode(work.h_ppath.dentry);
11841 +       ent = au_dr_hino_find(dr, h_inode->i_ino);
11842 +       BUG_ON(!ent);
11843 +       au_dr_hino_del(dr, ent);
11844 +       au_kfree_rcu(ent);
11845 +
11846 +out:
11847 +       au_kfree_try_rcu(rev);
11848 +       if (unlikely(err))
11849 +               pr_err("failed to remove dirren info\n");
11850 +}
11851 +
11852 +/* ---------------------------------------------------------------------- */
11853 +
11854 +static struct au_drinfo *au_drinfo_do_load(struct path *h_ppath,
11855 +                                          char *whname, int whnamelen,
11856 +                                          struct dentry **info_dentry)
11857 +{
11858 +       struct au_drinfo *drinfo;
11859 +       struct file *f;
11860 +       struct inode *h_dir;
11861 +       struct path infopath;
11862 +       int unlocked;
11863 +
11864 +       AuDbg("%pd/%.*s\n", h_ppath->dentry, whnamelen, whname);
11865 +
11866 +       *info_dentry = NULL;
11867 +       drinfo = NULL;
11868 +       unlocked = 0;
11869 +       h_dir = d_inode(h_ppath->dentry);
11870 +       inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
11871 +       infopath.dentry = vfsub_lookup_one_len(whname, h_ppath, whnamelen);
11872 +       if (IS_ERR(infopath.dentry)) {
11873 +               drinfo = (void *)infopath.dentry;
11874 +               goto out;
11875 +       }
11876 +
11877 +       if (d_is_negative(infopath.dentry))
11878 +               goto out_dput; /* success */
11879 +
11880 +       infopath.mnt = h_ppath->mnt;
11881 +       f = vfsub_dentry_open(&infopath, O_RDONLY);
11882 +       inode_unlock_shared(h_dir);
11883 +       unlocked = 1;
11884 +       if (IS_ERR(f)) {
11885 +               drinfo = (void *)f;
11886 +               goto out_dput;
11887 +       }
11888 +
11889 +       drinfo = au_drinfo_read_k(f, /*h_ino*/0);
11890 +       if (IS_ERR_OR_NULL(drinfo))
11891 +               goto out_fput;
11892 +
11893 +       AuDbg("oldname %.*s\n", drinfo->oldnamelen, drinfo->oldname);
11894 +       *info_dentry = dget(infopath.dentry); /* keep it alive */
11895 +
11896 +out_fput:
11897 +       fput(f);
11898 +out_dput:
11899 +       dput(infopath.dentry);
11900 +out:
11901 +       if (!unlocked)
11902 +               inode_unlock_shared(h_dir);
11903 +       AuTraceErrPtr(drinfo);
11904 +       return drinfo;
11905 +}
11906 +
11907 +struct au_drinfo_do_load_args {
11908 +       struct au_drinfo **drinfop;
11909 +       struct path *h_ppath;
11910 +       char *whname;
11911 +       int whnamelen;
11912 +       struct dentry **info_dentry;
11913 +};
11914 +
11915 +static void au_call_drinfo_do_load(void *args)
11916 +{
11917 +       struct au_drinfo_do_load_args *a = args;
11918 +
11919 +       *a->drinfop = au_drinfo_do_load(a->h_ppath, a->whname, a->whnamelen,
11920 +                                       a->info_dentry);
11921 +}
11922 +
11923 +struct au_drinfo_load {
11924 +       struct path h_ppath;
11925 +       struct qstr *qname;
11926 +       unsigned char no_sio;
11927 +
11928 +       aufs_bindex_t ninfo;
11929 +       struct au_drinfo **drinfo;
11930 +};
11931 +
11932 +static int au_drinfo_load(struct au_drinfo_load *w, aufs_bindex_t bindex,
11933 +                         struct au_branch *br)
11934 +{
11935 +       int err, wkq_err, whnamelen, e;
11936 +       char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ]
11937 +               = AUFS_WH_DR_INFO_PFX;
11938 +       struct au_drinfo *drinfo;
11939 +       struct qstr oldname;
11940 +       struct inode *h_dir, *delegated;
11941 +       struct dentry *info_dentry;
11942 +       struct path infopath;
11943 +
11944 +       whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11945 +       whnamelen += au_drinfo_name(br, whname + whnamelen,
11946 +                                   sizeof(whname) - whnamelen);
11947 +       if (w->no_sio)
11948 +               drinfo = au_drinfo_do_load(&w->h_ppath, whname, whnamelen,
11949 +                                          &info_dentry);
11950 +       else {
11951 +               struct au_drinfo_do_load_args args = {
11952 +                       .drinfop        = &drinfo,
11953 +                       .h_ppath        = &w->h_ppath,
11954 +                       .whname         = whname,
11955 +                       .whnamelen      = whnamelen,
11956 +                       .info_dentry    = &info_dentry
11957 +               };
11958 +               wkq_err = au_wkq_wait(au_call_drinfo_do_load, &args);
11959 +               if (unlikely(wkq_err))
11960 +                       drinfo = ERR_PTR(wkq_err);
11961 +       }
11962 +       err = PTR_ERR(drinfo);
11963 +       if (IS_ERR_OR_NULL(drinfo))
11964 +               goto out;
11965 +
11966 +       err = 0;
11967 +       oldname.len = drinfo->oldnamelen;
11968 +       oldname.name = drinfo->oldname;
11969 +       if (au_qstreq(w->qname, &oldname)) {
11970 +               /* the name is renamed back */
11971 +               au_kfree_rcu(drinfo);
11972 +               drinfo = NULL;
11973 +
11974 +               infopath.dentry = info_dentry;
11975 +               infopath.mnt = w->h_ppath.mnt;
11976 +               h_dir = d_inode(w->h_ppath.dentry);
11977 +               delegated = NULL;
11978 +               inode_lock_nested(h_dir, AuLsc_I_PARENT);
11979 +               e = vfsub_unlink(h_dir, &infopath, &delegated, !w->no_sio);
11980 +               inode_unlock(h_dir);
11981 +               if (unlikely(e))
11982 +                       AuIOErr("ignored %d, %pd2\n", e, &infopath.dentry);
11983 +               if (unlikely(e == -EWOULDBLOCK))
11984 +                       iput(delegated);
11985 +       }
11986 +       au_kfree_rcu(w->drinfo[bindex]);
11987 +       w->drinfo[bindex] = drinfo;
11988 +       dput(info_dentry);
11989 +
11990 +out:
11991 +       AuTraceErr(err);
11992 +       return err;
11993 +}
11994 +
11995 +/* ---------------------------------------------------------------------- */
11996 +
11997 +static void au_dr_lkup_free(struct au_drinfo **drinfo, int n)
11998 +{
11999 +       struct au_drinfo **p = drinfo;
12000 +
12001 +       while (n-- > 0)
12002 +               au_kfree_rcu(*drinfo++);
12003 +       au_kfree_try_rcu(p);
12004 +}
12005 +
12006 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12007 +              aufs_bindex_t btgt)
12008 +{
12009 +       int err, ninfo;
12010 +       struct au_drinfo_load w;
12011 +       aufs_bindex_t bindex, bbot;
12012 +       struct au_branch *br;
12013 +       struct inode *h_dir;
12014 +       struct au_dr_hino *ent;
12015 +       struct super_block *sb;
12016 +
12017 +       AuDbg("%.*s, name %.*s, whname %.*s, b%d\n",
12018 +             AuLNPair(&dentry->d_name), AuLNPair(&lkup->dirren.dr_name),
12019 +             AuLNPair(&lkup->whname), btgt);
12020 +
12021 +       sb = dentry->d_sb;
12022 +       bbot = au_sbbot(sb);
12023 +       w.ninfo = bbot + 1;
12024 +       if (!lkup->dirren.drinfo) {
12025 +               lkup->dirren.drinfo = kcalloc(w.ninfo,
12026 +                                             sizeof(*lkup->dirren.drinfo),
12027 +                                             GFP_NOFS);
12028 +               if (unlikely(!lkup->dirren.drinfo)) {
12029 +                       err = -ENOMEM;
12030 +                       goto out;
12031 +               }
12032 +               lkup->dirren.ninfo = w.ninfo;
12033 +       }
12034 +       w.drinfo = lkup->dirren.drinfo;
12035 +       w.no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
12036 +       w.h_ppath.dentry = au_h_dptr(dentry, btgt);
12037 +       AuDebugOn(!w.h_ppath.dentry);
12038 +       w.h_ppath.mnt = au_sbr_mnt(sb, btgt);
12039 +       w.qname = &dentry->d_name;
12040 +
12041 +       ninfo = 0;
12042 +       for (bindex = btgt + 1; bindex <= bbot; bindex++) {
12043 +               br = au_sbr(sb, bindex);
12044 +               err = au_drinfo_load(&w, bindex, br);
12045 +               if (unlikely(err))
12046 +                       goto out_free;
12047 +               if (w.drinfo[bindex])
12048 +                       ninfo++;
12049 +       }
12050 +       if (!ninfo) {
12051 +               br = au_sbr(sb, btgt);
12052 +               h_dir = d_inode(w.h_ppath.dentry);
12053 +               ent = au_dr_hino_find(&br->br_dirren, h_dir->i_ino);
12054 +               AuDebugOn(!ent);
12055 +               au_dr_hino_del(&br->br_dirren, ent);
12056 +               au_kfree_rcu(ent);
12057 +       }
12058 +       goto out; /* success */
12059 +
12060 +out_free:
12061 +       au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12062 +       lkup->dirren.ninfo = 0;
12063 +       lkup->dirren.drinfo = NULL;
12064 +out:
12065 +       AuTraceErr(err);
12066 +       return err;
12067 +}
12068 +
12069 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup)
12070 +{
12071 +       au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12072 +}
12073 +
12074 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt)
12075 +{
12076 +       int err;
12077 +       struct au_drinfo *drinfo;
12078 +
12079 +       err = 0;
12080 +       if (!lkup->dirren.drinfo)
12081 +               goto out;
12082 +       AuDebugOn(lkup->dirren.ninfo <= btgt);
12083 +       drinfo = lkup->dirren.drinfo[btgt];
12084 +       if (!drinfo)
12085 +               goto out;
12086 +
12087 +       au_kfree_try_rcu(lkup->whname.name);
12088 +       lkup->whname.name = NULL;
12089 +       lkup->dirren.dr_name.len = drinfo->oldnamelen;
12090 +       lkup->dirren.dr_name.name = drinfo->oldname;
12091 +       lkup->name = &lkup->dirren.dr_name;
12092 +       err = au_wh_name_alloc(&lkup->whname, lkup->name);
12093 +       if (!err)
12094 +               AuDbg("name %.*s, whname %.*s, b%d\n",
12095 +                     AuLNPair(lkup->name), AuLNPair(&lkup->whname),
12096 +                     btgt);
12097 +
12098 +out:
12099 +       AuTraceErr(err);
12100 +       return err;
12101 +}
12102 +
12103 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12104 +                    ino_t h_ino)
12105 +{
12106 +       int match;
12107 +       struct au_drinfo *drinfo;
12108 +
12109 +       match = 1;
12110 +       if (!lkup->dirren.drinfo)
12111 +               goto out;
12112 +       AuDebugOn(lkup->dirren.ninfo <= bindex);
12113 +       drinfo = lkup->dirren.drinfo[bindex];
12114 +       if (!drinfo)
12115 +               goto out;
12116 +
12117 +       match = (drinfo->ino == h_ino);
12118 +       AuDbg("match %d\n", match);
12119 +
12120 +out:
12121 +       return match;
12122 +}
12123 +
12124 +/* ---------------------------------------------------------------------- */
12125 +
12126 +int au_dr_opt_set(struct super_block *sb)
12127 +{
12128 +       int err;
12129 +       aufs_bindex_t bindex, bbot;
12130 +       struct au_branch *br;
12131 +
12132 +       err = 0;
12133 +       bbot = au_sbbot(sb);
12134 +       for (bindex = 0; !err && bindex <= bbot; bindex++) {
12135 +               br = au_sbr(sb, bindex);
12136 +               err = au_dr_hino(sb, bindex, /*br*/NULL, &br->br_path);
12137 +       }
12138 +
12139 +       return err;
12140 +}
12141 +
12142 +int au_dr_opt_flush(struct super_block *sb)
12143 +{
12144 +       int err;
12145 +       aufs_bindex_t bindex, bbot;
12146 +       struct au_branch *br;
12147 +
12148 +       err = 0;
12149 +       bbot = au_sbbot(sb);
12150 +       for (bindex = 0; !err && bindex <= bbot; bindex++) {
12151 +               br = au_sbr(sb, bindex);
12152 +               if (au_br_writable(br->br_perm))
12153 +                       err = au_dr_hino(sb, bindex, /*br*/NULL, /*path*/NULL);
12154 +       }
12155 +
12156 +       return err;
12157 +}
12158 +
12159 +int au_dr_opt_clr(struct super_block *sb, int no_flush)
12160 +{
12161 +       int err;
12162 +       aufs_bindex_t bindex, bbot;
12163 +       struct au_branch *br;
12164 +
12165 +       err = 0;
12166 +       if (!no_flush) {
12167 +               err = au_dr_opt_flush(sb);
12168 +               if (unlikely(err))
12169 +                       goto out;
12170 +       }
12171 +
12172 +       bbot = au_sbbot(sb);
12173 +       for (bindex = 0; bindex <= bbot; bindex++) {
12174 +               br = au_sbr(sb, bindex);
12175 +               au_dr_hino_free(&br->br_dirren);
12176 +       }
12177 +
12178 +out:
12179 +       return err;
12180 +}
12181 diff -urN /usr/share/empty/fs/aufs/dirren.h linux/fs/aufs/dirren.h
12182 --- /usr/share/empty/fs/aufs/dirren.h   1970-01-01 01:00:00.000000000 +0100
12183 +++ linux/fs/aufs/dirren.h      2022-11-05 23:02:18.962555950 +0100
12184 @@ -0,0 +1,140 @@
12185 +/* SPDX-License-Identifier: GPL-2.0 */
12186 +/*
12187 + * Copyright (C) 2017-2022 Junjiro R. Okajima
12188 + *
12189 + * This program is free software; you can redistribute it and/or modify
12190 + * it under the terms of the GNU General Public License as published by
12191 + * the Free Software Foundation; either version 2 of the License, or
12192 + * (at your option) any later version.
12193 + *
12194 + * This program is distributed in the hope that it will be useful,
12195 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12196 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12197 + * GNU General Public License for more details.
12198 + *
12199 + * You should have received a copy of the GNU General Public License
12200 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12201 + */
12202 +
12203 +/*
12204 + * renamed dir info
12205 + */
12206 +
12207 +#ifndef __AUFS_DIRREN_H__
12208 +#define __AUFS_DIRREN_H__
12209 +
12210 +#ifdef __KERNEL__
12211 +
12212 +#include <linux/dcache.h>
12213 +#include <linux/statfs.h>
12214 +#include <linux/uuid.h>
12215 +#include "hbl.h"
12216 +
12217 +#define AuDirren_NHASH 100
12218 +
12219 +#ifdef CONFIG_AUFS_DIRREN
12220 +enum au_brid_type {
12221 +       AuBrid_Unset,
12222 +       AuBrid_UUID,
12223 +       AuBrid_FSID,
12224 +       AuBrid_DEV
12225 +};
12226 +
12227 +struct au_dr_brid {
12228 +       enum au_brid_type       type;
12229 +       union {
12230 +               uuid_t  uuid;   /* unimplemented yet */
12231 +               fsid_t  fsid;
12232 +               dev_t   dev;
12233 +       };
12234 +};
12235 +
12236 +/* 20 is the max digits length of ulong 64 */
12237 +/* brid-type "_" uuid "_" inum */
12238 +#define AUFS_DIRREN_FNAME_SZ   (1 + 1 + UUID_STRING_LEN + 20)
12239 +#define AUFS_DIRREN_ENV_VAL_SZ (AUFS_DIRREN_FNAME_SZ + 1 + 20)
12240 +
12241 +struct au_dr_hino {
12242 +       struct hlist_bl_node    dr_hnode;
12243 +       ino_t                   dr_h_ino;
12244 +};
12245 +
12246 +struct au_dr_br {
12247 +       struct hlist_bl_head    dr_h_ino[AuDirren_NHASH];
12248 +       struct au_dr_brid       dr_brid;
12249 +};
12250 +
12251 +struct au_dr_lookup {
12252 +       /* dr_name is pointed by struct au_do_lookup_args.name */
12253 +       struct qstr             dr_name; /* subset of dr_info */
12254 +       aufs_bindex_t           ninfo;
12255 +       struct au_drinfo        **drinfo;
12256 +};
12257 +#else
12258 +struct au_dr_hino;
12259 +/* empty */
12260 +struct au_dr_br { };
12261 +struct au_dr_lookup { };
12262 +#endif
12263 +
12264 +/* ---------------------------------------------------------------------- */
12265 +
12266 +struct au_branch;
12267 +struct au_do_lookup_args;
12268 +struct au_hinode;
12269 +#ifdef CONFIG_AUFS_DIRREN
12270 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t h_ino,
12271 +                       struct au_dr_hino *add_ent);
12272 +void au_dr_hino_free(struct au_dr_br *dr);
12273 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
12274 +                 const struct path *path);
12275 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br);
12276 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
12277 +                struct qstr *dst_name, void *_rev);
12278 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *rev);
12279 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t bindex, void *rev);
12280 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12281 +              aufs_bindex_t bindex);
12282 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12283 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12284 +                    ino_t h_ino);
12285 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup);
12286 +int au_dr_opt_set(struct super_block *sb);
12287 +int au_dr_opt_flush(struct super_block *sb);
12288 +int au_dr_opt_clr(struct super_block *sb, int no_flush);
12289 +#else
12290 +AuStubInt0(au_dr_hino_test_add, struct au_dr_br *dr, ino_t h_ino,
12291 +          struct au_dr_hino *add_ent);
12292 +AuStubVoid(au_dr_hino_free, struct au_dr_br *dr);
12293 +AuStubInt0(au_dr_br_init, struct super_block *sb, struct au_branch *br,
12294 +          const struct path *path);
12295 +AuStubInt0(au_dr_br_fin, struct super_block *sb, struct au_branch *br);
12296 +AuStubInt0(au_dr_rename, struct dentry *src, aufs_bindex_t bindex,
12297 +          struct qstr *dst_name, void *_rev);
12298 +AuStubVoid(au_dr_rename_fin, struct dentry *src, aufs_bindex_t btgt, void *rev);
12299 +AuStubVoid(au_dr_rename_rev, struct dentry *src, aufs_bindex_t bindex,
12300 +          void *rev);
12301 +AuStubInt0(au_dr_lkup, struct au_do_lookup_args *lkup, struct dentry *dentry,
12302 +          aufs_bindex_t bindex);
12303 +AuStubInt0(au_dr_lkup_name, struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12304 +AuStubInt0(au_dr_lkup_h_ino, struct au_do_lookup_args *lkup,
12305 +          aufs_bindex_t bindex, ino_t h_ino);
12306 +AuStubVoid(au_dr_lkup_fin, struct au_do_lookup_args *lkup);
12307 +AuStubInt0(au_dr_opt_set, struct super_block *sb);
12308 +AuStubInt0(au_dr_opt_flush, struct super_block *sb);
12309 +AuStubInt0(au_dr_opt_clr, struct super_block *sb, int no_flush);
12310 +#endif
12311 +
12312 +/* ---------------------------------------------------------------------- */
12313 +
12314 +#ifdef CONFIG_AUFS_DIRREN
12315 +static inline int au_dr_ihash(ino_t h_ino)
12316 +{
12317 +       return h_ino % AuDirren_NHASH;
12318 +}
12319 +#else
12320 +AuStubInt0(au_dr_ihash, ino_t h_ino);
12321 +#endif
12322 +
12323 +#endif /* __KERNEL__ */
12324 +#endif /* __AUFS_DIRREN_H__ */
12325 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
12326 --- /usr/share/empty/fs/aufs/dynop.c    1970-01-01 01:00:00.000000000 +0100
12327 +++ linux/fs/aufs/dynop.c       2022-11-05 23:02:18.962555950 +0100
12328 @@ -0,0 +1,366 @@
12329 +// SPDX-License-Identifier: GPL-2.0
12330 +/*
12331 + * Copyright (C) 2010-2022 Junjiro R. Okajima
12332 + *
12333 + * This program is free software; you can redistribute it and/or modify
12334 + * it under the terms of the GNU General Public License as published by
12335 + * the Free Software Foundation; either version 2 of the License, or
12336 + * (at your option) any later version.
12337 + *
12338 + * This program is distributed in the hope that it will be useful,
12339 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12340 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12341 + * GNU General Public License for more details.
12342 + *
12343 + * You should have received a copy of the GNU General Public License
12344 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12345 + */
12346 +
12347 +/*
12348 + * dynamically customizable operations for regular files
12349 + */
12350 +
12351 +#include "aufs.h"
12352 +
12353 +#define DyPrSym(key)   AuDbgSym(key->dk_op.dy_hop)
12354 +
12355 +/*
12356 + * How large will these lists be?
12357 + * Usually just a few elements, 20-30 at most for each, I guess.
12358 + */
12359 +static struct hlist_bl_head dynop[AuDyLast];
12360 +
12361 +static struct au_dykey *dy_gfind_get(struct hlist_bl_head *hbl,
12362 +                                    const void *h_op)
12363 +{
12364 +       struct au_dykey *key, *tmp;
12365 +       struct hlist_bl_node *pos;
12366 +
12367 +       key = NULL;
12368 +       hlist_bl_lock(hbl);
12369 +       hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12370 +               if (tmp->dk_op.dy_hop == h_op) {
12371 +                       if (kref_get_unless_zero(&tmp->dk_kref))
12372 +                               key = tmp;
12373 +                       break;
12374 +               }
12375 +       hlist_bl_unlock(hbl);
12376 +
12377 +       return key;
12378 +}
12379 +
12380 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
12381 +{
12382 +       struct au_dykey **k, *found;
12383 +       const void *h_op = key->dk_op.dy_hop;
12384 +       int i;
12385 +
12386 +       found = NULL;
12387 +       k = br->br_dykey;
12388 +       for (i = 0; i < AuBrDynOp; i++)
12389 +               if (k[i]) {
12390 +                       if (k[i]->dk_op.dy_hop == h_op) {
12391 +                               found = k[i];
12392 +                               break;
12393 +                       }
12394 +               } else
12395 +                       break;
12396 +       if (!found) {
12397 +               spin_lock(&br->br_dykey_lock);
12398 +               for (; i < AuBrDynOp; i++)
12399 +                       if (k[i]) {
12400 +                               if (k[i]->dk_op.dy_hop == h_op) {
12401 +                                       found = k[i];
12402 +                                       break;
12403 +                               }
12404 +                       } else {
12405 +                               k[i] = key;
12406 +                               break;
12407 +                       }
12408 +               spin_unlock(&br->br_dykey_lock);
12409 +               BUG_ON(i == AuBrDynOp); /* expand the array */
12410 +       }
12411 +
12412 +       return found;
12413 +}
12414 +
12415 +/* kref_get() if @key is already added */
12416 +static struct au_dykey *dy_gadd(struct hlist_bl_head *hbl, struct au_dykey *key)
12417 +{
12418 +       struct au_dykey *tmp, *found;
12419 +       struct hlist_bl_node *pos;
12420 +       const void *h_op = key->dk_op.dy_hop;
12421 +
12422 +       found = NULL;
12423 +       hlist_bl_lock(hbl);
12424 +       hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12425 +               if (tmp->dk_op.dy_hop == h_op) {
12426 +                       if (kref_get_unless_zero(&tmp->dk_kref))
12427 +                               found = tmp;
12428 +                       break;
12429 +               }
12430 +       if (!found)
12431 +               hlist_bl_add_head(&key->dk_hnode, hbl);
12432 +       hlist_bl_unlock(hbl);
12433 +
12434 +       if (!found)
12435 +               DyPrSym(key);
12436 +       return found;
12437 +}
12438 +
12439 +static void dy_free_rcu(struct rcu_head *rcu)
12440 +{
12441 +       struct au_dykey *key;
12442 +
12443 +       key = container_of(rcu, struct au_dykey, dk_rcu);
12444 +       DyPrSym(key);
12445 +       kfree(key);
12446 +}
12447 +
12448 +static void dy_free(struct kref *kref)
12449 +{
12450 +       struct au_dykey *key;
12451 +       struct hlist_bl_head *hbl;
12452 +
12453 +       key = container_of(kref, struct au_dykey, dk_kref);
12454 +       hbl = dynop + key->dk_op.dy_type;
12455 +       au_hbl_del(&key->dk_hnode, hbl);
12456 +       call_rcu(&key->dk_rcu, dy_free_rcu);
12457 +}
12458 +
12459 +void au_dy_put(struct au_dykey *key)
12460 +{
12461 +       kref_put(&key->dk_kref, dy_free);
12462 +}
12463 +
12464 +/* ---------------------------------------------------------------------- */
12465 +
12466 +#define DyDbgSize(cnt, op)     AuDebugOn(cnt != sizeof(op)/sizeof(void *))
12467 +
12468 +#ifdef CONFIG_AUFS_DEBUG
12469 +#define DyDbgDeclare(cnt)      unsigned int cnt = 0
12470 +#define DyDbgInc(cnt)          do { cnt++; } while (0)
12471 +#else
12472 +#define DyDbgDeclare(cnt)      do {} while (0)
12473 +#define DyDbgInc(cnt)          do {} while (0)
12474 +#endif
12475 +
12476 +#define DySet(func, dst, src, h_op, h_sb) do {                         \
12477 +       DyDbgInc(cnt);                                                  \
12478 +       if (h_op->func) {                                               \
12479 +               if (src.func)                                           \
12480 +                       dst.func = src.func;                            \
12481 +               else                                                    \
12482 +                       AuDbg("%s %s\n", au_sbtype(h_sb), #func);       \
12483 +       }                                                               \
12484 +} while (0)
12485 +
12486 +#define DySetForce(func, dst, src) do {                \
12487 +       AuDebugOn(!src.func);                   \
12488 +       DyDbgInc(cnt);                          \
12489 +       dst.func = src.func;                    \
12490 +} while (0)
12491 +
12492 +#define DySetAop(func) \
12493 +       DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
12494 +#define DySetAopForce(func) \
12495 +       DySetForce(func, dyaop->da_op, aufs_aop)
12496 +
12497 +static void dy_aop(struct au_dykey *key, const void *h_op,
12498 +                  struct super_block *h_sb __maybe_unused)
12499 +{
12500 +       struct au_dyaop *dyaop = (void *)key;
12501 +       const struct address_space_operations *h_aop = h_op;
12502 +       DyDbgDeclare(cnt);
12503 +
12504 +       AuDbg("%s\n", au_sbtype(h_sb));
12505 +
12506 +       DySetAop(writepage);
12507 +       DySetAopForce(read_folio);      /* force */
12508 +       DySetAop(writepages);
12509 +       DySetAop(dirty_folio);
12510 +       DySetAop(invalidate_folio);
12511 +       DySetAop(readahead);
12512 +       DySetAop(write_begin);
12513 +       DySetAop(write_end);
12514 +       DySetAop(bmap);
12515 +       DySetAop(release_folio);
12516 +       DySetAop(free_folio);
12517 +       /* this one will be changed according to an aufs mount option */
12518 +       DySetAop(direct_IO);
12519 +       DySetAop(migrate_folio);
12520 +       DySetAop(launder_folio);
12521 +       DySetAop(is_partially_uptodate);
12522 +       DySetAop(is_dirty_writeback);
12523 +       DySetAop(error_remove_page);
12524 +       DySetAop(swap_activate);
12525 +       DySetAop(swap_deactivate);
12526 +       DySetAop(swap_rw);
12527 +
12528 +       DyDbgSize(cnt, *h_aop);
12529 +}
12530 +
12531 +/* ---------------------------------------------------------------------- */
12532 +
12533 +static void dy_bug(struct kref *kref)
12534 +{
12535 +       BUG();
12536 +}
12537 +
12538 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
12539 +{
12540 +       struct au_dykey *key, *old;
12541 +       struct hlist_bl_head *hbl;
12542 +       struct op {
12543 +               unsigned int sz;
12544 +               void (*set)(struct au_dykey *key, const void *h_op,
12545 +                           struct super_block *h_sb __maybe_unused);
12546 +       };
12547 +       static const struct op a[] = {
12548 +               [AuDy_AOP] = {
12549 +                       .sz     = sizeof(struct au_dyaop),
12550 +                       .set    = dy_aop
12551 +               }
12552 +       };
12553 +       const struct op *p;
12554 +
12555 +       hbl = dynop + op->dy_type;
12556 +       key = dy_gfind_get(hbl, op->dy_hop);
12557 +       if (key)
12558 +               goto out_add; /* success */
12559 +
12560 +       p = a + op->dy_type;
12561 +       key = kzalloc(p->sz, GFP_NOFS);
12562 +       if (unlikely(!key)) {
12563 +               key = ERR_PTR(-ENOMEM);
12564 +               goto out;
12565 +       }
12566 +
12567 +       key->dk_op.dy_hop = op->dy_hop;
12568 +       kref_init(&key->dk_kref);
12569 +       p->set(key, op->dy_hop, au_br_sb(br));
12570 +       old = dy_gadd(hbl, key);
12571 +       if (old) {
12572 +               au_kfree_rcu(key);
12573 +               key = old;
12574 +       }
12575 +
12576 +out_add:
12577 +       old = dy_bradd(br, key);
12578 +       if (old)
12579 +               /* its ref-count should never be zero here */
12580 +               kref_put(&key->dk_kref, dy_bug);
12581 +out:
12582 +       return key;
12583 +}
12584 +
12585 +/* ---------------------------------------------------------------------- */
12586 +/*
12587 + * Aufs prohibits O_DIRECT by default even if the branch supports it.
12588 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
12589 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
12590 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
12591 + * See the aufs manual in detail.
12592 + */
12593 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
12594 +{
12595 +       if (!do_dx)
12596 +               dyaop->da_op.direct_IO = NULL;
12597 +       else
12598 +               dyaop->da_op.direct_IO = aufs_aop.direct_IO;
12599 +}
12600 +
12601 +static struct au_dyaop *dy_aget(struct au_branch *br,
12602 +                               const struct address_space_operations *h_aop,
12603 +                               int do_dx)
12604 +{
12605 +       struct au_dyaop *dyaop;
12606 +       struct au_dynop op;
12607 +
12608 +       op.dy_type = AuDy_AOP;
12609 +       op.dy_haop = h_aop;
12610 +       dyaop = (void *)dy_get(&op, br);
12611 +       if (IS_ERR(dyaop))
12612 +               goto out;
12613 +       dy_adx(dyaop, do_dx);
12614 +
12615 +out:
12616 +       return dyaop;
12617 +}
12618 +
12619 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12620 +               struct inode *h_inode)
12621 +{
12622 +       int err, do_dx;
12623 +       struct super_block *sb;
12624 +       struct au_branch *br;
12625 +       struct au_dyaop *dyaop;
12626 +
12627 +       AuDebugOn(!S_ISREG(h_inode->i_mode));
12628 +       IiMustWriteLock(inode);
12629 +
12630 +       sb = inode->i_sb;
12631 +       br = au_sbr(sb, bindex);
12632 +       do_dx = !!au_opt_test(au_mntflags(sb), DIO);
12633 +       dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
12634 +       err = PTR_ERR(dyaop);
12635 +       if (IS_ERR(dyaop))
12636 +               /* unnecessary to call dy_fput() */
12637 +               goto out;
12638 +
12639 +       err = 0;
12640 +       inode->i_mapping->a_ops = &dyaop->da_op;
12641 +
12642 +out:
12643 +       return err;
12644 +}
12645 +
12646 +/*
12647 + * Is it safe to replace a_ops during the inode/file is in operation?
12648 + * Yes, I hope so.
12649 + */
12650 +int au_dy_irefresh(struct inode *inode)
12651 +{
12652 +       int err;
12653 +       aufs_bindex_t btop;
12654 +       struct inode *h_inode;
12655 +
12656 +       err = 0;
12657 +       if (S_ISREG(inode->i_mode)) {
12658 +               btop = au_ibtop(inode);
12659 +               h_inode = au_h_iptr(inode, btop);
12660 +               err = au_dy_iaop(inode, btop, h_inode);
12661 +       }
12662 +       return err;
12663 +}
12664 +
12665 +void au_dy_arefresh(int do_dx)
12666 +{
12667 +       struct hlist_bl_head *hbl;
12668 +       struct hlist_bl_node *pos;
12669 +       struct au_dykey *key;
12670 +
12671 +       hbl = dynop + AuDy_AOP;
12672 +       hlist_bl_lock(hbl);
12673 +       hlist_bl_for_each_entry(key, pos, hbl, dk_hnode)
12674 +               dy_adx((void *)key, do_dx);
12675 +       hlist_bl_unlock(hbl);
12676 +}
12677 +
12678 +/* ---------------------------------------------------------------------- */
12679 +
12680 +void __init au_dy_init(void)
12681 +{
12682 +       int i;
12683 +
12684 +       for (i = 0; i < AuDyLast; i++)
12685 +               INIT_HLIST_BL_HEAD(dynop + i);
12686 +}
12687 +
12688 +void au_dy_fin(void)
12689 +{
12690 +       int i;
12691 +
12692 +       for (i = 0; i < AuDyLast; i++)
12693 +               WARN_ON(!hlist_bl_empty(dynop + i));
12694 +}
12695 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
12696 --- /usr/share/empty/fs/aufs/dynop.h    1970-01-01 01:00:00.000000000 +0100
12697 +++ linux/fs/aufs/dynop.h       2022-11-05 23:02:18.962555950 +0100
12698 @@ -0,0 +1,77 @@
12699 +/* SPDX-License-Identifier: GPL-2.0 */
12700 +/*
12701 + * Copyright (C) 2010-2022 Junjiro R. Okajima
12702 + *
12703 + * This program is free software; you can redistribute it and/or modify
12704 + * it under the terms of the GNU General Public License as published by
12705 + * the Free Software Foundation; either version 2 of the License, or
12706 + * (at your option) any later version.
12707 + *
12708 + * This program is distributed in the hope that it will be useful,
12709 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12710 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12711 + * GNU General Public License for more details.
12712 + *
12713 + * You should have received a copy of the GNU General Public License
12714 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12715 + */
12716 +
12717 +/*
12718 + * dynamically customizable operations (for regular files only)
12719 + */
12720 +
12721 +#ifndef __AUFS_DYNOP_H__
12722 +#define __AUFS_DYNOP_H__
12723 +
12724 +#ifdef __KERNEL__
12725 +
12726 +#include <linux/fs.h>
12727 +#include <linux/kref.h>
12728 +
12729 +enum {AuDy_AOP, AuDyLast};
12730 +
12731 +struct au_dynop {
12732 +       int                                             dy_type;
12733 +       union {
12734 +               const void                              *dy_hop;
12735 +               const struct address_space_operations   *dy_haop;
12736 +       };
12737 +};
12738 +
12739 +struct au_dykey {
12740 +       union {
12741 +               struct hlist_bl_node    dk_hnode;
12742 +               struct rcu_head         dk_rcu;
12743 +       };
12744 +       struct au_dynop         dk_op;
12745 +
12746 +       /*
12747 +        * during I am in the branch local array, kref is gotten. when the
12748 +        * branch is removed, kref is put.
12749 +        */
12750 +       struct kref             dk_kref;
12751 +};
12752 +
12753 +/* stop unioning since their sizes are very different from each other */
12754 +struct au_dyaop {
12755 +       struct au_dykey                 da_key;
12756 +       struct address_space_operations da_op; /* not const */
12757 +};
12758 +/* make sure that 'struct au_dykey *' can be any type */
12759 +static_assert(!offsetof(struct au_dyaop, da_key));
12760 +
12761 +/* ---------------------------------------------------------------------- */
12762 +
12763 +/* dynop.c */
12764 +struct au_branch;
12765 +void au_dy_put(struct au_dykey *key);
12766 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12767 +               struct inode *h_inode);
12768 +int au_dy_irefresh(struct inode *inode);
12769 +void au_dy_arefresh(int do_dio);
12770 +
12771 +void __init au_dy_init(void);
12772 +void au_dy_fin(void);
12773 +
12774 +#endif /* __KERNEL__ */
12775 +#endif /* __AUFS_DYNOP_H__ */
12776 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
12777 --- /usr/share/empty/fs/aufs/export.c   1970-01-01 01:00:00.000000000 +0100
12778 +++ linux/fs/aufs/export.c      2022-11-05 23:02:18.962555950 +0100
12779 @@ -0,0 +1,830 @@
12780 +// SPDX-License-Identifier: GPL-2.0
12781 +/*
12782 + * Copyright (C) 2005-2022 Junjiro R. Okajima
12783 + *
12784 + * This program is free software; you can redistribute it and/or modify
12785 + * it under the terms of the GNU General Public License as published by
12786 + * the Free Software Foundation; either version 2 of the License, or
12787 + * (at your option) any later version.
12788 + *
12789 + * This program is distributed in the hope that it will be useful,
12790 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12791 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12792 + * GNU General Public License for more details.
12793 + *
12794 + * You should have received a copy of the GNU General Public License
12795 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
12796 + */
12797 +
12798 +/*
12799 + * export via nfs
12800 + */
12801 +
12802 +#include <linux/exportfs.h>
12803 +#include <linux/fs_struct.h>
12804 +#include <linux/nsproxy.h>
12805 +#include <linux/random.h>
12806 +#include <linux/writeback.h>
12807 +#include "aufs.h"
12808 +
12809 +union conv {
12810 +#ifdef CONFIG_AUFS_INO_T_64
12811 +       __u32 a[2];
12812 +#else
12813 +       __u32 a[1];
12814 +#endif
12815 +       ino_t ino;
12816 +};
12817 +
12818 +static ino_t decode_ino(__u32 *a)
12819 +{
12820 +       union conv u;
12821 +
12822 +       BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
12823 +       u.a[0] = a[0];
12824 +#ifdef CONFIG_AUFS_INO_T_64
12825 +       u.a[1] = a[1];
12826 +#endif
12827 +       return u.ino;
12828 +}
12829 +
12830 +static void encode_ino(__u32 *a, ino_t ino)
12831 +{
12832 +       union conv u;
12833 +
12834 +       u.ino = ino;
12835 +       a[0] = u.a[0];
12836 +#ifdef CONFIG_AUFS_INO_T_64
12837 +       a[1] = u.a[1];
12838 +#endif
12839 +}
12840 +
12841 +/* NFS file handle */
12842 +enum {
12843 +       Fh_br_id,
12844 +       Fh_sigen,
12845 +#ifdef CONFIG_AUFS_INO_T_64
12846 +       /* support 64bit inode number */
12847 +       Fh_ino1,
12848 +       Fh_ino2,
12849 +       Fh_dir_ino1,
12850 +       Fh_dir_ino2,
12851 +#else
12852 +       Fh_ino1,
12853 +       Fh_dir_ino1,
12854 +#endif
12855 +       Fh_igen,
12856 +       Fh_h_type,
12857 +       Fh_tail,
12858 +
12859 +       Fh_ino = Fh_ino1,
12860 +       Fh_dir_ino = Fh_dir_ino1
12861 +};
12862 +
12863 +static int au_test_anon(struct dentry *dentry)
12864 +{
12865 +       /* note: read d_flags without d_lock */
12866 +       return !!(dentry->d_flags & DCACHE_DISCONNECTED);
12867 +}
12868 +
12869 +int au_test_nfsd(void)
12870 +{
12871 +       int ret;
12872 +       struct task_struct *tsk = current;
12873 +       char comm[sizeof(tsk->comm)];
12874 +
12875 +       ret = 0;
12876 +       if (tsk->flags & PF_KTHREAD) {
12877 +               get_task_comm(comm, tsk);
12878 +               ret = !strcmp(comm, "nfsd");
12879 +       }
12880 +
12881 +       return ret;
12882 +}
12883 +
12884 +/* ---------------------------------------------------------------------- */
12885 +/* inode generation external table */
12886 +
12887 +void au_xigen_inc(struct inode *inode)
12888 +{
12889 +       loff_t pos;
12890 +       ssize_t sz;
12891 +       __u32 igen;
12892 +       struct super_block *sb;
12893 +       struct au_sbinfo *sbinfo;
12894 +
12895 +       sb = inode->i_sb;
12896 +       AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
12897 +
12898 +       sbinfo = au_sbi(sb);
12899 +       pos = inode->i_ino;
12900 +       pos *= sizeof(igen);
12901 +       igen = inode->i_generation + 1;
12902 +       sz = xino_fwrite(sbinfo->si_xigen, &igen, sizeof(igen), &pos);
12903 +       if (sz == sizeof(igen))
12904 +               return; /* success */
12905 +
12906 +       if (unlikely(sz >= 0))
12907 +               AuIOErr("xigen error (%zd)\n", sz);
12908 +}
12909 +
12910 +int au_xigen_new(struct inode *inode)
12911 +{
12912 +       int err;
12913 +       loff_t pos;
12914 +       ssize_t sz;
12915 +       struct super_block *sb;
12916 +       struct au_sbinfo *sbinfo;
12917 +       struct file *file;
12918 +
12919 +       err = 0;
12920 +       /* todo: dirty, at mount time */
12921 +       if (inode->i_ino == AUFS_ROOT_INO)
12922 +               goto out;
12923 +       sb = inode->i_sb;
12924 +       SiMustAnyLock(sb);
12925 +       if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
12926 +               goto out;
12927 +
12928 +       err = -EFBIG;
12929 +       pos = inode->i_ino;
12930 +       if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
12931 +               AuIOErr1("too large i%lld\n", pos);
12932 +               goto out;
12933 +       }
12934 +       pos *= sizeof(inode->i_generation);
12935 +
12936 +       err = 0;
12937 +       sbinfo = au_sbi(sb);
12938 +       file = sbinfo->si_xigen;
12939 +       BUG_ON(!file);
12940 +
12941 +       if (vfsub_f_size_read(file)
12942 +           < pos + sizeof(inode->i_generation)) {
12943 +               inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
12944 +               sz = xino_fwrite(file, &inode->i_generation,
12945 +                                sizeof(inode->i_generation), &pos);
12946 +       } else
12947 +               sz = xino_fread(file, &inode->i_generation,
12948 +                               sizeof(inode->i_generation), &pos);
12949 +       if (sz == sizeof(inode->i_generation))
12950 +               goto out; /* success */
12951 +
12952 +       err = sz;
12953 +       if (unlikely(sz >= 0)) {
12954 +               err = -EIO;
12955 +               AuIOErr("xigen error (%zd)\n", sz);
12956 +       }
12957 +
12958 +out:
12959 +       return err;
12960 +}
12961 +
12962 +int au_xigen_set(struct super_block *sb, struct path *path)
12963 +{
12964 +       int err;
12965 +       struct au_sbinfo *sbinfo;
12966 +       struct file *file;
12967 +
12968 +       SiMustWriteLock(sb);
12969 +
12970 +       sbinfo = au_sbi(sb);
12971 +       file = au_xino_create2(sb, path, sbinfo->si_xigen);
12972 +       err = PTR_ERR(file);
12973 +       if (IS_ERR(file))
12974 +               goto out;
12975 +       err = 0;
12976 +       if (sbinfo->si_xigen)
12977 +               fput(sbinfo->si_xigen);
12978 +       sbinfo->si_xigen = file;
12979 +
12980 +out:
12981 +       AuTraceErr(err);
12982 +       return err;
12983 +}
12984 +
12985 +void au_xigen_clr(struct super_block *sb)
12986 +{
12987 +       struct au_sbinfo *sbinfo;
12988 +
12989 +       SiMustWriteLock(sb);
12990 +
12991 +       sbinfo = au_sbi(sb);
12992 +       if (sbinfo->si_xigen) {
12993 +               fput(sbinfo->si_xigen);
12994 +               sbinfo->si_xigen = NULL;
12995 +       }
12996 +}
12997 +
12998 +/* ---------------------------------------------------------------------- */
12999 +
13000 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
13001 +                                   ino_t dir_ino)
13002 +{
13003 +       struct dentry *dentry, *d;
13004 +       struct inode *inode;
13005 +       unsigned int sigen;
13006 +
13007 +       dentry = NULL;
13008 +       inode = ilookup(sb, ino);
13009 +       if (!inode)
13010 +               goto out;
13011 +
13012 +       dentry = ERR_PTR(-ESTALE);
13013 +       sigen = au_sigen(sb);
13014 +       if (unlikely(au_is_bad_inode(inode)
13015 +                    || IS_DEADDIR(inode)
13016 +                    || sigen != au_iigen(inode, NULL)))
13017 +               goto out_iput;
13018 +
13019 +       dentry = NULL;
13020 +       if (!dir_ino || S_ISDIR(inode->i_mode))
13021 +               dentry = d_find_alias(inode);
13022 +       else {
13023 +               spin_lock(&inode->i_lock);
13024 +               hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
13025 +                       spin_lock(&d->d_lock);
13026 +                       if (!au_test_anon(d)
13027 +                           && d_inode(d->d_parent)->i_ino == dir_ino) {
13028 +                               dentry = dget_dlock(d);
13029 +                               spin_unlock(&d->d_lock);
13030 +                               break;
13031 +                       }
13032 +                       spin_unlock(&d->d_lock);
13033 +               }
13034 +               spin_unlock(&inode->i_lock);
13035 +       }
13036 +       if (unlikely(dentry && au_digen_test(dentry, sigen))) {
13037 +               /* need to refresh */
13038 +               dput(dentry);
13039 +               dentry = NULL;
13040 +       }
13041 +
13042 +out_iput:
13043 +       iput(inode);
13044 +out:
13045 +       AuTraceErrPtr(dentry);
13046 +       return dentry;
13047 +}
13048 +
13049 +/* ---------------------------------------------------------------------- */
13050 +
13051 +/* todo: dirty? */
13052 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
13053 +
13054 +struct au_compare_mnt_args {
13055 +       /* input */
13056 +       struct super_block *sb;
13057 +
13058 +       /* output */
13059 +       struct vfsmount *mnt;
13060 +};
13061 +
13062 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
13063 +{
13064 +       struct au_compare_mnt_args *a = arg;
13065 +
13066 +       if (mnt->mnt_sb != a->sb)
13067 +               return 0;
13068 +       a->mnt = mntget(mnt);
13069 +       return 1;
13070 +}
13071 +
13072 +static struct vfsmount *au_mnt_get(struct super_block *sb)
13073 +{
13074 +       int err;
13075 +       struct path root;
13076 +       struct au_compare_mnt_args args = {
13077 +               .sb = sb
13078 +       };
13079 +
13080 +       get_fs_root(current->fs, &root);
13081 +       rcu_read_lock();
13082 +       err = iterate_mounts(au_compare_mnt, &args, root.mnt);
13083 +       rcu_read_unlock();
13084 +       path_put(&root);
13085 +       AuDebugOn(!err);
13086 +       AuDebugOn(!args.mnt);
13087 +       return args.mnt;
13088 +}
13089 +
13090 +struct au_nfsd_si_lock {
13091 +       unsigned int sigen;
13092 +       aufs_bindex_t bindex, br_id;
13093 +       unsigned char force_lock;
13094 +};
13095 +
13096 +static int si_nfsd_read_lock(struct super_block *sb,
13097 +                            struct au_nfsd_si_lock *nsi_lock)
13098 +{
13099 +       int err;
13100 +       aufs_bindex_t bindex;
13101 +
13102 +       si_read_lock(sb, AuLock_FLUSH);
13103 +
13104 +       /* branch id may be wrapped around */
13105 +       err = 0;
13106 +       bindex = au_br_index(sb, nsi_lock->br_id);
13107 +       if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
13108 +               goto out; /* success */
13109 +
13110 +       err = -ESTALE;
13111 +       bindex = -1;
13112 +       if (!nsi_lock->force_lock)
13113 +               si_read_unlock(sb);
13114 +
13115 +out:
13116 +       nsi_lock->bindex = bindex;
13117 +       return err;
13118 +}
13119 +
13120 +struct find_name_by_ino {
13121 +       struct dir_context ctx;
13122 +       int called, found;
13123 +       ino_t ino;
13124 +       char *name;
13125 +       int namelen;
13126 +};
13127 +
13128 +static int
13129 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
13130 +                loff_t offset, u64 ino, unsigned int d_type)
13131 +{
13132 +       struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
13133 +                                                 ctx);
13134 +
13135 +       a->called++;
13136 +       if (a->ino != ino)
13137 +               return 0;
13138 +
13139 +       memcpy(a->name, name, namelen);
13140 +       a->namelen = namelen;
13141 +       a->found = 1;
13142 +       return 1;
13143 +}
13144 +
13145 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
13146 +                                    struct au_nfsd_si_lock *nsi_lock)
13147 +{
13148 +       struct dentry *dentry, *parent;
13149 +       struct file *file;
13150 +       struct inode *dir;
13151 +       struct find_name_by_ino arg = {
13152 +               .ctx = {
13153 +                       .actor = find_name_by_ino
13154 +               }
13155 +       };
13156 +       int err;
13157 +
13158 +       parent = path->dentry;
13159 +       if (nsi_lock)
13160 +               si_read_unlock(parent->d_sb);
13161 +       file = vfsub_dentry_open(path, au_dir_roflags);
13162 +       dentry = (void *)file;
13163 +       if (IS_ERR(file))
13164 +               goto out;
13165 +
13166 +       dentry = ERR_PTR(-ENOMEM);
13167 +       arg.name = (void *)__get_free_page(GFP_NOFS);
13168 +       if (unlikely(!arg.name))
13169 +               goto out_file;
13170 +       arg.ino = ino;
13171 +       arg.found = 0;
13172 +       do {
13173 +               arg.called = 0;
13174 +               /* smp_mb(); */
13175 +               err = vfsub_iterate_dir(file, &arg.ctx);
13176 +       } while (!err && !arg.found && arg.called);
13177 +       dentry = ERR_PTR(err);
13178 +       if (unlikely(err))
13179 +               goto out_name;
13180 +       /* instead of ENOENT */
13181 +       dentry = ERR_PTR(-ESTALE);
13182 +       if (!arg.found)
13183 +               goto out_name;
13184 +
13185 +       /* do not call vfsub_lkup_one() */
13186 +       dir = d_inode(parent);
13187 +       dentry = vfsub_lookup_one_len_unlocked(arg.name, path, arg.namelen);
13188 +       AuTraceErrPtr(dentry);
13189 +       if (IS_ERR(dentry))
13190 +               goto out_name;
13191 +       AuDebugOn(au_test_anon(dentry));
13192 +       if (unlikely(d_really_is_negative(dentry))) {
13193 +               dput(dentry);
13194 +               dentry = ERR_PTR(-ENOENT);
13195 +       }
13196 +
13197 +out_name:
13198 +       free_page((unsigned long)arg.name);
13199 +out_file:
13200 +       fput(file);
13201 +out:
13202 +       if (unlikely(nsi_lock
13203 +                    && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
13204 +               if (!IS_ERR(dentry)) {
13205 +                       dput(dentry);
13206 +                       dentry = ERR_PTR(-ESTALE);
13207 +               }
13208 +       AuTraceErrPtr(dentry);
13209 +       return dentry;
13210 +}
13211 +
13212 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
13213 +                                       ino_t dir_ino,
13214 +                                       struct au_nfsd_si_lock *nsi_lock)
13215 +{
13216 +       struct dentry *dentry;
13217 +       struct path path;
13218 +
13219 +       if (dir_ino != AUFS_ROOT_INO) {
13220 +               path.dentry = decode_by_ino(sb, dir_ino, 0);
13221 +               dentry = path.dentry;
13222 +               if (!path.dentry || IS_ERR(path.dentry))
13223 +                       goto out;
13224 +               AuDebugOn(au_test_anon(path.dentry));
13225 +       } else
13226 +               path.dentry = dget(sb->s_root);
13227 +
13228 +       path.mnt = au_mnt_get(sb);
13229 +       dentry = au_lkup_by_ino(&path, ino, nsi_lock);
13230 +       path_put(&path);
13231 +
13232 +out:
13233 +       AuTraceErrPtr(dentry);
13234 +       return dentry;
13235 +}
13236 +
13237 +/* ---------------------------------------------------------------------- */
13238 +
13239 +static int h_acceptable(void *expv, struct dentry *dentry)
13240 +{
13241 +       return 1;
13242 +}
13243 +
13244 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
13245 +                          char *buf, int len, struct super_block *sb)
13246 +{
13247 +       char *p;
13248 +       int n;
13249 +       struct path path;
13250 +
13251 +       p = d_path(h_rootpath, buf, len);
13252 +       if (IS_ERR(p))
13253 +               goto out;
13254 +       n = strlen(p);
13255 +
13256 +       path.mnt = h_rootpath->mnt;
13257 +       path.dentry = h_parent;
13258 +       p = d_path(&path, buf, len);
13259 +       if (IS_ERR(p))
13260 +               goto out;
13261 +       if (n != 1)
13262 +               p += n;
13263 +
13264 +       path.mnt = au_mnt_get(sb);
13265 +       path.dentry = sb->s_root;
13266 +       p = d_path(&path, buf, len - strlen(p));
13267 +       mntput(path.mnt);
13268 +       if (IS_ERR(p))
13269 +               goto out;
13270 +       if (n != 1)
13271 +               p[strlen(p)] = '/';
13272 +
13273 +out:
13274 +       AuTraceErrPtr(p);
13275 +       return p;
13276 +}
13277 +
13278 +static
13279 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
13280 +                             int fh_len, struct au_nfsd_si_lock *nsi_lock)
13281 +{
13282 +       struct dentry *dentry, *h_parent, *root;
13283 +       struct super_block *h_sb;
13284 +       char *pathname, *p;
13285 +       struct vfsmount *h_mnt;
13286 +       struct au_branch *br;
13287 +       int err;
13288 +       struct path path;
13289 +
13290 +       br = au_sbr(sb, nsi_lock->bindex);
13291 +       h_mnt = au_br_mnt(br);
13292 +       h_sb = h_mnt->mnt_sb;
13293 +       /* todo: call lower fh_to_dentry()? fh_to_parent()? */
13294 +       lockdep_off();
13295 +       h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
13296 +                                     fh_len - Fh_tail, fh[Fh_h_type],
13297 +                                     h_acceptable, /*context*/NULL);
13298 +       lockdep_on();
13299 +       dentry = h_parent;
13300 +       if (unlikely(!h_parent || IS_ERR(h_parent))) {
13301 +               AuWarn1("%s decode_fh failed, %ld\n",
13302 +                       au_sbtype(h_sb), PTR_ERR(h_parent));
13303 +               goto out;
13304 +       }
13305 +       dentry = NULL;
13306 +       if (unlikely(au_test_anon(h_parent))) {
13307 +               AuWarn1("%s decode_fh returned a disconnected dentry\n",
13308 +                       au_sbtype(h_sb));
13309 +               goto out_h_parent;
13310 +       }
13311 +
13312 +       dentry = ERR_PTR(-ENOMEM);
13313 +       pathname = (void *)__get_free_page(GFP_NOFS);
13314 +       if (unlikely(!pathname))
13315 +               goto out_h_parent;
13316 +
13317 +       root = sb->s_root;
13318 +       path.mnt = h_mnt;
13319 +       di_read_lock_parent(root, !AuLock_IR);
13320 +       path.dentry = au_h_dptr(root, nsi_lock->bindex);
13321 +       di_read_unlock(root, !AuLock_IR);
13322 +       p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
13323 +       dentry = (void *)p;
13324 +       if (IS_ERR(p))
13325 +               goto out_pathname;
13326 +
13327 +       si_read_unlock(sb);
13328 +       err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
13329 +       dentry = ERR_PTR(err);
13330 +       if (unlikely(err))
13331 +               goto out_relock;
13332 +
13333 +       dentry = ERR_PTR(-ENOENT);
13334 +       AuDebugOn(au_test_anon(path.dentry));
13335 +       if (unlikely(d_really_is_negative(path.dentry)))
13336 +               goto out_path;
13337 +
13338 +       if (ino != d_inode(path.dentry)->i_ino)
13339 +               dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
13340 +       else
13341 +               dentry = dget(path.dentry);
13342 +
13343 +out_path:
13344 +       path_put(&path);
13345 +out_relock:
13346 +       if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
13347 +               if (!IS_ERR(dentry)) {
13348 +                       dput(dentry);
13349 +                       dentry = ERR_PTR(-ESTALE);
13350 +               }
13351 +out_pathname:
13352 +       free_page((unsigned long)pathname);
13353 +out_h_parent:
13354 +       dput(h_parent);
13355 +out:
13356 +       AuTraceErrPtr(dentry);
13357 +       return dentry;
13358 +}
13359 +
13360 +/* ---------------------------------------------------------------------- */
13361 +
13362 +static struct dentry *
13363 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
13364 +                 int fh_type)
13365 +{
13366 +       struct dentry *dentry;
13367 +       __u32 *fh = fid->raw;
13368 +       struct au_branch *br;
13369 +       ino_t ino, dir_ino;
13370 +       struct au_nfsd_si_lock nsi_lock = {
13371 +               .force_lock     = 0
13372 +       };
13373 +
13374 +       dentry = ERR_PTR(-ESTALE);
13375 +       /* it should never happen, but the file handle is unreliable */
13376 +       if (unlikely(fh_len < Fh_tail))
13377 +               goto out;
13378 +       nsi_lock.sigen = fh[Fh_sigen];
13379 +       nsi_lock.br_id = fh[Fh_br_id];
13380 +
13381 +       /* branch id may be wrapped around */
13382 +       br = NULL;
13383 +       if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
13384 +               goto out;
13385 +       nsi_lock.force_lock = 1;
13386 +
13387 +       /* is this inode still cached? */
13388 +       ino = decode_ino(fh + Fh_ino);
13389 +       /* it should never happen */
13390 +       if (unlikely(ino == AUFS_ROOT_INO))
13391 +               goto out_unlock;
13392 +
13393 +       dir_ino = decode_ino(fh + Fh_dir_ino);
13394 +       dentry = decode_by_ino(sb, ino, dir_ino);
13395 +       if (IS_ERR(dentry))
13396 +               goto out_unlock;
13397 +       if (dentry)
13398 +               goto accept;
13399 +
13400 +       /* is the parent dir cached? */
13401 +       br = au_sbr(sb, nsi_lock.bindex);
13402 +       au_lcnt_inc(&br->br_nfiles);
13403 +       dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
13404 +       if (IS_ERR(dentry))
13405 +               goto out_unlock;
13406 +       if (dentry)
13407 +               goto accept;
13408 +
13409 +       /* lookup path */
13410 +       dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
13411 +       if (IS_ERR(dentry))
13412 +               goto out_unlock;
13413 +       if (unlikely(!dentry))
13414 +               /* todo?: make it ESTALE */
13415 +               goto out_unlock;
13416 +
13417 +accept:
13418 +       if (!au_digen_test(dentry, au_sigen(sb))
13419 +           && d_inode(dentry)->i_generation == fh[Fh_igen])
13420 +               goto out_unlock; /* success */
13421 +
13422 +       dput(dentry);
13423 +       dentry = ERR_PTR(-ESTALE);
13424 +out_unlock:
13425 +       if (br)
13426 +               au_lcnt_dec(&br->br_nfiles);
13427 +       si_read_unlock(sb);
13428 +out:
13429 +       AuTraceErrPtr(dentry);
13430 +       return dentry;
13431 +}
13432 +
13433 +#if 0 /* reserved for future use */
13434 +/* support subtreecheck option */
13435 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
13436 +                                       int fh_len, int fh_type)
13437 +{
13438 +       struct dentry *parent;
13439 +       __u32 *fh = fid->raw;
13440 +       ino_t dir_ino;
13441 +
13442 +       dir_ino = decode_ino(fh + Fh_dir_ino);
13443 +       parent = decode_by_ino(sb, dir_ino, 0);
13444 +       if (IS_ERR(parent))
13445 +               goto out;
13446 +       if (!parent)
13447 +               parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
13448 +                                       dir_ino, fh, fh_len);
13449 +
13450 +out:
13451 +       AuTraceErrPtr(parent);
13452 +       return parent;
13453 +}
13454 +#endif
13455 +
13456 +/* ---------------------------------------------------------------------- */
13457 +
13458 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
13459 +                         struct inode *dir)
13460 +{
13461 +       int err;
13462 +       aufs_bindex_t bindex;
13463 +       struct super_block *sb, *h_sb;
13464 +       struct dentry *dentry, *parent, *h_parent;
13465 +       struct inode *h_dir;
13466 +       struct au_branch *br;
13467 +
13468 +       err = -ENOSPC;
13469 +       if (unlikely(*max_len <= Fh_tail)) {
13470 +               AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
13471 +               goto out;
13472 +       }
13473 +
13474 +       err = FILEID_ROOT;
13475 +       if (inode->i_ino == AUFS_ROOT_INO) {
13476 +               AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
13477 +               goto out;
13478 +       }
13479 +
13480 +       h_parent = NULL;
13481 +       sb = inode->i_sb;
13482 +       err = si_read_lock(sb, AuLock_FLUSH);
13483 +       if (unlikely(err))
13484 +               goto out;
13485 +
13486 +#ifdef CONFIG_AUFS_DEBUG
13487 +       if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
13488 +               AuWarn1("NFS-exporting requires xino\n");
13489 +#endif
13490 +       err = -EIO;
13491 +       parent = NULL;
13492 +       ii_read_lock_child(inode);
13493 +       bindex = au_ibtop(inode);
13494 +       if (!dir) {
13495 +               dentry = d_find_any_alias(inode);
13496 +               if (unlikely(!dentry))
13497 +                       goto out_unlock;
13498 +               AuDebugOn(au_test_anon(dentry));
13499 +               parent = dget_parent(dentry);
13500 +               dput(dentry);
13501 +               if (unlikely(!parent))
13502 +                       goto out_unlock;
13503 +               if (d_really_is_positive(parent))
13504 +                       dir = d_inode(parent);
13505 +       }
13506 +
13507 +       ii_read_lock_parent(dir);
13508 +       h_dir = au_h_iptr(dir, bindex);
13509 +       ii_read_unlock(dir);
13510 +       if (unlikely(!h_dir))
13511 +               goto out_parent;
13512 +       h_parent = d_find_any_alias(h_dir);
13513 +       if (unlikely(!h_parent))
13514 +               goto out_hparent;
13515 +
13516 +       err = -EPERM;
13517 +       br = au_sbr(sb, bindex);
13518 +       h_sb = au_br_sb(br);
13519 +       if (unlikely(!h_sb->s_export_op)) {
13520 +               AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
13521 +               goto out_hparent;
13522 +       }
13523 +
13524 +       fh[Fh_br_id] = br->br_id;
13525 +       fh[Fh_sigen] = au_sigen(sb);
13526 +       encode_ino(fh + Fh_ino, inode->i_ino);
13527 +       encode_ino(fh + Fh_dir_ino, dir->i_ino);
13528 +       fh[Fh_igen] = inode->i_generation;
13529 +
13530 +       *max_len -= Fh_tail;
13531 +       fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
13532 +                                          max_len,
13533 +                                          /*connectable or subtreecheck*/0);
13534 +       err = fh[Fh_h_type];
13535 +       *max_len += Fh_tail;
13536 +       /* todo: macros? */
13537 +       if (err != FILEID_INVALID)
13538 +               err = 99;
13539 +       else
13540 +               AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
13541 +
13542 +out_hparent:
13543 +       dput(h_parent);
13544 +out_parent:
13545 +       dput(parent);
13546 +out_unlock:
13547 +       ii_read_unlock(inode);
13548 +       si_read_unlock(sb);
13549 +out:
13550 +       if (unlikely(err < 0))
13551 +               err = FILEID_INVALID;
13552 +       return err;
13553 +}
13554 +
13555 +/* ---------------------------------------------------------------------- */
13556 +
13557 +static int aufs_commit_metadata(struct inode *inode)
13558 +{
13559 +       int err;
13560 +       aufs_bindex_t bindex;
13561 +       struct super_block *sb;
13562 +       struct inode *h_inode;
13563 +       int (*f)(struct inode *inode);
13564 +
13565 +       sb = inode->i_sb;
13566 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13567 +       ii_write_lock_child(inode);
13568 +       bindex = au_ibtop(inode);
13569 +       AuDebugOn(bindex < 0);
13570 +       h_inode = au_h_iptr(inode, bindex);
13571 +
13572 +       f = h_inode->i_sb->s_export_op->commit_metadata;
13573 +       if (f)
13574 +               err = f(h_inode);
13575 +       else
13576 +               err = sync_inode_metadata(h_inode, /*wait*/1);
13577 +
13578 +       au_cpup_attr_timesizes(inode);
13579 +       ii_write_unlock(inode);
13580 +       si_read_unlock(sb);
13581 +       return err;
13582 +}
13583 +
13584 +/* ---------------------------------------------------------------------- */
13585 +
13586 +static struct export_operations aufs_export_op = {
13587 +       .fh_to_dentry           = aufs_fh_to_dentry,
13588 +       /* .fh_to_parent        = aufs_fh_to_parent, */
13589 +       .encode_fh              = aufs_encode_fh,
13590 +       .commit_metadata        = aufs_commit_metadata
13591 +};
13592 +
13593 +void au_export_init(struct super_block *sb)
13594 +{
13595 +       struct au_sbinfo *sbinfo;
13596 +       __u32 u;
13597 +
13598 +       BUILD_BUG_ON_MSG(IS_BUILTIN(CONFIG_AUFS_FS)
13599 +                        && IS_MODULE(CONFIG_EXPORTFS),
13600 +                        AUFS_NAME ": unsupported configuration "
13601 +                        "CONFIG_EXPORTFS=m and CONFIG_AUFS_FS=y");
13602 +
13603 +       sb->s_export_op = &aufs_export_op;
13604 +       sbinfo = au_sbi(sb);
13605 +       sbinfo->si_xigen = NULL;
13606 +       get_random_bytes(&u, sizeof(u));
13607 +       BUILD_BUG_ON(sizeof(u) != sizeof(int));
13608 +       atomic_set(&sbinfo->si_xigen_next, u);
13609 +}
13610 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
13611 --- /usr/share/empty/fs/aufs/fhsm.c     1970-01-01 01:00:00.000000000 +0100
13612 +++ linux/fs/aufs/fhsm.c        2022-11-05 23:02:18.962555950 +0100
13613 @@ -0,0 +1,426 @@
13614 +// SPDX-License-Identifier: GPL-2.0
13615 +/*
13616 + * Copyright (C) 2011-2022 Junjiro R. Okajima
13617 + *
13618 + * This program is free software; you can redistribute it and/or modify
13619 + * it under the terms of the GNU General Public License as published by
13620 + * the Free Software Foundation; either version 2 of the License, or
13621 + * (at your option) any later version.
13622 + *
13623 + * This program is distributed in the hope that it will be useful,
13624 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13625 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
13626 + * GNU General Public License for more details.
13627 + *
13628 + * You should have received a copy of the GNU General Public License
13629 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
13630 + */
13631 +
13632 +/*
13633 + * File-based Hierarchy Storage Management
13634 + */
13635 +
13636 +#include <linux/anon_inodes.h>
13637 +#include <linux/poll.h>
13638 +#include <linux/seq_file.h>
13639 +#include <linux/statfs.h>
13640 +#include "aufs.h"
13641 +
13642 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
13643 +{
13644 +       struct au_sbinfo *sbinfo;
13645 +       struct au_fhsm *fhsm;
13646 +
13647 +       SiMustAnyLock(sb);
13648 +
13649 +       sbinfo = au_sbi(sb);
13650 +       fhsm = &sbinfo->si_fhsm;
13651 +       AuDebugOn(!fhsm);
13652 +       return fhsm->fhsm_bottom;
13653 +}
13654 +
13655 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
13656 +{
13657 +       struct au_sbinfo *sbinfo;
13658 +       struct au_fhsm *fhsm;
13659 +
13660 +       SiMustWriteLock(sb);
13661 +
13662 +       sbinfo = au_sbi(sb);
13663 +       fhsm = &sbinfo->si_fhsm;
13664 +       AuDebugOn(!fhsm);
13665 +       fhsm->fhsm_bottom = bindex;
13666 +}
13667 +
13668 +/* ---------------------------------------------------------------------- */
13669 +
13670 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
13671 +{
13672 +       struct au_br_fhsm *bf;
13673 +
13674 +       bf = br->br_fhsm;
13675 +       MtxMustLock(&bf->bf_lock);
13676 +
13677 +       return !bf->bf_readable
13678 +               || time_after(jiffies,
13679 +                             bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
13680 +}
13681 +
13682 +/* ---------------------------------------------------------------------- */
13683 +
13684 +static void au_fhsm_notify(struct super_block *sb, int val)
13685 +{
13686 +       struct au_sbinfo *sbinfo;
13687 +       struct au_fhsm *fhsm;
13688 +
13689 +       SiMustAnyLock(sb);
13690 +
13691 +       sbinfo = au_sbi(sb);
13692 +       fhsm = &sbinfo->si_fhsm;
13693 +       if (au_fhsm_pid(fhsm)
13694 +           && atomic_read(&fhsm->fhsm_readable) != -1) {
13695 +               atomic_set(&fhsm->fhsm_readable, val);
13696 +               if (val)
13697 +                       wake_up(&fhsm->fhsm_wqh);
13698 +       }
13699 +}
13700 +
13701 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
13702 +                       struct aufs_stfs *rstfs, int do_lock, int do_notify)
13703 +{
13704 +       int err;
13705 +       struct au_branch *br;
13706 +       struct au_br_fhsm *bf;
13707 +
13708 +       br = au_sbr(sb, bindex);
13709 +       AuDebugOn(au_br_rdonly(br));
13710 +       bf = br->br_fhsm;
13711 +       AuDebugOn(!bf);
13712 +
13713 +       if (do_lock)
13714 +               mutex_lock(&bf->bf_lock);
13715 +       else
13716 +               MtxMustLock(&bf->bf_lock);
13717 +
13718 +       /* sb->s_root for NFS is unreliable */
13719 +       err = au_br_stfs(br, &bf->bf_stfs);
13720 +       if (unlikely(err)) {
13721 +               AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
13722 +               goto out;
13723 +       }
13724 +
13725 +       bf->bf_jiffy = jiffies;
13726 +       bf->bf_readable = 1;
13727 +       if (do_notify)
13728 +               au_fhsm_notify(sb, /*val*/1);
13729 +       if (rstfs)
13730 +               *rstfs = bf->bf_stfs;
13731 +
13732 +out:
13733 +       if (do_lock)
13734 +               mutex_unlock(&bf->bf_lock);
13735 +       au_fhsm_notify(sb, /*val*/1);
13736 +
13737 +       return err;
13738 +}
13739 +
13740 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
13741 +{
13742 +       int err;
13743 +       struct au_sbinfo *sbinfo;
13744 +       struct au_fhsm *fhsm;
13745 +       struct au_branch *br;
13746 +       struct au_br_fhsm *bf;
13747 +
13748 +       AuDbg("b%d, force %d\n", bindex, force);
13749 +       SiMustAnyLock(sb);
13750 +
13751 +       sbinfo = au_sbi(sb);
13752 +       fhsm = &sbinfo->si_fhsm;
13753 +       if (!au_ftest_si(sbinfo, FHSM)
13754 +           || fhsm->fhsm_bottom == bindex)
13755 +               return;
13756 +
13757 +       br = au_sbr(sb, bindex);
13758 +       bf = br->br_fhsm;
13759 +       AuDebugOn(!bf);
13760 +       mutex_lock(&bf->bf_lock);
13761 +       if (force
13762 +           || au_fhsm_pid(fhsm)
13763 +           || au_fhsm_test_jiffy(sbinfo, br))
13764 +               err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
13765 +                                 /*do_notify*/1);
13766 +       mutex_unlock(&bf->bf_lock);
13767 +}
13768 +
13769 +void au_fhsm_wrote_all(struct super_block *sb, int force)
13770 +{
13771 +       aufs_bindex_t bindex, bbot;
13772 +       struct au_branch *br;
13773 +
13774 +       /* exclude the bottom */
13775 +       bbot = au_fhsm_bottom(sb);
13776 +       for (bindex = 0; bindex < bbot; bindex++) {
13777 +               br = au_sbr(sb, bindex);
13778 +               if (au_br_fhsm(br->br_perm))
13779 +                       au_fhsm_wrote(sb, bindex, force);
13780 +       }
13781 +}
13782 +
13783 +/* ---------------------------------------------------------------------- */
13784 +
13785 +static __poll_t au_fhsm_poll(struct file *file, struct poll_table_struct *wait)
13786 +{
13787 +       __poll_t mask;
13788 +       struct au_sbinfo *sbinfo;
13789 +       struct au_fhsm *fhsm;
13790 +
13791 +       mask = 0;
13792 +       sbinfo = file->private_data;
13793 +       fhsm = &sbinfo->si_fhsm;
13794 +       poll_wait(file, &fhsm->fhsm_wqh, wait);
13795 +       if (atomic_read(&fhsm->fhsm_readable))
13796 +               mask = EPOLLIN /* | EPOLLRDNORM */;
13797 +
13798 +       if (!mask)
13799 +               AuDbg("mask 0x%x\n", mask);
13800 +       return mask;
13801 +}
13802 +
13803 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
13804 +                             struct aufs_stfs *stfs, __s16 brid)
13805 +{
13806 +       int err;
13807 +
13808 +       err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
13809 +       if (!err)
13810 +               err = __put_user(brid, &stbr->brid);
13811 +       if (unlikely(err))
13812 +               err = -EFAULT;
13813 +
13814 +       return err;
13815 +}
13816 +
13817 +static ssize_t au_fhsm_do_read(struct super_block *sb,
13818 +                              struct aufs_stbr __user *stbr, size_t count)
13819 +{
13820 +       ssize_t err;
13821 +       int nstbr;
13822 +       aufs_bindex_t bindex, bbot;
13823 +       struct au_branch *br;
13824 +       struct au_br_fhsm *bf;
13825 +
13826 +       /* except the bottom branch */
13827 +       err = 0;
13828 +       nstbr = 0;
13829 +       bbot = au_fhsm_bottom(sb);
13830 +       for (bindex = 0; !err && bindex < bbot; bindex++) {
13831 +               br = au_sbr(sb, bindex);
13832 +               if (!au_br_fhsm(br->br_perm))
13833 +                       continue;
13834 +
13835 +               bf = br->br_fhsm;
13836 +               mutex_lock(&bf->bf_lock);
13837 +               if (bf->bf_readable) {
13838 +                       err = -EFAULT;
13839 +                       if (count >= sizeof(*stbr))
13840 +                               err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
13841 +                                                         br->br_id);
13842 +                       if (!err) {
13843 +                               bf->bf_readable = 0;
13844 +                               count -= sizeof(*stbr);
13845 +                               nstbr++;
13846 +                       }
13847 +               }
13848 +               mutex_unlock(&bf->bf_lock);
13849 +       }
13850 +       if (!err)
13851 +               err = sizeof(*stbr) * nstbr;
13852 +
13853 +       return err;
13854 +}
13855 +
13856 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
13857 +                          loff_t *pos)
13858 +{
13859 +       ssize_t err;
13860 +       int readable;
13861 +       aufs_bindex_t nfhsm, bindex, bbot;
13862 +       struct au_sbinfo *sbinfo;
13863 +       struct au_fhsm *fhsm;
13864 +       struct au_branch *br;
13865 +       struct super_block *sb;
13866 +
13867 +       err = 0;
13868 +       sbinfo = file->private_data;
13869 +       fhsm = &sbinfo->si_fhsm;
13870 +need_data:
13871 +       spin_lock_irq(&fhsm->fhsm_wqh.lock);
13872 +       if (!atomic_read(&fhsm->fhsm_readable)) {
13873 +               if (vfsub_file_flags(file) & O_NONBLOCK)
13874 +                       err = -EAGAIN;
13875 +               else
13876 +                       err = wait_event_interruptible_locked_irq
13877 +                               (fhsm->fhsm_wqh,
13878 +                                atomic_read(&fhsm->fhsm_readable));
13879 +       }
13880 +       spin_unlock_irq(&fhsm->fhsm_wqh.lock);
13881 +       if (unlikely(err))
13882 +               goto out;
13883 +
13884 +       /* sb may already be dead */
13885 +       au_rw_read_lock(&sbinfo->si_rwsem);
13886 +       readable = atomic_read(&fhsm->fhsm_readable);
13887 +       if (readable > 0) {
13888 +               sb = sbinfo->si_sb;
13889 +               AuDebugOn(!sb);
13890 +               /* exclude the bottom branch */
13891 +               nfhsm = 0;
13892 +               bbot = au_fhsm_bottom(sb);
13893 +               for (bindex = 0; bindex < bbot; bindex++) {
13894 +                       br = au_sbr(sb, bindex);
13895 +                       if (au_br_fhsm(br->br_perm))
13896 +                               nfhsm++;
13897 +               }
13898 +               err = -EMSGSIZE;
13899 +               if (nfhsm * sizeof(struct aufs_stbr) <= count) {
13900 +                       atomic_set(&fhsm->fhsm_readable, 0);
13901 +                       err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
13902 +                                            count);
13903 +               }
13904 +       }
13905 +       au_rw_read_unlock(&sbinfo->si_rwsem);
13906 +       if (!readable)
13907 +               goto need_data;
13908 +
13909 +out:
13910 +       return err;
13911 +}
13912 +
13913 +static int au_fhsm_release(struct inode *inode, struct file *file)
13914 +{
13915 +       struct au_sbinfo *sbinfo;
13916 +       struct au_fhsm *fhsm;
13917 +
13918 +       /* sb may already be dead */
13919 +       sbinfo = file->private_data;
13920 +       fhsm = &sbinfo->si_fhsm;
13921 +       spin_lock(&fhsm->fhsm_spin);
13922 +       fhsm->fhsm_pid = 0;
13923 +       spin_unlock(&fhsm->fhsm_spin);
13924 +       kobject_put(&sbinfo->si_kobj);
13925 +
13926 +       return 0;
13927 +}
13928 +
13929 +static const struct file_operations au_fhsm_fops = {
13930 +       .owner          = THIS_MODULE,
13931 +       .llseek         = noop_llseek,
13932 +       .read           = au_fhsm_read,
13933 +       .poll           = au_fhsm_poll,
13934 +       .release        = au_fhsm_release
13935 +};
13936 +
13937 +int au_fhsm_fd(struct super_block *sb, int oflags)
13938 +{
13939 +       int err, fd;
13940 +       struct au_sbinfo *sbinfo;
13941 +       struct au_fhsm *fhsm;
13942 +
13943 +       err = -EPERM;
13944 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
13945 +               goto out;
13946 +
13947 +       err = -EINVAL;
13948 +       if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
13949 +               goto out;
13950 +
13951 +       err = 0;
13952 +       sbinfo = au_sbi(sb);
13953 +       fhsm = &sbinfo->si_fhsm;
13954 +       spin_lock(&fhsm->fhsm_spin);
13955 +       if (!fhsm->fhsm_pid)
13956 +               fhsm->fhsm_pid = current->pid;
13957 +       else
13958 +               err = -EBUSY;
13959 +       spin_unlock(&fhsm->fhsm_spin);
13960 +       if (unlikely(err))
13961 +               goto out;
13962 +
13963 +       oflags |= O_RDONLY;
13964 +       /* oflags |= FMODE_NONOTIFY; */
13965 +       fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
13966 +       err = fd;
13967 +       if (unlikely(fd < 0))
13968 +               goto out_pid;
13969 +
13970 +       /* succeed regardless 'fhsm' status */
13971 +       kobject_get(&sbinfo->si_kobj);
13972 +       si_noflush_read_lock(sb);
13973 +       if (au_ftest_si(sbinfo, FHSM))
13974 +               au_fhsm_wrote_all(sb, /*force*/0);
13975 +       si_read_unlock(sb);
13976 +       goto out; /* success */
13977 +
13978 +out_pid:
13979 +       spin_lock(&fhsm->fhsm_spin);
13980 +       fhsm->fhsm_pid = 0;
13981 +       spin_unlock(&fhsm->fhsm_spin);
13982 +out:
13983 +       AuTraceErr(err);
13984 +       return err;
13985 +}
13986 +
13987 +/* ---------------------------------------------------------------------- */
13988 +
13989 +int au_fhsm_br_alloc(struct au_branch *br)
13990 +{
13991 +       int err;
13992 +
13993 +       err = 0;
13994 +       br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
13995 +       if (br->br_fhsm)
13996 +               au_br_fhsm_init(br->br_fhsm);
13997 +       else
13998 +               err = -ENOMEM;
13999 +
14000 +       return err;
14001 +}
14002 +
14003 +/* ---------------------------------------------------------------------- */
14004 +
14005 +void au_fhsm_fin(struct super_block *sb)
14006 +{
14007 +       au_fhsm_notify(sb, /*val*/-1);
14008 +}
14009 +
14010 +void au_fhsm_init(struct au_sbinfo *sbinfo)
14011 +{
14012 +       struct au_fhsm *fhsm;
14013 +
14014 +       fhsm = &sbinfo->si_fhsm;
14015 +       spin_lock_init(&fhsm->fhsm_spin);
14016 +       init_waitqueue_head(&fhsm->fhsm_wqh);
14017 +       atomic_set(&fhsm->fhsm_readable, 0);
14018 +       fhsm->fhsm_expire
14019 +               = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
14020 +       fhsm->fhsm_bottom = -1;
14021 +}
14022 +
14023 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
14024 +{
14025 +       sbinfo->si_fhsm.fhsm_expire
14026 +               = msecs_to_jiffies(sec * MSEC_PER_SEC);
14027 +}
14028 +
14029 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
14030 +{
14031 +       unsigned int u;
14032 +
14033 +       if (!au_ftest_si(sbinfo, FHSM))
14034 +               return;
14035 +
14036 +       u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
14037 +       if (u != AUFS_FHSM_CACHE_DEF_SEC)
14038 +               seq_printf(seq, ",fhsm_sec=%u", u);
14039 +}
14040 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
14041 --- /usr/share/empty/fs/aufs/file.c     1970-01-01 01:00:00.000000000 +0100
14042 +++ linux/fs/aufs/file.c        2022-11-05 23:02:18.965889284 +0100
14043 @@ -0,0 +1,858 @@
14044 +// SPDX-License-Identifier: GPL-2.0
14045 +/*
14046 + * Copyright (C) 2005-2022 Junjiro R. Okajima
14047 + *
14048 + * This program is free software; you can redistribute it and/or modify
14049 + * it under the terms of the GNU General Public License as published by
14050 + * the Free Software Foundation; either version 2 of the License, or
14051 + * (at your option) any later version.
14052 + *
14053 + * This program is distributed in the hope that it will be useful,
14054 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14055 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14056 + * GNU General Public License for more details.
14057 + *
14058 + * You should have received a copy of the GNU General Public License
14059 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
14060 + */
14061 +
14062 +/*
14063 + * handling file/dir, and address_space operation
14064 + */
14065 +
14066 +#ifdef CONFIG_AUFS_DEBUG
14067 +#include <linux/migrate.h>
14068 +#endif
14069 +#include <linux/pagemap.h>
14070 +#include "aufs.h"
14071 +
14072 +/* drop flags for writing */
14073 +unsigned int au_file_roflags(unsigned int flags)
14074 +{
14075 +       flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
14076 +       flags |= O_RDONLY | O_NOATIME;
14077 +       return flags;
14078 +}
14079 +
14080 +/* common functions to regular file and dir */
14081 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14082 +                      struct file *file, int force_wr)
14083 +{
14084 +       struct file *h_file;
14085 +       struct dentry *h_dentry;
14086 +       struct inode *h_inode;
14087 +       struct super_block *sb;
14088 +       struct au_branch *br;
14089 +       struct path h_path;
14090 +       int err;
14091 +
14092 +       /* a race condition can happen between open and unlink/rmdir */
14093 +       h_file = ERR_PTR(-ENOENT);
14094 +       h_dentry = au_h_dptr(dentry, bindex);
14095 +       if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
14096 +               goto out;
14097 +       h_inode = d_inode(h_dentry);
14098 +       spin_lock(&h_dentry->d_lock);
14099 +       err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
14100 +               /* || !d_inode(dentry)->i_nlink */
14101 +               ;
14102 +       spin_unlock(&h_dentry->d_lock);
14103 +       if (unlikely(err))
14104 +               goto out;
14105 +
14106 +       sb = dentry->d_sb;
14107 +       br = au_sbr(sb, bindex);
14108 +       err = au_br_test_oflag(flags, br);
14109 +       h_file = ERR_PTR(err);
14110 +       if (unlikely(err))
14111 +               goto out;
14112 +
14113 +       /* drop flags for writing */
14114 +       if (au_test_ro(sb, bindex, d_inode(dentry))) {
14115 +               if (force_wr && !(flags & O_WRONLY))
14116 +                       force_wr = 0;
14117 +               flags = au_file_roflags(flags);
14118 +               if (force_wr) {
14119 +                       h_file = ERR_PTR(-EROFS);
14120 +                       flags = au_file_roflags(flags);
14121 +                       if (unlikely(vfsub_native_ro(h_inode)
14122 +                                    || IS_APPEND(h_inode)))
14123 +                               goto out;
14124 +                       flags &= ~O_ACCMODE;
14125 +                       flags |= O_WRONLY;
14126 +               }
14127 +       }
14128 +       flags &= ~O_CREAT;
14129 +       au_lcnt_inc(&br->br_nfiles);
14130 +       h_path.dentry = h_dentry;
14131 +       h_path.mnt = au_br_mnt(br);
14132 +       h_file = vfsub_dentry_open(&h_path, flags);
14133 +       if (IS_ERR(h_file))
14134 +               goto out_br;
14135 +
14136 +       if (flags & __FMODE_EXEC) {
14137 +               err = deny_write_access(h_file);
14138 +               if (unlikely(err)) {
14139 +                       fput(h_file);
14140 +                       h_file = ERR_PTR(err);
14141 +                       goto out_br;
14142 +               }
14143 +       }
14144 +       fsnotify_open(h_file);
14145 +       goto out; /* success */
14146 +
14147 +out_br:
14148 +       au_lcnt_dec(&br->br_nfiles);
14149 +out:
14150 +       return h_file;
14151 +}
14152 +
14153 +static int au_cmoo(struct dentry *dentry)
14154 +{
14155 +       int err, cmoo, matched;
14156 +       unsigned int udba;
14157 +       struct path h_path;
14158 +       struct au_pin pin;
14159 +       struct au_cp_generic cpg = {
14160 +               .dentry = dentry,
14161 +               .bdst   = -1,
14162 +               .bsrc   = -1,
14163 +               .len    = -1,
14164 +               .pin    = &pin,
14165 +               .flags  = AuCpup_DTIME | AuCpup_HOPEN
14166 +       };
14167 +       struct inode *delegated;
14168 +       struct super_block *sb;
14169 +       struct au_sbinfo *sbinfo;
14170 +       struct au_fhsm *fhsm;
14171 +       pid_t pid;
14172 +       struct au_branch *br;
14173 +       struct dentry *parent;
14174 +       struct au_hinode *hdir;
14175 +
14176 +       DiMustWriteLock(dentry);
14177 +       IiMustWriteLock(d_inode(dentry));
14178 +
14179 +       err = 0;
14180 +       if (IS_ROOT(dentry))
14181 +               goto out;
14182 +       cpg.bsrc = au_dbtop(dentry);
14183 +       if (!cpg.bsrc)
14184 +               goto out;
14185 +
14186 +       sb = dentry->d_sb;
14187 +       sbinfo = au_sbi(sb);
14188 +       fhsm = &sbinfo->si_fhsm;
14189 +       pid = au_fhsm_pid(fhsm);
14190 +       rcu_read_lock();
14191 +       matched = (pid
14192 +                  && (current->pid == pid
14193 +                      || rcu_dereference(current->real_parent)->pid == pid));
14194 +       rcu_read_unlock();
14195 +       if (matched)
14196 +               goto out;
14197 +
14198 +       br = au_sbr(sb, cpg.bsrc);
14199 +       cmoo = au_br_cmoo(br->br_perm);
14200 +       if (!cmoo)
14201 +               goto out;
14202 +       if (!d_is_reg(dentry))
14203 +               cmoo &= AuBrAttr_COO_ALL;
14204 +       if (!cmoo)
14205 +               goto out;
14206 +
14207 +       parent = dget_parent(dentry);
14208 +       di_write_lock_parent(parent);
14209 +       err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
14210 +       cpg.bdst = err;
14211 +       if (unlikely(err < 0)) {
14212 +               err = 0;        /* there is no upper writable branch */
14213 +               goto out_dgrade;
14214 +       }
14215 +       AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
14216 +
14217 +       /* do not respect the coo attrib for the target branch */
14218 +       err = au_cpup_dirs(dentry, cpg.bdst);
14219 +       if (unlikely(err))
14220 +               goto out_dgrade;
14221 +
14222 +       di_downgrade_lock(parent, AuLock_IR);
14223 +       udba = au_opt_udba(sb);
14224 +       err = au_pin(&pin, dentry, cpg.bdst, udba,
14225 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14226 +       if (unlikely(err))
14227 +               goto out_parent;
14228 +
14229 +       err = au_sio_cpup_simple(&cpg);
14230 +       au_unpin(&pin);
14231 +       if (unlikely(err))
14232 +               goto out_parent;
14233 +       if (!(cmoo & AuBrWAttr_MOO))
14234 +               goto out_parent; /* success */
14235 +
14236 +       err = au_pin(&pin, dentry, cpg.bsrc, udba,
14237 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14238 +       if (unlikely(err))
14239 +               goto out_parent;
14240 +
14241 +       h_path.mnt = au_br_mnt(br);
14242 +       h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
14243 +       hdir = au_hi(d_inode(parent), cpg.bsrc);
14244 +       delegated = NULL;
14245 +       err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
14246 +       au_unpin(&pin);
14247 +       /* todo: keep h_dentry or not? */
14248 +       if (unlikely(err == -EWOULDBLOCK)) {
14249 +               pr_warn("cannot retry for NFSv4 delegation"
14250 +                       " for an internal unlink\n");
14251 +               iput(delegated);
14252 +       }
14253 +       if (unlikely(err)) {
14254 +               pr_err("unlink %pd after coo failed (%d), ignored\n",
14255 +                      dentry, err);
14256 +               err = 0;
14257 +       }
14258 +       goto out_parent; /* success */
14259 +
14260 +out_dgrade:
14261 +       di_downgrade_lock(parent, AuLock_IR);
14262 +out_parent:
14263 +       di_read_unlock(parent, AuLock_IR);
14264 +       dput(parent);
14265 +out:
14266 +       AuTraceErr(err);
14267 +       return err;
14268 +}
14269 +
14270 +int au_do_open(struct file *file, struct au_do_open_args *args)
14271 +{
14272 +       int err, aopen = args->aopen;
14273 +       struct dentry *dentry;
14274 +       struct au_finfo *finfo;
14275 +
14276 +       if (!aopen)
14277 +               err = au_finfo_init(file, args->fidir);
14278 +       else {
14279 +               lockdep_off();
14280 +               err = au_finfo_init(file, args->fidir);
14281 +               lockdep_on();
14282 +       }
14283 +       if (unlikely(err))
14284 +               goto out;
14285 +
14286 +       dentry = file->f_path.dentry;
14287 +       AuDebugOn(IS_ERR_OR_NULL(dentry));
14288 +       di_write_lock_child(dentry);
14289 +       err = au_cmoo(dentry);
14290 +       di_downgrade_lock(dentry, AuLock_IR);
14291 +       if (!err) {
14292 +               if (!aopen)
14293 +                       err = args->open(file, vfsub_file_flags(file), NULL);
14294 +               else {
14295 +                       lockdep_off();
14296 +                       err = args->open(file, vfsub_file_flags(file),
14297 +                                        args->h_file);
14298 +                       lockdep_on();
14299 +               }
14300 +       }
14301 +       di_read_unlock(dentry, AuLock_IR);
14302 +
14303 +       finfo = au_fi(file);
14304 +       if (!err) {
14305 +               finfo->fi_file = file;
14306 +               au_hbl_add(&finfo->fi_hlist,
14307 +                          &au_sbi(file->f_path.dentry->d_sb)->si_files);
14308 +       }
14309 +       if (!aopen)
14310 +               fi_write_unlock(file);
14311 +       else {
14312 +               lockdep_off();
14313 +               fi_write_unlock(file);
14314 +               lockdep_on();
14315 +       }
14316 +       if (unlikely(err)) {
14317 +               finfo->fi_hdir = NULL;
14318 +               au_finfo_fin(file);
14319 +       }
14320 +
14321 +out:
14322 +       AuTraceErr(err);
14323 +       return err;
14324 +}
14325 +
14326 +int au_reopen_nondir(struct file *file)
14327 +{
14328 +       int err;
14329 +       aufs_bindex_t btop;
14330 +       struct dentry *dentry;
14331 +       struct au_branch *br;
14332 +       struct file *h_file, *h_file_tmp;
14333 +
14334 +       dentry = file->f_path.dentry;
14335 +       btop = au_dbtop(dentry);
14336 +       br = au_sbr(dentry->d_sb, btop);
14337 +       h_file_tmp = NULL;
14338 +       if (au_fbtop(file) == btop) {
14339 +               h_file = au_hf_top(file);
14340 +               if (file->f_mode == h_file->f_mode)
14341 +                       return 0; /* success */
14342 +               h_file_tmp = h_file;
14343 +               get_file(h_file_tmp);
14344 +               au_lcnt_inc(&br->br_nfiles);
14345 +               au_set_h_fptr(file, btop, NULL);
14346 +       }
14347 +       AuDebugOn(au_fi(file)->fi_hdir);
14348 +       /*
14349 +        * it can happen
14350 +        * file exists on both of rw and ro
14351 +        * open --> dbtop and fbtop are both 0
14352 +        * prepend a branch as rw, "rw" become ro
14353 +        * remove rw/file
14354 +        * delete the top branch, "rw" becomes rw again
14355 +        *      --> dbtop is 1, fbtop is still 0
14356 +        * write --> fbtop is 0 but dbtop is 1
14357 +        */
14358 +       /* AuDebugOn(au_fbtop(file) < btop); */
14359 +
14360 +       h_file = au_h_open(dentry, btop, vfsub_file_flags(file) & ~O_TRUNC,
14361 +                          file, /*force_wr*/0);
14362 +       err = PTR_ERR(h_file);
14363 +       if (IS_ERR(h_file)) {
14364 +               if (h_file_tmp) {
14365 +                       /* revert */
14366 +                       au_set_h_fptr(file, btop, h_file_tmp);
14367 +                       h_file_tmp = NULL;
14368 +               }
14369 +               goto out; /* todo: close all? */
14370 +       }
14371 +
14372 +       err = 0;
14373 +       au_set_fbtop(file, btop);
14374 +       au_set_h_fptr(file, btop, h_file);
14375 +       au_update_figen(file);
14376 +       /* todo: necessary? */
14377 +       /* file->f_ra = h_file->f_ra; */
14378 +
14379 +out:
14380 +       if (h_file_tmp) {
14381 +               fput(h_file_tmp);
14382 +               au_lcnt_dec(&br->br_nfiles);
14383 +       }
14384 +       return err;
14385 +}
14386 +
14387 +/* ---------------------------------------------------------------------- */
14388 +
14389 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
14390 +                       struct dentry *hi_wh)
14391 +{
14392 +       int err;
14393 +       aufs_bindex_t btop;
14394 +       struct au_dinfo *dinfo;
14395 +       struct dentry *h_dentry;
14396 +       struct au_hdentry *hdp;
14397 +
14398 +       dinfo = au_di(file->f_path.dentry);
14399 +       AuRwMustWriteLock(&dinfo->di_rwsem);
14400 +
14401 +       btop = dinfo->di_btop;
14402 +       dinfo->di_btop = btgt;
14403 +       hdp = au_hdentry(dinfo, btgt);
14404 +       h_dentry = hdp->hd_dentry;
14405 +       hdp->hd_dentry = hi_wh;
14406 +       err = au_reopen_nondir(file);
14407 +       hdp->hd_dentry = h_dentry;
14408 +       dinfo->di_btop = btop;
14409 +
14410 +       return err;
14411 +}
14412 +
14413 +static int au_ready_to_write_wh(struct file *file, loff_t len,
14414 +                               aufs_bindex_t bcpup, struct au_pin *pin)
14415 +{
14416 +       int err;
14417 +       struct inode *inode, *h_inode;
14418 +       struct dentry *h_dentry, *hi_wh;
14419 +       struct au_cp_generic cpg = {
14420 +               .dentry = file->f_path.dentry,
14421 +               .bdst   = bcpup,
14422 +               .bsrc   = -1,
14423 +               .len    = len,
14424 +               .pin    = pin
14425 +       };
14426 +
14427 +       au_update_dbtop(cpg.dentry);
14428 +       inode = d_inode(cpg.dentry);
14429 +       h_inode = NULL;
14430 +       if (au_dbtop(cpg.dentry) <= bcpup
14431 +           && au_dbbot(cpg.dentry) >= bcpup) {
14432 +               h_dentry = au_h_dptr(cpg.dentry, bcpup);
14433 +               if (h_dentry && d_is_positive(h_dentry))
14434 +                       h_inode = d_inode(h_dentry);
14435 +       }
14436 +       hi_wh = au_hi_wh(inode, bcpup);
14437 +       if (!hi_wh && !h_inode)
14438 +               err = au_sio_cpup_wh(&cpg, file);
14439 +       else
14440 +               /* already copied-up after unlink */
14441 +               err = au_reopen_wh(file, bcpup, hi_wh);
14442 +
14443 +       if (!err
14444 +           && (inode->i_nlink > 1
14445 +               || (inode->i_state & I_LINKABLE))
14446 +           && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
14447 +               au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
14448 +
14449 +       return err;
14450 +}
14451 +
14452 +/*
14453 + * prepare the @file for writing.
14454 + */
14455 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
14456 +{
14457 +       int err;
14458 +       aufs_bindex_t dbtop;
14459 +       struct dentry *parent;
14460 +       struct inode *inode;
14461 +       struct super_block *sb;
14462 +       struct file *h_file;
14463 +       struct au_cp_generic cpg = {
14464 +               .dentry = file->f_path.dentry,
14465 +               .bdst   = -1,
14466 +               .bsrc   = -1,
14467 +               .len    = len,
14468 +               .pin    = pin,
14469 +               .flags  = AuCpup_DTIME
14470 +       };
14471 +
14472 +       sb = cpg.dentry->d_sb;
14473 +       inode = d_inode(cpg.dentry);
14474 +       cpg.bsrc = au_fbtop(file);
14475 +       err = au_test_ro(sb, cpg.bsrc, inode);
14476 +       if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
14477 +               err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
14478 +                            /*flags*/0);
14479 +               goto out;
14480 +       }
14481 +
14482 +       /* need to cpup or reopen */
14483 +       parent = dget_parent(cpg.dentry);
14484 +       di_write_lock_parent(parent);
14485 +       err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14486 +       cpg.bdst = err;
14487 +       if (unlikely(err < 0))
14488 +               goto out_dgrade;
14489 +       err = 0;
14490 +
14491 +       if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
14492 +               err = au_cpup_dirs(cpg.dentry, cpg.bdst);
14493 +               if (unlikely(err))
14494 +                       goto out_dgrade;
14495 +       }
14496 +
14497 +       err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14498 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14499 +       if (unlikely(err))
14500 +               goto out_dgrade;
14501 +
14502 +       dbtop = au_dbtop(cpg.dentry);
14503 +       if (dbtop <= cpg.bdst)
14504 +               cpg.bsrc = cpg.bdst;
14505 +
14506 +       if (dbtop <= cpg.bdst           /* just reopen */
14507 +           || !d_unhashed(cpg.dentry)  /* copyup and reopen */
14508 +               ) {
14509 +               h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
14510 +               if (IS_ERR(h_file))
14511 +                       err = PTR_ERR(h_file);
14512 +               else {
14513 +                       di_downgrade_lock(parent, AuLock_IR);
14514 +                       if (dbtop > cpg.bdst)
14515 +                               err = au_sio_cpup_simple(&cpg);
14516 +                       if (!err)
14517 +                               err = au_reopen_nondir(file);
14518 +                       au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
14519 +               }
14520 +       } else {                        /* copyup as wh and reopen */
14521 +               /*
14522 +                * since writable hfsplus branch is not supported,
14523 +                * h_open_pre/post() are unnecessary.
14524 +                */
14525 +               err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
14526 +               di_downgrade_lock(parent, AuLock_IR);
14527 +       }
14528 +
14529 +       if (!err) {
14530 +               au_pin_set_parent_lflag(pin, /*lflag*/0);
14531 +               goto out_dput; /* success */
14532 +       }
14533 +       au_unpin(pin);
14534 +       goto out_unlock;
14535 +
14536 +out_dgrade:
14537 +       di_downgrade_lock(parent, AuLock_IR);
14538 +out_unlock:
14539 +       di_read_unlock(parent, AuLock_IR);
14540 +out_dput:
14541 +       dput(parent);
14542 +out:
14543 +       return err;
14544 +}
14545 +
14546 +/* ---------------------------------------------------------------------- */
14547 +
14548 +int au_do_flush(struct file *file, fl_owner_t id,
14549 +               int (*flush)(struct file *file, fl_owner_t id))
14550 +{
14551 +       int err;
14552 +       struct super_block *sb;
14553 +       struct inode *inode;
14554 +
14555 +       inode = file_inode(file);
14556 +       sb = inode->i_sb;
14557 +       si_noflush_read_lock(sb);
14558 +       fi_read_lock(file);
14559 +       ii_read_lock_child(inode);
14560 +
14561 +       err = flush(file, id);
14562 +       au_cpup_attr_timesizes(inode);
14563 +
14564 +       ii_read_unlock(inode);
14565 +       fi_read_unlock(file);
14566 +       si_read_unlock(sb);
14567 +       return err;
14568 +}
14569 +
14570 +/* ---------------------------------------------------------------------- */
14571 +
14572 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
14573 +{
14574 +       int err;
14575 +       struct au_pin pin;
14576 +       struct au_finfo *finfo;
14577 +       struct dentry *parent, *hi_wh;
14578 +       struct inode *inode;
14579 +       struct super_block *sb;
14580 +       struct au_cp_generic cpg = {
14581 +               .dentry = file->f_path.dentry,
14582 +               .bdst   = -1,
14583 +               .bsrc   = -1,
14584 +               .len    = -1,
14585 +               .pin    = &pin,
14586 +               .flags  = AuCpup_DTIME
14587 +       };
14588 +
14589 +       FiMustWriteLock(file);
14590 +
14591 +       err = 0;
14592 +       finfo = au_fi(file);
14593 +       sb = cpg.dentry->d_sb;
14594 +       inode = d_inode(cpg.dentry);
14595 +       cpg.bdst = au_ibtop(inode);
14596 +       if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
14597 +               goto out;
14598 +
14599 +       parent = dget_parent(cpg.dentry);
14600 +       if (au_test_ro(sb, cpg.bdst, inode)) {
14601 +               di_read_lock_parent(parent, !AuLock_IR);
14602 +               err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14603 +               cpg.bdst = err;
14604 +               di_read_unlock(parent, !AuLock_IR);
14605 +               if (unlikely(err < 0))
14606 +                       goto out_parent;
14607 +               err = 0;
14608 +       }
14609 +
14610 +       di_read_lock_parent(parent, AuLock_IR);
14611 +       hi_wh = au_hi_wh(inode, cpg.bdst);
14612 +       if (!S_ISDIR(inode->i_mode)
14613 +           && au_opt_test(au_mntflags(sb), PLINK)
14614 +           && au_plink_test(inode)
14615 +           && !d_unhashed(cpg.dentry)
14616 +           && cpg.bdst < au_dbtop(cpg.dentry)) {
14617 +               err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
14618 +               if (unlikely(err))
14619 +                       goto out_unlock;
14620 +
14621 +               /* always superio. */
14622 +               err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14623 +                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14624 +               if (!err) {
14625 +                       err = au_sio_cpup_simple(&cpg);
14626 +                       au_unpin(&pin);
14627 +               }
14628 +       } else if (hi_wh) {
14629 +               /* already copied-up after unlink */
14630 +               err = au_reopen_wh(file, cpg.bdst, hi_wh);
14631 +               *need_reopen = 0;
14632 +       }
14633 +
14634 +out_unlock:
14635 +       di_read_unlock(parent, AuLock_IR);
14636 +out_parent:
14637 +       dput(parent);
14638 +out:
14639 +       return err;
14640 +}
14641 +
14642 +static void au_do_refresh_dir(struct file *file)
14643 +{
14644 +       aufs_bindex_t bindex, bbot, new_bindex, brid;
14645 +       struct au_hfile *p, tmp, *q;
14646 +       struct au_finfo *finfo;
14647 +       struct super_block *sb;
14648 +       struct au_fidir *fidir;
14649 +
14650 +       FiMustWriteLock(file);
14651 +
14652 +       sb = file->f_path.dentry->d_sb;
14653 +       finfo = au_fi(file);
14654 +       fidir = finfo->fi_hdir;
14655 +       AuDebugOn(!fidir);
14656 +       p = fidir->fd_hfile + finfo->fi_btop;
14657 +       brid = p->hf_br->br_id;
14658 +       bbot = fidir->fd_bbot;
14659 +       for (bindex = finfo->fi_btop; bindex <= bbot; bindex++, p++) {
14660 +               if (!p->hf_file)
14661 +                       continue;
14662 +
14663 +               new_bindex = au_br_index(sb, p->hf_br->br_id);
14664 +               if (new_bindex == bindex)
14665 +                       continue;
14666 +               if (new_bindex < 0) {
14667 +                       au_set_h_fptr(file, bindex, NULL);
14668 +                       continue;
14669 +               }
14670 +
14671 +               /* swap two lower inode, and loop again */
14672 +               q = fidir->fd_hfile + new_bindex;
14673 +               tmp = *q;
14674 +               *q = *p;
14675 +               *p = tmp;
14676 +               if (tmp.hf_file) {
14677 +                       bindex--;
14678 +                       p--;
14679 +               }
14680 +       }
14681 +
14682 +       p = fidir->fd_hfile;
14683 +       if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
14684 +               bbot = au_sbbot(sb);
14685 +               for (finfo->fi_btop = 0; finfo->fi_btop <= bbot;
14686 +                    finfo->fi_btop++, p++)
14687 +                       if (p->hf_file) {
14688 +                               if (file_inode(p->hf_file))
14689 +                                       break;
14690 +                               au_hfput(p, /*execed*/0);
14691 +                       }
14692 +       } else {
14693 +               bbot = au_br_index(sb, brid);
14694 +               for (finfo->fi_btop = 0; finfo->fi_btop < bbot;
14695 +                    finfo->fi_btop++, p++)
14696 +                       if (p->hf_file)
14697 +                               au_hfput(p, /*execed*/0);
14698 +               bbot = au_sbbot(sb);
14699 +       }
14700 +
14701 +       p = fidir->fd_hfile + bbot;
14702 +       for (fidir->fd_bbot = bbot; fidir->fd_bbot >= finfo->fi_btop;
14703 +            fidir->fd_bbot--, p--)
14704 +               if (p->hf_file) {
14705 +                       if (file_inode(p->hf_file))
14706 +                               break;
14707 +                       au_hfput(p, /*execed*/0);
14708 +               }
14709 +       AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
14710 +}
14711 +
14712 +/*
14713 + * after branch manipulating, refresh the file.
14714 + */
14715 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
14716 +{
14717 +       int err, need_reopen, nbr;
14718 +       aufs_bindex_t bbot, bindex;
14719 +       struct dentry *dentry;
14720 +       struct super_block *sb;
14721 +       struct au_finfo *finfo;
14722 +       struct au_hfile *hfile;
14723 +
14724 +       dentry = file->f_path.dentry;
14725 +       sb = dentry->d_sb;
14726 +       nbr = au_sbbot(sb) + 1;
14727 +       finfo = au_fi(file);
14728 +       if (!finfo->fi_hdir) {
14729 +               hfile = &finfo->fi_htop;
14730 +               AuDebugOn(!hfile->hf_file);
14731 +               bindex = au_br_index(sb, hfile->hf_br->br_id);
14732 +               AuDebugOn(bindex < 0);
14733 +               if (bindex != finfo->fi_btop)
14734 +                       au_set_fbtop(file, bindex);
14735 +       } else {
14736 +               err = au_fidir_realloc(finfo, nbr, /*may_shrink*/0);
14737 +               if (unlikely(err))
14738 +                       goto out;
14739 +               au_do_refresh_dir(file);
14740 +       }
14741 +
14742 +       err = 0;
14743 +       need_reopen = 1;
14744 +       if (!au_test_mmapped(file))
14745 +               err = au_file_refresh_by_inode(file, &need_reopen);
14746 +       if (finfo->fi_hdir)
14747 +               /* harmless if err */
14748 +               au_fidir_realloc(finfo, nbr, /*may_shrink*/1);
14749 +       if (!err && need_reopen && !d_unlinked(dentry))
14750 +               err = reopen(file);
14751 +       if (!err) {
14752 +               au_update_figen(file);
14753 +               goto out; /* success */
14754 +       }
14755 +
14756 +       /* error, close all lower files */
14757 +       if (finfo->fi_hdir) {
14758 +               bbot = au_fbbot_dir(file);
14759 +               for (bindex = au_fbtop(file); bindex <= bbot; bindex++)
14760 +                       au_set_h_fptr(file, bindex, NULL);
14761 +       }
14762 +
14763 +out:
14764 +       return err;
14765 +}
14766 +
14767 +/* common function to regular file and dir */
14768 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14769 +                         int wlock, unsigned int fi_lsc)
14770 +{
14771 +       int err;
14772 +       unsigned int sigen, figen;
14773 +       aufs_bindex_t btop;
14774 +       unsigned char pseudo_link;
14775 +       struct dentry *dentry;
14776 +       struct inode *inode;
14777 +
14778 +       err = 0;
14779 +       dentry = file->f_path.dentry;
14780 +       inode = d_inode(dentry);
14781 +       sigen = au_sigen(dentry->d_sb);
14782 +       fi_write_lock_nested(file, fi_lsc);
14783 +       figen = au_figen(file);
14784 +       if (!fi_lsc)
14785 +               di_write_lock_child(dentry);
14786 +       else
14787 +               di_write_lock_child2(dentry);
14788 +       btop = au_dbtop(dentry);
14789 +       pseudo_link = (btop != au_ibtop(inode));
14790 +       if (sigen == figen && !pseudo_link && au_fbtop(file) == btop) {
14791 +               if (!wlock) {
14792 +                       di_downgrade_lock(dentry, AuLock_IR);
14793 +                       fi_downgrade_lock(file);
14794 +               }
14795 +               goto out; /* success */
14796 +       }
14797 +
14798 +       AuDbg("sigen %d, figen %d\n", sigen, figen);
14799 +       if (au_digen_test(dentry, sigen)) {
14800 +               err = au_reval_dpath(dentry, sigen);
14801 +               AuDebugOn(!err && au_digen_test(dentry, sigen));
14802 +       }
14803 +
14804 +       if (!err)
14805 +               err = refresh_file(file, reopen);
14806 +       if (!err) {
14807 +               if (!wlock) {
14808 +                       di_downgrade_lock(dentry, AuLock_IR);
14809 +                       fi_downgrade_lock(file);
14810 +               }
14811 +       } else {
14812 +               di_write_unlock(dentry);
14813 +               fi_write_unlock(file);
14814 +       }
14815 +
14816 +out:
14817 +       return err;
14818 +}
14819 +
14820 +/* ---------------------------------------------------------------------- */
14821 +
14822 +/* cf. aufs_nopage() */
14823 +/* for madvise(2) */
14824 +static int aufs_read_folio(struct file *file __maybe_unused, struct folio *folio)
14825 +{
14826 +       folio_unlock(folio);
14827 +       return 0;
14828 +}
14829 +
14830 +/* it will never be called, but necessary to support O_DIRECT */
14831 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
14832 +{ BUG(); return 0; }
14833 +
14834 +/* they will never be called. */
14835 +#ifdef CONFIG_AUFS_DEBUG
14836 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
14837 +                           loff_t pos, unsigned len,
14838 +                           struct page **pagep, void **fsdata)
14839 +{ AuUnsupport(); return 0; }
14840 +static int aufs_write_end(struct file *file, struct address_space *mapping,
14841 +                         loff_t pos, unsigned len, unsigned copied,
14842 +                         struct page *page, void *fsdata)
14843 +{ AuUnsupport(); return 0; }
14844 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
14845 +{ AuUnsupport(); return 0; }
14846 +
14847 +static bool aufs_dirty_folio(struct address_space *mapping, struct folio *folio)
14848 +{ AuUnsupport(); return true; }
14849 +static void aufs_invalidate_folio(struct folio *folio, size_t offset, size_t len)
14850 +{ AuUnsupport(); }
14851 +static bool aufs_release_folio(struct folio *folio, gfp_t gfp)
14852 +{ AuUnsupport(); return true; }
14853 +#if 0 /* called by memory compaction regardless file */
14854 +static int aufs_migrate_folio(struct address_space *mapping, struct folio *dst,
14855 +                             struct folio *src, enum migrate_mode mode)
14856 +{ AuUnsupport(); return 0; }
14857 +#endif
14858 +static int aufs_launder_folio(struct folio *folio)
14859 +{ AuUnsupport(); return 0; }
14860 +static bool aufs_is_partially_uptodate(struct folio *folio, size_t from,
14861 +                                     size_t count)
14862 +{ AuUnsupport(); return true; }
14863 +static void aufs_is_dirty_writeback(struct folio *folio, bool *dirty,
14864 +                                   bool *writeback)
14865 +{ AuUnsupport(); }
14866 +static int aufs_error_remove_page(struct address_space *mapping,
14867 +                                 struct page *page)
14868 +{ AuUnsupport(); return 0; }
14869 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
14870 +                             sector_t *span)
14871 +{ AuUnsupport(); return 0; }
14872 +static void aufs_swap_deactivate(struct file *file)
14873 +{ AuUnsupport(); }
14874 +static int aufs_swap_rw(struct kiocb *iocb, struct iov_iter *iter)
14875 +{ AuUnsupport(); return 0; }
14876 +#endif /* CONFIG_AUFS_DEBUG */
14877 +
14878 +const struct address_space_operations aufs_aop = {
14879 +       .read_folio             = aufs_read_folio,
14880 +       .direct_IO              = aufs_direct_IO,
14881 +#ifdef CONFIG_AUFS_DEBUG
14882 +       .writepage              = aufs_writepage,
14883 +       /* no writepages, because of writepage */
14884 +       .dirty_folio            = aufs_dirty_folio,
14885 +       /* no readpages, because of readpage */
14886 +       .write_begin            = aufs_write_begin,
14887 +       .write_end              = aufs_write_end,
14888 +       /* no bmap, no block device */
14889 +       .invalidate_folio       = aufs_invalidate_folio,
14890 +       .release_folio          = aufs_release_folio,
14891 +       /* is fallback_migrate_page ok? */
14892 +       /* .migrate_folio       = aufs_migrate_folio, */
14893 +       .launder_folio          = aufs_launder_folio,
14894 +       .is_partially_uptodate  = aufs_is_partially_uptodate,
14895 +       .is_dirty_writeback     = aufs_is_dirty_writeback,
14896 +       .error_remove_page      = aufs_error_remove_page,
14897 +       .swap_activate          = aufs_swap_activate,
14898 +       .swap_deactivate        = aufs_swap_deactivate,
14899 +       .swap_rw                = aufs_swap_rw
14900 +#endif /* CONFIG_AUFS_DEBUG */
14901 +};
14902 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
14903 --- /usr/share/empty/fs/aufs/file.h     1970-01-01 01:00:00.000000000 +0100
14904 +++ linux/fs/aufs/file.h        2022-11-05 23:02:18.965889284 +0100
14905 @@ -0,0 +1,342 @@
14906 +/* SPDX-License-Identifier: GPL-2.0 */
14907 +/*
14908 + * Copyright (C) 2005-2022 Junjiro R. Okajima
14909 + *
14910 + * This program is free software; you can redistribute it and/or modify
14911 + * it under the terms of the GNU General Public License as published by
14912 + * the Free Software Foundation; either version 2 of the License, or
14913 + * (at your option) any later version.
14914 + *
14915 + * This program is distributed in the hope that it will be useful,
14916 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14917 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14918 + * GNU General Public License for more details.
14919 + *
14920 + * You should have received a copy of the GNU General Public License
14921 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
14922 + */
14923 +
14924 +/*
14925 + * file operations
14926 + */
14927 +
14928 +#ifndef __AUFS_FILE_H__
14929 +#define __AUFS_FILE_H__
14930 +
14931 +#ifdef __KERNEL__
14932 +
14933 +#include <linux/file.h>
14934 +#include <linux/fs.h>
14935 +#include <linux/mm_types.h>
14936 +#include <linux/poll.h>
14937 +#include "rwsem.h"
14938 +
14939 +struct au_branch;
14940 +struct au_hfile {
14941 +       struct file             *hf_file;
14942 +       struct au_branch        *hf_br;
14943 +};
14944 +
14945 +struct au_vdir;
14946 +struct au_fidir {
14947 +       aufs_bindex_t           fd_bbot;
14948 +       aufs_bindex_t           fd_nent;
14949 +       struct au_vdir          *fd_vdir_cache;
14950 +       struct au_hfile         fd_hfile[];
14951 +};
14952 +
14953 +static inline int au_fidir_sz(int nent)
14954 +{
14955 +       AuDebugOn(nent < 0);
14956 +       return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
14957 +}
14958 +
14959 +struct au_finfo {
14960 +       atomic_t                fi_generation;
14961 +
14962 +       struct au_rwsem         fi_rwsem;
14963 +       aufs_bindex_t           fi_btop;
14964 +
14965 +       /* do not union them */
14966 +       struct {                                /* for non-dir */
14967 +               struct au_hfile                 fi_htop;
14968 +               atomic_t                        fi_mmapped;
14969 +       };
14970 +       struct au_fidir         *fi_hdir;       /* for dir only */
14971 +
14972 +       struct hlist_bl_node    fi_hlist;
14973 +       struct file             *fi_file;       /* very ugly */
14974 +       struct rcu_head         rcu;
14975 +} ____cacheline_aligned_in_smp;
14976 +
14977 +/* ---------------------------------------------------------------------- */
14978 +
14979 +/* file.c */
14980 +extern const struct address_space_operations aufs_aop;
14981 +unsigned int au_file_roflags(unsigned int flags);
14982 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14983 +                      struct file *file, int force_wr);
14984 +struct au_do_open_args {
14985 +       int             aopen;
14986 +       int             (*open)(struct file *file, int flags,
14987 +                               struct file *h_file);
14988 +       struct au_fidir *fidir;
14989 +       struct file     *h_file;
14990 +};
14991 +int au_do_open(struct file *file, struct au_do_open_args *args);
14992 +int au_reopen_nondir(struct file *file);
14993 +struct au_pin;
14994 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
14995 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14996 +                         int wlock, unsigned int fi_lsc);
14997 +int au_do_flush(struct file *file, fl_owner_t id,
14998 +               int (*flush)(struct file *file, fl_owner_t id));
14999 +
15000 +/* poll.c */
15001 +#ifdef CONFIG_AUFS_POLL
15002 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt);
15003 +#endif
15004 +
15005 +#ifdef CONFIG_AUFS_BR_HFSPLUS
15006 +/* hfsplus.c */
15007 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
15008 +                          int force_wr);
15009 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
15010 +                   struct file *h_file);
15011 +#else
15012 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
15013 +       aufs_bindex_t bindex, int force_wr)
15014 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
15015 +          struct file *h_file);
15016 +#endif
15017 +
15018 +/* f_op.c */
15019 +extern const struct file_operations aufs_file_fop;
15020 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
15021 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
15022 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc);
15023 +
15024 +/* finfo.c */
15025 +void au_hfput(struct au_hfile *hf, int execed);
15026 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
15027 +                  struct file *h_file);
15028 +
15029 +void au_update_figen(struct file *file);
15030 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
15031 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink);
15032 +
15033 +void au_fi_init_once(void *_fi);
15034 +void au_finfo_fin(struct file *file);
15035 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
15036 +
15037 +/* ioctl.c */
15038 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
15039 +#ifdef CONFIG_COMPAT
15040 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15041 +                          unsigned long arg);
15042 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15043 +                             unsigned long arg);
15044 +#endif
15045 +
15046 +/* ---------------------------------------------------------------------- */
15047 +
15048 +static inline struct au_finfo *au_fi(struct file *file)
15049 +{
15050 +       return file->private_data;
15051 +}
15052 +
15053 +/* ---------------------------------------------------------------------- */
15054 +
15055 +#define fi_read_lock(f)        au_rw_read_lock(&au_fi(f)->fi_rwsem)
15056 +#define fi_write_lock(f)       au_rw_write_lock(&au_fi(f)->fi_rwsem)
15057 +#define fi_read_trylock(f)     au_rw_read_trylock(&au_fi(f)->fi_rwsem)
15058 +#define fi_write_trylock(f)    au_rw_write_trylock(&au_fi(f)->fi_rwsem)
15059 +/*
15060 +#define fi_read_trylock_nested(f) \
15061 +       au_rw_read_trylock_nested(&au_fi(f)->fi_rwsem)
15062 +#define fi_write_trylock_nested(f) \
15063 +       au_rw_write_trylock_nested(&au_fi(f)->fi_rwsem)
15064 +*/
15065 +
15066 +#define fi_read_unlock(f)      au_rw_read_unlock(&au_fi(f)->fi_rwsem)
15067 +#define fi_write_unlock(f)     au_rw_write_unlock(&au_fi(f)->fi_rwsem)
15068 +#define fi_downgrade_lock(f)   au_rw_dgrade_lock(&au_fi(f)->fi_rwsem)
15069 +
15070 +/* lock subclass for finfo */
15071 +enum {
15072 +       AuLsc_FI_1,
15073 +       AuLsc_FI_2
15074 +};
15075 +
15076 +static inline void fi_read_lock_nested(struct file *f, unsigned int lsc)
15077 +{
15078 +       au_rw_read_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15079 +}
15080 +
15081 +static inline void fi_write_lock_nested(struct file *f, unsigned int lsc)
15082 +{
15083 +       au_rw_write_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15084 +}
15085 +
15086 +/*
15087 + * fi_read_lock_1, fi_write_lock_1,
15088 + * fi_read_lock_2, fi_write_lock_2
15089 + */
15090 +#define AuReadLockFunc(name) \
15091 +static inline void fi_read_lock_##name(struct file *f) \
15092 +{ fi_read_lock_nested(f, AuLsc_FI_##name); }
15093 +
15094 +#define AuWriteLockFunc(name) \
15095 +static inline void fi_write_lock_##name(struct file *f) \
15096 +{ fi_write_lock_nested(f, AuLsc_FI_##name); }
15097 +
15098 +#define AuRWLockFuncs(name) \
15099 +       AuReadLockFunc(name) \
15100 +       AuWriteLockFunc(name)
15101 +
15102 +AuRWLockFuncs(1);
15103 +AuRWLockFuncs(2);
15104 +
15105 +#undef AuReadLockFunc
15106 +#undef AuWriteLockFunc
15107 +#undef AuRWLockFuncs
15108 +
15109 +#define FiMustNoWaiters(f)     AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
15110 +#define FiMustAnyLock(f)       AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
15111 +#define FiMustWriteLock(f)     AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
15112 +
15113 +/* ---------------------------------------------------------------------- */
15114 +
15115 +/* todo: hard/soft set? */
15116 +static inline aufs_bindex_t au_fbtop(struct file *file)
15117 +{
15118 +       FiMustAnyLock(file);
15119 +       return au_fi(file)->fi_btop;
15120 +}
15121 +
15122 +static inline aufs_bindex_t au_fbbot_dir(struct file *file)
15123 +{
15124 +       FiMustAnyLock(file);
15125 +       AuDebugOn(!au_fi(file)->fi_hdir);
15126 +       return au_fi(file)->fi_hdir->fd_bbot;
15127 +}
15128 +
15129 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
15130 +{
15131 +       FiMustAnyLock(file);
15132 +       AuDebugOn(!au_fi(file)->fi_hdir);
15133 +       return au_fi(file)->fi_hdir->fd_vdir_cache;
15134 +}
15135 +
15136 +static inline void au_set_fbtop(struct file *file, aufs_bindex_t bindex)
15137 +{
15138 +       FiMustWriteLock(file);
15139 +       au_fi(file)->fi_btop = bindex;
15140 +}
15141 +
15142 +static inline void au_set_fbbot_dir(struct file *file, aufs_bindex_t bindex)
15143 +{
15144 +       FiMustWriteLock(file);
15145 +       AuDebugOn(!au_fi(file)->fi_hdir);
15146 +       au_fi(file)->fi_hdir->fd_bbot = bindex;
15147 +}
15148 +
15149 +static inline void au_set_fvdir_cache(struct file *file,
15150 +                                     struct au_vdir *vdir_cache)
15151 +{
15152 +       FiMustWriteLock(file);
15153 +       AuDebugOn(!au_fi(file)->fi_hdir);
15154 +       au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
15155 +}
15156 +
15157 +static inline struct file *au_hf_top(struct file *file)
15158 +{
15159 +       FiMustAnyLock(file);
15160 +       AuDebugOn(au_fi(file)->fi_hdir);
15161 +       return au_fi(file)->fi_htop.hf_file;
15162 +}
15163 +
15164 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
15165 +{
15166 +       FiMustAnyLock(file);
15167 +       AuDebugOn(!au_fi(file)->fi_hdir);
15168 +       return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
15169 +}
15170 +
15171 +/* todo: memory barrier? */
15172 +static inline unsigned int au_figen(struct file *f)
15173 +{
15174 +       return atomic_read(&au_fi(f)->fi_generation);
15175 +}
15176 +
15177 +static inline void au_set_mmapped(struct file *f)
15178 +{
15179 +       if (atomic_inc_return(&au_fi(f)->fi_mmapped))
15180 +               return;
15181 +       pr_warn("fi_mmapped wrapped around\n");
15182 +       while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
15183 +               ;
15184 +}
15185 +
15186 +static inline void au_unset_mmapped(struct file *f)
15187 +{
15188 +       atomic_dec(&au_fi(f)->fi_mmapped);
15189 +}
15190 +
15191 +static inline int au_test_mmapped(struct file *f)
15192 +{
15193 +       return atomic_read(&au_fi(f)->fi_mmapped);
15194 +}
15195 +
15196 +/* customize vma->vm_file */
15197 +
15198 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
15199 +                                      struct file *file)
15200 +{
15201 +       struct file *f;
15202 +
15203 +       f = vma->vm_file;
15204 +       get_file(file);
15205 +       vma->vm_file = file;
15206 +       fput(f);
15207 +}
15208 +
15209 +#ifdef CONFIG_MMU
15210 +#define AuDbgVmRegion(file, vma) do {} while (0)
15211 +
15212 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15213 +                                   struct file *file)
15214 +{
15215 +       au_do_vm_file_reset(vma, file);
15216 +}
15217 +#else
15218 +#define AuDbgVmRegion(file, vma) \
15219 +       AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
15220 +
15221 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15222 +                                   struct file *file)
15223 +{
15224 +       struct file *f;
15225 +
15226 +       au_do_vm_file_reset(vma, file);
15227 +       f = vma->vm_region->vm_file;
15228 +       get_file(file);
15229 +       vma->vm_region->vm_file = file;
15230 +       fput(f);
15231 +}
15232 +#endif /* CONFIG_MMU */
15233 +
15234 +/* handle vma->vm_prfile */
15235 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
15236 +                                   struct file *file)
15237 +{
15238 +       get_file(file);
15239 +       vma->vm_prfile = file;
15240 +#ifndef CONFIG_MMU
15241 +       get_file(file);
15242 +       vma->vm_region->vm_prfile = file;
15243 +#endif
15244 +}
15245 +
15246 +#endif /* __KERNEL__ */
15247 +#endif /* __AUFS_FILE_H__ */
15248 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
15249 --- /usr/share/empty/fs/aufs/finfo.c    1970-01-01 01:00:00.000000000 +0100
15250 +++ linux/fs/aufs/finfo.c       2022-11-05 23:02:18.965889284 +0100
15251 @@ -0,0 +1,149 @@
15252 +// SPDX-License-Identifier: GPL-2.0
15253 +/*
15254 + * Copyright (C) 2005-2022 Junjiro R. Okajima
15255 + *
15256 + * This program is free software; you can redistribute it and/or modify
15257 + * it under the terms of the GNU General Public License as published by
15258 + * the Free Software Foundation; either version 2 of the License, or
15259 + * (at your option) any later version.
15260 + *
15261 + * This program is distributed in the hope that it will be useful,
15262 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15263 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15264 + * GNU General Public License for more details.
15265 + *
15266 + * You should have received a copy of the GNU General Public License
15267 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
15268 + */
15269 +
15270 +/*
15271 + * file private data
15272 + */
15273 +
15274 +#include "aufs.h"
15275 +
15276 +void au_hfput(struct au_hfile *hf, int execed)
15277 +{
15278 +       if (execed)
15279 +               allow_write_access(hf->hf_file);
15280 +       fput(hf->hf_file);
15281 +       hf->hf_file = NULL;
15282 +       au_lcnt_dec(&hf->hf_br->br_nfiles);
15283 +       hf->hf_br = NULL;
15284 +}
15285 +
15286 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
15287 +{
15288 +       struct au_finfo *finfo = au_fi(file);
15289 +       struct au_hfile *hf;
15290 +       struct au_fidir *fidir;
15291 +
15292 +       fidir = finfo->fi_hdir;
15293 +       if (!fidir) {
15294 +               AuDebugOn(finfo->fi_btop != bindex);
15295 +               hf = &finfo->fi_htop;
15296 +       } else
15297 +               hf = fidir->fd_hfile + bindex;
15298 +
15299 +       if (hf && hf->hf_file)
15300 +               au_hfput(hf, vfsub_file_execed(file));
15301 +       if (val) {
15302 +               FiMustWriteLock(file);
15303 +               AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
15304 +               hf->hf_file = val;
15305 +               hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
15306 +       }
15307 +}
15308 +
15309 +void au_update_figen(struct file *file)
15310 +{
15311 +       atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
15312 +       /* smp_mb(); */ /* atomic_set */
15313 +}
15314 +
15315 +/* ---------------------------------------------------------------------- */
15316 +
15317 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
15318 +{
15319 +       struct au_fidir *fidir;
15320 +       int nbr;
15321 +
15322 +       nbr = au_sbbot(sb) + 1;
15323 +       if (nbr < 2)
15324 +               nbr = 2; /* initial allocate for 2 branches */
15325 +       fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
15326 +       if (fidir) {
15327 +               fidir->fd_bbot = -1;
15328 +               fidir->fd_nent = nbr;
15329 +       }
15330 +
15331 +       return fidir;
15332 +}
15333 +
15334 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink)
15335 +{
15336 +       int err;
15337 +       struct au_fidir *fidir, *p;
15338 +
15339 +       AuRwMustWriteLock(&finfo->fi_rwsem);
15340 +       fidir = finfo->fi_hdir;
15341 +       AuDebugOn(!fidir);
15342 +
15343 +       err = -ENOMEM;
15344 +       p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
15345 +                        GFP_NOFS, may_shrink);
15346 +       if (p) {
15347 +               p->fd_nent = nbr;
15348 +               finfo->fi_hdir = p;
15349 +               err = 0;
15350 +       }
15351 +
15352 +       return err;
15353 +}
15354 +
15355 +/* ---------------------------------------------------------------------- */
15356 +
15357 +void au_finfo_fin(struct file *file)
15358 +{
15359 +       struct au_finfo *finfo;
15360 +
15361 +       au_lcnt_dec(&au_sbi(file->f_path.dentry->d_sb)->si_nfiles);
15362 +
15363 +       finfo = au_fi(file);
15364 +       AuDebugOn(finfo->fi_hdir);
15365 +       AuRwDestroy(&finfo->fi_rwsem);
15366 +       au_cache_free_finfo(finfo);
15367 +}
15368 +
15369 +void au_fi_init_once(void *_finfo)
15370 +{
15371 +       struct au_finfo *finfo = _finfo;
15372 +
15373 +       au_rw_init(&finfo->fi_rwsem);
15374 +}
15375 +
15376 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
15377 +{
15378 +       int err;
15379 +       struct au_finfo *finfo;
15380 +       struct dentry *dentry;
15381 +
15382 +       err = -ENOMEM;
15383 +       dentry = file->f_path.dentry;
15384 +       finfo = au_cache_alloc_finfo();
15385 +       if (unlikely(!finfo))
15386 +               goto out;
15387 +
15388 +       err = 0;
15389 +       au_lcnt_inc(&au_sbi(dentry->d_sb)->si_nfiles);
15390 +       au_rw_write_lock(&finfo->fi_rwsem);
15391 +       finfo->fi_btop = -1;
15392 +       finfo->fi_hdir = fidir;
15393 +       atomic_set(&finfo->fi_generation, au_digen(dentry));
15394 +       /* smp_mb(); */ /* atomic_set */
15395 +
15396 +       file->private_data = finfo;
15397 +
15398 +out:
15399 +       return err;
15400 +}
15401 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
15402 --- /usr/share/empty/fs/aufs/f_op.c     1970-01-01 01:00:00.000000000 +0100
15403 +++ linux/fs/aufs/f_op.c        2022-11-05 23:02:18.962555950 +0100
15404 @@ -0,0 +1,771 @@
15405 +// SPDX-License-Identifier: GPL-2.0
15406 +/*
15407 + * Copyright (C) 2005-2022 Junjiro R. Okajima
15408 + *
15409 + * This program is free software; you can redistribute it and/or modify
15410 + * it under the terms of the GNU General Public License as published by
15411 + * the Free Software Foundation; either version 2 of the License, or
15412 + * (at your option) any later version.
15413 + *
15414 + * This program is distributed in the hope that it will be useful,
15415 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15416 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15417 + * GNU General Public License for more details.
15418 + *
15419 + * You should have received a copy of the GNU General Public License
15420 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
15421 + */
15422 +
15423 +/*
15424 + * file and vm operations
15425 + */
15426 +
15427 +#include <linux/aio.h>
15428 +#include <linux/fs_stack.h>
15429 +#include <linux/mman.h>
15430 +#include <linux/security.h>
15431 +#include "aufs.h"
15432 +
15433 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
15434 +{
15435 +       int err;
15436 +       aufs_bindex_t bindex;
15437 +       struct dentry *dentry, *h_dentry;
15438 +       struct au_finfo *finfo;
15439 +       struct inode *h_inode;
15440 +
15441 +       FiMustWriteLock(file);
15442 +
15443 +       err = 0;
15444 +       dentry = file->f_path.dentry;
15445 +       AuDebugOn(IS_ERR_OR_NULL(dentry));
15446 +       finfo = au_fi(file);
15447 +       memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
15448 +       atomic_set(&finfo->fi_mmapped, 0);
15449 +       bindex = au_dbtop(dentry);
15450 +       if (!h_file) {
15451 +               h_dentry = au_h_dptr(dentry, bindex);
15452 +               err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15453 +               if (unlikely(err))
15454 +                       goto out;
15455 +               h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
15456 +               if (IS_ERR(h_file)) {
15457 +                       err = PTR_ERR(h_file);
15458 +                       goto out;
15459 +               }
15460 +       } else {
15461 +               h_dentry = h_file->f_path.dentry;
15462 +               err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15463 +               if (unlikely(err))
15464 +                       goto out;
15465 +               /* br ref is already inc-ed */
15466 +       }
15467 +
15468 +       if ((flags & __O_TMPFILE)
15469 +           && !(flags & O_EXCL)) {
15470 +               h_inode = file_inode(h_file);
15471 +               spin_lock(&h_inode->i_lock);
15472 +               h_inode->i_state |= I_LINKABLE;
15473 +               spin_unlock(&h_inode->i_lock);
15474 +       }
15475 +       au_set_fbtop(file, bindex);
15476 +       au_set_h_fptr(file, bindex, h_file);
15477 +       au_update_figen(file);
15478 +       /* todo: necessary? */
15479 +       /* file->f_ra = h_file->f_ra; */
15480 +
15481 +out:
15482 +       return err;
15483 +}
15484 +
15485 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
15486 +                           struct file *file)
15487 +{
15488 +       int err;
15489 +       struct super_block *sb;
15490 +       struct au_do_open_args args = {
15491 +               .open   = au_do_open_nondir
15492 +       };
15493 +
15494 +       AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
15495 +             file, vfsub_file_flags(file), file->f_mode);
15496 +
15497 +       sb = file->f_path.dentry->d_sb;
15498 +       si_read_lock(sb, AuLock_FLUSH);
15499 +       err = au_do_open(file, &args);
15500 +       si_read_unlock(sb);
15501 +       return err;
15502 +}
15503 +
15504 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
15505 +{
15506 +       struct au_finfo *finfo;
15507 +       aufs_bindex_t bindex;
15508 +
15509 +       finfo = au_fi(file);
15510 +       au_hbl_del(&finfo->fi_hlist,
15511 +                  &au_sbi(file->f_path.dentry->d_sb)->si_files);
15512 +       bindex = finfo->fi_btop;
15513 +       if (bindex >= 0)
15514 +               au_set_h_fptr(file, bindex, NULL);
15515 +
15516 +       au_finfo_fin(file);
15517 +       return 0;
15518 +}
15519 +
15520 +/* ---------------------------------------------------------------------- */
15521 +
15522 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
15523 +{
15524 +       int err;
15525 +       struct file *h_file;
15526 +
15527 +       err = 0;
15528 +       h_file = au_hf_top(file);
15529 +       if (h_file)
15530 +               err = vfsub_flush(h_file, id);
15531 +       return err;
15532 +}
15533 +
15534 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
15535 +{
15536 +       return au_do_flush(file, id, au_do_flush_nondir);
15537 +}
15538 +
15539 +/* ---------------------------------------------------------------------- */
15540 +/*
15541 + * read and write functions acquire [fdi]_rwsem once, but release before
15542 + * mmap_sem. This is because to stop a race condition between mmap(2).
15543 + * Releasing these aufs-rwsem should be safe, no branch-management (by keeping
15544 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
15545 + * read functions after [fdi]_rwsem are released, but it should be harmless.
15546 + */
15547 +
15548 +/* Callers should call au_read_post() or fput() in the end */
15549 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc)
15550 +{
15551 +       struct file *h_file;
15552 +       int err;
15553 +
15554 +       err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0, lsc);
15555 +       if (!err) {
15556 +               di_read_unlock(file->f_path.dentry, AuLock_IR);
15557 +               h_file = au_hf_top(file);
15558 +               get_file(h_file);
15559 +               if (!keep_fi)
15560 +                       fi_read_unlock(file);
15561 +       } else
15562 +               h_file = ERR_PTR(err);
15563 +
15564 +       return h_file;
15565 +}
15566 +
15567 +static void au_read_post(struct inode *inode, struct file *h_file)
15568 +{
15569 +       /* update without lock, I don't think it a problem */
15570 +       fsstack_copy_attr_atime(inode, file_inode(h_file));
15571 +       fput(h_file);
15572 +}
15573 +
15574 +struct au_write_pre {
15575 +       /* input */
15576 +       unsigned int lsc;
15577 +
15578 +       /* output */
15579 +       blkcnt_t blks;
15580 +       aufs_bindex_t btop;
15581 +};
15582 +
15583 +/*
15584 + * return with iinfo is write-locked
15585 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
15586 + * end
15587 + */
15588 +static struct file *au_write_pre(struct file *file, int do_ready,
15589 +                                struct au_write_pre *wpre)
15590 +{
15591 +       struct file *h_file;
15592 +       struct dentry *dentry;
15593 +       int err;
15594 +       unsigned int lsc;
15595 +       struct au_pin pin;
15596 +
15597 +       lsc = 0;
15598 +       if (wpre)
15599 +               lsc = wpre->lsc;
15600 +       err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1, lsc);
15601 +       h_file = ERR_PTR(err);
15602 +       if (unlikely(err))
15603 +               goto out;
15604 +
15605 +       dentry = file->f_path.dentry;
15606 +       if (do_ready) {
15607 +               err = au_ready_to_write(file, -1, &pin);
15608 +               if (unlikely(err)) {
15609 +                       h_file = ERR_PTR(err);
15610 +                       di_write_unlock(dentry);
15611 +                       goto out_fi;
15612 +               }
15613 +       }
15614 +
15615 +       di_downgrade_lock(dentry, /*flags*/0);
15616 +       if (wpre)
15617 +               wpre->btop = au_fbtop(file);
15618 +       h_file = au_hf_top(file);
15619 +       get_file(h_file);
15620 +       if (wpre)
15621 +               wpre->blks = file_inode(h_file)->i_blocks;
15622 +       if (do_ready)
15623 +               au_unpin(&pin);
15624 +       di_read_unlock(dentry, /*flags*/0);
15625 +
15626 +out_fi:
15627 +       fi_write_unlock(file);
15628 +out:
15629 +       return h_file;
15630 +}
15631 +
15632 +static void au_write_post(struct inode *inode, struct file *h_file,
15633 +                         struct au_write_pre *wpre, ssize_t written)
15634 +{
15635 +       struct inode *h_inode;
15636 +
15637 +       au_cpup_attr_timesizes(inode);
15638 +       AuDebugOn(au_ibtop(inode) != wpre->btop);
15639 +       h_inode = file_inode(h_file);
15640 +       inode->i_mode = h_inode->i_mode;
15641 +       ii_write_unlock(inode);
15642 +       /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
15643 +       if (written > 0)
15644 +               au_fhsm_wrote(inode->i_sb, wpre->btop,
15645 +                             /*force*/h_inode->i_blocks > wpre->blks);
15646 +       fput(h_file);
15647 +}
15648 +
15649 +/*
15650 + * todo: very ugly
15651 + * it locks both of i_mutex and si_rwsem for read in safe.
15652 + * if the plink maintenance mode continues forever (that is the problem),
15653 + * may loop forever.
15654 + */
15655 +static void au_mtx_and_read_lock(struct inode *inode)
15656 +{
15657 +       int err;
15658 +       struct super_block *sb = inode->i_sb;
15659 +
15660 +       while (1) {
15661 +               inode_lock(inode);
15662 +               err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
15663 +               if (!err)
15664 +                       break;
15665 +               inode_unlock(inode);
15666 +               si_read_lock(sb, AuLock_NOPLMW);
15667 +               si_read_unlock(sb);
15668 +       }
15669 +}
15670 +
15671 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
15672 +                         struct iov_iter *iov_iter)
15673 +{
15674 +       ssize_t err;
15675 +       struct file *file;
15676 +       ssize_t (*iter)(struct kiocb *, struct iov_iter *);
15677 +
15678 +       err = security_file_permission(h_file, rw);
15679 +       if (unlikely(err))
15680 +               goto out;
15681 +
15682 +       err = -ENOSYS;  /* the branch doesn't have its ->(read|write)_iter() */
15683 +       iter = NULL;
15684 +       if (rw == MAY_READ)
15685 +               iter = h_file->f_op->read_iter;
15686 +       else if (rw == MAY_WRITE)
15687 +               iter = h_file->f_op->write_iter;
15688 +
15689 +       file = kio->ki_filp;
15690 +       kio->ki_filp = h_file;
15691 +       if (iter) {
15692 +               lockdep_off();
15693 +               err = iter(kio, iov_iter);
15694 +               lockdep_on();
15695 +       } else
15696 +               /* currently there is no such fs */
15697 +               WARN_ON_ONCE(1);
15698 +       kio->ki_filp = file;
15699 +
15700 +out:
15701 +       return err;
15702 +}
15703 +
15704 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15705 +{
15706 +       ssize_t err;
15707 +       struct file *file, *h_file;
15708 +       struct inode *inode;
15709 +       struct super_block *sb;
15710 +
15711 +       file = kio->ki_filp;
15712 +       inode = file_inode(file);
15713 +       sb = inode->i_sb;
15714 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15715 +
15716 +       h_file = au_read_pre(file, /*keep_fi*/1, /*lsc*/0);
15717 +       err = PTR_ERR(h_file);
15718 +       if (IS_ERR(h_file))
15719 +               goto out;
15720 +
15721 +       if (au_test_loopback_kthread()) {
15722 +               au_warn_loopback(h_file->f_path.dentry->d_sb);
15723 +               if (file->f_mapping != h_file->f_mapping) {
15724 +                       file->f_mapping = h_file->f_mapping;
15725 +                       smp_mb(); /* unnecessary? */
15726 +               }
15727 +       }
15728 +       fi_read_unlock(file);
15729 +
15730 +       err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
15731 +       /* todo: necessary? */
15732 +       /* file->f_ra = h_file->f_ra; */
15733 +       au_read_post(inode, h_file);
15734 +
15735 +out:
15736 +       si_read_unlock(sb);
15737 +       return err;
15738 +}
15739 +
15740 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15741 +{
15742 +       ssize_t err;
15743 +       struct au_write_pre wpre;
15744 +       struct inode *inode;
15745 +       struct file *file, *h_file;
15746 +
15747 +       file = kio->ki_filp;
15748 +       inode = file_inode(file);
15749 +       au_mtx_and_read_lock(inode);
15750 +
15751 +       wpre.lsc = 0;
15752 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15753 +       err = PTR_ERR(h_file);
15754 +       if (IS_ERR(h_file))
15755 +               goto out;
15756 +
15757 +       err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
15758 +       au_write_post(inode, h_file, &wpre, err);
15759 +
15760 +out:
15761 +       si_read_unlock(inode->i_sb);
15762 +       inode_unlock(inode);
15763 +       return err;
15764 +}
15765 +
15766 +/*
15767 + * We may be able to remove aufs_splice_{read,write}() since almost all FSes
15768 + * don't have their own .splice_{read,write} implimentations, and they use
15769 + * generic_file_splice_read() and iter_file_splice_write() who can act like the
15770 + * simple converters to f_op->iter_read() and ->iter_write().
15771 + * But we keep our own implementations because some non-mainlined FSes may have
15772 + * their own .splice_{read,write} implimentations and aufs doesn't want to take
15773 + * away an opportunity to co-work with aufs from them.
15774 + */
15775 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
15776 +                               struct pipe_inode_info *pipe, size_t len,
15777 +                               unsigned int flags)
15778 +{
15779 +       ssize_t err;
15780 +       struct file *h_file;
15781 +       struct inode *inode;
15782 +       struct super_block *sb;
15783 +
15784 +       inode = file_inode(file);
15785 +       sb = inode->i_sb;
15786 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15787 +
15788 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15789 +       err = PTR_ERR(h_file);
15790 +       if (IS_ERR(h_file))
15791 +               goto out;
15792 +
15793 +       err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
15794 +       /* todo: necessary? */
15795 +       /* file->f_ra = h_file->f_ra; */
15796 +       au_read_post(inode, h_file);
15797 +
15798 +out:
15799 +       si_read_unlock(sb);
15800 +       return err;
15801 +}
15802 +
15803 +static ssize_t
15804 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
15805 +                 size_t len, unsigned int flags)
15806 +{
15807 +       ssize_t err;
15808 +       struct au_write_pre wpre;
15809 +       struct inode *inode;
15810 +       struct file *h_file;
15811 +
15812 +       inode = file_inode(file);
15813 +       au_mtx_and_read_lock(inode);
15814 +
15815 +       wpre.lsc = 0;
15816 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15817 +       err = PTR_ERR(h_file);
15818 +       if (IS_ERR(h_file))
15819 +               goto out;
15820 +
15821 +       err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
15822 +       au_write_post(inode, h_file, &wpre, err);
15823 +
15824 +out:
15825 +       si_read_unlock(inode->i_sb);
15826 +       inode_unlock(inode);
15827 +       return err;
15828 +}
15829 +
15830 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
15831 +                          loff_t len)
15832 +{
15833 +       long err;
15834 +       struct au_write_pre wpre;
15835 +       struct inode *inode;
15836 +       struct file *h_file;
15837 +
15838 +       inode = file_inode(file);
15839 +       au_mtx_and_read_lock(inode);
15840 +
15841 +       wpre.lsc = 0;
15842 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15843 +       err = PTR_ERR(h_file);
15844 +       if (IS_ERR(h_file))
15845 +               goto out;
15846 +
15847 +       lockdep_off();
15848 +       err = vfs_fallocate(h_file, mode, offset, len);
15849 +       lockdep_on();
15850 +       au_write_post(inode, h_file, &wpre, /*written*/1);
15851 +
15852 +out:
15853 +       si_read_unlock(inode->i_sb);
15854 +       inode_unlock(inode);
15855 +       return err;
15856 +}
15857 +
15858 +static ssize_t aufs_copy_file_range(struct file *src, loff_t src_pos,
15859 +                                   struct file *dst, loff_t dst_pos,
15860 +                                   size_t len, unsigned int flags)
15861 +{
15862 +       ssize_t err;
15863 +       struct au_write_pre wpre;
15864 +       enum { SRC, DST };
15865 +       struct {
15866 +               struct inode *inode;
15867 +               struct file *h_file;
15868 +               struct super_block *h_sb;
15869 +       } a[2];
15870 +#define a_src  a[SRC]
15871 +#define a_dst  a[DST]
15872 +
15873 +       err = -EINVAL;
15874 +       a_src.inode = file_inode(src);
15875 +       if (unlikely(!S_ISREG(a_src.inode->i_mode)))
15876 +               goto out;
15877 +       a_dst.inode = file_inode(dst);
15878 +       if (unlikely(!S_ISREG(a_dst.inode->i_mode)))
15879 +               goto out;
15880 +
15881 +       au_mtx_and_read_lock(a_dst.inode);
15882 +       /*
15883 +        * in order to match the order in di_write_lock2_{child,parent}(),
15884 +        * use f_path.dentry for this comparison.
15885 +        */
15886 +       if (src->f_path.dentry < dst->f_path.dentry) {
15887 +               a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_1);
15888 +               err = PTR_ERR(a_src.h_file);
15889 +               if (IS_ERR(a_src.h_file))
15890 +                       goto out_si;
15891 +
15892 +               wpre.lsc = AuLsc_FI_2;
15893 +               a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15894 +               err = PTR_ERR(a_dst.h_file);
15895 +               if (IS_ERR(a_dst.h_file)) {
15896 +                       au_read_post(a_src.inode, a_src.h_file);
15897 +                       goto out_si;
15898 +               }
15899 +       } else {
15900 +               wpre.lsc = AuLsc_FI_1;
15901 +               a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15902 +               err = PTR_ERR(a_dst.h_file);
15903 +               if (IS_ERR(a_dst.h_file))
15904 +                       goto out_si;
15905 +
15906 +               a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_2);
15907 +               err = PTR_ERR(a_src.h_file);
15908 +               if (IS_ERR(a_src.h_file)) {
15909 +                       au_write_post(a_dst.inode, a_dst.h_file, &wpre,
15910 +                                     /*written*/0);
15911 +                       goto out_si;
15912 +               }
15913 +       }
15914 +
15915 +       err = -EXDEV;
15916 +       a_src.h_sb = file_inode(a_src.h_file)->i_sb;
15917 +       a_dst.h_sb = file_inode(a_dst.h_file)->i_sb;
15918 +       if (unlikely(a_src.h_sb != a_dst.h_sb)) {
15919 +               AuDbgFile(src);
15920 +               AuDbgFile(dst);
15921 +               goto out_file;
15922 +       }
15923 +
15924 +       err = vfsub_copy_file_range(a_src.h_file, src_pos, a_dst.h_file,
15925 +                                   dst_pos, len, flags);
15926 +
15927 +out_file:
15928 +       au_write_post(a_dst.inode, a_dst.h_file, &wpre, err);
15929 +       fi_read_unlock(src);
15930 +       au_read_post(a_src.inode, a_src.h_file);
15931 +out_si:
15932 +       si_read_unlock(a_dst.inode->i_sb);
15933 +       inode_unlock(a_dst.inode);
15934 +out:
15935 +       return err;
15936 +#undef a_src
15937 +#undef a_dst
15938 +}
15939 +
15940 +/* ---------------------------------------------------------------------- */
15941 +
15942 +/*
15943 + * The locking order around current->mmap_sem.
15944 + * - in most and regular cases
15945 + *   file I/O syscall -- aufs_read() or something
15946 + *     -- si_rwsem for read -- mmap_sem
15947 + *     (Note that [fdi]i_rwsem are released before mmap_sem).
15948 + * - in mmap case
15949 + *   mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
15950 + * This AB-BA order is definitely bad, but is not a problem since "si_rwsem for
15951 + * read" allows multiple processes to acquire it and [fdi]i_rwsem are not held
15952 + * in file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
15953 + * It means that when aufs acquires si_rwsem for write, the process should never
15954 + * acquire mmap_sem.
15955 + *
15956 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
15957 + * problem either since any directory is not able to be mmap-ed.
15958 + * The similar scenario is applied to aufs_readlink() too.
15959 + */
15960 +
15961 +#if 0 /* stop calling security_file_mmap() */
15962 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
15963 +#define AuConv_VM_PROT(f, b)   _calc_vm_trans(f, VM_##b, PROT_##b)
15964 +
15965 +static unsigned long au_arch_prot_conv(unsigned long flags)
15966 +{
15967 +       /* currently ppc64 only */
15968 +#ifdef CONFIG_PPC64
15969 +       /* cf. linux/arch/powerpc/include/asm/mman.h */
15970 +       AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
15971 +       return AuConv_VM_PROT(flags, SAO);
15972 +#else
15973 +       AuDebugOn(arch_calc_vm_prot_bits(-1));
15974 +       return 0;
15975 +#endif
15976 +}
15977 +
15978 +static unsigned long au_prot_conv(unsigned long flags)
15979 +{
15980 +       return AuConv_VM_PROT(flags, READ)
15981 +               | AuConv_VM_PROT(flags, WRITE)
15982 +               | AuConv_VM_PROT(flags, EXEC)
15983 +               | au_arch_prot_conv(flags);
15984 +}
15985 +
15986 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
15987 +#define AuConv_VM_MAP(f, b)    _calc_vm_trans(f, VM_##b, MAP_##b)
15988 +
15989 +static unsigned long au_flag_conv(unsigned long flags)
15990 +{
15991 +       return AuConv_VM_MAP(flags, GROWSDOWN)
15992 +               | AuConv_VM_MAP(flags, DENYWRITE)
15993 +               | AuConv_VM_MAP(flags, LOCKED);
15994 +}
15995 +#endif
15996 +
15997 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
15998 +{
15999 +       int err;
16000 +       const unsigned char wlock
16001 +               = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
16002 +       struct super_block *sb;
16003 +       struct file *h_file;
16004 +       struct inode *inode;
16005 +
16006 +       AuDbgVmRegion(file, vma);
16007 +
16008 +       inode = file_inode(file);
16009 +       sb = inode->i_sb;
16010 +       lockdep_off();
16011 +       si_read_lock(sb, AuLock_NOPLMW);
16012 +
16013 +       h_file = au_write_pre(file, wlock, /*wpre*/NULL);
16014 +       lockdep_on();
16015 +       err = PTR_ERR(h_file);
16016 +       if (IS_ERR(h_file))
16017 +               goto out;
16018 +
16019 +       err = 0;
16020 +       au_set_mmapped(file);
16021 +       au_vm_file_reset(vma, h_file);
16022 +       /*
16023 +        * we cannot call security_mmap_file() here since it may acquire
16024 +        * mmap_sem or i_mutex.
16025 +        *
16026 +        * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
16027 +        *                       au_flag_conv(vma->vm_flags));
16028 +        */
16029 +       if (!err)
16030 +               err = call_mmap(h_file, vma);
16031 +       if (!err) {
16032 +               au_vm_prfile_set(vma, file);
16033 +               fsstack_copy_attr_atime(inode, file_inode(h_file));
16034 +               goto out_fput; /* success */
16035 +       }
16036 +       au_unset_mmapped(file);
16037 +       au_vm_file_reset(vma, file);
16038 +
16039 +out_fput:
16040 +       lockdep_off();
16041 +       ii_write_unlock(inode);
16042 +       lockdep_on();
16043 +       fput(h_file);
16044 +out:
16045 +       lockdep_off();
16046 +       si_read_unlock(sb);
16047 +       lockdep_on();
16048 +       AuTraceErr(err);
16049 +       return err;
16050 +}
16051 +
16052 +/* ---------------------------------------------------------------------- */
16053 +
16054 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
16055 +                            int datasync)
16056 +{
16057 +       int err;
16058 +       struct au_write_pre wpre;
16059 +       struct inode *inode;
16060 +       struct file *h_file;
16061 +
16062 +       err = 0; /* -EBADF; */ /* posix? */
16063 +       if (unlikely(!(file->f_mode & FMODE_WRITE)))
16064 +               goto out;
16065 +
16066 +       inode = file_inode(file);
16067 +       au_mtx_and_read_lock(inode);
16068 +
16069 +       wpre.lsc = 0;
16070 +       h_file = au_write_pre(file, /*do_ready*/1, &wpre);
16071 +       err = PTR_ERR(h_file);
16072 +       if (IS_ERR(h_file))
16073 +               goto out_unlock;
16074 +
16075 +       err = vfsub_fsync(h_file, &h_file->f_path, datasync);
16076 +       au_write_post(inode, h_file, &wpre, /*written*/0);
16077 +
16078 +out_unlock:
16079 +       si_read_unlock(inode->i_sb);
16080 +       inode_unlock(inode);
16081 +out:
16082 +       return err;
16083 +}
16084 +
16085 +static int aufs_fasync(int fd, struct file *file, int flag)
16086 +{
16087 +       int err;
16088 +       struct file *h_file;
16089 +       struct super_block *sb;
16090 +
16091 +       sb = file->f_path.dentry->d_sb;
16092 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16093 +
16094 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16095 +       err = PTR_ERR(h_file);
16096 +       if (IS_ERR(h_file))
16097 +               goto out;
16098 +
16099 +       if (h_file->f_op->fasync)
16100 +               err = h_file->f_op->fasync(fd, h_file, flag);
16101 +       fput(h_file); /* instead of au_read_post() */
16102 +
16103 +out:
16104 +       si_read_unlock(sb);
16105 +       return err;
16106 +}
16107 +
16108 +static int aufs_setfl(struct file *file, unsigned long arg)
16109 +{
16110 +       int err;
16111 +       struct file *h_file;
16112 +       struct super_block *sb;
16113 +
16114 +       sb = file->f_path.dentry->d_sb;
16115 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16116 +
16117 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16118 +       err = PTR_ERR(h_file);
16119 +       if (IS_ERR(h_file))
16120 +               goto out;
16121 +
16122 +       /* stop calling h_file->fasync */
16123 +       arg |= vfsub_file_flags(file) & FASYNC;
16124 +       err = setfl(/*unused fd*/-1, h_file, arg);
16125 +       fput(h_file); /* instead of au_read_post() */
16126 +
16127 +out:
16128 +       si_read_unlock(sb);
16129 +       return err;
16130 +}
16131 +
16132 +/* ---------------------------------------------------------------------- */
16133 +
16134 +/* no one supports this operation, currently */
16135 +#if 0 /* reserved for future use */
16136 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
16137 +                            size_t len, loff_t *pos, int more)
16138 +{
16139 +}
16140 +#endif
16141 +
16142 +/* ---------------------------------------------------------------------- */
16143 +
16144 +const struct file_operations aufs_file_fop = {
16145 +       .owner          = THIS_MODULE,
16146 +
16147 +       .llseek         = default_llseek,
16148 +
16149 +       .read_iter      = aufs_read_iter,
16150 +       .write_iter     = aufs_write_iter,
16151 +
16152 +#ifdef CONFIG_AUFS_POLL
16153 +       .poll           = aufs_poll,
16154 +#endif
16155 +       .unlocked_ioctl = aufs_ioctl_nondir,
16156 +#ifdef CONFIG_COMPAT
16157 +       .compat_ioctl   = aufs_compat_ioctl_nondir,
16158 +#endif
16159 +       .mmap           = aufs_mmap,
16160 +       .open           = aufs_open_nondir,
16161 +       .flush          = aufs_flush_nondir,
16162 +       .release        = aufs_release_nondir,
16163 +       .fsync          = aufs_fsync_nondir,
16164 +       .fasync         = aufs_fasync,
16165 +       /* .sendpage    = aufs_sendpage, */
16166 +       .setfl          = aufs_setfl,
16167 +       .splice_write   = aufs_splice_write,
16168 +       .splice_read    = aufs_splice_read,
16169 +#if 0 /* reserved for future use */
16170 +       .aio_splice_write = aufs_aio_splice_write,
16171 +       .aio_splice_read  = aufs_aio_splice_read,
16172 +#endif
16173 +       .fallocate      = aufs_fallocate,
16174 +       .copy_file_range = aufs_copy_file_range
16175 +};
16176 diff -urN /usr/share/empty/fs/aufs/fsctx.c linux/fs/aufs/fsctx.c
16177 --- /usr/share/empty/fs/aufs/fsctx.c    1970-01-01 01:00:00.000000000 +0100
16178 +++ linux/fs/aufs/fsctx.c       2022-11-05 23:02:18.965889284 +0100
16179 @@ -0,0 +1,1242 @@
16180 +// SPDX-License-Identifier: GPL-2.0
16181 +/*
16182 + * Copyright (C) 2022 Junjiro R. Okajima
16183 + *
16184 + * This program is free software; you can redistribute it and/or modify
16185 + * it under the terms of the GNU General Public License as published by
16186 + * the Free Software Foundation; either version 2 of the License, or
16187 + * (at your option) any later version.
16188 + *
16189 + * This program is distributed in the hope that it will be useful,
16190 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16191 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16192 + * GNU General Public License for more details.
16193 + *
16194 + * You should have received a copy of the GNU General Public License
16195 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
16196 + */
16197 +
16198 +/*
16199 + * fs context, aka new mount api
16200 + */
16201 +
16202 +#include <linux/fs_context.h>
16203 +#include "aufs.h"
16204 +
16205 +struct au_fsctx_opts {
16206 +       aufs_bindex_t bindex;
16207 +       unsigned char skipped;
16208 +       struct au_opt *opt, *opt_tail;
16209 +       struct super_block *sb;
16210 +       struct au_sbinfo *sbinfo;
16211 +       struct au_opts opts;
16212 +};
16213 +
16214 +/* stop extra interpretation of errno in mount(8), and strange error messages */
16215 +static int cvt_err(int err)
16216 +{
16217 +       AuTraceErr(err);
16218 +
16219 +       switch (err) {
16220 +       case -ENOENT:
16221 +       case -ENOTDIR:
16222 +       case -EEXIST:
16223 +       case -EIO:
16224 +               err = -EINVAL;
16225 +       }
16226 +       return err;
16227 +}
16228 +
16229 +static int au_fsctx_reconfigure(struct fs_context *fc)
16230 +{
16231 +       int err, do_dx;
16232 +       unsigned int mntflags;
16233 +       struct dentry *root;
16234 +       struct super_block *sb;
16235 +       struct inode *inode;
16236 +       struct au_fsctx_opts *a = fc->fs_private;
16237 +
16238 +       AuDbg("fc %p\n", fc);
16239 +
16240 +       root = fc->root;
16241 +       sb = root->d_sb;
16242 +       err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16243 +       if (!err) {
16244 +               di_write_lock_child(root);
16245 +               err = au_opts_verify(sb, fc->sb_flags, /*pending*/0);
16246 +               aufs_write_unlock(root);
16247 +       }
16248 +
16249 +       inode = d_inode(root);
16250 +       inode_lock(inode);
16251 +       err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16252 +       if (unlikely(err))
16253 +               goto out;
16254 +       di_write_lock_child(root);
16255 +
16256 +       /* au_opts_remount() may return an error */
16257 +       err = au_opts_remount(sb, &a->opts);
16258 +
16259 +       if (au_ftest_opts(a->opts.flags, REFRESH))
16260 +               au_remount_refresh(sb, au_ftest_opts(a->opts.flags,
16261 +                                                    REFRESH_IDOP));
16262 +
16263 +       if (au_ftest_opts(a->opts.flags, REFRESH_DYAOP)) {
16264 +               mntflags = au_mntflags(sb);
16265 +               do_dx = !!au_opt_test(mntflags, DIO);
16266 +               au_dy_arefresh(do_dx);
16267 +       }
16268 +
16269 +       au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
16270 +       aufs_write_unlock(root);
16271 +
16272 +out:
16273 +       inode_unlock(inode);
16274 +       err = cvt_err(err);
16275 +       AuTraceErr(err);
16276 +
16277 +       return err;
16278 +}
16279 +
16280 +/* ---------------------------------------------------------------------- */
16281 +
16282 +static int au_fsctx_fill_super(struct super_block *sb, struct fs_context *fc)
16283 +{
16284 +       int err;
16285 +       struct au_fsctx_opts *a = fc->fs_private;
16286 +       struct au_sbinfo *sbinfo = a->sbinfo;
16287 +       struct dentry *root;
16288 +       struct inode *inode;
16289 +
16290 +       sbinfo->si_sb = sb;
16291 +       sb->s_fs_info = sbinfo;
16292 +       kobject_get(&sbinfo->si_kobj);
16293 +
16294 +       __si_write_lock(sb);
16295 +       si_pid_set(sb);
16296 +       au_sbilist_add(sb);
16297 +
16298 +       /* all timestamps always follow the ones on the branch */
16299 +       sb->s_flags |= SB_NOATIME | SB_NODIRATIME;
16300 +       sb->s_flags |= SB_I_VERSION; /* do we really need this? */
16301 +       sb->s_op = &aufs_sop;
16302 +       sb->s_d_op = &aufs_dop;
16303 +       sb->s_magic = AUFS_SUPER_MAGIC;
16304 +       sb->s_maxbytes = 0;
16305 +       sb->s_stack_depth = 1;
16306 +       au_export_init(sb);
16307 +       au_xattr_init(sb);
16308 +
16309 +       err = au_alloc_root(sb);
16310 +       if (unlikely(err)) {
16311 +               si_write_unlock(sb);
16312 +               goto out;
16313 +       }
16314 +       root = sb->s_root;
16315 +       inode = d_inode(root);
16316 +       ii_write_lock_parent(inode);
16317 +       aufs_write_unlock(root);
16318 +
16319 +       /* lock vfs_inode first, then aufs. */
16320 +       inode_lock(inode);
16321 +       aufs_write_lock(root);
16322 +       err = au_opts_mount(sb, &a->opts);
16323 +       AuTraceErr(err);
16324 +       if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
16325 +               sb->s_d_op = &aufs_dop_noreval;
16326 +               /* infofc(fc, "%ps", sb->s_d_op); */
16327 +               pr_info("%ps\n", sb->s_d_op);
16328 +               au_refresh_dop(root, /*force_reval*/0);
16329 +               sbinfo->si_iop_array = aufs_iop_nogetattr;
16330 +               au_refresh_iop(inode, /*force_getattr*/0);
16331 +       }
16332 +       aufs_write_unlock(root);
16333 +       inode_unlock(inode);
16334 +       if (!err)
16335 +               goto out; /* success */
16336 +
16337 +       dput(root);
16338 +       sb->s_root = NULL;
16339 +
16340 +out:
16341 +       if (unlikely(err))
16342 +               kobject_put(&sbinfo->si_kobj);
16343 +       AuTraceErr(err);
16344 +       err = cvt_err(err);
16345 +       AuTraceErr(err);
16346 +       return err;
16347 +}
16348 +
16349 +static int au_fsctx_get_tree(struct fs_context *fc)
16350 +{
16351 +       int err;
16352 +
16353 +       AuDbg("fc %p\n", fc);
16354 +       err = get_tree_nodev(fc, au_fsctx_fill_super);
16355 +
16356 +       AuTraceErr(err);
16357 +       return err;
16358 +}
16359 +
16360 +/* ---------------------------------------------------------------------- */
16361 +
16362 +static void au_fsctx_dump(struct au_opts *opts)
16363 +{
16364 +#ifdef CONFIG_AUFS_DEBUG
16365 +       /* reduce stack space */
16366 +       union {
16367 +               struct au_opt_add *add;
16368 +               struct au_opt_del *del;
16369 +               struct au_opt_mod *mod;
16370 +               struct au_opt_xino *xino;
16371 +               struct au_opt_xino_itrunc *xino_itrunc;
16372 +               struct au_opt_wbr_create *create;
16373 +       } u;
16374 +       struct au_opt *opt;
16375 +
16376 +       opt = opts->opt;
16377 +       while (opt->type != Opt_tail) {
16378 +               switch (opt->type) {
16379 +               case Opt_add:
16380 +                       u.add = &opt->add;
16381 +                       AuDbg("add {b%d, %s, 0x%x, %p}\n",
16382 +                                 u.add->bindex, u.add->pathname, u.add->perm,
16383 +                                 u.add->path.dentry);
16384 +                       break;
16385 +               case Opt_del:
16386 +                       fallthrough;
16387 +               case Opt_idel:
16388 +                       u.del = &opt->del;
16389 +                       AuDbg("del {%s, %p}\n",
16390 +                             u.del->pathname, u.del->h_path.dentry);
16391 +                       break;
16392 +               case Opt_mod:
16393 +                       fallthrough;
16394 +               case Opt_imod:
16395 +                       u.mod = &opt->mod;
16396 +                       AuDbg("mod {%s, 0x%x, %p}\n",
16397 +                                 u.mod->path, u.mod->perm, u.mod->h_root);
16398 +                       break;
16399 +               case Opt_append:
16400 +                       u.add = &opt->add;
16401 +                       AuDbg("append {b%d, %s, 0x%x, %p}\n",
16402 +                                 u.add->bindex, u.add->pathname, u.add->perm,
16403 +                                 u.add->path.dentry);
16404 +                       break;
16405 +               case Opt_prepend:
16406 +                       u.add = &opt->add;
16407 +                       AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
16408 +                                 u.add->bindex, u.add->pathname, u.add->perm,
16409 +                                 u.add->path.dentry);
16410 +                       break;
16411 +
16412 +               case Opt_dirwh:
16413 +                       AuDbg("dirwh %d\n", opt->dirwh);
16414 +                       break;
16415 +               case Opt_rdcache:
16416 +                       AuDbg("rdcache %d\n", opt->rdcache);
16417 +                       break;
16418 +               case Opt_rdblk:
16419 +                       AuDbg("rdblk %d\n", opt->rdblk);
16420 +                       break;
16421 +               case Opt_rdhash:
16422 +                       AuDbg("rdhash %u\n", opt->rdhash);
16423 +                       break;
16424 +
16425 +               case Opt_xino:
16426 +                       u.xino = &opt->xino;
16427 +                       AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
16428 +                       break;
16429 +
16430 +#define au_fsctx_TF(name)                                        \
16431 +                       case Opt_##name:                          \
16432 +                               if (opt->tf)                      \
16433 +                                       AuLabel(name);            \
16434 +                               else                              \
16435 +                                       AuLabel(no##name);        \
16436 +                               break;
16437 +
16438 +               /* simple true/false flag */
16439 +               au_fsctx_TF(trunc_xino);
16440 +               au_fsctx_TF(trunc_xib);
16441 +               au_fsctx_TF(dirperm1);
16442 +               au_fsctx_TF(plink);
16443 +               au_fsctx_TF(shwh);
16444 +               au_fsctx_TF(dio);
16445 +               au_fsctx_TF(warn_perm);
16446 +               au_fsctx_TF(verbose);
16447 +               au_fsctx_TF(sum);
16448 +               au_fsctx_TF(dirren);
16449 +               au_fsctx_TF(acl);
16450 +#undef au_fsctx_TF
16451 +
16452 +               case Opt_trunc_xino_path:
16453 +                       fallthrough;
16454 +               case Opt_itrunc_xino:
16455 +                       u.xino_itrunc = &opt->xino_itrunc;
16456 +                       AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
16457 +                       break;
16458 +               case Opt_noxino:
16459 +                       AuLabel(noxino);
16460 +                       break;
16461 +
16462 +               case Opt_list_plink:
16463 +                       AuLabel(list_plink);
16464 +                       break;
16465 +               case Opt_udba:
16466 +                       AuDbg("udba %d, %s\n",
16467 +                                 opt->udba, au_optstr_udba(opt->udba));
16468 +                       break;
16469 +               case Opt_diropq_a:
16470 +                       AuLabel(diropq_a);
16471 +                       break;
16472 +               case Opt_diropq_w:
16473 +                       AuLabel(diropq_w);
16474 +                       break;
16475 +               case Opt_wsum:
16476 +                       AuLabel(wsum);
16477 +                       break;
16478 +               case Opt_wbr_create:
16479 +                       u.create = &opt->wbr_create;
16480 +                       AuDbg("create %d, %s\n", u.create->wbr_create,
16481 +                                 au_optstr_wbr_create(u.create->wbr_create));
16482 +                       switch (u.create->wbr_create) {
16483 +                       case AuWbrCreate_MFSV:
16484 +                               fallthrough;
16485 +                       case AuWbrCreate_PMFSV:
16486 +                               AuDbg("%d sec\n", u.create->mfs_second);
16487 +                               break;
16488 +                       case AuWbrCreate_MFSRR:
16489 +                               fallthrough;
16490 +                       case AuWbrCreate_TDMFS:
16491 +                               AuDbg("%llu watermark\n",
16492 +                                         u.create->mfsrr_watermark);
16493 +                               break;
16494 +                       case AuWbrCreate_MFSRRV:
16495 +                               fallthrough;
16496 +                       case AuWbrCreate_TDMFSV:
16497 +                               fallthrough;
16498 +                       case AuWbrCreate_PMFSRRV:
16499 +                               AuDbg("%llu watermark, %d sec\n",
16500 +                                         u.create->mfsrr_watermark,
16501 +                                         u.create->mfs_second);
16502 +                               break;
16503 +                       }
16504 +                       break;
16505 +               case Opt_wbr_copyup:
16506 +                       AuDbg("copyup %d, %s\n", opt->wbr_copyup,
16507 +                                 au_optstr_wbr_copyup(opt->wbr_copyup));
16508 +                       break;
16509 +               case Opt_fhsm_sec:
16510 +                       AuDbg("fhsm_sec %u\n", opt->fhsm_second);
16511 +                       break;
16512 +
16513 +               default:
16514 +                       AuDbg("type %d\n", opt->type);
16515 +                       BUG();
16516 +               }
16517 +               opt++;
16518 +       }
16519 +#endif
16520 +}
16521 +
16522 +/* ---------------------------------------------------------------------- */
16523 +
16524 +/*
16525 + * For conditionally compiled mount options.
16526 + * Instead of fsparam_flag_no(), use this macro to distinguish ignore_silent.
16527 + */
16528 +#define au_ignore_flag(name, action)           \
16529 +       fsparam_flag(name, action),             \
16530 +       fsparam_flag("no" name, Opt_ignore_silent)
16531 +
16532 +const struct fs_parameter_spec aufs_fsctx_paramspec[] = {
16533 +       fsparam_string("br", Opt_br),
16534 +
16535 +       /* "add=%d:%s" or "ins=%d:%s" */
16536 +       fsparam_string("add", Opt_add),
16537 +       fsparam_string("ins", Opt_add),
16538 +       fsparam_path("append", Opt_append),
16539 +       fsparam_path("prepend", Opt_prepend),
16540 +
16541 +       fsparam_path("del", Opt_del),
16542 +       /* fsparam_s32("idel", Opt_idel), */
16543 +       fsparam_path("mod", Opt_mod),
16544 +       /* fsparam_string("imod", Opt_imod), */
16545 +
16546 +       fsparam_s32("dirwh", Opt_dirwh),
16547 +
16548 +       fsparam_path("xino", Opt_xino),
16549 +       fsparam_flag("noxino", Opt_noxino),
16550 +       fsparam_flag_no("trunc_xino", Opt_trunc_xino),
16551 +       /* "trunc_xino_v=%d:%d" */
16552 +       /* fsparam_string("trunc_xino_v", Opt_trunc_xino_v), */
16553 +       fsparam_path("trunc_xino", Opt_trunc_xino_path),
16554 +       fsparam_s32("itrunc_xino", Opt_itrunc_xino),
16555 +       /* fsparam_path("zxino", Opt_zxino), */
16556 +       fsparam_flag_no("trunc_xib", Opt_trunc_xib),
16557 +
16558 +#ifdef CONFIG_PROC_FS
16559 +       fsparam_flag_no("plink", Opt_plink),
16560 +#else
16561 +       au_ignore_flag("plink", Opt_ignore),
16562 +#endif
16563 +
16564 +#ifdef CONFIG_AUFS_DEBUG
16565 +       fsparam_flag("list_plink", Opt_list_plink),
16566 +#endif
16567 +
16568 +       fsparam_string("udba", Opt_udba),
16569 +
16570 +       fsparam_flag_no("dio", Opt_dio),
16571 +
16572 +#ifdef CONFIG_AUFS_DIRREN
16573 +       fsparam_flag_no("dirren", Opt_dirren),
16574 +#else
16575 +       au_ignore_flag("dirren", Opt_ignore),
16576 +#endif
16577 +
16578 +#ifdef CONFIG_AUFS_FHSM
16579 +       fsparam_s32("fhsm_sec", Opt_fhsm_sec),
16580 +#else
16581 +       fsparam_s32("fhsm_sec", Opt_ignore),
16582 +#endif
16583 +
16584 +       /* always | a | whiteouted | w */
16585 +       fsparam_string("diropq", Opt_diropq),
16586 +
16587 +       fsparam_flag_no("warn_perm", Opt_warn_perm),
16588 +
16589 +#ifdef CONFIG_AUFS_SHWH
16590 +       fsparam_flag_no("shwh", Opt_shwh),
16591 +#else
16592 +       au_ignore_flag("shwh", Opt_err),
16593 +#endif
16594 +
16595 +       fsparam_flag_no("dirperm1", Opt_dirperm1),
16596 +
16597 +       fsparam_flag_no("verbose", Opt_verbose),
16598 +       fsparam_flag("v", Opt_verbose),
16599 +       fsparam_flag("quiet", Opt_noverbose),
16600 +       fsparam_flag("q", Opt_noverbose),
16601 +       /* user-space may handle this */
16602 +       fsparam_flag("silent", Opt_noverbose),
16603 +
16604 +       fsparam_flag_no("sum", Opt_sum),
16605 +       fsparam_flag("wsum", Opt_wsum),
16606 +
16607 +       fsparam_s32("rdcache", Opt_rdcache),
16608 +       /* "def" or s32 */
16609 +       fsparam_string("rdblk", Opt_rdblk),
16610 +       /* "def" or s32 */
16611 +       fsparam_string("rdhash", Opt_rdhash),
16612 +
16613 +       fsparam_string("create", Opt_wbr_create),
16614 +       fsparam_string("create_policy", Opt_wbr_create),
16615 +       fsparam_string("cpup", Opt_wbr_copyup),
16616 +       fsparam_string("copyup", Opt_wbr_copyup),
16617 +       fsparam_string("copyup_policy", Opt_wbr_copyup),
16618 +
16619 +       /* generic VFS flag */
16620 +#ifdef CONFIG_FS_POSIX_ACL
16621 +       fsparam_flag_no("acl", Opt_acl),
16622 +#else
16623 +       au_ignore_flag("acl", Opt_ignore),
16624 +#endif
16625 +
16626 +       /* internal use for the scripts */
16627 +       fsparam_string("si", Opt_ignore_silent),
16628 +
16629 +       /* obsoleted, keep them temporary */
16630 +       fsparam_flag("nodlgt", Opt_ignore_silent),
16631 +       fsparam_flag("clean_plink", Opt_ignore),
16632 +       fsparam_string("dirs", Opt_br),
16633 +       fsparam_u32("debug", Opt_ignore),
16634 +       /* "whiteout" or "all" */
16635 +       fsparam_string("delete", Opt_ignore),
16636 +       fsparam_string("imap", Opt_ignore),
16637 +
16638 +       /* temporary workaround, due to old mount(8)? */
16639 +       fsparam_flag("relatime", Opt_ignore_silent),
16640 +
16641 +       {}
16642 +};
16643 +
16644 +static int au_fsctx_parse_do_add(struct fs_context *fc, struct au_opt *opt,
16645 +                                char *brspec, size_t speclen,
16646 +                                aufs_bindex_t bindex)
16647 +{
16648 +       int err;
16649 +       char *p;
16650 +
16651 +       AuDbg("brspec %s\n", brspec);
16652 +
16653 +       err = -ENOMEM;
16654 +       if (!speclen)
16655 +               speclen = strlen(brspec);
16656 +       /* will be freed by au_fsctx_free() */
16657 +       p = kmemdup_nul(brspec, speclen, GFP_NOFS);
16658 +       if (unlikely(!p)) {
16659 +               errorfc(fc, "failed in %s", brspec);
16660 +               goto out;
16661 +       }
16662 +       err = au_opt_add(opt, p, fc->sb_flags, bindex);
16663 +
16664 +out:
16665 +       AuTraceErr(err);
16666 +       return err;
16667 +}
16668 +
16669 +static int au_fsctx_parse_br(struct fs_context *fc, char *brspec)
16670 +{
16671 +       int err;
16672 +       char *p;
16673 +       struct au_fsctx_opts *a = fc->fs_private;
16674 +       struct au_opt *opt = a->opt;
16675 +       aufs_bindex_t bindex = a->bindex;
16676 +
16677 +       AuDbg("brspec %s\n", brspec);
16678 +
16679 +       err = -EINVAL;
16680 +       while ((p = strsep(&brspec, ":")) && *p) {
16681 +               err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, bindex);
16682 +               AuTraceErr(err);
16683 +               if (unlikely(err))
16684 +                       break;
16685 +               bindex++;
16686 +               opt++;
16687 +               if (unlikely(opt > a->opt_tail)) {
16688 +                       err = -E2BIG;
16689 +                       bindex--;
16690 +                       opt--;
16691 +                       break;
16692 +               }
16693 +               opt->type = Opt_tail;
16694 +               a->skipped = 1;
16695 +       }
16696 +       a->bindex = bindex;
16697 +       a->opt = opt;
16698 +
16699 +       AuTraceErr(err);
16700 +       return err;
16701 +}
16702 +
16703 +static int au_fsctx_parse_add(struct fs_context *fc, char *addspec)
16704 +{
16705 +       int err, n;
16706 +       char *p;
16707 +       struct au_fsctx_opts *a = fc->fs_private;
16708 +       struct au_opt *opt = a->opt;
16709 +
16710 +       err = -EINVAL;
16711 +       p = strchr(addspec, ':');
16712 +       if (unlikely(!p)) {
16713 +               errorfc(fc, "bad arg in %s", addspec);
16714 +               goto out;
16715 +       }
16716 +       *p++ = '\0';
16717 +       err = kstrtoint(addspec, 0, &n);
16718 +       if (unlikely(err)) {
16719 +               errorfc(fc, "bad integer in %s", addspec);
16720 +               goto out;
16721 +       }
16722 +       AuDbg("n %d\n", n);
16723 +       err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, n);
16724 +
16725 +out:
16726 +       AuTraceErr(err);
16727 +       return err;
16728 +}
16729 +
16730 +static int au_fsctx_parse_del(struct fs_context *fc, struct au_opt_del *del,
16731 +                             struct fs_parameter *param)
16732 +{
16733 +       int err;
16734 +
16735 +       err = -ENOMEM;
16736 +       /* will be freed by au_fsctx_free() */
16737 +       del->pathname = kmemdup_nul(param->string, param->size, GFP_NOFS);
16738 +       if (unlikely(!del->pathname))
16739 +               goto out;
16740 +       AuDbg("del %s\n", del->pathname);
16741 +       err = vfsub_kern_path(del->pathname, AuOpt_LkupDirFlags, &del->h_path);
16742 +       if (unlikely(err))
16743 +               errorfc(fc, "lookup failed %s (%d)", del->pathname, err);
16744 +
16745 +out:
16746 +       AuTraceErr(err);
16747 +       return err;
16748 +}
16749 +
16750 +#if 0 /* reserved for future use */
16751 +static int au_fsctx_parse_idel(struct fs_context *fc, struct au_opt_del *del,
16752 +                              aufs_bindex_t bindex)
16753 +{
16754 +       int err;
16755 +       struct super_block *sb;
16756 +       struct dentry *root;
16757 +       struct au_fsctx_opts *a = fc->fs_private;
16758 +
16759 +       sb = a->sb;
16760 +       AuDebugOn(!sb);
16761 +
16762 +       err = -EINVAL;
16763 +       root = sb->s_root;
16764 +       aufs_read_lock(root, AuLock_FLUSH);
16765 +       if (bindex < 0 || au_sbbot(sb) < bindex) {
16766 +               errorfc(fc, "out of bounds, %d", bindex);
16767 +               goto out;
16768 +       }
16769 +
16770 +       err = 0;
16771 +       del->h_path.dentry = dget(au_h_dptr(root, bindex));
16772 +       del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
16773 +
16774 +out:
16775 +       aufs_read_unlock(root, !AuLock_IR);
16776 +       AuTraceErr(err);
16777 +       return err;
16778 +}
16779 +#endif
16780 +
16781 +static int au_fsctx_parse_mod(struct fs_context *fc, struct au_opt_mod *mod,
16782 +                             struct fs_parameter *param)
16783 +{
16784 +       int err;
16785 +       struct path path;
16786 +       char *p;
16787 +
16788 +       err = -ENOMEM;
16789 +       /* will be freed by au_fsctx_free() */
16790 +       mod->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16791 +       if (unlikely(!mod->path))
16792 +               goto out;
16793 +
16794 +       err = -EINVAL;
16795 +       p = strchr(mod->path, '=');
16796 +       if (unlikely(!p)) {
16797 +               errorfc(fc, "no permission %s", mod->path);
16798 +               goto out;
16799 +       }
16800 +
16801 +       *p++ = 0;
16802 +       err = vfsub_kern_path(mod->path, AuOpt_LkupDirFlags, &path);
16803 +       if (unlikely(err)) {
16804 +               errorfc(fc, "lookup failed %s (%d)", mod->path, err);
16805 +               goto out;
16806 +       }
16807 +
16808 +       mod->perm = au_br_perm_val(p);
16809 +       AuDbg("mod path %s, perm 0x%x, %s", mod->path, mod->perm, p);
16810 +       mod->h_root = dget(path.dentry);
16811 +       path_put(&path);
16812 +
16813 +out:
16814 +       AuTraceErr(err);
16815 +       return err;
16816 +}
16817 +
16818 +#if 0 /* reserved for future use */
16819 +static int au_fsctx_parse_imod(struct fs_context *fc, struct au_opt_mod *mod,
16820 +                              char *ibrspec)
16821 +{
16822 +       int err, n;
16823 +       char *p;
16824 +       struct super_block *sb;
16825 +       struct dentry *root;
16826 +       struct au_fsctx_opts *a = fc->fs_private;
16827 +
16828 +       sb = a->sb;
16829 +       AuDebugOn(!sb);
16830 +
16831 +       err = -EINVAL;
16832 +       p = strchr(ibrspec, ':');
16833 +       if (unlikely(!p)) {
16834 +               errorfc(fc, "no index, %s", ibrspec);
16835 +               goto out;
16836 +       }
16837 +       *p++ = '\0';
16838 +       err = kstrtoint(ibrspec, 0, &n);
16839 +       if (unlikely(err)) {
16840 +               errorfc(fc, "bad integer in %s", ibrspec);
16841 +               goto out;
16842 +       }
16843 +       AuDbg("n %d\n", n);
16844 +
16845 +       root = sb->s_root;
16846 +       aufs_read_lock(root, AuLock_FLUSH);
16847 +       if (n < 0 || au_sbbot(sb) < n) {
16848 +               errorfc(fc, "out of bounds, %d", bindex);
16849 +               goto out_root;
16850 +       }
16851 +
16852 +       err = 0;
16853 +       mod->perm = au_br_perm_val(p);
16854 +       AuDbg("mod path %s, perm 0x%x, %s\n",
16855 +             mod->path, mod->perm, p);
16856 +       mod->h_root = dget(au_h_dptr(root, bindex));
16857 +
16858 +out_root:
16859 +       aufs_read_unlock(root, !AuLock_IR);
16860 +out:
16861 +       AuTraceErr(err);
16862 +       return err;
16863 +}
16864 +#endif
16865 +
16866 +static int au_fsctx_parse_xino(struct fs_context *fc,
16867 +                              struct au_opt_xino *xino,
16868 +                              struct fs_parameter *param)
16869 +{
16870 +       int err;
16871 +       struct au_fsctx_opts *a = fc->fs_private;
16872 +
16873 +       err = -ENOMEM;
16874 +       /* will be freed by au_opts_free() */
16875 +       xino->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16876 +       if (unlikely(!xino->path))
16877 +               goto out;
16878 +       AuDbg("path %s\n", xino->path);
16879 +
16880 +       xino->file = au_xino_create(a->sb, xino->path, /*silent*/0,
16881 +                                   /*wbrtop*/0);
16882 +       err = PTR_ERR(xino->file);
16883 +       if (IS_ERR(xino->file)) {
16884 +               xino->file = NULL;
16885 +               goto out;
16886 +       }
16887 +
16888 +       err = 0;
16889 +       if (unlikely(a->sb && xino->file->f_path.dentry->d_sb == a->sb)) {
16890 +               err = -EINVAL;
16891 +               errorfc(fc, "%s must be outside", xino->path);
16892 +       }
16893 +
16894 +out:
16895 +       AuTraceErr(err);
16896 +       return err;
16897 +}
16898 +
16899 +static
16900 +int au_fsctx_parse_xino_itrunc_path(struct fs_context *fc,
16901 +                                   struct au_opt_xino_itrunc *xino_itrunc,
16902 +                                   char *pathname)
16903 +{
16904 +       int err;
16905 +       aufs_bindex_t bbot, bindex;
16906 +       struct path path;
16907 +       struct dentry *root;
16908 +       struct au_fsctx_opts *a = fc->fs_private;
16909 +
16910 +       AuDebugOn(!a->sb);
16911 +
16912 +       err = vfsub_kern_path(pathname, AuOpt_LkupDirFlags, &path);
16913 +       if (unlikely(err)) {
16914 +               errorfc(fc, "lookup failed %s (%d)", pathname, err);
16915 +               goto out;
16916 +       }
16917 +
16918 +       xino_itrunc->bindex = -1;
16919 +       root = a->sb->s_root;
16920 +       aufs_read_lock(root, AuLock_FLUSH);
16921 +       bbot = au_sbbot(a->sb);
16922 +       for (bindex = 0; bindex <= bbot; bindex++) {
16923 +               if (au_h_dptr(root, bindex) == path.dentry) {
16924 +                       xino_itrunc->bindex = bindex;
16925 +                       break;
16926 +               }
16927 +       }
16928 +       aufs_read_unlock(root, !AuLock_IR);
16929 +       path_put(&path);
16930 +
16931 +       if (unlikely(xino_itrunc->bindex < 0)) {
16932 +               err = -EINVAL;
16933 +               errorfc(fc, "no such branch %s", pathname);
16934 +       }
16935 +
16936 +out:
16937 +       AuTraceErr(err);
16938 +       return err;
16939 +}
16940 +
16941 +static int au_fsctx_parse_xino_itrunc(struct fs_context *fc,
16942 +                                     struct au_opt_xino_itrunc *xino_itrunc,
16943 +                                     unsigned int bindex)
16944 +{
16945 +       int err;
16946 +       aufs_bindex_t bbot;
16947 +       struct super_block *sb;
16948 +       struct au_fsctx_opts *a = fc->fs_private;
16949 +
16950 +       sb = a->sb;
16951 +       AuDebugOn(!sb);
16952 +
16953 +       err = 0;
16954 +       si_noflush_read_lock(sb);
16955 +       bbot = au_sbbot(sb);
16956 +       si_read_unlock(sb);
16957 +       if (bindex <= bbot)
16958 +               xino_itrunc->bindex = bindex;
16959 +       else {
16960 +               err = -EINVAL;
16961 +               errorfc(fc, "out of bounds, %u", bindex);
16962 +       }
16963 +
16964 +       AuTraceErr(err);
16965 +       return err;
16966 +}
16967 +
16968 +static int au_fsctx_parse_param(struct fs_context *fc, struct fs_parameter *param)
16969 +{
16970 +       int err, token;
16971 +       struct fs_parse_result result;
16972 +       struct au_fsctx_opts *a = fc->fs_private;
16973 +       struct au_opt *opt = a->opt;
16974 +
16975 +       AuDbg("fc %p, param {key %s, string %s}\n",
16976 +             fc, param->key, param->string);
16977 +       err = fs_parse(fc, aufs_fsctx_paramspec, param, &result);
16978 +       if (unlikely(err < 0))
16979 +               goto out;
16980 +       token = err;
16981 +       AuDbg("token %d, res{negated %d, uint64 %llu}\n",
16982 +             token, result.negated, result.uint_64);
16983 +
16984 +       err = -EINVAL;
16985 +       a->skipped = 0;
16986 +       switch (token) {
16987 +       case Opt_br:
16988 +               err = au_fsctx_parse_br(fc, param->string);
16989 +               break;
16990 +       case Opt_add:
16991 +               err = au_fsctx_parse_add(fc, param->string);
16992 +               break;
16993 +       case Opt_append:
16994 +               err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
16995 +                                           /*dummy bindex*/1);
16996 +               break;
16997 +       case Opt_prepend:
16998 +               err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
16999 +                                           /*bindex*/0);
17000 +               break;
17001 +
17002 +       case Opt_del:
17003 +               err = au_fsctx_parse_del(fc, &opt->del, param);
17004 +               break;
17005 +#if 0 /* reserved for future use */
17006 +       case Opt_idel:
17007 +               if (!a->sb) {
17008 +                       err = 0;
17009 +                       a->skipped = 1;
17010 +                       break;
17011 +               }
17012 +               del->pathname = "(indexed)";
17013 +               err = au_opts_parse_idel(fc, &opt->del, result.uint_32);
17014 +               break;
17015 +#endif
17016 +
17017 +       case Opt_mod:
17018 +               err = au_fsctx_parse_mod(fc, &opt->mod, param);
17019 +               break;
17020 +#ifdef IMOD /* reserved for future use */
17021 +       case Opt_imod:
17022 +               if (!a->sb) {
17023 +                       err = 0;
17024 +                       a->skipped = 1;
17025 +                       break;
17026 +               }
17027 +               u.mod->path = "(indexed)";
17028 +               err = au_opts_parse_imod(fc, &opt->mod, param->string);
17029 +               break;
17030 +#endif
17031 +
17032 +       case Opt_xino:
17033 +               err = au_fsctx_parse_xino(fc, &opt->xino, param);
17034 +               break;
17035 +       case Opt_trunc_xino_path:
17036 +               if (!a->sb) {
17037 +                       errorfc(fc, "no such branch %s", param->string);
17038 +                       break;
17039 +               }
17040 +               err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17041 +                                                     param->string);
17042 +               break;
17043 +#if 0
17044 +       case Opt_trunc_xino_v:
17045 +               if (!a->sb) {
17046 +                       err = 0;
17047 +                       a->skipped = 1;
17048 +                       break;
17049 +               }
17050 +               err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17051 +                                                     param->string);
17052 +               break;
17053 +#endif
17054 +       case Opt_itrunc_xino:
17055 +               if (!a->sb) {
17056 +                       errorfc(fc, "out of bounds %s", param->string);
17057 +                       break;
17058 +               }
17059 +               err = au_fsctx_parse_xino_itrunc(fc, &opt->xino_itrunc,
17060 +                                                result.int_32);
17061 +               break;
17062 +
17063 +       case Opt_dirwh:
17064 +               err = 0;
17065 +               opt->dirwh = result.int_32;
17066 +               break;
17067 +
17068 +       case Opt_rdcache:
17069 +               if (unlikely(result.int_32 > AUFS_RDCACHE_MAX)) {
17070 +                       errorfc(fc, "rdcache must be smaller than %d",
17071 +                               AUFS_RDCACHE_MAX);
17072 +                       break;
17073 +               }
17074 +               err = 0;
17075 +               opt->rdcache = result.int_32;
17076 +               break;
17077 +
17078 +       case Opt_rdblk:
17079 +               err = 0;
17080 +               opt->rdblk = AUFS_RDBLK_DEF;
17081 +               if (!strcmp(param->string, "def"))
17082 +                       break;
17083 +
17084 +               err = kstrtoint(param->string, 0, &result.int_32);
17085 +               if (unlikely(err)) {
17086 +                       errorfc(fc, "bad value in %s", param->key);
17087 +                       break;
17088 +               }
17089 +               err = -EINVAL;
17090 +               if (unlikely(result.int_32 < 0
17091 +                            || result.int_32 > KMALLOC_MAX_SIZE)) {
17092 +                       errorfc(fc, "bad value in %s", param->key);
17093 +                       break;
17094 +               }
17095 +               if (unlikely(result.int_32 && result.int_32 < NAME_MAX)) {
17096 +                       errorfc(fc, "rdblk must be larger than %d", NAME_MAX);
17097 +                       break;
17098 +               }
17099 +               err = 0;
17100 +               opt->rdblk = result.int_32;
17101 +               break;
17102 +
17103 +       case Opt_rdhash:
17104 +               err = 0;
17105 +               opt->rdhash = AUFS_RDHASH_DEF;
17106 +               if (!strcmp(param->string, "def"))
17107 +                       break;
17108 +
17109 +               err = kstrtoint(param->string, 0, &result.int_32);
17110 +               if (unlikely(err)) {
17111 +                       errorfc(fc, "bad value in %s", param->key);
17112 +                       break;
17113 +               }
17114 +               /* how about zero? */
17115 +               if (result.int_32 < 0
17116 +                   || result.int_32 * sizeof(struct hlist_head)
17117 +                   > KMALLOC_MAX_SIZE) {
17118 +                       err = -EINVAL;
17119 +                       errorfc(fc, "bad integer in %s", param->key);
17120 +                       break;
17121 +               }
17122 +               opt->rdhash = result.int_32;
17123 +               break;
17124 +
17125 +       case Opt_diropq:
17126 +               /*
17127 +                * As other options, fs/aufs/opts.c can handle these strings by
17128 +                * match_token().  But "diropq=" is deprecated now and will
17129 +                * never have other value.  So simple strcmp() is enough here.
17130 +                */
17131 +               if (!strcmp(param->string, "a") ||
17132 +                   !strcmp(param->string, "always")) {
17133 +                       err = 0;
17134 +                       opt->type = Opt_diropq_a;
17135 +               } else if (!strcmp(param->string, "w") ||
17136 +                          !strcmp(param->string, "whiteouted")) {
17137 +                       err = 0;
17138 +                       opt->type = Opt_diropq_w;
17139 +               } else
17140 +                       errorfc(fc, "unknown value %s", param->string);
17141 +               break;
17142 +
17143 +       case Opt_udba:
17144 +               opt->udba = au_udba_val(param->string);
17145 +               if (opt->udba >= 0)
17146 +                       err = 0;
17147 +               else
17148 +                       errorf(fc, "wrong value, %s", param->string);
17149 +               break;
17150 +
17151 +       case Opt_wbr_create:
17152 +               opt->wbr_create.wbr_create
17153 +                       = au_wbr_create_val(param->string, &opt->wbr_create);
17154 +               if (opt->wbr_create.wbr_create >= 0)
17155 +                       err = 0;
17156 +               else
17157 +                       errorf(fc, "wrong value, %s", param->key);
17158 +               break;
17159 +
17160 +       case Opt_wbr_copyup:
17161 +               opt->wbr_copyup = au_wbr_copyup_val(param->string);
17162 +               if (opt->wbr_copyup >= 0)
17163 +                       err = 0;
17164 +               else
17165 +                       errorfc(fc, "wrong value, %s", param->key);
17166 +               break;
17167 +
17168 +       case Opt_fhsm_sec:
17169 +               if (unlikely(result.int_32 < 0)) {
17170 +                       errorfc(fc, "bad integer in %s\n", param->key);
17171 +                       break;
17172 +               }
17173 +               err = 0;
17174 +               if (sysaufs_brs)
17175 +                       opt->fhsm_second = result.int_32;
17176 +               else
17177 +                       warnfc(fc, "ignored %s %s", param->key, param->string);
17178 +               break;
17179 +
17180 +       /* simple true/false flag */
17181 +#define au_fsctx_TF(name)                              \
17182 +               case Opt_##name:                        \
17183 +                       err = 0;                        \
17184 +                       opt->tf = !result.negated;      \
17185 +                       break
17186 +       au_fsctx_TF(trunc_xino);
17187 +       au_fsctx_TF(trunc_xib);
17188 +       au_fsctx_TF(dirperm1);
17189 +       au_fsctx_TF(plink);
17190 +       au_fsctx_TF(shwh);
17191 +       au_fsctx_TF(dio);
17192 +       au_fsctx_TF(warn_perm);
17193 +       au_fsctx_TF(verbose);
17194 +       au_fsctx_TF(sum);
17195 +       au_fsctx_TF(dirren);
17196 +       au_fsctx_TF(acl);
17197 +#undef au_fsctx_TF
17198 +
17199 +       case Opt_noverbose:
17200 +               err = 0;
17201 +               opt->type = Opt_verbose;
17202 +               opt->tf = false;
17203 +               break;
17204 +
17205 +       case Opt_noxino:
17206 +               fallthrough;
17207 +       case Opt_list_plink:
17208 +               fallthrough;
17209 +       case Opt_wsum:
17210 +               err = 0;
17211 +               break;
17212 +
17213 +       case Opt_ignore:
17214 +               warnfc(fc, "ignored %s", param->key);
17215 +               fallthrough;
17216 +       case Opt_ignore_silent:
17217 +               a->skipped = 1;
17218 +               err = 0;
17219 +               break;
17220 +       default:
17221 +               a->skipped = 1;
17222 +               err = -ENOPARAM;
17223 +               break;
17224 +       }
17225 +       if (unlikely(err))
17226 +               goto out;
17227 +       if (a->skipped)
17228 +               goto out;
17229 +
17230 +       switch (token) {
17231 +       case Opt_br:
17232 +               fallthrough;
17233 +       case Opt_noverbose:
17234 +               fallthrough;
17235 +       case Opt_diropq:
17236 +               break;
17237 +       default:
17238 +               opt->type = token;
17239 +               break;
17240 +       }
17241 +       opt++;
17242 +       if (unlikely(opt > a->opt_tail)) {
17243 +               err = -E2BIG;
17244 +               opt--;
17245 +       }
17246 +       opt->type = Opt_tail;
17247 +       a->opt = opt;
17248 +
17249 +out:
17250 +       return err;
17251 +}
17252 +
17253 +/*
17254 + * these options accept both 'name=val' and 'name:val' form.
17255 + * some accept optional '=' in its value.
17256 + * eg. br:/br1=rw:/br2=ro and br=/br1=rw:/br2=ro
17257 + */
17258 +static inline unsigned int is_colonopt(char *str)
17259 +{
17260 +#define do_test(name)                                  \
17261 +       if (!strncmp(str, name ":", sizeof(name)))      \
17262 +               return sizeof(name) - 1
17263 +       do_test("br");
17264 +       do_test("add");
17265 +       do_test("ins");
17266 +       do_test("append");
17267 +       do_test("prepend");
17268 +       do_test("del");
17269 +       /* do_test("idel"); */
17270 +       do_test("mod");
17271 +       /* do_test("imod"); */
17272 +#undef do_test
17273 +
17274 +       return 0;
17275 +}
17276 +
17277 +static int au_fsctx_parse_monolithic(struct fs_context *fc, void *data)
17278 +{
17279 +       int err;
17280 +       unsigned int u;
17281 +       char *str;
17282 +       struct au_fsctx_opts *a = fc->fs_private;
17283 +
17284 +       str = data;
17285 +       AuDbg("str %s\n", str);
17286 +       while (str) {
17287 +               u = is_colonopt(str);
17288 +               if (u)
17289 +                       str[u] = '=';
17290 +               str = strchr(str, ',');
17291 +               if (!str)
17292 +                       break;
17293 +               str++;
17294 +       }
17295 +       str = data;
17296 +       AuDbg("str %s\n", str);
17297 +
17298 +       err = generic_parse_monolithic(fc, str);
17299 +       AuTraceErr(err);
17300 +       au_fsctx_dump(&a->opts);
17301 +
17302 +       return err;
17303 +}
17304 +
17305 +/* ---------------------------------------------------------------------- */
17306 +
17307 +static void au_fsctx_opts_free(struct au_opts *opts)
17308 +{
17309 +       struct au_opt *opt;
17310 +
17311 +       opt = opts->opt;
17312 +       while (opt->type != Opt_tail) {
17313 +               switch (opt->type) {
17314 +               case Opt_add:
17315 +                       fallthrough;
17316 +               case Opt_append:
17317 +                       fallthrough;
17318 +               case Opt_prepend:
17319 +                       kfree(opt->add.pathname);
17320 +                       path_put(&opt->add.path);
17321 +                       break;
17322 +               case Opt_del:
17323 +                       kfree(opt->del.pathname);
17324 +                       fallthrough;
17325 +               case Opt_idel:
17326 +                       path_put(&opt->del.h_path);
17327 +                       break;
17328 +               case Opt_mod:
17329 +                       kfree(opt->mod.path);
17330 +                       fallthrough;
17331 +               case Opt_imod:
17332 +                       dput(opt->mod.h_root);
17333 +                       break;
17334 +               case Opt_xino:
17335 +                       kfree(opt->xino.path);
17336 +                       fput(opt->xino.file);
17337 +                       break;
17338 +               }
17339 +               opt++;
17340 +       }
17341 +}
17342 +
17343 +static void au_fsctx_free(struct fs_context *fc)
17344 +{
17345 +       struct au_fsctx_opts *a = fc->fs_private;
17346 +
17347 +       /* fs_type=%p, root=%pD */
17348 +       AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17349 +             fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17350 +
17351 +       kobject_put(&a->sbinfo->si_kobj);
17352 +       au_fsctx_opts_free(&a->opts);
17353 +       free_page((unsigned long)a->opts.opt);
17354 +       au_kfree_rcu(a);
17355 +}
17356 +
17357 +static const struct fs_context_operations au_fsctx_ops = {
17358 +       .free                   = au_fsctx_free,
17359 +       .parse_param            = au_fsctx_parse_param,
17360 +       .parse_monolithic       = au_fsctx_parse_monolithic,
17361 +       .get_tree               = au_fsctx_get_tree,
17362 +       .reconfigure            = au_fsctx_reconfigure
17363 +       /*
17364 +        * nfs4 requires ->dup()? No.
17365 +        * I don't know what is this ->dup() for.
17366 +        */
17367 +};
17368 +
17369 +int aufs_fsctx_init(struct fs_context *fc)
17370 +{
17371 +       int err;
17372 +       struct au_fsctx_opts *a;
17373 +
17374 +       /* fs_type=%p, root=%pD */
17375 +       AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17376 +             fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17377 +
17378 +       /* they will be freed by au_fsctx_free() */
17379 +       err = -ENOMEM;
17380 +       a = kzalloc(sizeof(*a), GFP_NOFS);
17381 +       if (unlikely(!a))
17382 +               goto out;
17383 +       a->bindex = 0;
17384 +       a->opts.opt = (void *)__get_free_page(GFP_NOFS);
17385 +       if (unlikely(!a->opts.opt))
17386 +               goto out_a;
17387 +       a->opt = a->opts.opt;
17388 +       a->opt->type = Opt_tail;
17389 +       a->opts.max_opt = PAGE_SIZE / sizeof(*a->opts.opt);
17390 +       a->opt_tail = a->opt + a->opts.max_opt - 1;
17391 +       a->opts.sb_flags = fc->sb_flags;
17392 +
17393 +       a->sb = NULL;
17394 +       if (fc->root) {
17395 +               AuDebugOn(fc->purpose != FS_CONTEXT_FOR_RECONFIGURE);
17396 +               a->opts.flags = AuOpts_REMOUNT;
17397 +               a->sb = fc->root->d_sb;
17398 +               a->sbinfo = au_sbi(a->sb);
17399 +               kobject_get(&a->sbinfo->si_kobj);
17400 +       } else {
17401 +               a->sbinfo = au_si_alloc(a->sb);
17402 +               AuDebugOn(!a->sbinfo);
17403 +               err = PTR_ERR(a->sbinfo);
17404 +               if (IS_ERR(a->sbinfo))
17405 +                       goto out_opt;
17406 +               au_rw_write_unlock(&a->sbinfo->si_rwsem);
17407 +       }
17408 +
17409 +       err = 0;
17410 +       fc->fs_private = a;
17411 +       fc->ops = &au_fsctx_ops;
17412 +       goto out; /* success */
17413 +
17414 +out_opt:
17415 +       free_page((unsigned long)a->opts.opt);
17416 +out_a:
17417 +       au_kfree_rcu(a);
17418 +out:
17419 +       AuTraceErr(err);
17420 +       return err;
17421 +}
17422 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
17423 --- /usr/share/empty/fs/aufs/fstype.h   1970-01-01 01:00:00.000000000 +0100
17424 +++ linux/fs/aufs/fstype.h      2022-11-05 23:02:18.965889284 +0100
17425 @@ -0,0 +1,401 @@
17426 +/* SPDX-License-Identifier: GPL-2.0 */
17427 +/*
17428 + * Copyright (C) 2005-2022 Junjiro R. Okajima
17429 + *
17430 + * This program is free software; you can redistribute it and/or modify
17431 + * it under the terms of the GNU General Public License as published by
17432 + * the Free Software Foundation; either version 2 of the License, or
17433 + * (at your option) any later version.
17434 + *
17435 + * This program is distributed in the hope that it will be useful,
17436 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17437 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17438 + * GNU General Public License for more details.
17439 + *
17440 + * You should have received a copy of the GNU General Public License
17441 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
17442 + */
17443 +
17444 +/*
17445 + * judging filesystem type
17446 + */
17447 +
17448 +#ifndef __AUFS_FSTYPE_H__
17449 +#define __AUFS_FSTYPE_H__
17450 +
17451 +#ifdef __KERNEL__
17452 +
17453 +#include <linux/fs.h>
17454 +#include <linux/magic.h>
17455 +#include <linux/nfs_fs.h>
17456 +#include <linux/romfs_fs.h>
17457 +
17458 +static inline int au_test_aufs(struct super_block *sb)
17459 +{
17460 +       return sb->s_magic == AUFS_SUPER_MAGIC;
17461 +}
17462 +
17463 +static inline const char *au_sbtype(struct super_block *sb)
17464 +{
17465 +       return sb->s_type->name;
17466 +}
17467 +
17468 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
17469 +{
17470 +#if IS_ENABLED(CONFIG_ISO9660_FS)
17471 +       return sb->s_magic == ISOFS_SUPER_MAGIC;
17472 +#else
17473 +       return 0;
17474 +#endif
17475 +}
17476 +
17477 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
17478 +{
17479 +#if IS_ENABLED(CONFIG_ROMFS_FS)
17480 +       return sb->s_magic == ROMFS_MAGIC;
17481 +#else
17482 +       return 0;
17483 +#endif
17484 +}
17485 +
17486 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
17487 +{
17488 +#if IS_ENABLED(CONFIG_CRAMFS)
17489 +       return sb->s_magic == CRAMFS_MAGIC;
17490 +#endif
17491 +       return 0;
17492 +}
17493 +
17494 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
17495 +{
17496 +#if IS_ENABLED(CONFIG_NFS_FS)
17497 +       return sb->s_magic == NFS_SUPER_MAGIC;
17498 +#else
17499 +       return 0;
17500 +#endif
17501 +}
17502 +
17503 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
17504 +{
17505 +#if IS_ENABLED(CONFIG_FUSE_FS)
17506 +       return sb->s_magic == FUSE_SUPER_MAGIC;
17507 +#else
17508 +       return 0;
17509 +#endif
17510 +}
17511 +
17512 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
17513 +{
17514 +#if IS_ENABLED(CONFIG_XFS_FS)
17515 +       return sb->s_magic == XFS_SB_MAGIC;
17516 +#else
17517 +       return 0;
17518 +#endif
17519 +}
17520 +
17521 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
17522 +{
17523 +#ifdef CONFIG_TMPFS
17524 +       return sb->s_magic == TMPFS_MAGIC;
17525 +#else
17526 +       return 0;
17527 +#endif
17528 +}
17529 +
17530 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
17531 +{
17532 +#if IS_ENABLED(CONFIG_ECRYPT_FS)
17533 +       return !strcmp(au_sbtype(sb), "ecryptfs");
17534 +#else
17535 +       return 0;
17536 +#endif
17537 +}
17538 +
17539 +static inline int au_test_ramfs(struct super_block *sb)
17540 +{
17541 +       return sb->s_magic == RAMFS_MAGIC;
17542 +}
17543 +
17544 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
17545 +{
17546 +#if IS_ENABLED(CONFIG_UBIFS_FS)
17547 +       return sb->s_magic == UBIFS_SUPER_MAGIC;
17548 +#else
17549 +       return 0;
17550 +#endif
17551 +}
17552 +
17553 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
17554 +{
17555 +#ifdef CONFIG_PROC_FS
17556 +       return sb->s_magic == PROC_SUPER_MAGIC;
17557 +#else
17558 +       return 0;
17559 +#endif
17560 +}
17561 +
17562 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
17563 +{
17564 +#ifdef CONFIG_SYSFS
17565 +       return sb->s_magic == SYSFS_MAGIC;
17566 +#else
17567 +       return 0;
17568 +#endif
17569 +}
17570 +
17571 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
17572 +{
17573 +#if IS_ENABLED(CONFIG_CONFIGFS_FS)
17574 +       return sb->s_magic == CONFIGFS_MAGIC;
17575 +#else
17576 +       return 0;
17577 +#endif
17578 +}
17579 +
17580 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
17581 +{
17582 +#if IS_ENABLED(CONFIG_MINIX_FS)
17583 +       return sb->s_magic == MINIX3_SUPER_MAGIC
17584 +               || sb->s_magic == MINIX2_SUPER_MAGIC
17585 +               || sb->s_magic == MINIX2_SUPER_MAGIC2
17586 +               || sb->s_magic == MINIX_SUPER_MAGIC
17587 +               || sb->s_magic == MINIX_SUPER_MAGIC2;
17588 +#else
17589 +       return 0;
17590 +#endif
17591 +}
17592 +
17593 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
17594 +{
17595 +#if IS_ENABLED(CONFIG_FAT_FS)
17596 +       return sb->s_magic == MSDOS_SUPER_MAGIC;
17597 +#else
17598 +       return 0;
17599 +#endif
17600 +}
17601 +
17602 +static inline int au_test_msdos(struct super_block *sb)
17603 +{
17604 +       return au_test_fat(sb);
17605 +}
17606 +
17607 +static inline int au_test_vfat(struct super_block *sb)
17608 +{
17609 +       return au_test_fat(sb);
17610 +}
17611 +
17612 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
17613 +{
17614 +#ifdef CONFIG_SECURITYFS
17615 +       return sb->s_magic == SECURITYFS_MAGIC;
17616 +#else
17617 +       return 0;
17618 +#endif
17619 +}
17620 +
17621 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
17622 +{
17623 +#if IS_ENABLED(CONFIG_SQUASHFS)
17624 +       return sb->s_magic == SQUASHFS_MAGIC;
17625 +#else
17626 +       return 0;
17627 +#endif
17628 +}
17629 +
17630 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
17631 +{
17632 +#if IS_ENABLED(CONFIG_BTRFS_FS)
17633 +       return sb->s_magic == BTRFS_SUPER_MAGIC;
17634 +#else
17635 +       return 0;
17636 +#endif
17637 +}
17638 +
17639 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
17640 +{
17641 +#if IS_ENABLED(CONFIG_XENFS)
17642 +       return sb->s_magic == XENFS_SUPER_MAGIC;
17643 +#else
17644 +       return 0;
17645 +#endif
17646 +}
17647 +
17648 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
17649 +{
17650 +#ifdef CONFIG_DEBUG_FS
17651 +       return sb->s_magic == DEBUGFS_MAGIC;
17652 +#else
17653 +       return 0;
17654 +#endif
17655 +}
17656 +
17657 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
17658 +{
17659 +#if IS_ENABLED(CONFIG_NILFS)
17660 +       return sb->s_magic == NILFS_SUPER_MAGIC;
17661 +#else
17662 +       return 0;
17663 +#endif
17664 +}
17665 +
17666 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
17667 +{
17668 +#if IS_ENABLED(CONFIG_HFSPLUS_FS)
17669 +       return sb->s_magic == HFSPLUS_SUPER_MAGIC;
17670 +#else
17671 +       return 0;
17672 +#endif
17673 +}
17674 +
17675 +/* ---------------------------------------------------------------------- */
17676 +/*
17677 + * they can't be an aufs branch.
17678 + */
17679 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
17680 +{
17681 +       return
17682 +#ifndef CONFIG_AUFS_BR_RAMFS
17683 +               au_test_ramfs(sb) ||
17684 +#endif
17685 +               au_test_procfs(sb)
17686 +               || au_test_sysfs(sb)
17687 +               || au_test_configfs(sb)
17688 +               || au_test_debugfs(sb)
17689 +               || au_test_securityfs(sb)
17690 +               || au_test_xenfs(sb)
17691 +               || au_test_ecryptfs(sb)
17692 +               /* || !strcmp(au_sbtype(sb), "unionfs") */
17693 +               || au_test_aufs(sb); /* will be supported in next version */
17694 +}
17695 +
17696 +static inline int au_test_fs_remote(struct super_block *sb)
17697 +{
17698 +       return !au_test_tmpfs(sb)
17699 +#ifdef CONFIG_AUFS_BR_RAMFS
17700 +               && !au_test_ramfs(sb)
17701 +#endif
17702 +               && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
17703 +}
17704 +
17705 +/* ---------------------------------------------------------------------- */
17706 +
17707 +/*
17708 + * Note: these functions (below) are created after reading ->getattr() in all
17709 + * filesystems under linux/fs. it means we have to do so in every update...
17710 + */
17711 +
17712 +/*
17713 + * some filesystems require getattr to refresh the inode attributes before
17714 + * referencing.
17715 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
17716 + * and leave the work for d_revalidate()
17717 + */
17718 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
17719 +{
17720 +       return au_test_nfs(sb)
17721 +               || au_test_fuse(sb)
17722 +               /* || au_test_btrfs(sb) */      /* untested */
17723 +               ;
17724 +}
17725 +
17726 +/*
17727 + * filesystems which don't maintain i_size or i_blocks.
17728 + */
17729 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
17730 +{
17731 +       return au_test_xfs(sb)
17732 +               || au_test_btrfs(sb)
17733 +               || au_test_ubifs(sb)
17734 +               || au_test_hfsplus(sb)  /* maintained, but incorrect */
17735 +               /* || au_test_minix(sb) */      /* untested */
17736 +               ;
17737 +}
17738 +
17739 +/*
17740 + * filesystems which don't store the correct value in some of their inode
17741 + * attributes.
17742 + */
17743 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
17744 +{
17745 +       return au_test_fs_bad_iattr_size(sb)
17746 +               || au_test_fat(sb)
17747 +               || au_test_msdos(sb)
17748 +               || au_test_vfat(sb);
17749 +}
17750 +
17751 +/* they don't check i_nlink in link(2) */
17752 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
17753 +{
17754 +       return au_test_tmpfs(sb)
17755 +#ifdef CONFIG_AUFS_BR_RAMFS
17756 +               || au_test_ramfs(sb)
17757 +#endif
17758 +               || au_test_ubifs(sb)
17759 +               || au_test_hfsplus(sb);
17760 +}
17761 +
17762 +/*
17763 + * filesystems which sets S_NOATIME and S_NOCMTIME.
17764 + */
17765 +static inline int au_test_fs_notime(struct super_block *sb)
17766 +{
17767 +       return au_test_nfs(sb)
17768 +               || au_test_fuse(sb)
17769 +               || au_test_ubifs(sb)
17770 +               ;
17771 +}
17772 +
17773 +/* temporary support for i#1 in cramfs */
17774 +static inline int au_test_fs_unique_ino(struct inode *inode)
17775 +{
17776 +       if (au_test_cramfs(inode->i_sb))
17777 +               return inode->i_ino != 1;
17778 +       return 1;
17779 +}
17780 +
17781 +/* ---------------------------------------------------------------------- */
17782 +
17783 +/*
17784 + * the filesystem where the xino files placed must support i/o after unlink and
17785 + * maintain i_size and i_blocks.
17786 + */
17787 +static inline int au_test_fs_bad_xino(struct super_block *sb)
17788 +{
17789 +       return au_test_fs_remote(sb)
17790 +               || au_test_fs_bad_iattr_size(sb)
17791 +               /* don't want unnecessary work for xino */
17792 +               || au_test_aufs(sb)
17793 +               || au_test_ecryptfs(sb)
17794 +               || au_test_nilfs(sb);
17795 +}
17796 +
17797 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
17798 +{
17799 +       return au_test_tmpfs(sb)
17800 +               || au_test_ramfs(sb);
17801 +}
17802 +
17803 +/*
17804 + * test if the @sb is real-readonly.
17805 + */
17806 +static inline int au_test_fs_rr(struct super_block *sb)
17807 +{
17808 +       return au_test_squashfs(sb)
17809 +               || au_test_iso9660(sb)
17810 +               || au_test_cramfs(sb)
17811 +               || au_test_romfs(sb);
17812 +}
17813 +
17814 +/*
17815 + * test if the @inode is nfs with 'noacl' option
17816 + * NFS always sets SB_POSIXACL regardless its mount option 'noacl.'
17817 + */
17818 +static inline int au_test_nfs_noacl(struct inode *inode)
17819 +{
17820 +       return au_test_nfs(inode->i_sb)
17821 +               /* && IS_POSIXACL(inode) */
17822 +               && !nfs_server_capable(inode, NFS_CAP_ACLS);
17823 +}
17824 +
17825 +#endif /* __KERNEL__ */
17826 +#endif /* __AUFS_FSTYPE_H__ */
17827 diff -urN /usr/share/empty/fs/aufs/hbl.h linux/fs/aufs/hbl.h
17828 --- /usr/share/empty/fs/aufs/hbl.h      1970-01-01 01:00:00.000000000 +0100
17829 +++ linux/fs/aufs/hbl.h 2022-11-05 23:02:18.965889284 +0100
17830 @@ -0,0 +1,65 @@
17831 +/* SPDX-License-Identifier: GPL-2.0 */
17832 +/*
17833 + * Copyright (C) 2017-2022 Junjiro R. Okajima
17834 + *
17835 + * This program is free software; you can redistribute it and/or modify
17836 + * it under the terms of the GNU General Public License as published by
17837 + * the Free Software Foundation; either version 2 of the License, or
17838 + * (at your option) any later version.
17839 + *
17840 + * This program is distributed in the hope that it will be useful,
17841 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17842 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17843 + * GNU General Public License for more details.
17844 + *
17845 + * You should have received a copy of the GNU General Public License
17846 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
17847 + */
17848 +
17849 +/*
17850 + * helpers for hlist_bl.h
17851 + */
17852 +
17853 +#ifndef __AUFS_HBL_H__
17854 +#define __AUFS_HBL_H__
17855 +
17856 +#ifdef __KERNEL__
17857 +
17858 +#include <linux/list_bl.h>
17859 +
17860 +static inline void au_hbl_add(struct hlist_bl_node *node,
17861 +                             struct hlist_bl_head *hbl)
17862 +{
17863 +       hlist_bl_lock(hbl);
17864 +       hlist_bl_add_head(node, hbl);
17865 +       hlist_bl_unlock(hbl);
17866 +}
17867 +
17868 +static inline void au_hbl_del(struct hlist_bl_node *node,
17869 +                             struct hlist_bl_head *hbl)
17870 +{
17871 +       hlist_bl_lock(hbl);
17872 +       hlist_bl_del(node);
17873 +       hlist_bl_unlock(hbl);
17874 +}
17875 +
17876 +#define au_hbl_for_each(pos, head)                                     \
17877 +       for (pos = hlist_bl_first(head);                                \
17878 +            pos;                                                       \
17879 +            pos = pos->next)
17880 +
17881 +static inline unsigned long au_hbl_count(struct hlist_bl_head *hbl)
17882 +{
17883 +       unsigned long cnt;
17884 +       struct hlist_bl_node *pos;
17885 +
17886 +       cnt = 0;
17887 +       hlist_bl_lock(hbl);
17888 +       au_hbl_for_each(pos, hbl)
17889 +               cnt++;
17890 +       hlist_bl_unlock(hbl);
17891 +       return cnt;
17892 +}
17893 +
17894 +#endif /* __KERNEL__ */
17895 +#endif /* __AUFS_HBL_H__ */
17896 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
17897 --- /usr/share/empty/fs/aufs/hfsnotify.c        1970-01-01 01:00:00.000000000 +0100
17898 +++ linux/fs/aufs/hfsnotify.c   2022-11-05 23:02:18.965889284 +0100
17899 @@ -0,0 +1,290 @@
17900 +// SPDX-License-Identifier: GPL-2.0
17901 +/*
17902 + * Copyright (C) 2005-2022 Junjiro R. Okajima
17903 + *
17904 + * This program is free software; you can redistribute it and/or modify
17905 + * it under the terms of the GNU General Public License as published by
17906 + * the Free Software Foundation; either version 2 of the License, or
17907 + * (at your option) any later version.
17908 + *
17909 + * This program is distributed in the hope that it will be useful,
17910 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17911 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17912 + * GNU General Public License for more details.
17913 + *
17914 + * You should have received a copy of the GNU General Public License
17915 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
17916 + */
17917 +
17918 +/*
17919 + * fsnotify for the lower directories
17920 + */
17921 +
17922 +#include "aufs.h"
17923 +
17924 +/* FS_IN_IGNORED is unnecessary */
17925 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
17926 +                                | FS_CREATE | FS_EVENT_ON_CHILD);
17927 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
17928 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
17929 +
17930 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
17931 +{
17932 +       struct au_hnotify *hn = container_of(mark, struct au_hnotify,
17933 +                                            hn_mark);
17934 +       /* AuDbg("here\n"); */
17935 +       au_cache_free_hnotify(hn);
17936 +       smp_mb__before_atomic(); /* for atomic64_dec */
17937 +       if (atomic64_dec_and_test(&au_hfsn_ifree))
17938 +               wake_up(&au_hfsn_wq);
17939 +}
17940 +
17941 +static int au_hfsn_alloc(struct au_hinode *hinode)
17942 +{
17943 +       int err;
17944 +       struct au_hnotify *hn;
17945 +       struct super_block *sb;
17946 +       struct au_branch *br;
17947 +       struct fsnotify_mark *mark;
17948 +       aufs_bindex_t bindex;
17949 +
17950 +       hn = hinode->hi_notify;
17951 +       sb = hn->hn_aufs_inode->i_sb;
17952 +       bindex = au_br_index(sb, hinode->hi_id);
17953 +       br = au_sbr(sb, bindex);
17954 +       AuDebugOn(!br->br_hfsn);
17955 +
17956 +       mark = &hn->hn_mark;
17957 +       fsnotify_init_mark(mark, br->br_hfsn->hfsn_group);
17958 +       mark->mask = AuHfsnMask;
17959 +       /*
17960 +        * by udba rename or rmdir, aufs assign a new inode to the known
17961 +        * h_inode, so specify 1 to allow dups.
17962 +        */
17963 +       lockdep_off();
17964 +       err = fsnotify_add_inode_mark(mark, hinode->hi_inode, /*allow_dups*/1);
17965 +       lockdep_on();
17966 +
17967 +       return err;
17968 +}
17969 +
17970 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
17971 +{
17972 +       struct fsnotify_mark *mark;
17973 +       unsigned long long ull;
17974 +       struct fsnotify_group *group;
17975 +
17976 +       ull = atomic64_inc_return(&au_hfsn_ifree);
17977 +       BUG_ON(!ull);
17978 +
17979 +       mark = &hn->hn_mark;
17980 +       spin_lock(&mark->lock);
17981 +       group = mark->group;
17982 +       fsnotify_get_group(group);
17983 +       spin_unlock(&mark->lock);
17984 +       lockdep_off();
17985 +       fsnotify_destroy_mark(mark, group);
17986 +       fsnotify_put_mark(mark);
17987 +       fsnotify_put_group(group);
17988 +       lockdep_on();
17989 +
17990 +       /* free hn by myself */
17991 +       return 0;
17992 +}
17993 +
17994 +/* ---------------------------------------------------------------------- */
17995 +
17996 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
17997 +{
17998 +       struct fsnotify_mark *mark;
17999 +
18000 +       mark = &hinode->hi_notify->hn_mark;
18001 +       spin_lock(&mark->lock);
18002 +       if (do_set) {
18003 +               AuDebugOn(mark->mask & AuHfsnMask);
18004 +               mark->mask |= AuHfsnMask;
18005 +       } else {
18006 +               AuDebugOn(!(mark->mask & AuHfsnMask));
18007 +               mark->mask &= ~AuHfsnMask;
18008 +       }
18009 +       spin_unlock(&mark->lock);
18010 +       /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
18011 +}
18012 +
18013 +/* ---------------------------------------------------------------------- */
18014 +
18015 +/* #define AuDbgHnotify */
18016 +#ifdef AuDbgHnotify
18017 +static char *au_hfsn_name(u32 mask)
18018 +{
18019 +#ifdef CONFIG_AUFS_DEBUG
18020 +#define test_ret(flag)                         \
18021 +       do {                                    \
18022 +               if (mask & flag)                \
18023 +                       return #flag;           \
18024 +       } while (0)
18025 +       test_ret(FS_ACCESS);
18026 +       test_ret(FS_MODIFY);
18027 +       test_ret(FS_ATTRIB);
18028 +       test_ret(FS_CLOSE_WRITE);
18029 +       test_ret(FS_CLOSE_NOWRITE);
18030 +       test_ret(FS_OPEN);
18031 +       test_ret(FS_MOVED_FROM);
18032 +       test_ret(FS_MOVED_TO);
18033 +       test_ret(FS_CREATE);
18034 +       test_ret(FS_DELETE);
18035 +       test_ret(FS_DELETE_SELF);
18036 +       test_ret(FS_MOVE_SELF);
18037 +       test_ret(FS_UNMOUNT);
18038 +       test_ret(FS_Q_OVERFLOW);
18039 +       test_ret(FS_IN_IGNORED);
18040 +       test_ret(FS_ISDIR);
18041 +       test_ret(FS_IN_ONESHOT);
18042 +       test_ret(FS_EVENT_ON_CHILD);
18043 +       return "";
18044 +#undef test_ret
18045 +#else
18046 +       return "??";
18047 +#endif
18048 +}
18049 +#endif
18050 +
18051 +/* ---------------------------------------------------------------------- */
18052 +
18053 +static void au_hfsn_free_group(struct fsnotify_group *group)
18054 +{
18055 +       struct au_br_hfsnotify *hfsn = group->private;
18056 +
18057 +       /* AuDbg("here\n"); */
18058 +       au_kfree_try_rcu(hfsn);
18059 +}
18060 +
18061 +static int au_hfsn_handle_event(struct fsnotify_group *group,
18062 +                               u32 mask, const void *data, int data_type,
18063 +                               struct inode *dir,
18064 +                               const struct qstr *file_name, u32 cookie,
18065 +                               struct fsnotify_iter_info *iter_info)
18066 +{
18067 +       int err;
18068 +       struct au_hnotify *hnotify;
18069 +       struct inode *h_dir, *h_inode;
18070 +       struct fsnotify_mark *inode_mark;
18071 +
18072 +       AuDebugOn(!(data_type == FSNOTIFY_EVENT_INODE
18073 +                   || data_type == FSNOTIFY_EVENT_DENTRY));
18074 +
18075 +       err = 0;
18076 +       /* if FS_UNMOUNT happens, there must be another bug */
18077 +       AuDebugOn(mask & FS_UNMOUNT);
18078 +       if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
18079 +               goto out;
18080 +
18081 +       h_dir = dir;
18082 +       h_inode = NULL;
18083 +#ifdef AuDbgHnotify
18084 +       au_debug_on();
18085 +       if (1 || file_name.len != sizeof(AUFS_XINO_FNAME) - 1
18086 +           || strncmp(file_name.name, AUFS_XINO_FNAME, file_name.len)) {
18087 +               AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
18088 +                     h_dir->i_ino, mask, au_hfsn_name(mask),
18089 +                     AuLNPair(file_name), h_inode ? h_inode->i_ino : 0);
18090 +               /* WARN_ON(1); */
18091 +       }
18092 +       au_debug_off();
18093 +#endif
18094 +
18095 +       inode_mark = fsnotify_iter_inode_mark(iter_info);
18096 +       AuDebugOn(!inode_mark);
18097 +       hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
18098 +       err = au_hnotify(h_dir, hnotify, mask, file_name, h_inode);
18099 +
18100 +out:
18101 +       return err;
18102 +}
18103 +
18104 +static struct fsnotify_ops au_hfsn_ops = {
18105 +       .handle_event           = au_hfsn_handle_event,
18106 +       .free_group_priv        = au_hfsn_free_group,
18107 +       .free_mark              = au_hfsn_free_mark
18108 +};
18109 +
18110 +/* ---------------------------------------------------------------------- */
18111 +
18112 +static void au_hfsn_fin_br(struct au_branch *br)
18113 +{
18114 +       struct au_br_hfsnotify *hfsn;
18115 +
18116 +       hfsn = br->br_hfsn;
18117 +       if (hfsn) {
18118 +               lockdep_off();
18119 +               fsnotify_put_group(hfsn->hfsn_group);
18120 +               lockdep_on();
18121 +       }
18122 +}
18123 +
18124 +static int au_hfsn_init_br(struct au_branch *br, int perm)
18125 +{
18126 +       int err;
18127 +       struct fsnotify_group *group;
18128 +       struct au_br_hfsnotify *hfsn;
18129 +
18130 +       err = 0;
18131 +       br->br_hfsn = NULL;
18132 +       if (!au_br_hnotifyable(perm))
18133 +               goto out;
18134 +
18135 +       err = -ENOMEM;
18136 +       hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
18137 +       if (unlikely(!hfsn))
18138 +               goto out;
18139 +
18140 +       err = 0;
18141 +       group = fsnotify_alloc_group(&au_hfsn_ops,
18142 +                                    /*flags - not for userspace*/0);
18143 +       if (IS_ERR(group)) {
18144 +               err = PTR_ERR(group);
18145 +               pr_err("fsnotify_alloc_group() failed, %d\n", err);
18146 +               goto out_hfsn;
18147 +       }
18148 +
18149 +       group->private = hfsn;
18150 +       hfsn->hfsn_group = group;
18151 +       br->br_hfsn = hfsn;
18152 +       goto out; /* success */
18153 +
18154 +out_hfsn:
18155 +       au_kfree_try_rcu(hfsn);
18156 +out:
18157 +       return err;
18158 +}
18159 +
18160 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
18161 +{
18162 +       int err;
18163 +
18164 +       err = 0;
18165 +       if (!br->br_hfsn)
18166 +               err = au_hfsn_init_br(br, perm);
18167 +
18168 +       return err;
18169 +}
18170 +
18171 +/* ---------------------------------------------------------------------- */
18172 +
18173 +static void au_hfsn_fin(void)
18174 +{
18175 +       AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
18176 +       wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
18177 +}
18178 +
18179 +const struct au_hnotify_op au_hnotify_op = {
18180 +       .ctl            = au_hfsn_ctl,
18181 +       .alloc          = au_hfsn_alloc,
18182 +       .free           = au_hfsn_free,
18183 +
18184 +       .fin            = au_hfsn_fin,
18185 +
18186 +       .reset_br       = au_hfsn_reset_br,
18187 +       .fin_br         = au_hfsn_fin_br,
18188 +       .init_br        = au_hfsn_init_br
18189 +};
18190 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
18191 --- /usr/share/empty/fs/aufs/hfsplus.c  1970-01-01 01:00:00.000000000 +0100
18192 +++ linux/fs/aufs/hfsplus.c     2022-11-05 23:02:18.965889284 +0100
18193 @@ -0,0 +1,60 @@
18194 +// SPDX-License-Identifier: GPL-2.0
18195 +/*
18196 + * Copyright (C) 2010-2022 Junjiro R. Okajima
18197 + *
18198 + * This program is free software; you can redistribute it and/or modify
18199 + * it under the terms of the GNU General Public License as published by
18200 + * the Free Software Foundation; either version 2 of the License, or
18201 + * (at your option) any later version.
18202 + *
18203 + * This program is distributed in the hope that it will be useful,
18204 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18205 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18206 + * GNU General Public License for more details.
18207 + *
18208 + * You should have received a copy of the GNU General Public License
18209 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18210 + */
18211 +
18212 +/*
18213 + * special support for filesystems which acquires an inode mutex
18214 + * at final closing a file, eg, hfsplus.
18215 + *
18216 + * This trick is very simple and stupid, just to open the file before really
18217 + * necessary open to tell hfsplus that this is not the final closing.
18218 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
18219 + * and au_h_open_post() after releasing it.
18220 + */
18221 +
18222 +#include "aufs.h"
18223 +
18224 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
18225 +                          int force_wr)
18226 +{
18227 +       struct file *h_file;
18228 +       struct dentry *h_dentry;
18229 +
18230 +       h_dentry = au_h_dptr(dentry, bindex);
18231 +       AuDebugOn(!h_dentry);
18232 +       AuDebugOn(d_is_negative(h_dentry));
18233 +
18234 +       h_file = NULL;
18235 +       if (au_test_hfsplus(h_dentry->d_sb)
18236 +           && d_is_reg(h_dentry))
18237 +               h_file = au_h_open(dentry, bindex,
18238 +                                  O_RDONLY | O_NOATIME | O_LARGEFILE,
18239 +                                  /*file*/NULL, force_wr);
18240 +       return h_file;
18241 +}
18242 +
18243 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
18244 +                   struct file *h_file)
18245 +{
18246 +       struct au_branch *br;
18247 +
18248 +       if (h_file) {
18249 +               fput(h_file);
18250 +               br = au_sbr(dentry->d_sb, bindex);
18251 +               au_lcnt_dec(&br->br_nfiles);
18252 +       }
18253 +}
18254 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
18255 --- /usr/share/empty/fs/aufs/hnotify.c  1970-01-01 01:00:00.000000000 +0100
18256 +++ linux/fs/aufs/hnotify.c     2022-11-05 23:02:18.965889284 +0100
18257 @@ -0,0 +1,715 @@
18258 +// SPDX-License-Identifier: GPL-2.0
18259 +/*
18260 + * Copyright (C) 2005-2022 Junjiro R. Okajima
18261 + *
18262 + * This program is free software; you can redistribute it and/or modify
18263 + * it under the terms of the GNU General Public License as published by
18264 + * the Free Software Foundation; either version 2 of the License, or
18265 + * (at your option) any later version.
18266 + *
18267 + * This program is distributed in the hope that it will be useful,
18268 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18269 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18270 + * GNU General Public License for more details.
18271 + *
18272 + * You should have received a copy of the GNU General Public License
18273 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18274 + */
18275 +
18276 +/*
18277 + * abstraction to notify the direct changes on lower directories
18278 + */
18279 +
18280 +/* #include <linux/iversion.h> */
18281 +#include "aufs.h"
18282 +
18283 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
18284 +{
18285 +       int err;
18286 +       struct au_hnotify *hn;
18287 +
18288 +       err = -ENOMEM;
18289 +       hn = au_cache_alloc_hnotify();
18290 +       if (hn) {
18291 +               hn->hn_aufs_inode = inode;
18292 +               hinode->hi_notify = hn;
18293 +               err = au_hnotify_op.alloc(hinode);
18294 +               AuTraceErr(err);
18295 +               if (unlikely(err)) {
18296 +                       hinode->hi_notify = NULL;
18297 +                       au_cache_free_hnotify(hn);
18298 +                       /*
18299 +                        * The upper dir was removed by udba, but the same named
18300 +                        * dir left. In this case, aufs assigns a new inode
18301 +                        * number and set the monitor again.
18302 +                        * For the lower dir, the old monitor is still left.
18303 +                        */
18304 +                       if (err == -EEXIST)
18305 +                               err = 0;
18306 +               }
18307 +       }
18308 +
18309 +       AuTraceErr(err);
18310 +       return err;
18311 +}
18312 +
18313 +void au_hn_free(struct au_hinode *hinode)
18314 +{
18315 +       struct au_hnotify *hn;
18316 +
18317 +       hn = hinode->hi_notify;
18318 +       if (hn) {
18319 +               hinode->hi_notify = NULL;
18320 +               if (au_hnotify_op.free(hinode, hn))
18321 +                       au_cache_free_hnotify(hn);
18322 +       }
18323 +}
18324 +
18325 +/* ---------------------------------------------------------------------- */
18326 +
18327 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
18328 +{
18329 +       if (hinode->hi_notify)
18330 +               au_hnotify_op.ctl(hinode, do_set);
18331 +}
18332 +
18333 +void au_hn_reset(struct inode *inode, unsigned int flags)
18334 +{
18335 +       aufs_bindex_t bindex, bbot;
18336 +       struct inode *hi;
18337 +       struct dentry *iwhdentry;
18338 +
18339 +       bbot = au_ibbot(inode);
18340 +       for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
18341 +               hi = au_h_iptr(inode, bindex);
18342 +               if (!hi)
18343 +                       continue;
18344 +
18345 +               /* inode_lock_nested(hi, AuLsc_I_CHILD); */
18346 +               iwhdentry = au_hi_wh(inode, bindex);
18347 +               if (iwhdentry)
18348 +                       dget(iwhdentry);
18349 +               au_igrab(hi);
18350 +               au_set_h_iptr(inode, bindex, NULL, 0);
18351 +               au_set_h_iptr(inode, bindex, au_igrab(hi),
18352 +                             flags & ~AuHi_XINO);
18353 +               iput(hi);
18354 +               dput(iwhdentry);
18355 +               /* inode_unlock(hi); */
18356 +       }
18357 +}
18358 +
18359 +/* ---------------------------------------------------------------------- */
18360 +
18361 +static int hn_xino(struct inode *inode, struct inode *h_inode)
18362 +{
18363 +       int err;
18364 +       aufs_bindex_t bindex, bbot, bfound, btop;
18365 +       struct inode *h_i;
18366 +
18367 +       err = 0;
18368 +       if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18369 +               pr_warn("branch root dir was changed\n");
18370 +               goto out;
18371 +       }
18372 +
18373 +       bfound = -1;
18374 +       bbot = au_ibbot(inode);
18375 +       btop = au_ibtop(inode);
18376 +#if 0 /* reserved for future use */
18377 +       if (bindex == bbot) {
18378 +               /* keep this ino in rename case */
18379 +               goto out;
18380 +       }
18381 +#endif
18382 +       for (bindex = btop; bindex <= bbot; bindex++)
18383 +               if (au_h_iptr(inode, bindex) == h_inode) {
18384 +                       bfound = bindex;
18385 +                       break;
18386 +               }
18387 +       if (bfound < 0)
18388 +               goto out;
18389 +
18390 +       for (bindex = btop; bindex <= bbot; bindex++) {
18391 +               h_i = au_h_iptr(inode, bindex);
18392 +               if (!h_i)
18393 +                       continue;
18394 +
18395 +               err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
18396 +               /* ignore this error */
18397 +               /* bad action? */
18398 +       }
18399 +
18400 +       /* children inode number will be broken */
18401 +
18402 +out:
18403 +       AuTraceErr(err);
18404 +       return err;
18405 +}
18406 +
18407 +static int hn_gen_tree(struct dentry *dentry)
18408 +{
18409 +       int err, i, j, ndentry;
18410 +       struct au_dcsub_pages dpages;
18411 +       struct au_dpage *dpage;
18412 +       struct dentry **dentries;
18413 +
18414 +       err = au_dpages_init(&dpages, GFP_NOFS);
18415 +       if (unlikely(err))
18416 +               goto out;
18417 +       err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
18418 +       if (unlikely(err))
18419 +               goto out_dpages;
18420 +
18421 +       for (i = 0; i < dpages.ndpage; i++) {
18422 +               dpage = dpages.dpages + i;
18423 +               dentries = dpage->dentries;
18424 +               ndentry = dpage->ndentry;
18425 +               for (j = 0; j < ndentry; j++) {
18426 +                       struct dentry *d;
18427 +
18428 +                       d = dentries[j];
18429 +                       if (IS_ROOT(d))
18430 +                               continue;
18431 +
18432 +                       au_digen_dec(d);
18433 +                       if (d_really_is_positive(d))
18434 +                               /* todo: reset children xino?
18435 +                                  cached children only? */
18436 +                               au_iigen_dec(d_inode(d));
18437 +               }
18438 +       }
18439 +
18440 +out_dpages:
18441 +       au_dpages_free(&dpages);
18442 +out:
18443 +       return err;
18444 +}
18445 +
18446 +/*
18447 + * return 0 if processed.
18448 + */
18449 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
18450 +                          const unsigned int isdir)
18451 +{
18452 +       int err;
18453 +       struct dentry *d;
18454 +       struct qstr *dname;
18455 +
18456 +       err = 1;
18457 +       if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18458 +               pr_warn("branch root dir was changed\n");
18459 +               err = 0;
18460 +               goto out;
18461 +       }
18462 +
18463 +       if (!isdir) {
18464 +               AuDebugOn(!name);
18465 +               au_iigen_dec(inode);
18466 +               spin_lock(&inode->i_lock);
18467 +               hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
18468 +                       spin_lock(&d->d_lock);
18469 +                       dname = &d->d_name;
18470 +                       if (dname->len != nlen
18471 +                           && memcmp(dname->name, name, nlen)) {
18472 +                               spin_unlock(&d->d_lock);
18473 +                               continue;
18474 +                       }
18475 +                       err = 0;
18476 +                       au_digen_dec(d);
18477 +                       spin_unlock(&d->d_lock);
18478 +                       break;
18479 +               }
18480 +               spin_unlock(&inode->i_lock);
18481 +       } else {
18482 +               au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
18483 +               d = d_find_any_alias(inode);
18484 +               if (!d) {
18485 +                       au_iigen_dec(inode);
18486 +                       goto out;
18487 +               }
18488 +
18489 +               spin_lock(&d->d_lock);
18490 +               dname = &d->d_name;
18491 +               if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
18492 +                       spin_unlock(&d->d_lock);
18493 +                       err = hn_gen_tree(d);
18494 +                       spin_lock(&d->d_lock);
18495 +               }
18496 +               spin_unlock(&d->d_lock);
18497 +               dput(d);
18498 +       }
18499 +
18500 +out:
18501 +       AuTraceErr(err);
18502 +       return err;
18503 +}
18504 +
18505 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
18506 +{
18507 +       int err;
18508 +
18509 +       if (IS_ROOT(dentry)) {
18510 +               pr_warn("branch root dir was changed\n");
18511 +               return 0;
18512 +       }
18513 +
18514 +       err = 0;
18515 +       if (!isdir) {
18516 +               au_digen_dec(dentry);
18517 +               if (d_really_is_positive(dentry))
18518 +                       au_iigen_dec(d_inode(dentry));
18519 +       } else {
18520 +               au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
18521 +               if (d_really_is_positive(dentry))
18522 +                       err = hn_gen_tree(dentry);
18523 +       }
18524 +
18525 +       AuTraceErr(err);
18526 +       return err;
18527 +}
18528 +
18529 +/* ---------------------------------------------------------------------- */
18530 +
18531 +/* hnotify job flags */
18532 +#define AuHnJob_XINO0          1
18533 +#define AuHnJob_GEN            (1 << 1)
18534 +#define AuHnJob_DIRENT         (1 << 2)
18535 +#define AuHnJob_ISDIR          (1 << 3)
18536 +#define AuHnJob_TRYXINO0       (1 << 4)
18537 +#define AuHnJob_MNTPNT         (1 << 5)
18538 +#define au_ftest_hnjob(flags, name)    ((flags) & AuHnJob_##name)
18539 +#define au_fset_hnjob(flags, name) \
18540 +       do { (flags) |= AuHnJob_##name; } while (0)
18541 +#define au_fclr_hnjob(flags, name) \
18542 +       do { (flags) &= ~AuHnJob_##name; } while (0)
18543 +
18544 +enum {
18545 +       AuHn_CHILD,
18546 +       AuHn_PARENT,
18547 +       AuHnLast
18548 +};
18549 +
18550 +struct au_hnotify_args {
18551 +       struct inode *h_dir, *dir, *h_child_inode;
18552 +       u32 mask;
18553 +       unsigned int flags[AuHnLast];
18554 +       unsigned int h_child_nlen;
18555 +       char h_child_name[];
18556 +};
18557 +
18558 +struct hn_job_args {
18559 +       unsigned int flags;
18560 +       struct inode *inode, *h_inode, *dir, *h_dir;
18561 +       struct dentry *dentry;
18562 +       char *h_name;
18563 +       int h_nlen;
18564 +};
18565 +
18566 +static int hn_job(struct hn_job_args *a)
18567 +{
18568 +       const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
18569 +       int e;
18570 +
18571 +       /* reset xino */
18572 +       if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
18573 +               hn_xino(a->inode, a->h_inode); /* ignore this error */
18574 +
18575 +       if (au_ftest_hnjob(a->flags, TRYXINO0)
18576 +           && a->inode
18577 +           && a->h_inode) {
18578 +               inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
18579 +               if (!a->h_inode->i_nlink
18580 +                   && !(a->h_inode->i_state & I_LINKABLE))
18581 +                       hn_xino(a->inode, a->h_inode); /* ignore this error */
18582 +               inode_unlock_shared(a->h_inode);
18583 +       }
18584 +
18585 +       /* make the generation obsolete */
18586 +       if (au_ftest_hnjob(a->flags, GEN)) {
18587 +               e = -1;
18588 +               if (a->inode)
18589 +                       e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
18590 +                                             isdir);
18591 +               if (e && a->dentry)
18592 +                       hn_gen_by_name(a->dentry, isdir);
18593 +               /* ignore this error */
18594 +       }
18595 +
18596 +       /* make dir entries obsolete */
18597 +       if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
18598 +               struct au_vdir *vdir;
18599 +
18600 +               vdir = au_ivdir(a->inode);
18601 +               if (vdir)
18602 +                       vdir->vd_jiffy = 0;
18603 +               /* IMustLock(a->inode); */
18604 +               /* inode_inc_iversion(a->inode); */
18605 +       }
18606 +
18607 +       /* can do nothing but warn */
18608 +       if (au_ftest_hnjob(a->flags, MNTPNT)
18609 +           && a->dentry
18610 +           && d_mountpoint(a->dentry))
18611 +               pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
18612 +
18613 +       return 0;
18614 +}
18615 +
18616 +/* ---------------------------------------------------------------------- */
18617 +
18618 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
18619 +                                          struct inode *dir)
18620 +{
18621 +       struct dentry *dentry, *d, *parent;
18622 +       struct qstr *dname;
18623 +
18624 +       parent = d_find_any_alias(dir);
18625 +       if (!parent)
18626 +               return NULL;
18627 +
18628 +       dentry = NULL;
18629 +       spin_lock(&parent->d_lock);
18630 +       list_for_each_entry(d, &parent->d_subdirs, d_child) {
18631 +               /* AuDbg("%pd\n", d); */
18632 +               spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
18633 +               dname = &d->d_name;
18634 +               if (dname->len != nlen || memcmp(dname->name, name, nlen))
18635 +                       goto cont_unlock;
18636 +               if (au_di(d))
18637 +                       au_digen_dec(d);
18638 +               else
18639 +                       goto cont_unlock;
18640 +               if (au_dcount(d) > 0) {
18641 +                       dentry = dget_dlock(d);
18642 +                       spin_unlock(&d->d_lock);
18643 +                       break;
18644 +               }
18645 +
18646 +cont_unlock:
18647 +               spin_unlock(&d->d_lock);
18648 +       }
18649 +       spin_unlock(&parent->d_lock);
18650 +       dput(parent);
18651 +
18652 +       if (dentry)
18653 +               di_write_lock_child(dentry);
18654 +
18655 +       return dentry;
18656 +}
18657 +
18658 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
18659 +                                        aufs_bindex_t bindex, ino_t h_ino)
18660 +{
18661 +       struct inode *inode;
18662 +       ino_t ino;
18663 +       int err;
18664 +
18665 +       inode = NULL;
18666 +       err = au_xino_read(sb, bindex, h_ino, &ino);
18667 +       if (!err && ino)
18668 +               inode = ilookup(sb, ino);
18669 +       if (!inode)
18670 +               goto out;
18671 +
18672 +       if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18673 +               pr_warn("wrong root branch\n");
18674 +               iput(inode);
18675 +               inode = NULL;
18676 +               goto out;
18677 +       }
18678 +
18679 +       ii_write_lock_child(inode);
18680 +
18681 +out:
18682 +       return inode;
18683 +}
18684 +
18685 +static void au_hn_bh(void *_args)
18686 +{
18687 +       struct au_hnotify_args *a = _args;
18688 +       struct super_block *sb;
18689 +       aufs_bindex_t bindex, bbot, bfound;
18690 +       unsigned char xino, try_iput;
18691 +       int err;
18692 +       struct inode *inode;
18693 +       ino_t h_ino;
18694 +       struct hn_job_args args;
18695 +       struct dentry *dentry;
18696 +       struct au_sbinfo *sbinfo;
18697 +
18698 +       AuDebugOn(!_args);
18699 +       AuDebugOn(!a->h_dir);
18700 +       AuDebugOn(!a->dir);
18701 +       AuDebugOn(!a->mask);
18702 +       AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
18703 +             a->mask, a->dir->i_ino, a->h_dir->i_ino,
18704 +             a->h_child_inode ? a->h_child_inode->i_ino : 0);
18705 +
18706 +       inode = NULL;
18707 +       dentry = NULL;
18708 +       /*
18709 +        * do not lock a->dir->i_mutex here
18710 +        * because of d_revalidate() may cause a deadlock.
18711 +        */
18712 +       sb = a->dir->i_sb;
18713 +       AuDebugOn(!sb);
18714 +       sbinfo = au_sbi(sb);
18715 +       AuDebugOn(!sbinfo);
18716 +       si_write_lock(sb, AuLock_NOPLMW);
18717 +
18718 +       if (au_opt_test(sbinfo->si_mntflags, DIRREN))
18719 +               switch (a->mask & FS_EVENTS_POSS_ON_CHILD) {
18720 +               case FS_MOVED_FROM:
18721 +               case FS_MOVED_TO:
18722 +                       AuWarn1("DIRREN with UDBA may not work correctly "
18723 +                               "for the direct rename(2)\n");
18724 +               }
18725 +
18726 +       ii_read_lock_parent(a->dir);
18727 +       bfound = -1;
18728 +       bbot = au_ibbot(a->dir);
18729 +       for (bindex = au_ibtop(a->dir); bindex <= bbot; bindex++)
18730 +               if (au_h_iptr(a->dir, bindex) == a->h_dir) {
18731 +                       bfound = bindex;
18732 +                       break;
18733 +               }
18734 +       ii_read_unlock(a->dir);
18735 +       if (unlikely(bfound < 0))
18736 +               goto out;
18737 +
18738 +       xino = !!au_opt_test(au_mntflags(sb), XINO);
18739 +       h_ino = 0;
18740 +       if (a->h_child_inode)
18741 +               h_ino = a->h_child_inode->i_ino;
18742 +
18743 +       if (a->h_child_nlen
18744 +           && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
18745 +               || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
18746 +               dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
18747 +                                             a->dir);
18748 +       try_iput = 0;
18749 +       if (dentry && d_really_is_positive(dentry))
18750 +               inode = d_inode(dentry);
18751 +       if (xino && !inode && h_ino
18752 +           && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
18753 +               || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
18754 +               || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
18755 +               inode = lookup_wlock_by_ino(sb, bfound, h_ino);
18756 +               try_iput = 1;
18757 +       }
18758 +
18759 +       args.flags = a->flags[AuHn_CHILD];
18760 +       args.dentry = dentry;
18761 +       args.inode = inode;
18762 +       args.h_inode = a->h_child_inode;
18763 +       args.dir = a->dir;
18764 +       args.h_dir = a->h_dir;
18765 +       args.h_name = a->h_child_name;
18766 +       args.h_nlen = a->h_child_nlen;
18767 +       err = hn_job(&args);
18768 +       if (dentry) {
18769 +               if (au_di(dentry))
18770 +                       di_write_unlock(dentry);
18771 +               dput(dentry);
18772 +       }
18773 +       if (inode && try_iput) {
18774 +               ii_write_unlock(inode);
18775 +               iput(inode);
18776 +       }
18777 +
18778 +       ii_write_lock_parent(a->dir);
18779 +       args.flags = a->flags[AuHn_PARENT];
18780 +       args.dentry = NULL;
18781 +       args.inode = a->dir;
18782 +       args.h_inode = a->h_dir;
18783 +       args.dir = NULL;
18784 +       args.h_dir = NULL;
18785 +       args.h_name = NULL;
18786 +       args.h_nlen = 0;
18787 +       err = hn_job(&args);
18788 +       ii_write_unlock(a->dir);
18789 +
18790 +out:
18791 +       iput(a->h_child_inode);
18792 +       iput(a->h_dir);
18793 +       iput(a->dir);
18794 +       si_write_unlock(sb);
18795 +       au_nwt_done(&sbinfo->si_nowait);
18796 +       au_kfree_rcu(a);
18797 +}
18798 +
18799 +/* ---------------------------------------------------------------------- */
18800 +
18801 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
18802 +              const struct qstr *h_child_qstr, struct inode *h_child_inode)
18803 +{
18804 +       int err, len;
18805 +       unsigned int flags[AuHnLast], f;
18806 +       unsigned char isdir, isroot, wh;
18807 +       struct inode *dir;
18808 +       struct au_hnotify_args *args;
18809 +       char *p, *h_child_name;
18810 +
18811 +       err = 0;
18812 +       AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
18813 +       dir = igrab(hnotify->hn_aufs_inode);
18814 +       if (!dir)
18815 +               goto out;
18816 +
18817 +       isroot = (dir->i_ino == AUFS_ROOT_INO);
18818 +       wh = 0;
18819 +       h_child_name = (void *)h_child_qstr->name;
18820 +       len = h_child_qstr->len;
18821 +       if (h_child_name) {
18822 +               if (len > AUFS_WH_PFX_LEN
18823 +                   && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
18824 +                       h_child_name += AUFS_WH_PFX_LEN;
18825 +                       len -= AUFS_WH_PFX_LEN;
18826 +                       wh = 1;
18827 +               }
18828 +       }
18829 +
18830 +       isdir = 0;
18831 +       if (h_child_inode)
18832 +               isdir = !!S_ISDIR(h_child_inode->i_mode);
18833 +       flags[AuHn_PARENT] = AuHnJob_ISDIR;
18834 +       flags[AuHn_CHILD] = 0;
18835 +       if (isdir)
18836 +               flags[AuHn_CHILD] = AuHnJob_ISDIR;
18837 +       au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
18838 +       au_fset_hnjob(flags[AuHn_CHILD], GEN);
18839 +       switch (mask & ALL_FSNOTIFY_DIRENT_EVENTS) {
18840 +       case FS_MOVED_FROM:
18841 +       case FS_MOVED_TO:
18842 +               au_fset_hnjob(flags[AuHn_CHILD], XINO0);
18843 +               au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18844 +               fallthrough;
18845 +       case FS_CREATE:
18846 +               AuDebugOn(!h_child_name);
18847 +               break;
18848 +
18849 +       case FS_DELETE:
18850 +               /*
18851 +                * aufs never be able to get this child inode.
18852 +                * revalidation should be in d_revalidate()
18853 +                * by checking i_nlink, i_generation or d_unhashed().
18854 +                */
18855 +               AuDebugOn(!h_child_name);
18856 +               au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
18857 +               au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18858 +               break;
18859 +
18860 +       default:
18861 +               AuDebugOn(1);
18862 +       }
18863 +
18864 +       if (wh)
18865 +               h_child_inode = NULL;
18866 +
18867 +       err = -ENOMEM;
18868 +       /* iput() and kfree() will be called in au_hnotify() */
18869 +       args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
18870 +       if (unlikely(!args)) {
18871 +               AuErr1("no memory\n");
18872 +               iput(dir);
18873 +               goto out;
18874 +       }
18875 +       args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
18876 +       args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
18877 +       args->mask = mask;
18878 +       args->dir = dir;
18879 +       args->h_dir = igrab(h_dir);
18880 +       if (h_child_inode)
18881 +               h_child_inode = igrab(h_child_inode); /* can be NULL */
18882 +       args->h_child_inode = h_child_inode;
18883 +       args->h_child_nlen = len;
18884 +       if (len) {
18885 +               p = (void *)args;
18886 +               p += sizeof(*args);
18887 +               memcpy(p, h_child_name, len);
18888 +               p[len] = 0;
18889 +       }
18890 +
18891 +       /* NFS fires the event for silly-renamed one from kworker */
18892 +       f = 0;
18893 +       if (!dir->i_nlink
18894 +           || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
18895 +               f = AuWkq_NEST;
18896 +       err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
18897 +       if (unlikely(err)) {
18898 +               pr_err("wkq %d\n", err);
18899 +               iput(args->h_child_inode);
18900 +               iput(args->h_dir);
18901 +               iput(args->dir);
18902 +               au_kfree_rcu(args);
18903 +       }
18904 +
18905 +out:
18906 +       return err;
18907 +}
18908 +
18909 +/* ---------------------------------------------------------------------- */
18910 +
18911 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
18912 +{
18913 +       int err;
18914 +
18915 +       AuDebugOn(!(udba & AuOptMask_UDBA));
18916 +
18917 +       err = 0;
18918 +       if (au_hnotify_op.reset_br)
18919 +               err = au_hnotify_op.reset_br(udba, br, perm);
18920 +
18921 +       return err;
18922 +}
18923 +
18924 +int au_hnotify_init_br(struct au_branch *br, int perm)
18925 +{
18926 +       int err;
18927 +
18928 +       err = 0;
18929 +       if (au_hnotify_op.init_br)
18930 +               err = au_hnotify_op.init_br(br, perm);
18931 +
18932 +       return err;
18933 +}
18934 +
18935 +void au_hnotify_fin_br(struct au_branch *br)
18936 +{
18937 +       if (au_hnotify_op.fin_br)
18938 +               au_hnotify_op.fin_br(br);
18939 +}
18940 +
18941 +static void au_hn_destroy_cache(void)
18942 +{
18943 +       kmem_cache_destroy(au_cache[AuCache_HNOTIFY]);
18944 +       au_cache[AuCache_HNOTIFY] = NULL;
18945 +}
18946 +
18947 +int __init au_hnotify_init(void)
18948 +{
18949 +       int err;
18950 +
18951 +       err = -ENOMEM;
18952 +       au_cache[AuCache_HNOTIFY] = AuCache(au_hnotify);
18953 +       if (au_cache[AuCache_HNOTIFY]) {
18954 +               err = 0;
18955 +               if (au_hnotify_op.init)
18956 +                       err = au_hnotify_op.init();
18957 +               if (unlikely(err))
18958 +                       au_hn_destroy_cache();
18959 +       }
18960 +       AuTraceErr(err);
18961 +       return err;
18962 +}
18963 +
18964 +void au_hnotify_fin(void)
18965 +{
18966 +       if (au_hnotify_op.fin)
18967 +               au_hnotify_op.fin();
18968 +
18969 +       /* cf. au_cache_fin() */
18970 +       if (au_cache[AuCache_HNOTIFY])
18971 +               au_hn_destroy_cache();
18972 +}
18973 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
18974 --- /usr/share/empty/fs/aufs/iinfo.c    1970-01-01 01:00:00.000000000 +0100
18975 +++ linux/fs/aufs/iinfo.c       2022-11-05 23:02:18.965889284 +0100
18976 @@ -0,0 +1,286 @@
18977 +// SPDX-License-Identifier: GPL-2.0
18978 +/*
18979 + * Copyright (C) 2005-2022 Junjiro R. Okajima
18980 + *
18981 + * This program is free software; you can redistribute it and/or modify
18982 + * it under the terms of the GNU General Public License as published by
18983 + * the Free Software Foundation; either version 2 of the License, or
18984 + * (at your option) any later version.
18985 + *
18986 + * This program is distributed in the hope that it will be useful,
18987 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18988 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18989 + * GNU General Public License for more details.
18990 + *
18991 + * You should have received a copy of the GNU General Public License
18992 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18993 + */
18994 +
18995 +/*
18996 + * inode private data
18997 + */
18998 +
18999 +#include "aufs.h"
19000 +
19001 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
19002 +{
19003 +       struct inode *h_inode;
19004 +       struct au_hinode *hinode;
19005 +
19006 +       IiMustAnyLock(inode);
19007 +
19008 +       hinode = au_hinode(au_ii(inode), bindex);
19009 +       h_inode = hinode->hi_inode;
19010 +       AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19011 +       return h_inode;
19012 +}
19013 +
19014 +/* todo: hard/soft set? */
19015 +void au_hiput(struct au_hinode *hinode)
19016 +{
19017 +       au_hn_free(hinode);
19018 +       dput(hinode->hi_whdentry);
19019 +       iput(hinode->hi_inode);
19020 +}
19021 +
19022 +unsigned int au_hi_flags(struct inode *inode, int isdir)
19023 +{
19024 +       unsigned int flags;
19025 +       const unsigned int mnt_flags = au_mntflags(inode->i_sb);
19026 +
19027 +       flags = 0;
19028 +       if (au_opt_test(mnt_flags, XINO))
19029 +               au_fset_hi(flags, XINO);
19030 +       if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
19031 +               au_fset_hi(flags, HNOTIFY);
19032 +       return flags;
19033 +}
19034 +
19035 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
19036 +                  struct inode *h_inode, unsigned int flags)
19037 +{
19038 +       struct au_hinode *hinode;
19039 +       struct inode *hi;
19040 +       struct au_iinfo *iinfo = au_ii(inode);
19041 +
19042 +       IiMustWriteLock(inode);
19043 +
19044 +       hinode = au_hinode(iinfo, bindex);
19045 +       hi = hinode->hi_inode;
19046 +       AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19047 +
19048 +       if (hi)
19049 +               au_hiput(hinode);
19050 +       hinode->hi_inode = h_inode;
19051 +       if (h_inode) {
19052 +               int err;
19053 +               struct super_block *sb = inode->i_sb;
19054 +               struct au_branch *br;
19055 +
19056 +               AuDebugOn(inode->i_mode
19057 +                         && (h_inode->i_mode & S_IFMT)
19058 +                         != (inode->i_mode & S_IFMT));
19059 +               if (bindex == iinfo->ii_btop)
19060 +                       au_cpup_igen(inode, h_inode);
19061 +               br = au_sbr(sb, bindex);
19062 +               hinode->hi_id = br->br_id;
19063 +               if (au_ftest_hi(flags, XINO)) {
19064 +                       err = au_xino_write(sb, bindex, h_inode->i_ino,
19065 +                                           inode->i_ino);
19066 +                       if (unlikely(err))
19067 +                               AuIOErr1("failed au_xino_write() %d\n", err);
19068 +               }
19069 +
19070 +               if (au_ftest_hi(flags, HNOTIFY)
19071 +                   && au_br_hnotifyable(br->br_perm)) {
19072 +                       err = au_hn_alloc(hinode, inode);
19073 +                       if (unlikely(err))
19074 +                               AuIOErr1("au_hn_alloc() %d\n", err);
19075 +               }
19076 +       }
19077 +}
19078 +
19079 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
19080 +                 struct dentry *h_wh)
19081 +{
19082 +       struct au_hinode *hinode;
19083 +
19084 +       IiMustWriteLock(inode);
19085 +
19086 +       hinode = au_hinode(au_ii(inode), bindex);
19087 +       AuDebugOn(hinode->hi_whdentry);
19088 +       hinode->hi_whdentry = h_wh;
19089 +}
19090 +
19091 +void au_update_iigen(struct inode *inode, int half)
19092 +{
19093 +       struct au_iinfo *iinfo;
19094 +       struct au_iigen *iigen;
19095 +       unsigned int sigen;
19096 +
19097 +       sigen = au_sigen(inode->i_sb);
19098 +       iinfo = au_ii(inode);
19099 +       iigen = &iinfo->ii_generation;
19100 +       spin_lock(&iigen->ig_spin);
19101 +       iigen->ig_generation = sigen;
19102 +       if (half)
19103 +               au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
19104 +       else
19105 +               au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
19106 +       spin_unlock(&iigen->ig_spin);
19107 +}
19108 +
19109 +/* it may be called at remount time, too */
19110 +void au_update_ibrange(struct inode *inode, int do_put_zero)
19111 +{
19112 +       struct au_iinfo *iinfo;
19113 +       aufs_bindex_t bindex, bbot;
19114 +
19115 +       AuDebugOn(au_is_bad_inode(inode));
19116 +       IiMustWriteLock(inode);
19117 +
19118 +       iinfo = au_ii(inode);
19119 +       if (do_put_zero && iinfo->ii_btop >= 0) {
19120 +               for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19121 +                    bindex++) {
19122 +                       struct inode *h_i;
19123 +
19124 +                       h_i = au_hinode(iinfo, bindex)->hi_inode;
19125 +                       if (h_i
19126 +                           && !h_i->i_nlink
19127 +                           && !(h_i->i_state & I_LINKABLE))
19128 +                               au_set_h_iptr(inode, bindex, NULL, 0);
19129 +               }
19130 +       }
19131 +
19132 +       iinfo->ii_btop = -1;
19133 +       iinfo->ii_bbot = -1;
19134 +       bbot = au_sbbot(inode->i_sb);
19135 +       for (bindex = 0; bindex <= bbot; bindex++)
19136 +               if (au_hinode(iinfo, bindex)->hi_inode) {
19137 +                       iinfo->ii_btop = bindex;
19138 +                       break;
19139 +               }
19140 +       if (iinfo->ii_btop >= 0)
19141 +               for (bindex = bbot; bindex >= iinfo->ii_btop; bindex--)
19142 +                       if (au_hinode(iinfo, bindex)->hi_inode) {
19143 +                               iinfo->ii_bbot = bindex;
19144 +                               break;
19145 +                       }
19146 +       AuDebugOn(iinfo->ii_btop > iinfo->ii_bbot);
19147 +}
19148 +
19149 +/* ---------------------------------------------------------------------- */
19150 +
19151 +void au_icntnr_init_once(void *_c)
19152 +{
19153 +       struct au_icntnr *c = _c;
19154 +       struct au_iinfo *iinfo = &c->iinfo;
19155 +
19156 +       spin_lock_init(&iinfo->ii_generation.ig_spin);
19157 +       au_rw_init(&iinfo->ii_rwsem);
19158 +       inode_init_once(&c->vfs_inode);
19159 +}
19160 +
19161 +void au_hinode_init(struct au_hinode *hinode)
19162 +{
19163 +       hinode->hi_inode = NULL;
19164 +       hinode->hi_id = -1;
19165 +       au_hn_init(hinode);
19166 +       hinode->hi_whdentry = NULL;
19167 +}
19168 +
19169 +int au_iinfo_init(struct inode *inode)
19170 +{
19171 +       struct au_iinfo *iinfo;
19172 +       struct super_block *sb;
19173 +       struct au_hinode *hi;
19174 +       int nbr, i;
19175 +
19176 +       sb = inode->i_sb;
19177 +       iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19178 +       nbr = au_sbbot(sb) + 1;
19179 +       if (unlikely(nbr <= 0))
19180 +               nbr = 1;
19181 +       hi = kmalloc_array(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
19182 +       if (hi) {
19183 +               au_lcnt_inc(&au_sbi(sb)->si_ninodes);
19184 +
19185 +               iinfo->ii_hinode = hi;
19186 +               for (i = 0; i < nbr; i++, hi++)
19187 +                       au_hinode_init(hi);
19188 +
19189 +               iinfo->ii_generation.ig_generation = au_sigen(sb);
19190 +               iinfo->ii_btop = -1;
19191 +               iinfo->ii_bbot = -1;
19192 +               iinfo->ii_vdir = NULL;
19193 +               return 0;
19194 +       }
19195 +       return -ENOMEM;
19196 +}
19197 +
19198 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink)
19199 +{
19200 +       int err, i;
19201 +       struct au_hinode *hip;
19202 +
19203 +       AuRwMustWriteLock(&iinfo->ii_rwsem);
19204 +
19205 +       err = -ENOMEM;
19206 +       hip = au_krealloc(iinfo->ii_hinode, sizeof(*hip) * nbr, GFP_NOFS,
19207 +                         may_shrink);
19208 +       if (hip) {
19209 +               iinfo->ii_hinode = hip;
19210 +               i = iinfo->ii_bbot + 1;
19211 +               hip += i;
19212 +               for (; i < nbr; i++, hip++)
19213 +                       au_hinode_init(hip);
19214 +               err = 0;
19215 +       }
19216 +
19217 +       return err;
19218 +}
19219 +
19220 +void au_iinfo_fin(struct inode *inode)
19221 +{
19222 +       struct au_iinfo *iinfo;
19223 +       struct au_hinode *hi;
19224 +       struct super_block *sb;
19225 +       aufs_bindex_t bindex, bbot;
19226 +       const unsigned char unlinked = !inode->i_nlink;
19227 +
19228 +       AuDebugOn(au_is_bad_inode(inode));
19229 +
19230 +       sb = inode->i_sb;
19231 +       au_lcnt_dec(&au_sbi(sb)->si_ninodes);
19232 +       if (si_pid_test(sb))
19233 +               au_xino_delete_inode(inode, unlinked);
19234 +       else {
19235 +               /*
19236 +                * it is safe to hide the dependency between sbinfo and
19237 +                * sb->s_umount.
19238 +                */
19239 +               lockdep_off();
19240 +               si_noflush_read_lock(sb);
19241 +               au_xino_delete_inode(inode, unlinked);
19242 +               si_read_unlock(sb);
19243 +               lockdep_on();
19244 +       }
19245 +
19246 +       iinfo = au_ii(inode);
19247 +       if (iinfo->ii_vdir)
19248 +               au_vdir_free(iinfo->ii_vdir);
19249 +
19250 +       bindex = iinfo->ii_btop;
19251 +       if (bindex >= 0) {
19252 +               hi = au_hinode(iinfo, bindex);
19253 +               bbot = iinfo->ii_bbot;
19254 +               while (bindex++ <= bbot) {
19255 +                       if (hi->hi_inode)
19256 +                               au_hiput(hi);
19257 +                       hi++;
19258 +               }
19259 +       }
19260 +       au_kfree_rcu(iinfo->ii_hinode);
19261 +       AuRwDestroy(&iinfo->ii_rwsem);
19262 +}
19263 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
19264 --- /usr/share/empty/fs/aufs/inode.c    1970-01-01 01:00:00.000000000 +0100
19265 +++ linux/fs/aufs/inode.c       2022-11-05 23:02:18.965889284 +0100
19266 @@ -0,0 +1,531 @@
19267 +// SPDX-License-Identifier: GPL-2.0
19268 +/*
19269 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19270 + *
19271 + * This program is free software; you can redistribute it and/or modify
19272 + * it under the terms of the GNU General Public License as published by
19273 + * the Free Software Foundation; either version 2 of the License, or
19274 + * (at your option) any later version.
19275 + *
19276 + * This program is distributed in the hope that it will be useful,
19277 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19278 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19279 + * GNU General Public License for more details.
19280 + *
19281 + * You should have received a copy of the GNU General Public License
19282 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
19283 + */
19284 +
19285 +/*
19286 + * inode functions
19287 + */
19288 +
19289 +#include <linux/iversion.h>
19290 +#include "aufs.h"
19291 +
19292 +struct inode *au_igrab(struct inode *inode)
19293 +{
19294 +       if (inode) {
19295 +               AuDebugOn(!atomic_read(&inode->i_count));
19296 +               ihold(inode);
19297 +       }
19298 +       return inode;
19299 +}
19300 +
19301 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
19302 +{
19303 +       au_cpup_attr_all(inode, /*force*/0);
19304 +       au_update_iigen(inode, /*half*/1);
19305 +       if (do_version)
19306 +               inode_inc_iversion(inode);
19307 +}
19308 +
19309 +static int au_ii_refresh(struct inode *inode, int *update)
19310 +{
19311 +       int err, e, nbr;
19312 +       umode_t type;
19313 +       aufs_bindex_t bindex, new_bindex;
19314 +       struct super_block *sb;
19315 +       struct au_iinfo *iinfo;
19316 +       struct au_hinode *p, *q, tmp;
19317 +
19318 +       AuDebugOn(au_is_bad_inode(inode));
19319 +       IiMustWriteLock(inode);
19320 +
19321 +       *update = 0;
19322 +       sb = inode->i_sb;
19323 +       nbr = au_sbbot(sb) + 1;
19324 +       type = inode->i_mode & S_IFMT;
19325 +       iinfo = au_ii(inode);
19326 +       err = au_hinode_realloc(iinfo, nbr, /*may_shrink*/0);
19327 +       if (unlikely(err))
19328 +               goto out;
19329 +
19330 +       AuDebugOn(iinfo->ii_btop < 0);
19331 +       p = au_hinode(iinfo, iinfo->ii_btop);
19332 +       for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19333 +            bindex++, p++) {
19334 +               if (!p->hi_inode)
19335 +                       continue;
19336 +
19337 +               AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
19338 +               new_bindex = au_br_index(sb, p->hi_id);
19339 +               if (new_bindex == bindex)
19340 +                       continue;
19341 +
19342 +               if (new_bindex < 0) {
19343 +                       *update = 1;
19344 +                       au_hiput(p);
19345 +                       p->hi_inode = NULL;
19346 +                       continue;
19347 +               }
19348 +
19349 +               if (new_bindex < iinfo->ii_btop)
19350 +                       iinfo->ii_btop = new_bindex;
19351 +               if (iinfo->ii_bbot < new_bindex)
19352 +                       iinfo->ii_bbot = new_bindex;
19353 +               /* swap two lower inode, and loop again */
19354 +               q = au_hinode(iinfo, new_bindex);
19355 +               tmp = *q;
19356 +               *q = *p;
19357 +               *p = tmp;
19358 +               if (tmp.hi_inode) {
19359 +                       bindex--;
19360 +                       p--;
19361 +               }
19362 +       }
19363 +       au_update_ibrange(inode, /*do_put_zero*/0);
19364 +       au_hinode_realloc(iinfo, nbr, /*may_shrink*/1); /* harmless if err */
19365 +       e = au_dy_irefresh(inode);
19366 +       if (unlikely(e && !err))
19367 +               err = e;
19368 +
19369 +out:
19370 +       AuTraceErr(err);
19371 +       return err;
19372 +}
19373 +
19374 +void au_refresh_iop(struct inode *inode, int force_getattr)
19375 +{
19376 +       int type;
19377 +       struct au_sbinfo *sbi = au_sbi(inode->i_sb);
19378 +       const struct inode_operations *iop
19379 +               = force_getattr ? aufs_iop : sbi->si_iop_array;
19380 +
19381 +       if (inode->i_op == iop)
19382 +               return;
19383 +
19384 +       switch (inode->i_mode & S_IFMT) {
19385 +       case S_IFDIR:
19386 +               type = AuIop_DIR;
19387 +               break;
19388 +       case S_IFLNK:
19389 +               type = AuIop_SYMLINK;
19390 +               break;
19391 +       default:
19392 +               type = AuIop_OTHER;
19393 +               break;
19394 +       }
19395 +
19396 +       inode->i_op = iop + type;
19397 +       /* unnecessary smp_wmb() */
19398 +}
19399 +
19400 +int au_refresh_hinode_self(struct inode *inode)
19401 +{
19402 +       int err, update;
19403 +
19404 +       err = au_ii_refresh(inode, &update);
19405 +       if (!err)
19406 +               au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
19407 +
19408 +       AuTraceErr(err);
19409 +       return err;
19410 +}
19411 +
19412 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
19413 +{
19414 +       int err, e, update;
19415 +       unsigned int flags;
19416 +       umode_t mode;
19417 +       aufs_bindex_t bindex, bbot;
19418 +       unsigned char isdir;
19419 +       struct au_hinode *p;
19420 +       struct au_iinfo *iinfo;
19421 +
19422 +       err = au_ii_refresh(inode, &update);
19423 +       if (unlikely(err))
19424 +               goto out;
19425 +
19426 +       update = 0;
19427 +       iinfo = au_ii(inode);
19428 +       p = au_hinode(iinfo, iinfo->ii_btop);
19429 +       mode = (inode->i_mode & S_IFMT);
19430 +       isdir = S_ISDIR(mode);
19431 +       flags = au_hi_flags(inode, isdir);
19432 +       bbot = au_dbbot(dentry);
19433 +       for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
19434 +               struct inode *h_i, *h_inode;
19435 +               struct dentry *h_d;
19436 +
19437 +               h_d = au_h_dptr(dentry, bindex);
19438 +               if (!h_d || d_is_negative(h_d))
19439 +                       continue;
19440 +
19441 +               h_inode = d_inode(h_d);
19442 +               AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
19443 +               if (iinfo->ii_btop <= bindex && bindex <= iinfo->ii_bbot) {
19444 +                       h_i = au_h_iptr(inode, bindex);
19445 +                       if (h_i) {
19446 +                               if (h_i == h_inode)
19447 +                                       continue;
19448 +                               err = -EIO;
19449 +                               break;
19450 +                       }
19451 +               }
19452 +               if (bindex < iinfo->ii_btop)
19453 +                       iinfo->ii_btop = bindex;
19454 +               if (iinfo->ii_bbot < bindex)
19455 +                       iinfo->ii_bbot = bindex;
19456 +               au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
19457 +               update = 1;
19458 +       }
19459 +       au_update_ibrange(inode, /*do_put_zero*/0);
19460 +       e = au_dy_irefresh(inode);
19461 +       if (unlikely(e && !err))
19462 +               err = e;
19463 +       if (!err)
19464 +               au_refresh_hinode_attr(inode, update && isdir);
19465 +
19466 +out:
19467 +       AuTraceErr(err);
19468 +       return err;
19469 +}
19470 +
19471 +static int set_inode(struct inode *inode, struct dentry *dentry)
19472 +{
19473 +       int err;
19474 +       unsigned int flags;
19475 +       umode_t mode;
19476 +       aufs_bindex_t bindex, btop, btail;
19477 +       unsigned char isdir;
19478 +       struct dentry *h_dentry;
19479 +       struct inode *h_inode;
19480 +       struct au_iinfo *iinfo;
19481 +       const struct inode_operations *iop;
19482 +
19483 +       IiMustWriteLock(inode);
19484 +
19485 +       err = 0;
19486 +       isdir = 0;
19487 +       iop = au_sbi(inode->i_sb)->si_iop_array;
19488 +       btop = au_dbtop(dentry);
19489 +       h_dentry = au_h_dptr(dentry, btop);
19490 +       h_inode = d_inode(h_dentry);
19491 +       mode = h_inode->i_mode;
19492 +       switch (mode & S_IFMT) {
19493 +       case S_IFREG:
19494 +               btail = au_dbtail(dentry);
19495 +               inode->i_op = iop + AuIop_OTHER;
19496 +               inode->i_fop = &aufs_file_fop;
19497 +               err = au_dy_iaop(inode, btop, h_inode);
19498 +               if (unlikely(err))
19499 +                       goto out;
19500 +               break;
19501 +       case S_IFDIR:
19502 +               isdir = 1;
19503 +               btail = au_dbtaildir(dentry);
19504 +               inode->i_op = iop + AuIop_DIR;
19505 +               inode->i_fop = &aufs_dir_fop;
19506 +               break;
19507 +       case S_IFLNK:
19508 +               btail = au_dbtail(dentry);
19509 +               inode->i_op = iop + AuIop_SYMLINK;
19510 +               break;
19511 +       case S_IFBLK:
19512 +       case S_IFCHR:
19513 +       case S_IFIFO:
19514 +       case S_IFSOCK:
19515 +               btail = au_dbtail(dentry);
19516 +               inode->i_op = iop + AuIop_OTHER;
19517 +               init_special_inode(inode, mode, h_inode->i_rdev);
19518 +               break;
19519 +       default:
19520 +               AuIOErr("Unknown file type 0%o\n", mode);
19521 +               err = -EIO;
19522 +               goto out;
19523 +       }
19524 +
19525 +       /* do not set hnotify for whiteouted dirs (SHWH mode) */
19526 +       flags = au_hi_flags(inode, isdir);
19527 +       if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
19528 +           && au_ftest_hi(flags, HNOTIFY)
19529 +           && dentry->d_name.len > AUFS_WH_PFX_LEN
19530 +           && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
19531 +               au_fclr_hi(flags, HNOTIFY);
19532 +       iinfo = au_ii(inode);
19533 +       iinfo->ii_btop = btop;
19534 +       iinfo->ii_bbot = btail;
19535 +       for (bindex = btop; bindex <= btail; bindex++) {
19536 +               h_dentry = au_h_dptr(dentry, bindex);
19537 +               if (h_dentry)
19538 +                       au_set_h_iptr(inode, bindex,
19539 +                                     au_igrab(d_inode(h_dentry)), flags);
19540 +       }
19541 +       au_cpup_attr_all(inode, /*force*/1);
19542 +       /*
19543 +        * to force calling aufs_get_acl() every time,
19544 +        * do not call cache_no_acl() for aufs inode.
19545 +        */
19546 +
19547 +out:
19548 +       return err;
19549 +}
19550 +
19551 +/*
19552 + * successful returns with iinfo write_locked
19553 + * minus: errno
19554 + * zero: success, matched
19555 + * plus: no error, but unmatched
19556 + */
19557 +static int reval_inode(struct inode *inode, struct dentry *dentry)
19558 +{
19559 +       int err;
19560 +       unsigned int gen, igflags;
19561 +       aufs_bindex_t bindex, bbot;
19562 +       struct inode *h_inode, *h_dinode;
19563 +       struct dentry *h_dentry;
19564 +
19565 +       /*
19566 +        * before this function, if aufs got any iinfo lock, it must be only
19567 +        * one, the parent dir.
19568 +        * it can happen by UDBA and the obsoleted inode number.
19569 +        */
19570 +       err = -EIO;
19571 +       if (unlikely(inode->i_ino == parent_ino(dentry)))
19572 +               goto out;
19573 +
19574 +       err = 1;
19575 +       ii_write_lock_new_child(inode);
19576 +       h_dentry = au_h_dptr(dentry, au_dbtop(dentry));
19577 +       h_dinode = d_inode(h_dentry);
19578 +       bbot = au_ibbot(inode);
19579 +       for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
19580 +               h_inode = au_h_iptr(inode, bindex);
19581 +               if (!h_inode || h_inode != h_dinode)
19582 +                       continue;
19583 +
19584 +               err = 0;
19585 +               gen = au_iigen(inode, &igflags);
19586 +               if (gen == au_digen(dentry)
19587 +                   && !au_ig_ftest(igflags, HALF_REFRESHED))
19588 +                       break;
19589 +
19590 +               /* fully refresh inode using dentry */
19591 +               err = au_refresh_hinode(inode, dentry);
19592 +               if (!err)
19593 +                       au_update_iigen(inode, /*half*/0);
19594 +               break;
19595 +       }
19596 +
19597 +       if (unlikely(err))
19598 +               ii_write_unlock(inode);
19599 +out:
19600 +       return err;
19601 +}
19602 +
19603 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19604 +          unsigned int d_type, ino_t *ino)
19605 +{
19606 +       int err, idx;
19607 +       const int isnondir = d_type != DT_DIR;
19608 +
19609 +       /* prevent hardlinked inode number from race condition */
19610 +       if (isnondir) {
19611 +               err = au_xinondir_enter(sb, bindex, h_ino, &idx);
19612 +               if (unlikely(err))
19613 +                       goto out;
19614 +       }
19615 +
19616 +       err = au_xino_read(sb, bindex, h_ino, ino);
19617 +       if (unlikely(err))
19618 +               goto out_xinondir;
19619 +
19620 +       if (!*ino) {
19621 +               err = -EIO;
19622 +               *ino = au_xino_new_ino(sb);
19623 +               if (unlikely(!*ino))
19624 +                       goto out_xinondir;
19625 +               err = au_xino_write(sb, bindex, h_ino, *ino);
19626 +               if (unlikely(err))
19627 +                       goto out_xinondir;
19628 +       }
19629 +
19630 +out_xinondir:
19631 +       if (isnondir && idx >= 0)
19632 +               au_xinondir_leave(sb, bindex, h_ino, idx);
19633 +out:
19634 +       return err;
19635 +}
19636 +
19637 +/* successful returns with iinfo write_locked */
19638 +/* todo: return with unlocked? */
19639 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
19640 +{
19641 +       struct inode *inode, *h_inode;
19642 +       struct dentry *h_dentry;
19643 +       struct super_block *sb;
19644 +       ino_t h_ino, ino;
19645 +       int err, idx, hlinked;
19646 +       aufs_bindex_t btop;
19647 +
19648 +       sb = dentry->d_sb;
19649 +       btop = au_dbtop(dentry);
19650 +       h_dentry = au_h_dptr(dentry, btop);
19651 +       h_inode = d_inode(h_dentry);
19652 +       h_ino = h_inode->i_ino;
19653 +       hlinked = !d_is_dir(h_dentry) && h_inode->i_nlink > 1;
19654 +
19655 +new_ino:
19656 +       /*
19657 +        * stop 'race'-ing between hardlinks under different
19658 +        * parents.
19659 +        */
19660 +       if (hlinked) {
19661 +               err = au_xinondir_enter(sb, btop, h_ino, &idx);
19662 +               inode = ERR_PTR(err);
19663 +               if (unlikely(err))
19664 +                       goto out;
19665 +       }
19666 +
19667 +       err = au_xino_read(sb, btop, h_ino, &ino);
19668 +       inode = ERR_PTR(err);
19669 +       if (unlikely(err))
19670 +               goto out_xinondir;
19671 +
19672 +       if (!ino) {
19673 +               ino = au_xino_new_ino(sb);
19674 +               if (unlikely(!ino)) {
19675 +                       inode = ERR_PTR(-EIO);
19676 +                       goto out_xinondir;
19677 +               }
19678 +       }
19679 +
19680 +       AuDbg("i%lu\n", (unsigned long)ino);
19681 +       inode = au_iget_locked(sb, ino);
19682 +       err = PTR_ERR(inode);
19683 +       if (IS_ERR(inode))
19684 +               goto out_xinondir;
19685 +
19686 +       AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
19687 +       if (inode->i_state & I_NEW) {
19688 +               ii_write_lock_new_child(inode);
19689 +               err = set_inode(inode, dentry);
19690 +               if (!err) {
19691 +                       unlock_new_inode(inode);
19692 +                       goto out_xinondir; /* success */
19693 +               }
19694 +
19695 +               /*
19696 +                * iget_failed() calls iput(), but we need to call
19697 +                * ii_write_unlock() after iget_failed(). so dirty hack for
19698 +                * i_count.
19699 +                */
19700 +               atomic_inc(&inode->i_count);
19701 +               iget_failed(inode);
19702 +               ii_write_unlock(inode);
19703 +               au_xino_write(sb, btop, h_ino, /*ino*/0);
19704 +               /* ignore this error */
19705 +               goto out_iput;
19706 +       } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
19707 +               /*
19708 +                * horrible race condition between lookup, readdir and copyup
19709 +                * (or something).
19710 +                */
19711 +               if (hlinked && idx >= 0)
19712 +                       au_xinondir_leave(sb, btop, h_ino, idx);
19713 +               err = reval_inode(inode, dentry);
19714 +               if (unlikely(err < 0)) {
19715 +                       hlinked = 0;
19716 +                       goto out_iput;
19717 +               }
19718 +               if (!err)
19719 +                       goto out; /* success */
19720 +               else if (hlinked && idx >= 0) {
19721 +                       err = au_xinondir_enter(sb, btop, h_ino, &idx);
19722 +                       if (unlikely(err)) {
19723 +                               iput(inode);
19724 +                               inode = ERR_PTR(err);
19725 +                               goto out;
19726 +                       }
19727 +               }
19728 +       }
19729 +
19730 +       if (unlikely(au_test_fs_unique_ino(h_inode)))
19731 +               AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
19732 +                       " b%d, %s, %pd, hi%lu, i%lu.\n",
19733 +                       btop, au_sbtype(h_dentry->d_sb), dentry,
19734 +                       (unsigned long)h_ino, (unsigned long)ino);
19735 +       ino = 0;
19736 +       err = au_xino_write(sb, btop, h_ino, /*ino*/0);
19737 +       if (!err) {
19738 +               iput(inode);
19739 +               if (hlinked && idx >= 0)
19740 +                       au_xinondir_leave(sb, btop, h_ino, idx);
19741 +               goto new_ino;
19742 +       }
19743 +
19744 +out_iput:
19745 +       iput(inode);
19746 +       inode = ERR_PTR(err);
19747 +out_xinondir:
19748 +       if (hlinked && idx >= 0)
19749 +               au_xinondir_leave(sb, btop, h_ino, idx);
19750 +out:
19751 +       return inode;
19752 +}
19753 +
19754 +/* ---------------------------------------------------------------------- */
19755 +
19756 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19757 +              struct inode *inode)
19758 +{
19759 +       int err;
19760 +       struct inode *hi;
19761 +
19762 +       err = au_br_rdonly(au_sbr(sb, bindex));
19763 +
19764 +       /* pseudo-link after flushed may happen out of bounds */
19765 +       if (!err
19766 +           && inode
19767 +           && au_ibtop(inode) <= bindex
19768 +           && bindex <= au_ibbot(inode)) {
19769 +               /*
19770 +                * permission check is unnecessary since vfsub routine
19771 +                * will be called later
19772 +                */
19773 +               hi = au_h_iptr(inode, bindex);
19774 +               if (hi)
19775 +                       err = IS_IMMUTABLE(hi) ? -EROFS : 0;
19776 +       }
19777 +
19778 +       return err;
19779 +}
19780 +
19781 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19782 +                  int mask)
19783 +{
19784 +       if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
19785 +               return 0;
19786 +       return inode_permission(h_userns, h_inode, mask);
19787 +}
19788 +
19789 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19790 +                      int mask)
19791 +{
19792 +       if (au_test_nfs(h_inode->i_sb)
19793 +           && (mask & MAY_WRITE)
19794 +           && S_ISDIR(h_inode->i_mode))
19795 +               mask |= MAY_READ; /* force permission check */
19796 +       return au_test_h_perm(h_userns, h_inode, mask);
19797 +}
19798 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
19799 --- /usr/share/empty/fs/aufs/inode.h    1970-01-01 01:00:00.000000000 +0100
19800 +++ linux/fs/aufs/inode.h       2022-11-05 23:02:18.965889284 +0100
19801 @@ -0,0 +1,705 @@
19802 +/* SPDX-License-Identifier: GPL-2.0 */
19803 +/*
19804 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19805 + *
19806 + * This program is free software; you can redistribute it and/or modify
19807 + * it under the terms of the GNU General Public License as published by
19808 + * the Free Software Foundation; either version 2 of the License, or
19809 + * (at your option) any later version.
19810 + *
19811 + * This program is distributed in the hope that it will be useful,
19812 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19813 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19814 + * GNU General Public License for more details.
19815 + *
19816 + * You should have received a copy of the GNU General Public License
19817 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
19818 + */
19819 +
19820 +/*
19821 + * inode operations
19822 + */
19823 +
19824 +#ifndef __AUFS_INODE_H__
19825 +#define __AUFS_INODE_H__
19826 +
19827 +#ifdef __KERNEL__
19828 +
19829 +#include <linux/fsnotify.h>
19830 +#include "rwsem.h"
19831 +
19832 +struct vfsmount;
19833 +
19834 +struct au_hnotify {
19835 +#ifdef CONFIG_AUFS_HNOTIFY
19836 +#ifdef CONFIG_AUFS_HFSNOTIFY
19837 +       /* never use fsnotify_add_vfsmount_mark() */
19838 +       struct fsnotify_mark            hn_mark;
19839 +#endif
19840 +       struct inode            *hn_aufs_inode; /* no get/put */
19841 +       struct rcu_head         rcu;
19842 +#endif
19843 +} ____cacheline_aligned_in_smp;
19844 +
19845 +struct au_hinode {
19846 +       struct inode            *hi_inode;
19847 +       aufs_bindex_t           hi_id;
19848 +#ifdef CONFIG_AUFS_HNOTIFY
19849 +       struct au_hnotify       *hi_notify;
19850 +#endif
19851 +
19852 +       /* reference to the copied-up whiteout with get/put */
19853 +       struct dentry           *hi_whdentry;
19854 +};
19855 +
19856 +/* ig_flags */
19857 +#define AuIG_HALF_REFRESHED            1
19858 +#define au_ig_ftest(flags, name)       ((flags) & AuIG_##name)
19859 +#define au_ig_fset(flags, name) \
19860 +       do { (flags) |= AuIG_##name; } while (0)
19861 +#define au_ig_fclr(flags, name) \
19862 +       do { (flags) &= ~AuIG_##name; } while (0)
19863 +
19864 +struct au_iigen {
19865 +       spinlock_t      ig_spin;
19866 +       __u32           ig_generation, ig_flags;
19867 +};
19868 +
19869 +struct au_vdir;
19870 +struct au_iinfo {
19871 +       struct au_iigen         ii_generation;
19872 +       struct super_block      *ii_hsb1;       /* no get/put */
19873 +
19874 +       struct au_rwsem         ii_rwsem;
19875 +       aufs_bindex_t           ii_btop, ii_bbot;
19876 +       __u32                   ii_higen;
19877 +       struct au_hinode        *ii_hinode;
19878 +       struct au_vdir          *ii_vdir;
19879 +};
19880 +
19881 +struct au_icntnr {
19882 +       struct au_iinfo         iinfo;
19883 +       struct inode            vfs_inode;
19884 +       struct hlist_bl_node    plink;
19885 +       struct rcu_head         rcu;
19886 +} ____cacheline_aligned_in_smp;
19887 +
19888 +/* au_pin flags */
19889 +#define AuPin_DI_LOCKED                1
19890 +#define AuPin_MNT_WRITE                (1 << 1)
19891 +#define au_ftest_pin(flags, name)      ((flags) & AuPin_##name)
19892 +#define au_fset_pin(flags, name) \
19893 +       do { (flags) |= AuPin_##name; } while (0)
19894 +#define au_fclr_pin(flags, name) \
19895 +       do { (flags) &= ~AuPin_##name; } while (0)
19896 +
19897 +struct au_pin {
19898 +       /* input */
19899 +       struct dentry *dentry;
19900 +       unsigned int udba;
19901 +       unsigned char lsc_di, lsc_hi, flags;
19902 +       aufs_bindex_t bindex;
19903 +
19904 +       /* output */
19905 +       struct dentry *parent;
19906 +       struct au_hinode *hdir;
19907 +       struct vfsmount *h_mnt;
19908 +
19909 +       /* temporary unlock/relock for copyup */
19910 +       struct dentry *h_dentry, *h_parent;
19911 +       struct au_branch *br;
19912 +       struct task_struct *task;
19913 +};
19914 +
19915 +void au_pin_hdir_unlock(struct au_pin *p);
19916 +int au_pin_hdir_lock(struct au_pin *p);
19917 +int au_pin_hdir_relock(struct au_pin *p);
19918 +void au_pin_hdir_acquire_nest(struct au_pin *p);
19919 +void au_pin_hdir_release(struct au_pin *p);
19920 +
19921 +/* ---------------------------------------------------------------------- */
19922 +
19923 +static inline struct au_iinfo *au_ii(struct inode *inode)
19924 +{
19925 +       BUG_ON(is_bad_inode(inode));
19926 +       return &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19927 +}
19928 +
19929 +/* ---------------------------------------------------------------------- */
19930 +
19931 +/* inode.c */
19932 +struct inode *au_igrab(struct inode *inode);
19933 +void au_refresh_iop(struct inode *inode, int force_getattr);
19934 +int au_refresh_hinode_self(struct inode *inode);
19935 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
19936 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19937 +          unsigned int d_type, ino_t *ino);
19938 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
19939 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19940 +              struct inode *inode);
19941 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19942 +                  int mask);
19943 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19944 +                      int mask);
19945 +
19946 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
19947 +                           ino_t h_ino, unsigned int d_type, ino_t *ino)
19948 +{
19949 +#ifdef CONFIG_AUFS_SHWH
19950 +       return au_ino(sb, bindex, h_ino, d_type, ino);
19951 +#else
19952 +       return 0;
19953 +#endif
19954 +}
19955 +
19956 +/* i_op.c */
19957 +enum {
19958 +       AuIop_SYMLINK,
19959 +       AuIop_DIR,
19960 +       AuIop_OTHER,
19961 +       AuIop_Last
19962 +};
19963 +extern struct inode_operations aufs_iop[AuIop_Last], /* not const */
19964 +       aufs_iop_nogetattr[AuIop_Last];
19965 +
19966 +/* au_wr_dir flags */
19967 +#define AuWrDir_ADD_ENTRY      1
19968 +#define AuWrDir_ISDIR          (1 << 1)
19969 +#define AuWrDir_TMPFILE                (1 << 2)
19970 +#define au_ftest_wrdir(flags, name)    ((flags) & AuWrDir_##name)
19971 +#define au_fset_wrdir(flags, name) \
19972 +       do { (flags) |= AuWrDir_##name; } while (0)
19973 +#define au_fclr_wrdir(flags, name) \
19974 +       do { (flags) &= ~AuWrDir_##name; } while (0)
19975 +
19976 +struct au_wr_dir_args {
19977 +       aufs_bindex_t force_btgt;
19978 +       unsigned char flags;
19979 +};
19980 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
19981 +             struct au_wr_dir_args *args);
19982 +
19983 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
19984 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
19985 +                aufs_bindex_t bindex, int lsc_di, int lsc_hi,
19986 +                unsigned int udba, unsigned char flags);
19987 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
19988 +          unsigned int udba, unsigned char flags) __must_check;
19989 +int au_do_pin(struct au_pin *pin) __must_check;
19990 +void au_unpin(struct au_pin *pin);
19991 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
19992 +
19993 +#define AuIcpup_DID_CPUP       1
19994 +#define au_ftest_icpup(flags, name)    ((flags) & AuIcpup_##name)
19995 +#define au_fset_icpup(flags, name) \
19996 +       do { (flags) |= AuIcpup_##name; } while (0)
19997 +#define au_fclr_icpup(flags, name) \
19998 +       do { (flags) &= ~AuIcpup_##name; } while (0)
19999 +
20000 +struct au_icpup_args {
20001 +       unsigned char flags;
20002 +       unsigned char pin_flags;
20003 +       aufs_bindex_t btgt;
20004 +       unsigned int udba;
20005 +       struct au_pin pin;
20006 +       struct path h_path;
20007 +       struct inode *h_inode;
20008 +};
20009 +
20010 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
20011 +                    struct au_icpup_args *a);
20012 +
20013 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
20014 +                     struct path *h_path, int locked);
20015 +
20016 +/* i_op_add.c */
20017 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20018 +              struct dentry *h_parent, int isdir);
20019 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
20020 +              struct dentry *dentry, umode_t mode, dev_t dev);
20021 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
20022 +                struct dentry *dentry, const char *symname);
20023 +int aufs_create(struct user_namespace *userns, struct inode *dir,
20024 +               struct dentry *dentry, umode_t mode, bool want_excl);
20025 +struct vfsub_aopen_args;
20026 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
20027 +                      struct vfsub_aopen_args *args);
20028 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
20029 +                struct dentry *dentry, umode_t mode);
20030 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
20031 +             struct dentry *dentry);
20032 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
20033 +              struct dentry *dentry, umode_t mode);
20034 +
20035 +/* i_op_del.c */
20036 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
20037 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
20038 +              struct dentry *h_parent, int isdir);
20039 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
20040 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
20041 +
20042 +/* i_op_ren.c */
20043 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
20044 +int aufs_rename(struct user_namespace *userns,
20045 +               struct inode *_src_dir, struct dentry *_src_dentry,
20046 +               struct inode *_dst_dir, struct dentry *_dst_dentry,
20047 +               unsigned int _flags);
20048 +
20049 +/* iinfo.c */
20050 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
20051 +void au_hiput(struct au_hinode *hinode);
20052 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
20053 +                 struct dentry *h_wh);
20054 +unsigned int au_hi_flags(struct inode *inode, int isdir);
20055 +
20056 +/* hinode flags */
20057 +#define AuHi_XINO      1
20058 +#define AuHi_HNOTIFY   (1 << 1)
20059 +#define au_ftest_hi(flags, name)       ((flags) & AuHi_##name)
20060 +#define au_fset_hi(flags, name) \
20061 +       do { (flags) |= AuHi_##name; } while (0)
20062 +#define au_fclr_hi(flags, name) \
20063 +       do { (flags) &= ~AuHi_##name; } while (0)
20064 +
20065 +#ifndef CONFIG_AUFS_HNOTIFY
20066 +#undef AuHi_HNOTIFY
20067 +#define AuHi_HNOTIFY   0
20068 +#endif
20069 +
20070 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
20071 +                  struct inode *h_inode, unsigned int flags);
20072 +
20073 +void au_update_iigen(struct inode *inode, int half);
20074 +void au_update_ibrange(struct inode *inode, int do_put_zero);
20075 +
20076 +void au_icntnr_init_once(void *_c);
20077 +void au_hinode_init(struct au_hinode *hinode);
20078 +int au_iinfo_init(struct inode *inode);
20079 +void au_iinfo_fin(struct inode *inode);
20080 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink);
20081 +
20082 +#ifdef CONFIG_PROC_FS
20083 +/* plink.c */
20084 +int au_plink_maint(struct super_block *sb, int flags);
20085 +struct au_sbinfo;
20086 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
20087 +int au_plink_maint_enter(struct super_block *sb);
20088 +#ifdef CONFIG_AUFS_DEBUG
20089 +void au_plink_list(struct super_block *sb);
20090 +#else
20091 +AuStubVoid(au_plink_list, struct super_block *sb)
20092 +#endif
20093 +int au_plink_test(struct inode *inode);
20094 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
20095 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
20096 +                    struct dentry *h_dentry);
20097 +void au_plink_put(struct super_block *sb, int verbose);
20098 +void au_plink_clean(struct super_block *sb, int verbose);
20099 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
20100 +#else
20101 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
20102 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
20103 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
20104 +AuStubVoid(au_plink_list, struct super_block *sb);
20105 +AuStubInt0(au_plink_test, struct inode *inode);
20106 +AuStub(struct dentry *, au_plink_lkup, return NULL,
20107 +       struct inode *inode, aufs_bindex_t bindex);
20108 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
20109 +          struct dentry *h_dentry);
20110 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
20111 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
20112 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
20113 +#endif /* CONFIG_PROC_FS */
20114 +
20115 +#ifdef CONFIG_AUFS_XATTR
20116 +/* xattr.c */
20117 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
20118 +                 unsigned int verbose);
20119 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
20120 +void au_xattr_init(struct super_block *sb);
20121 +#else
20122 +AuStubInt0(au_cpup_xattr, struct path *h_dst, struct path *h_src,
20123 +          int ignore_flags, unsigned int verbose);
20124 +AuStubVoid(au_xattr_init, struct super_block *sb);
20125 +#endif
20126 +
20127 +#ifdef CONFIG_FS_POSIX_ACL
20128 +struct posix_acl *aufs_get_acl(struct inode *inode, int type, bool rcu);
20129 +int aufs_set_acl(struct user_namespace *userns, struct inode *inode,
20130 +                struct posix_acl *acl, int type);
20131 +#endif
20132 +
20133 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
20134 +enum {
20135 +       AU_XATTR_SET,
20136 +       AU_ACL_SET
20137 +};
20138 +
20139 +struct au_sxattr {
20140 +       int type;
20141 +       union {
20142 +               struct {
20143 +                       const char      *name;
20144 +                       void            *value;
20145 +                       size_t          size;
20146 +                       int             flags;
20147 +               } set;
20148 +               struct {
20149 +                       struct posix_acl *acl;
20150 +                       int             type;
20151 +               } acl_set;
20152 +       } u;
20153 +};
20154 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
20155 +                 struct au_sxattr *arg);
20156 +#endif
20157 +
20158 +/* ---------------------------------------------------------------------- */
20159 +
20160 +/* lock subclass for iinfo */
20161 +enum {
20162 +       AuLsc_II_CHILD,         /* child first */
20163 +       AuLsc_II_CHILD2,        /* rename(2), link(2), and cpup at hnotify */
20164 +       AuLsc_II_CHILD3,        /* copyup dirs */
20165 +       AuLsc_II_PARENT,        /* see AuLsc_I_PARENT in vfsub.h */
20166 +       AuLsc_II_PARENT2,
20167 +       AuLsc_II_PARENT3,       /* copyup dirs */
20168 +       AuLsc_II_NEW_CHILD
20169 +};
20170 +
20171 +/*
20172 + * ii_read_lock_child, ii_write_lock_child,
20173 + * ii_read_lock_child2, ii_write_lock_child2,
20174 + * ii_read_lock_child3, ii_write_lock_child3,
20175 + * ii_read_lock_parent, ii_write_lock_parent,
20176 + * ii_read_lock_parent2, ii_write_lock_parent2,
20177 + * ii_read_lock_parent3, ii_write_lock_parent3,
20178 + * ii_read_lock_new_child, ii_write_lock_new_child,
20179 + */
20180 +#define AuReadLockFunc(name, lsc) \
20181 +static inline void ii_read_lock_##name(struct inode *i) \
20182 +{ \
20183 +       au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20184 +}
20185 +
20186 +#define AuWriteLockFunc(name, lsc) \
20187 +static inline void ii_write_lock_##name(struct inode *i) \
20188 +{ \
20189 +       au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20190 +}
20191 +
20192 +#define AuRWLockFuncs(name, lsc) \
20193 +       AuReadLockFunc(name, lsc) \
20194 +       AuWriteLockFunc(name, lsc)
20195 +
20196 +AuRWLockFuncs(child, CHILD);
20197 +AuRWLockFuncs(child2, CHILD2);
20198 +AuRWLockFuncs(child3, CHILD3);
20199 +AuRWLockFuncs(parent, PARENT);
20200 +AuRWLockFuncs(parent2, PARENT2);
20201 +AuRWLockFuncs(parent3, PARENT3);
20202 +AuRWLockFuncs(new_child, NEW_CHILD);
20203 +
20204 +#undef AuReadLockFunc
20205 +#undef AuWriteLockFunc
20206 +#undef AuRWLockFuncs
20207 +
20208 +#define ii_read_unlock(i)      au_rw_read_unlock(&au_ii(i)->ii_rwsem)
20209 +#define ii_write_unlock(i)     au_rw_write_unlock(&au_ii(i)->ii_rwsem)
20210 +#define ii_downgrade_lock(i)   au_rw_dgrade_lock(&au_ii(i)->ii_rwsem)
20211 +
20212 +#define IiMustNoWaiters(i)     AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
20213 +#define IiMustAnyLock(i)       AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
20214 +#define IiMustWriteLock(i)     AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
20215 +
20216 +/* ---------------------------------------------------------------------- */
20217 +
20218 +static inline void au_icntnr_init(struct au_icntnr *c)
20219 +{
20220 +#ifdef CONFIG_AUFS_DEBUG
20221 +       c->vfs_inode.i_mode = 0;
20222 +#endif
20223 +}
20224 +
20225 +static inline unsigned int au_iigen(struct inode *inode, unsigned int *igflags)
20226 +{
20227 +       unsigned int gen;
20228 +       struct au_iinfo *iinfo;
20229 +       struct au_iigen *iigen;
20230 +
20231 +       iinfo = au_ii(inode);
20232 +       iigen = &iinfo->ii_generation;
20233 +       spin_lock(&iigen->ig_spin);
20234 +       if (igflags)
20235 +               *igflags = iigen->ig_flags;
20236 +       gen = iigen->ig_generation;
20237 +       spin_unlock(&iigen->ig_spin);
20238 +
20239 +       return gen;
20240 +}
20241 +
20242 +/* tiny test for inode number */
20243 +/* tmpfs generation is too rough */
20244 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
20245 +{
20246 +       struct au_iinfo *iinfo;
20247 +
20248 +       iinfo = au_ii(inode);
20249 +       AuRwMustAnyLock(&iinfo->ii_rwsem);
20250 +       return !(iinfo->ii_hsb1 == h_inode->i_sb
20251 +                && iinfo->ii_higen == h_inode->i_generation);
20252 +}
20253 +
20254 +static inline void au_iigen_dec(struct inode *inode)
20255 +{
20256 +       struct au_iinfo *iinfo;
20257 +       struct au_iigen *iigen;
20258 +
20259 +       iinfo = au_ii(inode);
20260 +       iigen = &iinfo->ii_generation;
20261 +       spin_lock(&iigen->ig_spin);
20262 +       iigen->ig_generation--;
20263 +       spin_unlock(&iigen->ig_spin);
20264 +}
20265 +
20266 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
20267 +{
20268 +       int err;
20269 +
20270 +       err = 0;
20271 +       if (unlikely(inode && au_iigen(inode, NULL) != sigen))
20272 +               err = -EIO;
20273 +
20274 +       return err;
20275 +}
20276 +
20277 +/* ---------------------------------------------------------------------- */
20278 +
20279 +static inline struct au_hinode *au_hinode(struct au_iinfo *iinfo,
20280 +                                         aufs_bindex_t bindex)
20281 +{
20282 +       return iinfo->ii_hinode + bindex;
20283 +}
20284 +
20285 +static inline int au_is_bad_inode(struct inode *inode)
20286 +{
20287 +       return !!(is_bad_inode(inode) || !au_hinode(au_ii(inode), 0));
20288 +}
20289 +
20290 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
20291 +                                       aufs_bindex_t bindex)
20292 +{
20293 +       IiMustAnyLock(inode);
20294 +       return au_hinode(au_ii(inode), bindex)->hi_id;
20295 +}
20296 +
20297 +static inline aufs_bindex_t au_ibtop(struct inode *inode)
20298 +{
20299 +       IiMustAnyLock(inode);
20300 +       return au_ii(inode)->ii_btop;
20301 +}
20302 +
20303 +static inline aufs_bindex_t au_ibbot(struct inode *inode)
20304 +{
20305 +       IiMustAnyLock(inode);
20306 +       return au_ii(inode)->ii_bbot;
20307 +}
20308 +
20309 +static inline struct au_vdir *au_ivdir(struct inode *inode)
20310 +{
20311 +       IiMustAnyLock(inode);
20312 +       return au_ii(inode)->ii_vdir;
20313 +}
20314 +
20315 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
20316 +{
20317 +       IiMustAnyLock(inode);
20318 +       return au_hinode(au_ii(inode), bindex)->hi_whdentry;
20319 +}
20320 +
20321 +static inline void au_set_ibtop(struct inode *inode, aufs_bindex_t bindex)
20322 +{
20323 +       IiMustWriteLock(inode);
20324 +       au_ii(inode)->ii_btop = bindex;
20325 +}
20326 +
20327 +static inline void au_set_ibbot(struct inode *inode, aufs_bindex_t bindex)
20328 +{
20329 +       IiMustWriteLock(inode);
20330 +       au_ii(inode)->ii_bbot = bindex;
20331 +}
20332 +
20333 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
20334 +{
20335 +       IiMustWriteLock(inode);
20336 +       au_ii(inode)->ii_vdir = vdir;
20337 +}
20338 +
20339 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
20340 +{
20341 +       IiMustAnyLock(inode);
20342 +       return au_hinode(au_ii(inode), bindex);
20343 +}
20344 +
20345 +/* ---------------------------------------------------------------------- */
20346 +
20347 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
20348 +{
20349 +       if (pin)
20350 +               return pin->parent;
20351 +       return NULL;
20352 +}
20353 +
20354 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
20355 +{
20356 +       if (pin && pin->hdir)
20357 +               return pin->hdir->hi_inode;
20358 +       return NULL;
20359 +}
20360 +
20361 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
20362 +{
20363 +       if (pin)
20364 +               return pin->hdir;
20365 +       return NULL;
20366 +}
20367 +
20368 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
20369 +{
20370 +       if (pin)
20371 +               pin->dentry = dentry;
20372 +}
20373 +
20374 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
20375 +                                          unsigned char lflag)
20376 +{
20377 +       if (pin) {
20378 +               if (lflag)
20379 +                       au_fset_pin(pin->flags, DI_LOCKED);
20380 +               else
20381 +                       au_fclr_pin(pin->flags, DI_LOCKED);
20382 +       }
20383 +}
20384 +
20385 +#if 0 /* reserved */
20386 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
20387 +{
20388 +       if (pin) {
20389 +               dput(pin->parent);
20390 +               pin->parent = dget(parent);
20391 +       }
20392 +}
20393 +#endif
20394 +
20395 +/* ---------------------------------------------------------------------- */
20396 +
20397 +struct au_branch;
20398 +#ifdef CONFIG_AUFS_HNOTIFY
20399 +struct au_hnotify_op {
20400 +       void (*ctl)(struct au_hinode *hinode, int do_set);
20401 +       int (*alloc)(struct au_hinode *hinode);
20402 +
20403 +       /*
20404 +        * if it returns true, the caller should free hinode->hi_notify,
20405 +        * otherwise ->free() frees it.
20406 +        */
20407 +       int (*free)(struct au_hinode *hinode,
20408 +                   struct au_hnotify *hn) __must_check;
20409 +
20410 +       void (*fin)(void);
20411 +       int (*init)(void);
20412 +
20413 +       int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
20414 +       void (*fin_br)(struct au_branch *br);
20415 +       int (*init_br)(struct au_branch *br, int perm);
20416 +};
20417 +
20418 +/* hnotify.c */
20419 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
20420 +void au_hn_free(struct au_hinode *hinode);
20421 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
20422 +void au_hn_reset(struct inode *inode, unsigned int flags);
20423 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
20424 +              const struct qstr *h_child_qstr, struct inode *h_child_inode);
20425 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
20426 +int au_hnotify_init_br(struct au_branch *br, int perm);
20427 +void au_hnotify_fin_br(struct au_branch *br);
20428 +int __init au_hnotify_init(void);
20429 +void au_hnotify_fin(void);
20430 +
20431 +/* hfsnotify.c */
20432 +extern const struct au_hnotify_op au_hnotify_op;
20433 +
20434 +static inline
20435 +void au_hn_init(struct au_hinode *hinode)
20436 +{
20437 +       hinode->hi_notify = NULL;
20438 +}
20439 +
20440 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
20441 +{
20442 +       return hinode->hi_notify;
20443 +}
20444 +
20445 +#else
20446 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
20447 +       struct au_hinode *hinode __maybe_unused,
20448 +       struct inode *inode __maybe_unused)
20449 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
20450 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
20451 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
20452 +          int do_set __maybe_unused)
20453 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
20454 +          unsigned int flags __maybe_unused)
20455 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
20456 +          struct au_branch *br __maybe_unused,
20457 +          int perm __maybe_unused)
20458 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
20459 +          int perm __maybe_unused)
20460 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
20461 +AuStubInt0(__init au_hnotify_init, void)
20462 +AuStubVoid(au_hnotify_fin, void)
20463 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
20464 +#endif /* CONFIG_AUFS_HNOTIFY */
20465 +
20466 +static inline void au_hn_suspend(struct au_hinode *hdir)
20467 +{
20468 +       au_hn_ctl(hdir, /*do_set*/0);
20469 +}
20470 +
20471 +static inline void au_hn_resume(struct au_hinode *hdir)
20472 +{
20473 +       au_hn_ctl(hdir, /*do_set*/1);
20474 +}
20475 +
20476 +static inline void au_hn_inode_lock(struct au_hinode *hdir)
20477 +{
20478 +       inode_lock(hdir->hi_inode);
20479 +       au_hn_suspend(hdir);
20480 +}
20481 +
20482 +static inline void au_hn_inode_lock_nested(struct au_hinode *hdir,
20483 +                                         unsigned int sc __maybe_unused)
20484 +{
20485 +       inode_lock_nested(hdir->hi_inode, sc);
20486 +       au_hn_suspend(hdir);
20487 +}
20488 +
20489 +#if 0 /* unused */
20490 +#include "vfsub.h"
20491 +static inline void au_hn_inode_lock_shared_nested(struct au_hinode *hdir,
20492 +                                                 unsigned int sc)
20493 +{
20494 +       inode_lock_shared_nested(hdir->hi_inode, sc);
20495 +       au_hn_suspend(hdir);
20496 +}
20497 +#endif
20498 +
20499 +static inline void au_hn_inode_unlock(struct au_hinode *hdir)
20500 +{
20501 +       au_hn_resume(hdir);
20502 +       inode_unlock(hdir->hi_inode);
20503 +}
20504 +
20505 +#endif /* __KERNEL__ */
20506 +#endif /* __AUFS_INODE_H__ */
20507 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
20508 --- /usr/share/empty/fs/aufs/ioctl.c    1970-01-01 01:00:00.000000000 +0100
20509 +++ linux/fs/aufs/ioctl.c       2022-11-05 23:02:18.965889284 +0100
20510 @@ -0,0 +1,220 @@
20511 +// SPDX-License-Identifier: GPL-2.0
20512 +/*
20513 + * Copyright (C) 2005-2022 Junjiro R. Okajima
20514 + *
20515 + * This program is free software; you can redistribute it and/or modify
20516 + * it under the terms of the GNU General Public License as published by
20517 + * the Free Software Foundation; either version 2 of the License, or
20518 + * (at your option) any later version.
20519 + *
20520 + * This program is distributed in the hope that it will be useful,
20521 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20522 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20523 + * GNU General Public License for more details.
20524 + *
20525 + * You should have received a copy of the GNU General Public License
20526 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
20527 + */
20528 +
20529 +/*
20530 + * ioctl
20531 + * plink-management and readdir in userspace.
20532 + * assist the pathconf(3) wrapper library.
20533 + * move-down
20534 + * File-based Hierarchical Storage Management.
20535 + */
20536 +
20537 +#include <linux/compat.h>
20538 +#include <linux/file.h>
20539 +#include "aufs.h"
20540 +
20541 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
20542 +{
20543 +       int err, fd;
20544 +       aufs_bindex_t wbi, bindex, bbot;
20545 +       struct file *h_file;
20546 +       struct super_block *sb;
20547 +       struct dentry *root;
20548 +       struct au_branch *br;
20549 +       struct aufs_wbr_fd wbrfd = {
20550 +               .oflags = au_dir_roflags,
20551 +               .brid   = -1
20552 +       };
20553 +       const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
20554 +               | O_NOATIME | O_CLOEXEC;
20555 +
20556 +       AuDebugOn(wbrfd.oflags & ~valid);
20557 +
20558 +       if (arg) {
20559 +               err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
20560 +               if (unlikely(err)) {
20561 +                       err = -EFAULT;
20562 +                       goto out;
20563 +               }
20564 +
20565 +               err = -EINVAL;
20566 +               AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
20567 +               wbrfd.oflags |= au_dir_roflags;
20568 +               AuDbg("0%o\n", wbrfd.oflags);
20569 +               if (unlikely(wbrfd.oflags & ~valid))
20570 +                       goto out;
20571 +       }
20572 +
20573 +       fd = get_unused_fd_flags(0);
20574 +       err = fd;
20575 +       if (unlikely(fd < 0))
20576 +               goto out;
20577 +
20578 +       h_file = ERR_PTR(-EINVAL);
20579 +       wbi = 0;
20580 +       br = NULL;
20581 +       sb = path->dentry->d_sb;
20582 +       root = sb->s_root;
20583 +       aufs_read_lock(root, AuLock_IR);
20584 +       bbot = au_sbbot(sb);
20585 +       if (wbrfd.brid >= 0) {
20586 +               wbi = au_br_index(sb, wbrfd.brid);
20587 +               if (unlikely(wbi < 0 || wbi > bbot))
20588 +                       goto out_unlock;
20589 +       }
20590 +
20591 +       h_file = ERR_PTR(-ENOENT);
20592 +       br = au_sbr(sb, wbi);
20593 +       if (!au_br_writable(br->br_perm)) {
20594 +               if (arg)
20595 +                       goto out_unlock;
20596 +
20597 +               bindex = wbi + 1;
20598 +               wbi = -1;
20599 +               for (; bindex <= bbot; bindex++) {
20600 +                       br = au_sbr(sb, bindex);
20601 +                       if (au_br_writable(br->br_perm)) {
20602 +                               wbi = bindex;
20603 +                               br = au_sbr(sb, wbi);
20604 +                               break;
20605 +                       }
20606 +               }
20607 +       }
20608 +       AuDbg("wbi %d\n", wbi);
20609 +       if (wbi >= 0)
20610 +               h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
20611 +                                  /*force_wr*/0);
20612 +
20613 +out_unlock:
20614 +       aufs_read_unlock(root, AuLock_IR);
20615 +       err = PTR_ERR(h_file);
20616 +       if (IS_ERR(h_file))
20617 +               goto out_fd;
20618 +
20619 +       au_lcnt_dec(&br->br_nfiles); /* cf. au_h_open() */
20620 +       fd_install(fd, h_file);
20621 +       err = fd;
20622 +       goto out; /* success */
20623 +
20624 +out_fd:
20625 +       put_unused_fd(fd);
20626 +out:
20627 +       AuTraceErr(err);
20628 +       return err;
20629 +}
20630 +
20631 +/* ---------------------------------------------------------------------- */
20632 +
20633 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
20634 +{
20635 +       long err;
20636 +       struct dentry *dentry;
20637 +
20638 +       switch (cmd) {
20639 +       case AUFS_CTL_RDU:
20640 +       case AUFS_CTL_RDU_INO:
20641 +               err = au_rdu_ioctl(file, cmd, arg);
20642 +               break;
20643 +
20644 +       case AUFS_CTL_WBR_FD:
20645 +               err = au_wbr_fd(&file->f_path, (void __user *)arg);
20646 +               break;
20647 +
20648 +       case AUFS_CTL_IBUSY:
20649 +               err = au_ibusy_ioctl(file, arg);
20650 +               break;
20651 +
20652 +       case AUFS_CTL_BRINFO:
20653 +               err = au_brinfo_ioctl(file, arg);
20654 +               break;
20655 +
20656 +       case AUFS_CTL_FHSM_FD:
20657 +               dentry = file->f_path.dentry;
20658 +               if (IS_ROOT(dentry))
20659 +                       err = au_fhsm_fd(dentry->d_sb, arg);
20660 +               else
20661 +                       err = -ENOTTY;
20662 +               break;
20663 +
20664 +       default:
20665 +               /* do not call the lower */
20666 +               AuDbg("0x%x\n", cmd);
20667 +               err = -ENOTTY;
20668 +       }
20669 +
20670 +       AuTraceErr(err);
20671 +       return err;
20672 +}
20673 +
20674 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
20675 +{
20676 +       long err;
20677 +
20678 +       switch (cmd) {
20679 +       case AUFS_CTL_MVDOWN:
20680 +               err = au_mvdown(file->f_path.dentry, (void __user *)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 +       default:
20688 +               /* do not call the lower */
20689 +               AuDbg("0x%x\n", cmd);
20690 +               err = -ENOTTY;
20691 +       }
20692 +
20693 +       AuTraceErr(err);
20694 +       return err;
20695 +}
20696 +
20697 +#ifdef CONFIG_COMPAT
20698 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
20699 +                          unsigned long arg)
20700 +{
20701 +       long err;
20702 +
20703 +       switch (cmd) {
20704 +       case AUFS_CTL_RDU:
20705 +       case AUFS_CTL_RDU_INO:
20706 +               err = au_rdu_compat_ioctl(file, cmd, arg);
20707 +               break;
20708 +
20709 +       case AUFS_CTL_IBUSY:
20710 +               err = au_ibusy_compat_ioctl(file, arg);
20711 +               break;
20712 +
20713 +       case AUFS_CTL_BRINFO:
20714 +               err = au_brinfo_compat_ioctl(file, arg);
20715 +               break;
20716 +
20717 +       default:
20718 +               err = aufs_ioctl_dir(file, cmd, arg);
20719 +       }
20720 +
20721 +       AuTraceErr(err);
20722 +       return err;
20723 +}
20724 +
20725 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
20726 +                             unsigned long arg)
20727 +{
20728 +       return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
20729 +}
20730 +#endif
20731 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
20732 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
20733 +++ linux/fs/aufs/i_op_add.c    2022-11-05 23:02:18.965889284 +0100
20734 @@ -0,0 +1,941 @@
20735 +// SPDX-License-Identifier: GPL-2.0
20736 +/*
20737 + * Copyright (C) 2005-2022 Junjiro R. Okajima
20738 + *
20739 + * This program is free software; you can redistribute it and/or modify
20740 + * it under the terms of the GNU General Public License as published by
20741 + * the Free Software Foundation; either version 2 of the License, or
20742 + * (at your option) any later version.
20743 + *
20744 + * This program is distributed in the hope that it will be useful,
20745 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20746 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20747 + * GNU General Public License for more details.
20748 + *
20749 + * You should have received a copy of the GNU General Public License
20750 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
20751 + */
20752 +
20753 +/*
20754 + * inode operations (add entry)
20755 + */
20756 +
20757 +#include <linux/iversion.h>
20758 +#include "aufs.h"
20759 +
20760 +/*
20761 + * final procedure of adding a new entry, except link(2).
20762 + * remove whiteout, instantiate, copyup the parent dir's times and size
20763 + * and update version.
20764 + * if it failed, re-create the removed whiteout.
20765 + */
20766 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
20767 +                 struct dentry *wh_dentry, struct dentry *dentry)
20768 +{
20769 +       int err, rerr;
20770 +       aufs_bindex_t bwh;
20771 +       struct path h_path;
20772 +       struct super_block *sb;
20773 +       struct inode *inode, *h_dir;
20774 +       struct dentry *wh;
20775 +
20776 +       bwh = -1;
20777 +       sb = dir->i_sb;
20778 +       if (wh_dentry) {
20779 +               h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
20780 +               IMustLock(h_dir);
20781 +               AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
20782 +               bwh = au_dbwh(dentry);
20783 +               h_path.dentry = wh_dentry;
20784 +               h_path.mnt = au_sbr_mnt(sb, bindex);
20785 +               err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
20786 +                                         dentry);
20787 +               if (unlikely(err))
20788 +                       goto out;
20789 +       }
20790 +
20791 +       inode = au_new_inode(dentry, /*must_new*/1);
20792 +       if (!IS_ERR(inode)) {
20793 +               d_instantiate(dentry, inode);
20794 +               dir = d_inode(dentry->d_parent); /* dir inode is locked */
20795 +               IMustLock(dir);
20796 +               au_dir_ts(dir, bindex);
20797 +               inode_inc_iversion(dir);
20798 +               au_fhsm_wrote(sb, bindex, /*force*/0);
20799 +               return 0; /* success */
20800 +       }
20801 +
20802 +       err = PTR_ERR(inode);
20803 +       if (!wh_dentry)
20804 +               goto out;
20805 +
20806 +       /* revert */
20807 +       /* dir inode is locked */
20808 +       wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
20809 +       rerr = PTR_ERR(wh);
20810 +       if (IS_ERR(wh)) {
20811 +               AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
20812 +                       dentry, err, rerr);
20813 +               err = -EIO;
20814 +       } else
20815 +               dput(wh);
20816 +
20817 +out:
20818 +       return err;
20819 +}
20820 +
20821 +static int au_d_may_add(struct dentry *dentry)
20822 +{
20823 +       int err;
20824 +
20825 +       err = 0;
20826 +       if (unlikely(d_unhashed(dentry)))
20827 +               err = -ENOENT;
20828 +       if (unlikely(d_really_is_positive(dentry)))
20829 +               err = -EEXIST;
20830 +       return err;
20831 +}
20832 +
20833 +/*
20834 + * simple tests for the adding inode operations.
20835 + * following the checks in vfs, plus the parent-child relationship.
20836 + */
20837 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20838 +              struct dentry *h_parent, int isdir)
20839 +{
20840 +       int err;
20841 +       umode_t h_mode;
20842 +       struct dentry *h_dentry;
20843 +       struct inode *h_inode;
20844 +
20845 +       err = -ENAMETOOLONG;
20846 +       if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20847 +               goto out;
20848 +
20849 +       h_dentry = au_h_dptr(dentry, bindex);
20850 +       if (d_really_is_negative(dentry)) {
20851 +               err = -EEXIST;
20852 +               if (unlikely(d_is_positive(h_dentry)))
20853 +                       goto out;
20854 +       } else {
20855 +               /* rename(2) case */
20856 +               err = -EIO;
20857 +               if (unlikely(d_is_negative(h_dentry)))
20858 +                       goto out;
20859 +               h_inode = d_inode(h_dentry);
20860 +               if (unlikely(!h_inode->i_nlink))
20861 +                       goto out;
20862 +
20863 +               h_mode = h_inode->i_mode;
20864 +               if (!isdir) {
20865 +                       err = -EISDIR;
20866 +                       if (unlikely(S_ISDIR(h_mode)))
20867 +                               goto out;
20868 +               } else if (unlikely(!S_ISDIR(h_mode))) {
20869 +                       err = -ENOTDIR;
20870 +                       goto out;
20871 +               }
20872 +       }
20873 +
20874 +       err = 0;
20875 +       /* expected parent dir is locked */
20876 +       if (unlikely(h_parent != h_dentry->d_parent))
20877 +               err = -EIO;
20878 +
20879 +out:
20880 +       AuTraceErr(err);
20881 +       return err;
20882 +}
20883 +
20884 +/*
20885 + * initial procedure of adding a new entry.
20886 + * prepare writable branch and the parent dir, lock it,
20887 + * and lookup whiteout for the new entry.
20888 + */
20889 +static struct dentry*
20890 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
20891 +                 struct dentry *src_dentry, struct au_pin *pin,
20892 +                 struct au_wr_dir_args *wr_dir_args)
20893 +{
20894 +       struct dentry *wh_dentry, *h_parent;
20895 +       struct super_block *sb;
20896 +       struct au_branch *br;
20897 +       int err;
20898 +       unsigned int udba;
20899 +       aufs_bindex_t bcpup;
20900 +
20901 +       AuDbg("%pd\n", dentry);
20902 +
20903 +       err = au_wr_dir(dentry, src_dentry, wr_dir_args);
20904 +       bcpup = err;
20905 +       wh_dentry = ERR_PTR(err);
20906 +       if (unlikely(err < 0))
20907 +               goto out;
20908 +
20909 +       sb = dentry->d_sb;
20910 +       udba = au_opt_udba(sb);
20911 +       err = au_pin(pin, dentry, bcpup, udba,
20912 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20913 +       wh_dentry = ERR_PTR(err);
20914 +       if (unlikely(err))
20915 +               goto out;
20916 +
20917 +       h_parent = au_pinned_h_parent(pin);
20918 +       if (udba != AuOpt_UDBA_NONE
20919 +           && au_dbtop(dentry) == bcpup)
20920 +               err = au_may_add(dentry, bcpup, h_parent,
20921 +                                au_ftest_wrdir(wr_dir_args->flags, ISDIR));
20922 +       else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20923 +               err = -ENAMETOOLONG;
20924 +       wh_dentry = ERR_PTR(err);
20925 +       if (unlikely(err))
20926 +               goto out_unpin;
20927 +
20928 +       br = au_sbr(sb, bcpup);
20929 +       if (dt) {
20930 +               struct path tmp = {
20931 +                       .dentry = h_parent,
20932 +                       .mnt    = au_br_mnt(br)
20933 +               };
20934 +               au_dtime_store(dt, au_pinned_parent(pin), &tmp);
20935 +       }
20936 +
20937 +       wh_dentry = NULL;
20938 +       if (bcpup != au_dbwh(dentry))
20939 +               goto out; /* success */
20940 +
20941 +       /*
20942 +        * ENAMETOOLONG here means that if we allowed create such name, then it
20943 +        * would not be able to removed in the future. So we don't allow such
20944 +        * name here and we don't handle ENAMETOOLONG differently here.
20945 +        */
20946 +       wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
20947 +
20948 +out_unpin:
20949 +       if (IS_ERR(wh_dentry))
20950 +               au_unpin(pin);
20951 +out:
20952 +       return wh_dentry;
20953 +}
20954 +
20955 +/* ---------------------------------------------------------------------- */
20956 +
20957 +enum { Mknod, Symlink, Creat };
20958 +struct simple_arg {
20959 +       int type;
20960 +       union {
20961 +               struct {
20962 +                       umode_t                 mode;
20963 +                       bool                    want_excl;
20964 +                       bool                    try_aopen;
20965 +                       struct vfsub_aopen_args *aopen;
20966 +               } c;
20967 +               struct {
20968 +                       const char *symname;
20969 +               } s;
20970 +               struct {
20971 +                       umode_t mode;
20972 +                       dev_t dev;
20973 +               } m;
20974 +       } u;
20975 +};
20976 +
20977 +static int add_simple(struct inode *dir, struct dentry *dentry,
20978 +                     struct simple_arg *arg)
20979 +{
20980 +       int err, rerr;
20981 +       aufs_bindex_t btop;
20982 +       unsigned char created;
20983 +       const unsigned char try_aopen
20984 +               = (arg->type == Creat && arg->u.c.try_aopen);
20985 +       struct vfsub_aopen_args *aopen = arg->u.c.aopen;
20986 +       struct dentry *wh_dentry, *parent;
20987 +       struct inode *h_dir;
20988 +       struct super_block *sb;
20989 +       struct au_branch *br;
20990 +       /* to reduce stack size */
20991 +       struct {
20992 +               struct au_dtime dt;
20993 +               struct au_pin pin;
20994 +               struct path h_path;
20995 +               struct au_wr_dir_args wr_dir_args;
20996 +       } *a;
20997 +
20998 +       AuDbg("%pd\n", dentry);
20999 +       IMustLock(dir);
21000 +
21001 +       err = -ENOMEM;
21002 +       a = kmalloc(sizeof(*a), GFP_NOFS);
21003 +       if (unlikely(!a))
21004 +               goto out;
21005 +       a->wr_dir_args.force_btgt = -1;
21006 +       a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
21007 +
21008 +       parent = dentry->d_parent; /* dir inode is locked */
21009 +       if (!try_aopen) {
21010 +               err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21011 +               if (unlikely(err))
21012 +                       goto out_free;
21013 +       }
21014 +       err = au_d_may_add(dentry);
21015 +       if (unlikely(err))
21016 +               goto out_unlock;
21017 +       if (!try_aopen)
21018 +               di_write_lock_parent(parent);
21019 +       wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21020 +                                     &a->pin, &a->wr_dir_args);
21021 +       err = PTR_ERR(wh_dentry);
21022 +       if (IS_ERR(wh_dentry))
21023 +               goto out_parent;
21024 +
21025 +       btop = au_dbtop(dentry);
21026 +       sb = dentry->d_sb;
21027 +       br = au_sbr(sb, btop);
21028 +       a->h_path.dentry = au_h_dptr(dentry, btop);
21029 +       a->h_path.mnt = au_br_mnt(br);
21030 +       h_dir = au_pinned_h_dir(&a->pin);
21031 +       switch (arg->type) {
21032 +       case Creat:
21033 +               if (!try_aopen || !h_dir->i_op->atomic_open) {
21034 +                       err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
21035 +                                          arg->u.c.want_excl);
21036 +                       created = !err;
21037 +                       if (!err && try_aopen)
21038 +                               aopen->file->f_mode |= FMODE_CREATED;
21039 +               } else {
21040 +                       aopen->br = br;
21041 +                       err = vfsub_atomic_open(h_dir, a->h_path.dentry, aopen);
21042 +                       AuDbg("err %d\n", err);
21043 +                       AuDbgFile(aopen->file);
21044 +                       created = err >= 0
21045 +                               && !!(aopen->file->f_mode & FMODE_CREATED);
21046 +               }
21047 +               break;
21048 +       case Symlink:
21049 +               err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
21050 +               created = !err;
21051 +               break;
21052 +       case Mknod:
21053 +               err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
21054 +                                 arg->u.m.dev);
21055 +               created = !err;
21056 +               break;
21057 +       default:
21058 +               BUG();
21059 +       }
21060 +       if (unlikely(err < 0))
21061 +               goto out_unpin;
21062 +
21063 +       err = epilog(dir, btop, wh_dentry, dentry);
21064 +       if (!err)
21065 +               goto out_unpin; /* success */
21066 +
21067 +       /* revert */
21068 +       if (created /* && d_is_positive(a->h_path.dentry) */) {
21069 +               /* no delegation since it is just created */
21070 +               rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
21071 +                                   /*force*/0);
21072 +               if (rerr) {
21073 +                       AuIOErr("%pd revert failure(%d, %d)\n",
21074 +                               dentry, err, rerr);
21075 +                       err = -EIO;
21076 +               }
21077 +               au_dtime_revert(&a->dt);
21078 +       }
21079 +       if (try_aopen && h_dir->i_op->atomic_open
21080 +           && (aopen->file->f_mode & FMODE_OPENED))
21081 +               /* aopen->file is still opened */
21082 +               au_lcnt_dec(&aopen->br->br_nfiles);
21083 +
21084 +out_unpin:
21085 +       au_unpin(&a->pin);
21086 +       dput(wh_dentry);
21087 +out_parent:
21088 +       if (!try_aopen)
21089 +               di_write_unlock(parent);
21090 +out_unlock:
21091 +       if (unlikely(err)) {
21092 +               au_update_dbtop(dentry);
21093 +               d_drop(dentry);
21094 +       }
21095 +       if (!try_aopen)
21096 +               aufs_read_unlock(dentry, AuLock_DW);
21097 +out_free:
21098 +       au_kfree_rcu(a);
21099 +out:
21100 +       return err;
21101 +}
21102 +
21103 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
21104 +              struct dentry *dentry, umode_t mode, dev_t dev)
21105 +{
21106 +       struct simple_arg arg = {
21107 +               .type = Mknod,
21108 +               .u.m = {
21109 +                       .mode   = mode,
21110 +                       .dev    = dev
21111 +               }
21112 +       };
21113 +       return add_simple(dir, dentry, &arg);
21114 +}
21115 +
21116 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
21117 +                struct dentry *dentry, const char *symname)
21118 +{
21119 +       struct simple_arg arg = {
21120 +               .type = Symlink,
21121 +               .u.s.symname = symname
21122 +       };
21123 +       return add_simple(dir, dentry, &arg);
21124 +}
21125 +
21126 +int aufs_create(struct user_namespace *userns, struct inode *dir,
21127 +               struct dentry *dentry, umode_t mode, bool want_excl)
21128 +{
21129 +       struct simple_arg arg = {
21130 +               .type = Creat,
21131 +               .u.c = {
21132 +                       .mode           = mode,
21133 +                       .want_excl      = want_excl
21134 +               }
21135 +       };
21136 +       return add_simple(dir, dentry, &arg);
21137 +}
21138 +
21139 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
21140 +                      struct vfsub_aopen_args *aopen_args)
21141 +{
21142 +       struct simple_arg arg = {
21143 +               .type = Creat,
21144 +               .u.c = {
21145 +                       .mode           = aopen_args->create_mode,
21146 +                       .want_excl      = aopen_args->open_flag & O_EXCL,
21147 +                       .try_aopen      = true,
21148 +                       .aopen          = aopen_args
21149 +               }
21150 +       };
21151 +       return add_simple(dir, dentry, &arg);
21152 +}
21153 +
21154 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
21155 +                struct dentry *dentry, umode_t mode)
21156 +{
21157 +       int err;
21158 +       aufs_bindex_t bindex;
21159 +       struct super_block *sb;
21160 +       struct dentry *parent, *h_parent, *h_dentry;
21161 +       struct inode *h_dir, *inode;
21162 +       struct vfsmount *h_mnt;
21163 +       struct user_namespace *h_userns;
21164 +       struct au_wr_dir_args wr_dir_args = {
21165 +               .force_btgt     = -1,
21166 +               .flags          = AuWrDir_TMPFILE
21167 +       };
21168 +
21169 +       /* copy-up may happen */
21170 +       inode_lock(dir);
21171 +
21172 +       sb = dir->i_sb;
21173 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21174 +       if (unlikely(err))
21175 +               goto out;
21176 +
21177 +       err = au_di_init(dentry);
21178 +       if (unlikely(err))
21179 +               goto out_si;
21180 +
21181 +       err = -EBUSY;
21182 +       parent = d_find_any_alias(dir);
21183 +       AuDebugOn(!parent);
21184 +       di_write_lock_parent(parent);
21185 +       if (unlikely(d_inode(parent) != dir))
21186 +               goto out_parent;
21187 +
21188 +       err = au_digen_test(parent, au_sigen(sb));
21189 +       if (unlikely(err))
21190 +               goto out_parent;
21191 +
21192 +       bindex = au_dbtop(parent);
21193 +       au_set_dbtop(dentry, bindex);
21194 +       au_set_dbbot(dentry, bindex);
21195 +       err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
21196 +       bindex = err;
21197 +       if (unlikely(err < 0))
21198 +               goto out_parent;
21199 +
21200 +       err = -EOPNOTSUPP;
21201 +       h_dir = au_h_iptr(dir, bindex);
21202 +       if (unlikely(!h_dir->i_op->tmpfile))
21203 +               goto out_parent;
21204 +
21205 +       h_mnt = au_sbr_mnt(sb, bindex);
21206 +       err = vfsub_mnt_want_write(h_mnt);
21207 +       if (unlikely(err))
21208 +               goto out_parent;
21209 +
21210 +       h_userns = mnt_user_ns(h_mnt);
21211 +       h_parent = au_h_dptr(parent, bindex);
21212 +       h_dentry = vfs_tmpfile(h_userns, h_parent, mode, /*open_flag*/0);
21213 +       if (IS_ERR(h_dentry)) {
21214 +               err = PTR_ERR(h_dentry);
21215 +               goto out_mnt;
21216 +       }
21217 +
21218 +       au_set_dbtop(dentry, bindex);
21219 +       au_set_dbbot(dentry, bindex);
21220 +       au_set_h_dptr(dentry, bindex, dget(h_dentry));
21221 +       inode = au_new_inode(dentry, /*must_new*/1);
21222 +       if (IS_ERR(inode)) {
21223 +               err = PTR_ERR(inode);
21224 +               au_set_h_dptr(dentry, bindex, NULL);
21225 +               au_set_dbtop(dentry, -1);
21226 +               au_set_dbbot(dentry, -1);
21227 +       } else {
21228 +               if (!inode->i_nlink)
21229 +                       set_nlink(inode, 1);
21230 +               d_tmpfile(dentry, inode);
21231 +               au_di(dentry)->di_tmpfile = 1;
21232 +
21233 +               /* update without i_mutex */
21234 +               if (au_ibtop(dir) == au_dbtop(dentry))
21235 +                       au_cpup_attr_timesizes(dir);
21236 +       }
21237 +       dput(h_dentry);
21238 +
21239 +out_mnt:
21240 +       vfsub_mnt_drop_write(h_mnt);
21241 +out_parent:
21242 +       di_write_unlock(parent);
21243 +       dput(parent);
21244 +       di_write_unlock(dentry);
21245 +       if (unlikely(err)) {
21246 +               au_di_fin(dentry);
21247 +               dentry->d_fsdata = NULL;
21248 +       }
21249 +out_si:
21250 +       si_read_unlock(sb);
21251 +out:
21252 +       inode_unlock(dir);
21253 +       return err;
21254 +}
21255 +
21256 +/* ---------------------------------------------------------------------- */
21257 +
21258 +struct au_link_args {
21259 +       aufs_bindex_t bdst, bsrc;
21260 +       struct au_pin pin;
21261 +       struct path h_path;
21262 +       struct dentry *src_parent, *parent;
21263 +};
21264 +
21265 +static int au_cpup_before_link(struct dentry *src_dentry,
21266 +                              struct au_link_args *a)
21267 +{
21268 +       int err;
21269 +       struct dentry *h_src_dentry;
21270 +       struct au_cp_generic cpg = {
21271 +               .dentry = src_dentry,
21272 +               .bdst   = a->bdst,
21273 +               .bsrc   = a->bsrc,
21274 +               .len    = -1,
21275 +               .pin    = &a->pin,
21276 +               .flags  = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
21277 +       };
21278 +
21279 +       di_read_lock_parent(a->src_parent, AuLock_IR);
21280 +       err = au_test_and_cpup_dirs(src_dentry, a->bdst);
21281 +       if (unlikely(err))
21282 +               goto out;
21283 +
21284 +       h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
21285 +       err = au_pin(&a->pin, src_dentry, a->bdst,
21286 +                    au_opt_udba(src_dentry->d_sb),
21287 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21288 +       if (unlikely(err))
21289 +               goto out;
21290 +
21291 +       err = au_sio_cpup_simple(&cpg);
21292 +       au_unpin(&a->pin);
21293 +
21294 +out:
21295 +       di_read_unlock(a->src_parent, AuLock_IR);
21296 +       return err;
21297 +}
21298 +
21299 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
21300 +                          struct au_link_args *a)
21301 +{
21302 +       int err;
21303 +       unsigned char plink;
21304 +       aufs_bindex_t bbot;
21305 +       struct dentry *h_src_dentry;
21306 +       struct inode *h_inode, *inode, *delegated;
21307 +       struct super_block *sb;
21308 +       struct file *h_file;
21309 +
21310 +       plink = 0;
21311 +       h_inode = NULL;
21312 +       sb = src_dentry->d_sb;
21313 +       inode = d_inode(src_dentry);
21314 +       if (au_ibtop(inode) <= a->bdst)
21315 +               h_inode = au_h_iptr(inode, a->bdst);
21316 +       if (!h_inode || !h_inode->i_nlink) {
21317 +               /* copyup src_dentry as the name of dentry. */
21318 +               bbot = au_dbbot(dentry);
21319 +               if (bbot < a->bsrc)
21320 +                       au_set_dbbot(dentry, a->bsrc);
21321 +               au_set_h_dptr(dentry, a->bsrc,
21322 +                             dget(au_h_dptr(src_dentry, a->bsrc)));
21323 +               dget(a->h_path.dentry);
21324 +               au_set_h_dptr(dentry, a->bdst, NULL);
21325 +               AuDbg("temporary d_inode...\n");
21326 +               spin_lock(&dentry->d_lock);
21327 +               dentry->d_inode = d_inode(src_dentry); /* tmp */
21328 +               spin_unlock(&dentry->d_lock);
21329 +               h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
21330 +               if (IS_ERR(h_file))
21331 +                       err = PTR_ERR(h_file);
21332 +               else {
21333 +                       struct au_cp_generic cpg = {
21334 +                               .dentry = dentry,
21335 +                               .bdst   = a->bdst,
21336 +                               .bsrc   = -1,
21337 +                               .len    = -1,
21338 +                               .pin    = &a->pin,
21339 +                               .flags  = AuCpup_KEEPLINO
21340 +                       };
21341 +                       err = au_sio_cpup_simple(&cpg);
21342 +                       au_h_open_post(dentry, a->bsrc, h_file);
21343 +                       if (!err) {
21344 +                               dput(a->h_path.dentry);
21345 +                               a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21346 +                       } else
21347 +                               au_set_h_dptr(dentry, a->bdst,
21348 +                                             a->h_path.dentry);
21349 +               }
21350 +               spin_lock(&dentry->d_lock);
21351 +               dentry->d_inode = NULL; /* restore */
21352 +               spin_unlock(&dentry->d_lock);
21353 +               AuDbg("temporary d_inode...done\n");
21354 +               au_set_h_dptr(dentry, a->bsrc, NULL);
21355 +               au_set_dbbot(dentry, bbot);
21356 +       } else {
21357 +               /* the inode of src_dentry already exists on a.bdst branch */
21358 +               h_src_dentry = d_find_alias(h_inode);
21359 +               if (!h_src_dentry && au_plink_test(inode)) {
21360 +                       plink = 1;
21361 +                       h_src_dentry = au_plink_lkup(inode, a->bdst);
21362 +                       err = PTR_ERR(h_src_dentry);
21363 +                       if (IS_ERR(h_src_dentry))
21364 +                               goto out;
21365 +
21366 +                       if (unlikely(d_is_negative(h_src_dentry))) {
21367 +                               dput(h_src_dentry);
21368 +                               h_src_dentry = NULL;
21369 +                       }
21370 +
21371 +               }
21372 +               if (h_src_dentry) {
21373 +                       delegated = NULL;
21374 +                       err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21375 +                                        &a->h_path, &delegated);
21376 +                       if (unlikely(err == -EWOULDBLOCK)) {
21377 +                               pr_warn("cannot retry for NFSv4 delegation"
21378 +                                       " for an internal link\n");
21379 +                               iput(delegated);
21380 +                       }
21381 +                       dput(h_src_dentry);
21382 +               } else {
21383 +                       AuIOErr("no dentry found for hi%lu on b%d\n",
21384 +                               h_inode->i_ino, a->bdst);
21385 +                       err = -EIO;
21386 +               }
21387 +       }
21388 +
21389 +       if (!err && !plink)
21390 +               au_plink_append(inode, a->bdst, a->h_path.dentry);
21391 +
21392 +out:
21393 +       AuTraceErr(err);
21394 +       return err;
21395 +}
21396 +
21397 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
21398 +             struct dentry *dentry)
21399 +{
21400 +       int err, rerr;
21401 +       struct au_dtime dt;
21402 +       struct au_link_args *a;
21403 +       struct dentry *wh_dentry, *h_src_dentry;
21404 +       struct inode *inode, *delegated;
21405 +       struct super_block *sb;
21406 +       struct au_wr_dir_args wr_dir_args = {
21407 +               /* .force_btgt  = -1, */
21408 +               .flags          = AuWrDir_ADD_ENTRY
21409 +       };
21410 +
21411 +       IMustLock(dir);
21412 +       inode = d_inode(src_dentry);
21413 +       IMustLock(inode);
21414 +
21415 +       err = -ENOMEM;
21416 +       a = kzalloc(sizeof(*a), GFP_NOFS);
21417 +       if (unlikely(!a))
21418 +               goto out;
21419 +
21420 +       a->parent = dentry->d_parent; /* dir inode is locked */
21421 +       err = aufs_read_and_write_lock2(dentry, src_dentry,
21422 +                                       AuLock_NOPLM | AuLock_GEN);
21423 +       if (unlikely(err))
21424 +               goto out_kfree;
21425 +       err = au_d_linkable(src_dentry);
21426 +       if (unlikely(err))
21427 +               goto out_unlock;
21428 +       err = au_d_may_add(dentry);
21429 +       if (unlikely(err))
21430 +               goto out_unlock;
21431 +
21432 +       a->src_parent = dget_parent(src_dentry);
21433 +       wr_dir_args.force_btgt = au_ibtop(inode);
21434 +
21435 +       di_write_lock_parent(a->parent);
21436 +       wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
21437 +       wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
21438 +                                     &wr_dir_args);
21439 +       err = PTR_ERR(wh_dentry);
21440 +       if (IS_ERR(wh_dentry))
21441 +               goto out_parent;
21442 +
21443 +       err = 0;
21444 +       sb = dentry->d_sb;
21445 +       a->bdst = au_dbtop(dentry);
21446 +       a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21447 +       a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
21448 +       a->bsrc = au_ibtop(inode);
21449 +       h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21450 +       if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
21451 +               h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
21452 +       if (!h_src_dentry) {
21453 +               a->bsrc = au_dbtop(src_dentry);
21454 +               h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21455 +               AuDebugOn(!h_src_dentry);
21456 +       } else if (IS_ERR(h_src_dentry)) {
21457 +               err = PTR_ERR(h_src_dentry);
21458 +               goto out_parent;
21459 +       }
21460 +
21461 +       /*
21462 +        * aufs doesn't touch the credential so
21463 +        * security_dentry_create_files_as() is unnecessary.
21464 +        */
21465 +       if (au_opt_test(au_mntflags(sb), PLINK)) {
21466 +               if (a->bdst < a->bsrc
21467 +                   /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
21468 +                       err = au_cpup_or_link(src_dentry, dentry, a);
21469 +               else {
21470 +                       delegated = NULL;
21471 +                       err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21472 +                                        &a->h_path, &delegated);
21473 +                       if (unlikely(err == -EWOULDBLOCK)) {
21474 +                               pr_warn("cannot retry for NFSv4 delegation"
21475 +                                       " for an internal link\n");
21476 +                               iput(delegated);
21477 +                       }
21478 +               }
21479 +               dput(h_src_dentry);
21480 +       } else {
21481 +               /*
21482 +                * copyup src_dentry to the branch we process,
21483 +                * and then link(2) to it.
21484 +                */
21485 +               dput(h_src_dentry);
21486 +               if (a->bdst < a->bsrc
21487 +                   /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
21488 +                       au_unpin(&a->pin);
21489 +                       di_write_unlock(a->parent);
21490 +                       err = au_cpup_before_link(src_dentry, a);
21491 +                       di_write_lock_parent(a->parent);
21492 +                       if (!err)
21493 +                               err = au_pin(&a->pin, dentry, a->bdst,
21494 +                                            au_opt_udba(sb),
21495 +                                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21496 +                       if (unlikely(err))
21497 +                               goto out_wh;
21498 +               }
21499 +               if (!err) {
21500 +                       h_src_dentry = au_h_dptr(src_dentry, a->bdst);
21501 +                       err = -ENOENT;
21502 +                       if (h_src_dentry && d_is_positive(h_src_dentry)) {
21503 +                               delegated = NULL;
21504 +                               err = vfsub_link(h_src_dentry,
21505 +                                                au_pinned_h_dir(&a->pin),
21506 +                                                &a->h_path, &delegated);
21507 +                               if (unlikely(err == -EWOULDBLOCK)) {
21508 +                                       pr_warn("cannot retry"
21509 +                                               " for NFSv4 delegation"
21510 +                                               " for an internal link\n");
21511 +                                       iput(delegated);
21512 +                               }
21513 +                       }
21514 +               }
21515 +       }
21516 +       if (unlikely(err))
21517 +               goto out_unpin;
21518 +
21519 +       if (wh_dentry) {
21520 +               a->h_path.dentry = wh_dentry;
21521 +               err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
21522 +                                         dentry);
21523 +               if (unlikely(err))
21524 +                       goto out_revert;
21525 +       }
21526 +
21527 +       au_dir_ts(dir, a->bdst);
21528 +       inode_inc_iversion(dir);
21529 +       inc_nlink(inode);
21530 +       inode->i_ctime = dir->i_ctime;
21531 +       d_instantiate(dentry, au_igrab(inode));
21532 +       if (d_unhashed(a->h_path.dentry))
21533 +               /* some filesystem calls d_drop() */
21534 +               d_drop(dentry);
21535 +       /* some filesystems consume an inode even hardlink */
21536 +       au_fhsm_wrote(sb, a->bdst, /*force*/0);
21537 +       goto out_unpin; /* success */
21538 +
21539 +out_revert:
21540 +       /* no delegation since it is just created */
21541 +       rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
21542 +                           /*delegated*/NULL, /*force*/0);
21543 +       if (unlikely(rerr)) {
21544 +               AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
21545 +               err = -EIO;
21546 +       }
21547 +       au_dtime_revert(&dt);
21548 +out_unpin:
21549 +       au_unpin(&a->pin);
21550 +out_wh:
21551 +       dput(wh_dentry);
21552 +out_parent:
21553 +       di_write_unlock(a->parent);
21554 +       dput(a->src_parent);
21555 +out_unlock:
21556 +       if (unlikely(err)) {
21557 +               au_update_dbtop(dentry);
21558 +               d_drop(dentry);
21559 +       }
21560 +       aufs_read_and_write_unlock2(dentry, src_dentry);
21561 +out_kfree:
21562 +       au_kfree_rcu(a);
21563 +out:
21564 +       AuTraceErr(err);
21565 +       return err;
21566 +}
21567 +
21568 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
21569 +              struct dentry *dentry, umode_t mode)
21570 +{
21571 +       int err, rerr;
21572 +       aufs_bindex_t bindex;
21573 +       unsigned char diropq;
21574 +       struct path h_path;
21575 +       struct dentry *wh_dentry, *parent, *opq_dentry;
21576 +       struct inode *h_inode;
21577 +       struct super_block *sb;
21578 +       struct {
21579 +               struct au_pin pin;
21580 +               struct au_dtime dt;
21581 +       } *a; /* reduce the stack usage */
21582 +       struct au_wr_dir_args wr_dir_args = {
21583 +               .force_btgt     = -1,
21584 +               .flags          = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
21585 +       };
21586 +
21587 +       IMustLock(dir);
21588 +
21589 +       err = -ENOMEM;
21590 +       a = kmalloc(sizeof(*a), GFP_NOFS);
21591 +       if (unlikely(!a))
21592 +               goto out;
21593 +
21594 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21595 +       if (unlikely(err))
21596 +               goto out_free;
21597 +       err = au_d_may_add(dentry);
21598 +       if (unlikely(err))
21599 +               goto out_unlock;
21600 +
21601 +       parent = dentry->d_parent; /* dir inode is locked */
21602 +       di_write_lock_parent(parent);
21603 +       wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21604 +                                     &a->pin, &wr_dir_args);
21605 +       err = PTR_ERR(wh_dentry);
21606 +       if (IS_ERR(wh_dentry))
21607 +               goto out_parent;
21608 +
21609 +       sb = dentry->d_sb;
21610 +       bindex = au_dbtop(dentry);
21611 +       h_path.dentry = au_h_dptr(dentry, bindex);
21612 +       h_path.mnt = au_sbr_mnt(sb, bindex);
21613 +       err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
21614 +       if (unlikely(err))
21615 +               goto out_unpin;
21616 +
21617 +       /* make the dir opaque */
21618 +       diropq = 0;
21619 +       h_inode = d_inode(h_path.dentry);
21620 +       if (wh_dentry
21621 +           || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
21622 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
21623 +               opq_dentry = au_diropq_create(dentry, bindex);
21624 +               inode_unlock(h_inode);
21625 +               err = PTR_ERR(opq_dentry);
21626 +               if (IS_ERR(opq_dentry))
21627 +                       goto out_dir;
21628 +               dput(opq_dentry);
21629 +               diropq = 1;
21630 +       }
21631 +
21632 +       err = epilog(dir, bindex, wh_dentry, dentry);
21633 +       if (!err) {
21634 +               inc_nlink(dir);
21635 +               goto out_unpin; /* success */
21636 +       }
21637 +
21638 +       /* revert */
21639 +       if (diropq) {
21640 +               AuLabel(revert opq);
21641 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
21642 +               rerr = au_diropq_remove(dentry, bindex);
21643 +               inode_unlock(h_inode);
21644 +               if (rerr) {
21645 +                       AuIOErr("%pd reverting diropq failed(%d, %d)\n",
21646 +                               dentry, err, rerr);
21647 +                       err = -EIO;
21648 +               }
21649 +       }
21650 +
21651 +out_dir:
21652 +       AuLabel(revert dir);
21653 +       rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
21654 +       if (rerr) {
21655 +               AuIOErr("%pd reverting dir failed(%d, %d)\n",
21656 +                       dentry, err, rerr);
21657 +               err = -EIO;
21658 +       }
21659 +       au_dtime_revert(&a->dt);
21660 +out_unpin:
21661 +       au_unpin(&a->pin);
21662 +       dput(wh_dentry);
21663 +out_parent:
21664 +       di_write_unlock(parent);
21665 +out_unlock:
21666 +       if (unlikely(err)) {
21667 +               au_update_dbtop(dentry);
21668 +               d_drop(dentry);
21669 +       }
21670 +       aufs_read_unlock(dentry, AuLock_DW);
21671 +out_free:
21672 +       au_kfree_rcu(a);
21673 +out:
21674 +       return err;
21675 +}
21676 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
21677 --- /usr/share/empty/fs/aufs/i_op.c     1970-01-01 01:00:00.000000000 +0100
21678 +++ linux/fs/aufs/i_op.c        2022-11-05 23:02:18.965889284 +0100
21679 @@ -0,0 +1,1512 @@
21680 +// SPDX-License-Identifier: GPL-2.0
21681 +/*
21682 + * Copyright (C) 2005-2022 Junjiro R. Okajima
21683 + *
21684 + * This program is free software; you can redistribute it and/or modify
21685 + * it under the terms of the GNU General Public License as published by
21686 + * the Free Software Foundation; either version 2 of the License, or
21687 + * (at your option) any later version.
21688 + *
21689 + * This program is distributed in the hope that it will be useful,
21690 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21691 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
21692 + * GNU General Public License for more details.
21693 + *
21694 + * You should have received a copy of the GNU General Public License
21695 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
21696 + */
21697 +
21698 +/*
21699 + * inode operations (except add/del/rename)
21700 + */
21701 +
21702 +#include <linux/device_cgroup.h>
21703 +#include <linux/fs_stack.h>
21704 +#include <linux/iversion.h>
21705 +#include <linux/security.h>
21706 +#include "aufs.h"
21707 +
21708 +static int h_permission(struct inode *h_inode, int mask,
21709 +                       struct path *h_path, int brperm)
21710 +{
21711 +       int err;
21712 +       const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21713 +       struct user_namespace *h_userns;
21714 +
21715 +       err = -EPERM;
21716 +       if (write_mask && IS_IMMUTABLE(h_inode))
21717 +               goto out;
21718 +
21719 +       err = -EACCES;
21720 +       if (((mask & MAY_EXEC)
21721 +            && S_ISREG(h_inode->i_mode)
21722 +            && (path_noexec(h_path)
21723 +                || !(h_inode->i_mode & 0111))))
21724 +               goto out;
21725 +
21726 +       /*
21727 +        * - skip the lower fs test in the case of write to ro branch.
21728 +        * - nfs dir permission write check is optimized, but a policy for
21729 +        *   link/rename requires a real check.
21730 +        * - nfs always sets SB_POSIXACL regardless its mount option 'noacl.'
21731 +        *   in this case, generic_permission() returns -EOPNOTSUPP.
21732 +        */
21733 +       h_userns = mnt_user_ns(h_path->mnt);
21734 +       if ((write_mask && !au_br_writable(brperm))
21735 +           || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
21736 +               && write_mask && !(mask & MAY_READ))
21737 +           || !h_inode->i_op->permission) {
21738 +               /* AuLabel(generic_permission); */
21739 +               /* AuDbg("get_acl %ps\n", h_inode->i_op->get_acl); */
21740 +               err = generic_permission(h_userns, h_inode, mask);
21741 +               if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
21742 +                       err = h_inode->i_op->permission(h_userns, h_inode,
21743 +                                                       mask);
21744 +               AuTraceErr(err);
21745 +       } else {
21746 +               /* AuLabel(h_inode->permission); */
21747 +               err = h_inode->i_op->permission(h_userns, h_inode, mask);
21748 +               AuTraceErr(err);
21749 +       }
21750 +
21751 +       if (!err)
21752 +               err = devcgroup_inode_permission(h_inode, mask);
21753 +       if (!err)
21754 +               err = security_inode_permission(h_inode, mask);
21755 +
21756 +out:
21757 +       return err;
21758 +}
21759 +
21760 +static int aufs_permission(struct user_namespace *userns, struct inode *inode,
21761 +                          int mask)
21762 +{
21763 +       int err;
21764 +       aufs_bindex_t bindex, bbot;
21765 +       const unsigned char isdir = !!S_ISDIR(inode->i_mode),
21766 +               write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21767 +       struct inode *h_inode;
21768 +       struct super_block *sb;
21769 +       struct au_branch *br;
21770 +
21771 +       /* todo: support rcu-walk? */
21772 +       if (mask & MAY_NOT_BLOCK)
21773 +               return -ECHILD;
21774 +
21775 +       sb = inode->i_sb;
21776 +       si_read_lock(sb, AuLock_FLUSH);
21777 +       ii_read_lock_child(inode);
21778 +#if 0 /* reserved for future use */
21779 +       /*
21780 +        * This test may be rather 'too much' since the test is essentially done
21781 +        * in the aufs_lookup().  Theoretically it is possible that the inode
21782 +        * generation doesn't match to the superblock's here.  But it isn't a
21783 +        * big deal I suppose.
21784 +        */
21785 +       err = au_iigen_test(inode, au_sigen(sb));
21786 +       if (unlikely(err))
21787 +               goto out;
21788 +#endif
21789 +
21790 +       if (!isdir
21791 +           || write_mask
21792 +           || au_opt_test(au_mntflags(sb), DIRPERM1)) {
21793 +               err = au_busy_or_stale();
21794 +               h_inode = au_h_iptr(inode, au_ibtop(inode));
21795 +               if (unlikely(!h_inode
21796 +                            || (h_inode->i_mode & S_IFMT)
21797 +                            != (inode->i_mode & S_IFMT)))
21798 +                       goto out;
21799 +
21800 +               err = 0;
21801 +               bindex = au_ibtop(inode);
21802 +               br = au_sbr(sb, bindex);
21803 +               err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
21804 +               if (write_mask
21805 +                   && !err
21806 +                   && !special_file(h_inode->i_mode)) {
21807 +                       /* test whether the upper writable branch exists */
21808 +                       err = -EROFS;
21809 +                       for (; bindex >= 0; bindex--)
21810 +                               if (!au_br_rdonly(au_sbr(sb, bindex))) {
21811 +                                       err = 0;
21812 +                                       break;
21813 +                               }
21814 +               }
21815 +               goto out;
21816 +       }
21817 +
21818 +       /* non-write to dir */
21819 +       err = 0;
21820 +       bbot = au_ibbot(inode);
21821 +       for (bindex = au_ibtop(inode); !err && bindex <= bbot; bindex++) {
21822 +               h_inode = au_h_iptr(inode, bindex);
21823 +               if (h_inode) {
21824 +                       err = au_busy_or_stale();
21825 +                       if (unlikely(!S_ISDIR(h_inode->i_mode)))
21826 +                               break;
21827 +
21828 +                       br = au_sbr(sb, bindex);
21829 +                       err = h_permission(h_inode, mask, &br->br_path,
21830 +                                          br->br_perm);
21831 +               }
21832 +       }
21833 +
21834 +out:
21835 +       ii_read_unlock(inode);
21836 +       si_read_unlock(sb);
21837 +       return err;
21838 +}
21839 +
21840 +/* ---------------------------------------------------------------------- */
21841 +
21842 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
21843 +                                 unsigned int flags)
21844 +{
21845 +       struct dentry *ret, *parent;
21846 +       struct inode *inode;
21847 +       struct super_block *sb;
21848 +       int err, npositive;
21849 +
21850 +       IMustLock(dir);
21851 +
21852 +       /* todo: support rcu-walk? */
21853 +       ret = ERR_PTR(-ECHILD);
21854 +       if (flags & LOOKUP_RCU)
21855 +               goto out;
21856 +
21857 +       ret = ERR_PTR(-ENAMETOOLONG);
21858 +       if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
21859 +               goto out;
21860 +
21861 +       sb = dir->i_sb;
21862 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21863 +       ret = ERR_PTR(err);
21864 +       if (unlikely(err))
21865 +               goto out;
21866 +
21867 +       err = au_di_init(dentry);
21868 +       ret = ERR_PTR(err);
21869 +       if (unlikely(err))
21870 +               goto out_si;
21871 +
21872 +       inode = NULL;
21873 +       npositive = 0; /* suppress a warning */
21874 +       parent = dentry->d_parent; /* dir inode is locked */
21875 +       di_read_lock_parent(parent, AuLock_IR);
21876 +       err = au_alive_dir(parent);
21877 +       if (!err)
21878 +               err = au_digen_test(parent, au_sigen(sb));
21879 +       if (!err) {
21880 +               /* regardless LOOKUP_CREATE, always ALLOW_NEG */
21881 +               npositive = au_lkup_dentry(dentry, au_dbtop(parent),
21882 +                                          AuLkup_ALLOW_NEG);
21883 +               err = npositive;
21884 +       }
21885 +       di_read_unlock(parent, AuLock_IR);
21886 +       ret = ERR_PTR(err);
21887 +       if (unlikely(err < 0))
21888 +               goto out_unlock;
21889 +
21890 +       if (npositive) {
21891 +               inode = au_new_inode(dentry, /*must_new*/0);
21892 +               if (IS_ERR(inode)) {
21893 +                       ret = (void *)inode;
21894 +                       inode = NULL;
21895 +                       goto out_unlock;
21896 +               }
21897 +       }
21898 +
21899 +       if (inode)
21900 +               atomic_inc(&inode->i_count);
21901 +       ret = d_splice_alias(inode, dentry);
21902 +#if 0 /* reserved for future use */
21903 +       if (unlikely(d_need_lookup(dentry))) {
21904 +               spin_lock(&dentry->d_lock);
21905 +               dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
21906 +               spin_unlock(&dentry->d_lock);
21907 +       } else
21908 +#endif
21909 +       if (inode) {
21910 +               if (!IS_ERR(ret)) {
21911 +                       iput(inode);
21912 +                       if (ret && ret != dentry)
21913 +                               ii_write_unlock(inode);
21914 +               } else {
21915 +                       ii_write_unlock(inode);
21916 +                       iput(inode);
21917 +                       inode = NULL;
21918 +               }
21919 +       }
21920 +
21921 +out_unlock:
21922 +       di_write_unlock(dentry);
21923 +out_si:
21924 +       si_read_unlock(sb);
21925 +out:
21926 +       return ret;
21927 +}
21928 +
21929 +/* ---------------------------------------------------------------------- */
21930 +
21931 +/*
21932 + * very dirty and complicated aufs ->atomic_open().
21933 + * aufs_atomic_open()
21934 + * + au_aopen_or_create()
21935 + *   + add_simple()
21936 + *     + vfsub_atomic_open()
21937 + *       + branch fs ->atomic_open()
21938 + *        may call the actual 'open' for h_file
21939 + *       + inc br_nfiles only if opened
21940 + * + au_aopen_no_open() or au_aopen_do_open()
21941 + *
21942 + * au_aopen_do_open()
21943 + * + finish_open()
21944 + *   + au_do_aopen()
21945 + *     + au_do_open() the body of all 'open'
21946 + *       + au_do_open_nondir()
21947 + *        set the passed h_file
21948 + *
21949 + * au_aopen_no_open()
21950 + * + finish_no_open()
21951 + */
21952 +
21953 +struct aopen_node {
21954 +       struct hlist_bl_node hblist;
21955 +       struct file *file, *h_file;
21956 +};
21957 +
21958 +static int au_do_aopen(struct inode *inode, struct file *file)
21959 +{
21960 +       struct hlist_bl_head *aopen;
21961 +       struct hlist_bl_node *pos;
21962 +       struct aopen_node *node;
21963 +       struct au_do_open_args args = {
21964 +               .aopen  = 1,
21965 +               .open   = au_do_open_nondir
21966 +       };
21967 +
21968 +       aopen = &au_sbi(inode->i_sb)->si_aopen;
21969 +       hlist_bl_lock(aopen);
21970 +       hlist_bl_for_each_entry(node, pos, aopen, hblist)
21971 +               if (node->file == file) {
21972 +                       args.h_file = node->h_file;
21973 +                       break;
21974 +               }
21975 +       hlist_bl_unlock(aopen);
21976 +       /* AuDebugOn(!args.h_file); */
21977 +
21978 +       return au_do_open(file, &args);
21979 +}
21980 +
21981 +static int au_aopen_do_open(struct file *file, struct dentry *dentry,
21982 +                           struct aopen_node *aopen_node)
21983 +{
21984 +       int err;
21985 +       struct hlist_bl_head *aopen;
21986 +
21987 +       AuLabel(here);
21988 +       aopen = &au_sbi(dentry->d_sb)->si_aopen;
21989 +       au_hbl_add(&aopen_node->hblist, aopen);
21990 +       err = finish_open(file, dentry, au_do_aopen);
21991 +       au_hbl_del(&aopen_node->hblist, aopen);
21992 +       /* AuDbgFile(file); */
21993 +       AuDbg("%pd%s%s\n", dentry,
21994 +             (file->f_mode & FMODE_CREATED) ? " created" : "",
21995 +             (file->f_mode & FMODE_OPENED) ? " opened" : "");
21996 +
21997 +       AuTraceErr(err);
21998 +       return err;
21999 +}
22000 +
22001 +static int au_aopen_no_open(struct file *file, struct dentry *dentry)
22002 +{
22003 +       int err;
22004 +
22005 +       AuLabel(here);
22006 +       dget(dentry);
22007 +       err = finish_no_open(file, dentry);
22008 +
22009 +       AuTraceErr(err);
22010 +       return err;
22011 +}
22012 +
22013 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
22014 +                           struct file *file, unsigned int open_flag,
22015 +                           umode_t create_mode)
22016 +{
22017 +       int err, did_open;
22018 +       unsigned int lkup_flags;
22019 +       aufs_bindex_t bindex;
22020 +       struct super_block *sb;
22021 +       struct dentry *parent, *d;
22022 +       struct vfsub_aopen_args args = {
22023 +               .open_flag      = open_flag,
22024 +               .create_mode    = create_mode
22025 +       };
22026 +       struct aopen_node aopen_node = {
22027 +               .file   = file
22028 +       };
22029 +
22030 +       IMustLock(dir);
22031 +       AuDbg("open_flag 0%o\n", open_flag);
22032 +       AuDbgDentry(dentry);
22033 +
22034 +       err = 0;
22035 +       if (!au_di(dentry)) {
22036 +               lkup_flags = LOOKUP_OPEN;
22037 +               if (open_flag & O_CREAT)
22038 +                       lkup_flags |= LOOKUP_CREATE;
22039 +               d = aufs_lookup(dir, dentry, lkup_flags);
22040 +               if (IS_ERR(d)) {
22041 +                       err = PTR_ERR(d);
22042 +                       AuTraceErr(err);
22043 +                       goto out;
22044 +               } else if (d) {
22045 +                       /*
22046 +                        * obsoleted dentry found.
22047 +                        * another error will be returned later.
22048 +                        */
22049 +                       d_drop(d);
22050 +                       AuDbgDentry(d);
22051 +                       dput(d);
22052 +               }
22053 +               AuDbgDentry(dentry);
22054 +       }
22055 +
22056 +       if (d_is_positive(dentry)
22057 +           || d_unhashed(dentry)
22058 +           || d_unlinked(dentry)
22059 +           || !(open_flag & O_CREAT)) {
22060 +               err = au_aopen_no_open(file, dentry);
22061 +               goto out; /* success */
22062 +       }
22063 +
22064 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
22065 +       if (unlikely(err))
22066 +               goto out;
22067 +
22068 +       sb = dentry->d_sb;
22069 +       parent = dentry->d_parent;      /* dir is locked */
22070 +       di_write_lock_parent(parent);
22071 +       err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
22072 +       if (unlikely(err < 0))
22073 +               goto out_parent;
22074 +
22075 +       AuDbgDentry(dentry);
22076 +       if (d_is_positive(dentry)) {
22077 +               err = au_aopen_no_open(file, dentry);
22078 +               goto out_parent; /* success */
22079 +       }
22080 +
22081 +       args.file = alloc_empty_file(file->f_flags, current_cred());
22082 +       err = PTR_ERR(args.file);
22083 +       if (IS_ERR(args.file))
22084 +               goto out_parent;
22085 +
22086 +       bindex = au_dbtop(dentry);
22087 +       err = au_aopen_or_create(dir, dentry, &args);
22088 +       AuTraceErr(err);
22089 +       AuDbgFile(args.file);
22090 +       file->f_mode = args.file->f_mode & ~FMODE_OPENED;
22091 +       did_open = !!(args.file->f_mode & FMODE_OPENED);
22092 +       if (!did_open) {
22093 +               fput(args.file);
22094 +               args.file = NULL;
22095 +       }
22096 +       di_write_unlock(parent);
22097 +       di_write_unlock(dentry);
22098 +       if (unlikely(err < 0)) {
22099 +               if (args.file)
22100 +                       fput(args.file);
22101 +               goto out_sb;
22102 +       }
22103 +
22104 +       if (!did_open)
22105 +               err = au_aopen_no_open(file, dentry);
22106 +       else {
22107 +               aopen_node.h_file = args.file;
22108 +               err = au_aopen_do_open(file, dentry, &aopen_node);
22109 +       }
22110 +       if (unlikely(err < 0)) {
22111 +               if (args.file)
22112 +                       fput(args.file);
22113 +               if (did_open)
22114 +                       au_lcnt_dec(&args.br->br_nfiles);
22115 +       }
22116 +       goto out_sb; /* success */
22117 +
22118 +out_parent:
22119 +       di_write_unlock(parent);
22120 +       di_write_unlock(dentry);
22121 +out_sb:
22122 +       si_read_unlock(sb);
22123 +out:
22124 +       AuTraceErr(err);
22125 +       AuDbgFile(file);
22126 +       return err;
22127 +}
22128 +
22129 +
22130 +/* ---------------------------------------------------------------------- */
22131 +
22132 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
22133 +                         const unsigned char add_entry, aufs_bindex_t bcpup,
22134 +                         aufs_bindex_t btop)
22135 +{
22136 +       int err;
22137 +       struct dentry *h_parent;
22138 +       struct inode *h_dir;
22139 +
22140 +       if (add_entry)
22141 +               IMustLock(d_inode(parent));
22142 +       else
22143 +               di_write_lock_parent(parent);
22144 +
22145 +       err = 0;
22146 +       if (!au_h_dptr(parent, bcpup)) {
22147 +               if (btop > bcpup)
22148 +                       err = au_cpup_dirs(dentry, bcpup);
22149 +               else if (btop < bcpup)
22150 +                       err = au_cpdown_dirs(dentry, bcpup);
22151 +               else
22152 +                       BUG();
22153 +       }
22154 +       if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
22155 +               h_parent = au_h_dptr(parent, bcpup);
22156 +               h_dir = d_inode(h_parent);
22157 +               inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
22158 +               err = au_lkup_neg(dentry, bcpup, /*wh*/0);
22159 +               /* todo: no unlock here */
22160 +               inode_unlock_shared(h_dir);
22161 +
22162 +               AuDbg("bcpup %d\n", bcpup);
22163 +               if (!err) {
22164 +                       if (d_really_is_negative(dentry))
22165 +                               au_set_h_dptr(dentry, btop, NULL);
22166 +                       au_update_dbrange(dentry, /*do_put_zero*/0);
22167 +               }
22168 +       }
22169 +
22170 +       if (!add_entry)
22171 +               di_write_unlock(parent);
22172 +       if (!err)
22173 +               err = bcpup; /* success */
22174 +
22175 +       AuTraceErr(err);
22176 +       return err;
22177 +}
22178 +
22179 +/*
22180 + * decide the branch and the parent dir where we will create a new entry.
22181 + * returns new bindex or an error.
22182 + * copyup the parent dir if needed.
22183 + */
22184 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
22185 +             struct au_wr_dir_args *args)
22186 +{
22187 +       int err;
22188 +       unsigned int flags;
22189 +       aufs_bindex_t bcpup, btop, src_btop;
22190 +       const unsigned char add_entry
22191 +               = au_ftest_wrdir(args->flags, ADD_ENTRY)
22192 +               | au_ftest_wrdir(args->flags, TMPFILE);
22193 +       struct super_block *sb;
22194 +       struct dentry *parent;
22195 +       struct au_sbinfo *sbinfo;
22196 +
22197 +       sb = dentry->d_sb;
22198 +       sbinfo = au_sbi(sb);
22199 +       parent = dget_parent(dentry);
22200 +       btop = au_dbtop(dentry);
22201 +       bcpup = btop;
22202 +       if (args->force_btgt < 0) {
22203 +               if (src_dentry) {
22204 +                       src_btop = au_dbtop(src_dentry);
22205 +                       if (src_btop < btop)
22206 +                               bcpup = src_btop;
22207 +               } else if (add_entry) {
22208 +                       flags = 0;
22209 +                       if (au_ftest_wrdir(args->flags, ISDIR))
22210 +                               au_fset_wbr(flags, DIR);
22211 +                       err = AuWbrCreate(sbinfo, dentry, flags);
22212 +                       bcpup = err;
22213 +               }
22214 +
22215 +               if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
22216 +                       if (add_entry)
22217 +                               err = AuWbrCopyup(sbinfo, dentry);
22218 +                       else {
22219 +                               if (!IS_ROOT(dentry)) {
22220 +                                       di_read_lock_parent(parent, !AuLock_IR);
22221 +                                       err = AuWbrCopyup(sbinfo, dentry);
22222 +                                       di_read_unlock(parent, !AuLock_IR);
22223 +                               } else
22224 +                                       err = AuWbrCopyup(sbinfo, dentry);
22225 +                       }
22226 +                       bcpup = err;
22227 +                       if (unlikely(err < 0))
22228 +                               goto out;
22229 +               }
22230 +       } else {
22231 +               bcpup = args->force_btgt;
22232 +               AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
22233 +       }
22234 +
22235 +       AuDbg("btop %d, bcpup %d\n", btop, bcpup);
22236 +       err = bcpup;
22237 +       if (bcpup == btop)
22238 +               goto out; /* success */
22239 +
22240 +       /* copyup the new parent into the branch we process */
22241 +       err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, btop);
22242 +       if (err >= 0) {
22243 +               if (d_really_is_negative(dentry)) {
22244 +                       au_set_h_dptr(dentry, btop, NULL);
22245 +                       au_set_dbtop(dentry, bcpup);
22246 +                       au_set_dbbot(dentry, bcpup);
22247 +               }
22248 +               AuDebugOn(add_entry
22249 +                         && !au_ftest_wrdir(args->flags, TMPFILE)
22250 +                         && !au_h_dptr(dentry, bcpup));
22251 +       }
22252 +
22253 +out:
22254 +       dput(parent);
22255 +       return err;
22256 +}
22257 +
22258 +/* ---------------------------------------------------------------------- */
22259 +
22260 +void au_pin_hdir_unlock(struct au_pin *p)
22261 +{
22262 +       if (p->hdir)
22263 +               au_hn_inode_unlock(p->hdir);
22264 +}
22265 +
22266 +int au_pin_hdir_lock(struct au_pin *p)
22267 +{
22268 +       int err;
22269 +
22270 +       err = 0;
22271 +       if (!p->hdir)
22272 +               goto out;
22273 +
22274 +       /* even if an error happens later, keep this lock */
22275 +       au_hn_inode_lock_nested(p->hdir, p->lsc_hi);
22276 +
22277 +       err = -EBUSY;
22278 +       if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
22279 +               goto out;
22280 +
22281 +       err = 0;
22282 +       if (p->h_dentry)
22283 +               err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
22284 +                                 p->h_parent, p->br);
22285 +
22286 +out:
22287 +       return err;
22288 +}
22289 +
22290 +int au_pin_hdir_relock(struct au_pin *p)
22291 +{
22292 +       int err, i;
22293 +       struct inode *h_i;
22294 +       struct dentry *h_d[] = {
22295 +               p->h_dentry,
22296 +               p->h_parent
22297 +       };
22298 +
22299 +       err = au_pin_hdir_lock(p);
22300 +       if (unlikely(err))
22301 +               goto out;
22302 +
22303 +       for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
22304 +               if (!h_d[i])
22305 +                       continue;
22306 +               if (d_is_positive(h_d[i])) {
22307 +                       h_i = d_inode(h_d[i]);
22308 +                       err = !h_i->i_nlink;
22309 +               }
22310 +       }
22311 +
22312 +out:
22313 +       return err;
22314 +}
22315 +
22316 +static void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
22317 +{
22318 +       atomic_long_set(&p->hdir->hi_inode->i_rwsem.owner, (long)task);
22319 +}
22320 +
22321 +void au_pin_hdir_acquire_nest(struct au_pin *p)
22322 +{
22323 +       if (p->hdir) {
22324 +               rwsem_acquire_nest(&p->hdir->hi_inode->i_rwsem.dep_map,
22325 +                                  p->lsc_hi, 0, NULL, _RET_IP_);
22326 +               au_pin_hdir_set_owner(p, current);
22327 +       }
22328 +}
22329 +
22330 +void au_pin_hdir_release(struct au_pin *p)
22331 +{
22332 +       if (p->hdir) {
22333 +               au_pin_hdir_set_owner(p, p->task);
22334 +               rwsem_release(&p->hdir->hi_inode->i_rwsem.dep_map, _RET_IP_);
22335 +       }
22336 +}
22337 +
22338 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
22339 +{
22340 +       if (pin && pin->parent)
22341 +               return au_h_dptr(pin->parent, pin->bindex);
22342 +       return NULL;
22343 +}
22344 +
22345 +void au_unpin(struct au_pin *p)
22346 +{
22347 +       if (p->hdir)
22348 +               au_pin_hdir_unlock(p);
22349 +       if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
22350 +               vfsub_mnt_drop_write(p->h_mnt);
22351 +       if (!p->hdir)
22352 +               return;
22353 +
22354 +       if (!au_ftest_pin(p->flags, DI_LOCKED))
22355 +               di_read_unlock(p->parent, AuLock_IR);
22356 +       iput(p->hdir->hi_inode);
22357 +       dput(p->parent);
22358 +       p->parent = NULL;
22359 +       p->hdir = NULL;
22360 +       p->h_mnt = NULL;
22361 +       /* do not clear p->task */
22362 +}
22363 +
22364 +int au_do_pin(struct au_pin *p)
22365 +{
22366 +       int err;
22367 +       struct super_block *sb;
22368 +       struct inode *h_dir;
22369 +
22370 +       err = 0;
22371 +       sb = p->dentry->d_sb;
22372 +       p->br = au_sbr(sb, p->bindex);
22373 +       if (IS_ROOT(p->dentry)) {
22374 +               if (au_ftest_pin(p->flags, MNT_WRITE)) {
22375 +                       p->h_mnt = au_br_mnt(p->br);
22376 +                       err = vfsub_mnt_want_write(p->h_mnt);
22377 +                       if (unlikely(err)) {
22378 +                               au_fclr_pin(p->flags, MNT_WRITE);
22379 +                               goto out_err;
22380 +                       }
22381 +               }
22382 +               goto out;
22383 +       }
22384 +
22385 +       p->h_dentry = NULL;
22386 +       if (p->bindex <= au_dbbot(p->dentry))
22387 +               p->h_dentry = au_h_dptr(p->dentry, p->bindex);
22388 +
22389 +       p->parent = dget_parent(p->dentry);
22390 +       if (!au_ftest_pin(p->flags, DI_LOCKED))
22391 +               di_read_lock(p->parent, AuLock_IR, p->lsc_di);
22392 +
22393 +       h_dir = NULL;
22394 +       p->h_parent = au_h_dptr(p->parent, p->bindex);
22395 +       p->hdir = au_hi(d_inode(p->parent), p->bindex);
22396 +       if (p->hdir)
22397 +               h_dir = p->hdir->hi_inode;
22398 +
22399 +       /*
22400 +        * udba case, or
22401 +        * if DI_LOCKED is not set, then p->parent may be different
22402 +        * and h_parent can be NULL.
22403 +        */
22404 +       if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
22405 +               err = -EBUSY;
22406 +               if (!au_ftest_pin(p->flags, DI_LOCKED))
22407 +                       di_read_unlock(p->parent, AuLock_IR);
22408 +               dput(p->parent);
22409 +               p->parent = NULL;
22410 +               goto out_err;
22411 +       }
22412 +
22413 +       if (au_ftest_pin(p->flags, MNT_WRITE)) {
22414 +               p->h_mnt = au_br_mnt(p->br);
22415 +               err = vfsub_mnt_want_write(p->h_mnt);
22416 +               if (unlikely(err)) {
22417 +                       au_fclr_pin(p->flags, MNT_WRITE);
22418 +                       if (!au_ftest_pin(p->flags, DI_LOCKED))
22419 +                               di_read_unlock(p->parent, AuLock_IR);
22420 +                       dput(p->parent);
22421 +                       p->parent = NULL;
22422 +                       goto out_err;
22423 +               }
22424 +       }
22425 +
22426 +       au_igrab(h_dir);
22427 +       err = au_pin_hdir_lock(p);
22428 +       if (!err)
22429 +               goto out; /* success */
22430 +
22431 +       au_unpin(p);
22432 +
22433 +out_err:
22434 +       pr_err("err %d\n", err);
22435 +       err = au_busy_or_stale();
22436 +out:
22437 +       return err;
22438 +}
22439 +
22440 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
22441 +                aufs_bindex_t bindex, int lsc_di, int lsc_hi,
22442 +                unsigned int udba, unsigned char flags)
22443 +{
22444 +       p->dentry = dentry;
22445 +       p->udba = udba;
22446 +       p->lsc_di = lsc_di;
22447 +       p->lsc_hi = lsc_hi;
22448 +       p->flags = flags;
22449 +       p->bindex = bindex;
22450 +
22451 +       p->parent = NULL;
22452 +       p->hdir = NULL;
22453 +       p->h_mnt = NULL;
22454 +
22455 +       p->h_dentry = NULL;
22456 +       p->h_parent = NULL;
22457 +       p->br = NULL;
22458 +       p->task = current;
22459 +}
22460 +
22461 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
22462 +          unsigned int udba, unsigned char flags)
22463 +{
22464 +       au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
22465 +                   udba, flags);
22466 +       return au_do_pin(pin);
22467 +}
22468 +
22469 +/* ---------------------------------------------------------------------- */
22470 +
22471 +/*
22472 + * ->setattr() and ->getattr() are called in various cases.
22473 + * chmod, stat: dentry is revalidated.
22474 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
22475 + *               unhashed.
22476 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
22477 + */
22478 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
22479 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
22480 +{
22481 +       int err;
22482 +       struct dentry *parent;
22483 +
22484 +       err = 0;
22485 +       if (au_digen_test(dentry, sigen)) {
22486 +               parent = dget_parent(dentry);
22487 +               di_read_lock_parent(parent, AuLock_IR);
22488 +               err = au_refresh_dentry(dentry, parent);
22489 +               di_read_unlock(parent, AuLock_IR);
22490 +               dput(parent);
22491 +       }
22492 +
22493 +       AuTraceErr(err);
22494 +       return err;
22495 +}
22496 +
22497 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
22498 +                    struct au_icpup_args *a)
22499 +{
22500 +       int err;
22501 +       loff_t sz;
22502 +       aufs_bindex_t btop, ibtop;
22503 +       struct dentry *hi_wh, *parent;
22504 +       struct inode *inode;
22505 +       struct au_wr_dir_args wr_dir_args = {
22506 +               .force_btgt     = -1,
22507 +               .flags          = 0
22508 +       };
22509 +
22510 +       if (d_is_dir(dentry))
22511 +               au_fset_wrdir(wr_dir_args.flags, ISDIR);
22512 +       /* plink or hi_wh() case */
22513 +       btop = au_dbtop(dentry);
22514 +       inode = d_inode(dentry);
22515 +       ibtop = au_ibtop(inode);
22516 +       if (btop != ibtop && !au_test_ro(inode->i_sb, ibtop, inode))
22517 +               wr_dir_args.force_btgt = ibtop;
22518 +       err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
22519 +       if (unlikely(err < 0))
22520 +               goto out;
22521 +       a->btgt = err;
22522 +       if (err != btop)
22523 +               au_fset_icpup(a->flags, DID_CPUP);
22524 +
22525 +       err = 0;
22526 +       a->pin_flags = AuPin_MNT_WRITE;
22527 +       parent = NULL;
22528 +       if (!IS_ROOT(dentry)) {
22529 +               au_fset_pin(a->pin_flags, DI_LOCKED);
22530 +               parent = dget_parent(dentry);
22531 +               di_write_lock_parent(parent);
22532 +       }
22533 +
22534 +       err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
22535 +       if (unlikely(err))
22536 +               goto out_parent;
22537 +
22538 +       sz = -1;
22539 +       a->h_path.dentry = au_h_dptr(dentry, btop);
22540 +       a->h_inode = d_inode(a->h_path.dentry);
22541 +       if (ia && (ia->ia_valid & ATTR_SIZE)) {
22542 +               inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
22543 +               if (ia->ia_size < i_size_read(a->h_inode))
22544 +                       sz = ia->ia_size;
22545 +               inode_unlock_shared(a->h_inode);
22546 +       }
22547 +
22548 +       hi_wh = NULL;
22549 +       if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
22550 +               hi_wh = au_hi_wh(inode, a->btgt);
22551 +               if (!hi_wh) {
22552 +                       struct au_cp_generic cpg = {
22553 +                               .dentry = dentry,
22554 +                               .bdst   = a->btgt,
22555 +                               .bsrc   = -1,
22556 +                               .len    = sz,
22557 +                               .pin    = &a->pin
22558 +                       };
22559 +                       err = au_sio_cpup_wh(&cpg, /*file*/NULL);
22560 +                       if (unlikely(err))
22561 +                               goto out_unlock;
22562 +                       hi_wh = au_hi_wh(inode, a->btgt);
22563 +                       /* todo: revalidate hi_wh? */
22564 +               }
22565 +       }
22566 +
22567 +       if (parent) {
22568 +               au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
22569 +               di_downgrade_lock(parent, AuLock_IR);
22570 +               dput(parent);
22571 +               parent = NULL;
22572 +       }
22573 +       if (!au_ftest_icpup(a->flags, DID_CPUP))
22574 +               goto out; /* success */
22575 +
22576 +       if (!d_unhashed(dentry)) {
22577 +               struct au_cp_generic cpg = {
22578 +                       .dentry = dentry,
22579 +                       .bdst   = a->btgt,
22580 +                       .bsrc   = btop,
22581 +                       .len    = sz,
22582 +                       .pin    = &a->pin,
22583 +                       .flags  = AuCpup_DTIME | AuCpup_HOPEN
22584 +               };
22585 +               err = au_sio_cpup_simple(&cpg);
22586 +               if (!err)
22587 +                       a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22588 +       } else if (!hi_wh)
22589 +               a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22590 +       else
22591 +               a->h_path.dentry = hi_wh; /* do not dget here */
22592 +
22593 +out_unlock:
22594 +       a->h_inode = d_inode(a->h_path.dentry);
22595 +       if (!err)
22596 +               goto out; /* success */
22597 +       au_unpin(&a->pin);
22598 +out_parent:
22599 +       if (parent) {
22600 +               di_write_unlock(parent);
22601 +               dput(parent);
22602 +       }
22603 +out:
22604 +       if (!err)
22605 +               inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22606 +       return err;
22607 +}
22608 +
22609 +static int aufs_setattr(struct user_namespace *userns, struct dentry *dentry,
22610 +                       struct iattr *ia)
22611 +{
22612 +       int err;
22613 +       struct inode *inode, *delegated;
22614 +       struct super_block *sb;
22615 +       struct file *file;
22616 +       struct au_icpup_args *a;
22617 +       struct user_namespace *h_userns;
22618 +
22619 +       inode = d_inode(dentry);
22620 +       IMustLock(inode);
22621 +
22622 +       err = setattr_prepare(userns, dentry, ia);
22623 +       if (unlikely(err))
22624 +               goto out;
22625 +
22626 +       err = -ENOMEM;
22627 +       a = kzalloc(sizeof(*a), GFP_NOFS);
22628 +       if (unlikely(!a))
22629 +               goto out;
22630 +
22631 +       if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
22632 +               ia->ia_valid &= ~ATTR_MODE;
22633 +
22634 +       file = NULL;
22635 +       sb = dentry->d_sb;
22636 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22637 +       if (unlikely(err))
22638 +               goto out_kfree;
22639 +
22640 +       if (ia->ia_valid & ATTR_FILE) {
22641 +               /* currently ftruncate(2) only */
22642 +               AuDebugOn(!d_is_reg(dentry));
22643 +               file = ia->ia_file;
22644 +               err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1,
22645 +                                           /*fi_lsc*/0);
22646 +               if (unlikely(err))
22647 +                       goto out_si;
22648 +               ia->ia_file = au_hf_top(file);
22649 +               a->udba = AuOpt_UDBA_NONE;
22650 +       } else {
22651 +               /* fchmod() doesn't pass ia_file */
22652 +               a->udba = au_opt_udba(sb);
22653 +               di_write_lock_child(dentry);
22654 +               /* no d_unlinked(), to set UDBA_NONE for root */
22655 +               if (d_unhashed(dentry))
22656 +                       a->udba = AuOpt_UDBA_NONE;
22657 +               if (a->udba != AuOpt_UDBA_NONE) {
22658 +                       AuDebugOn(IS_ROOT(dentry));
22659 +                       err = au_reval_for_attr(dentry, au_sigen(sb));
22660 +                       if (unlikely(err))
22661 +                               goto out_dentry;
22662 +               }
22663 +       }
22664 +
22665 +       err = au_pin_and_icpup(dentry, ia, a);
22666 +       if (unlikely(err < 0))
22667 +               goto out_dentry;
22668 +       if (au_ftest_icpup(a->flags, DID_CPUP)) {
22669 +               ia->ia_file = NULL;
22670 +               ia->ia_valid &= ~ATTR_FILE;
22671 +       }
22672 +
22673 +       a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
22674 +       if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
22675 +           == (ATTR_MODE | ATTR_CTIME)) {
22676 +               err = security_path_chmod(&a->h_path, ia->ia_mode);
22677 +               if (unlikely(err))
22678 +                       goto out_unlock;
22679 +       } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
22680 +                  && (ia->ia_valid & ATTR_CTIME)) {
22681 +               err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
22682 +               if (unlikely(err))
22683 +                       goto out_unlock;
22684 +       }
22685 +
22686 +       if (ia->ia_valid & ATTR_SIZE) {
22687 +               struct file *f;
22688 +
22689 +               if (ia->ia_size < i_size_read(inode))
22690 +                       /* unmap only */
22691 +                       truncate_setsize(inode, ia->ia_size);
22692 +
22693 +               f = NULL;
22694 +               if (ia->ia_valid & ATTR_FILE)
22695 +                       f = ia->ia_file;
22696 +               inode_unlock(a->h_inode);
22697 +               err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
22698 +               inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22699 +       } else {
22700 +               delegated = NULL;
22701 +               while (1) {
22702 +                       err = vfsub_notify_change(&a->h_path, ia, &delegated);
22703 +                       if (delegated) {
22704 +                               err = break_deleg_wait(&delegated);
22705 +                               if (!err)
22706 +                                       continue;
22707 +                       }
22708 +                       break;
22709 +               }
22710 +       }
22711 +       /*
22712 +        * regardless aufs 'acl' option setting.
22713 +        * why don't all acl-aware fs call this func from their ->setattr()?
22714 +        */
22715 +       if (!err && (ia->ia_valid & ATTR_MODE)) {
22716 +               h_userns = mnt_user_ns(a->h_path.mnt);
22717 +               err = vfsub_acl_chmod(h_userns, a->h_inode, ia->ia_mode);
22718 +       }
22719 +       if (!err)
22720 +               au_cpup_attr_changeable(inode);
22721 +
22722 +out_unlock:
22723 +       inode_unlock(a->h_inode);
22724 +       au_unpin(&a->pin);
22725 +       if (unlikely(err))
22726 +               au_update_dbtop(dentry);
22727 +out_dentry:
22728 +       di_write_unlock(dentry);
22729 +       if (file) {
22730 +               fi_write_unlock(file);
22731 +               ia->ia_file = file;
22732 +               ia->ia_valid |= ATTR_FILE;
22733 +       }
22734 +out_si:
22735 +       si_read_unlock(sb);
22736 +out_kfree:
22737 +       au_kfree_rcu(a);
22738 +out:
22739 +       AuTraceErr(err);
22740 +       return err;
22741 +}
22742 +
22743 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
22744 +static int au_h_path_to_set_attr(struct dentry *dentry,
22745 +                                struct au_icpup_args *a, struct path *h_path)
22746 +{
22747 +       int err;
22748 +       struct super_block *sb;
22749 +
22750 +       sb = dentry->d_sb;
22751 +       a->udba = au_opt_udba(sb);
22752 +       /* no d_unlinked(), to set UDBA_NONE for root */
22753 +       if (d_unhashed(dentry))
22754 +               a->udba = AuOpt_UDBA_NONE;
22755 +       if (a->udba != AuOpt_UDBA_NONE) {
22756 +               AuDebugOn(IS_ROOT(dentry));
22757 +               err = au_reval_for_attr(dentry, au_sigen(sb));
22758 +               if (unlikely(err))
22759 +                       goto out;
22760 +       }
22761 +       err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
22762 +       if (unlikely(err < 0))
22763 +               goto out;
22764 +
22765 +       h_path->dentry = a->h_path.dentry;
22766 +       h_path->mnt = au_sbr_mnt(sb, a->btgt);
22767 +
22768 +out:
22769 +       return err;
22770 +}
22771 +
22772 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
22773 +                 struct au_sxattr *arg)
22774 +{
22775 +       int err;
22776 +       struct path h_path;
22777 +       struct super_block *sb;
22778 +       struct au_icpup_args *a;
22779 +       struct inode *h_inode;
22780 +       struct user_namespace *h_userns;
22781 +
22782 +       IMustLock(inode);
22783 +
22784 +       err = -ENOMEM;
22785 +       a = kzalloc(sizeof(*a), GFP_NOFS);
22786 +       if (unlikely(!a))
22787 +               goto out;
22788 +
22789 +       sb = dentry->d_sb;
22790 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22791 +       if (unlikely(err))
22792 +               goto out_kfree;
22793 +
22794 +       h_path.dentry = NULL;   /* silence gcc */
22795 +       di_write_lock_child(dentry);
22796 +       err = au_h_path_to_set_attr(dentry, a, &h_path);
22797 +       if (unlikely(err))
22798 +               goto out_di;
22799 +       h_userns = mnt_user_ns(h_path.mnt);
22800 +
22801 +       inode_unlock(a->h_inode);
22802 +       switch (arg->type) {
22803 +       case AU_XATTR_SET:
22804 +               AuDebugOn(d_is_negative(h_path.dentry));
22805 +               err = vfsub_setxattr(h_userns, h_path.dentry,
22806 +                                    arg->u.set.name, arg->u.set.value,
22807 +                                    arg->u.set.size, arg->u.set.flags);
22808 +               break;
22809 +       case AU_ACL_SET:
22810 +               err = -EOPNOTSUPP;
22811 +               h_inode = d_inode(h_path.dentry);
22812 +               if (h_inode->i_op->set_acl) {
22813 +                       /* this will call posix_acl_update_mode */
22814 +                       err = h_inode->i_op->set_acl(h_userns, h_inode,
22815 +                                                    arg->u.acl_set.acl,
22816 +                                                    arg->u.acl_set.type);
22817 +               }
22818 +               break;
22819 +       }
22820 +       if (!err)
22821 +               au_cpup_attr_timesizes(inode);
22822 +
22823 +       au_unpin(&a->pin);
22824 +       if (unlikely(err))
22825 +               au_update_dbtop(dentry);
22826 +
22827 +out_di:
22828 +       di_write_unlock(dentry);
22829 +       si_read_unlock(sb);
22830 +out_kfree:
22831 +       au_kfree_rcu(a);
22832 +out:
22833 +       AuTraceErr(err);
22834 +       return err;
22835 +}
22836 +#endif
22837 +
22838 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
22839 +                            unsigned int nlink)
22840 +{
22841 +       unsigned int n;
22842 +
22843 +       inode->i_mode = st->mode;
22844 +       /* don't i_[ug]id_write() here */
22845 +       inode->i_uid = st->uid;
22846 +       inode->i_gid = st->gid;
22847 +       inode->i_atime = st->atime;
22848 +       inode->i_mtime = st->mtime;
22849 +       inode->i_ctime = st->ctime;
22850 +
22851 +       au_cpup_attr_nlink(inode, /*force*/0);
22852 +       if (S_ISDIR(inode->i_mode)) {
22853 +               n = inode->i_nlink;
22854 +               n -= nlink;
22855 +               n += st->nlink;
22856 +               smp_mb(); /* for i_nlink */
22857 +               /* 0 can happen */
22858 +               set_nlink(inode, n);
22859 +       }
22860 +
22861 +       spin_lock(&inode->i_lock);
22862 +       inode->i_blocks = st->blocks;
22863 +       i_size_write(inode, st->size);
22864 +       spin_unlock(&inode->i_lock);
22865 +}
22866 +
22867 +/*
22868 + * common routine for aufs_getattr() and au_getxattr().
22869 + * returns zero or negative (an error).
22870 + * @dentry will be read-locked in success.
22871 + */
22872 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
22873 +                     struct path *h_path, int locked)
22874 +{
22875 +       int err;
22876 +       unsigned int mnt_flags, sigen;
22877 +       unsigned char udba_none;
22878 +       aufs_bindex_t bindex;
22879 +       struct super_block *sb, *h_sb;
22880 +
22881 +       h_path->mnt = NULL;
22882 +       h_path->dentry = NULL;
22883 +
22884 +       err = 0;
22885 +       sb = dentry->d_sb;
22886 +       mnt_flags = au_mntflags(sb);
22887 +       udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
22888 +
22889 +       if (unlikely(locked))
22890 +               goto body; /* skip locking dinfo */
22891 +
22892 +       /* support fstat(2) */
22893 +       if (!d_unlinked(dentry) && !udba_none) {
22894 +               sigen = au_sigen(sb);
22895 +               err = au_digen_test(dentry, sigen);
22896 +               if (!err) {
22897 +                       di_read_lock_child(dentry, AuLock_IR);
22898 +                       err = au_dbrange_test(dentry);
22899 +                       if (unlikely(err)) {
22900 +                               di_read_unlock(dentry, AuLock_IR);
22901 +                               goto out;
22902 +                       }
22903 +               } else {
22904 +                       AuDebugOn(IS_ROOT(dentry));
22905 +                       di_write_lock_child(dentry);
22906 +                       err = au_dbrange_test(dentry);
22907 +                       if (!err)
22908 +                               err = au_reval_for_attr(dentry, sigen);
22909 +                       if (!err)
22910 +                               di_downgrade_lock(dentry, AuLock_IR);
22911 +                       else {
22912 +                               di_write_unlock(dentry);
22913 +                               goto out;
22914 +                       }
22915 +               }
22916 +       } else
22917 +               di_read_lock_child(dentry, AuLock_IR);
22918 +
22919 +body:
22920 +       if (!inode) {
22921 +               inode = d_inode(dentry);
22922 +               if (unlikely(!inode))
22923 +                       goto out;
22924 +       }
22925 +       bindex = au_ibtop(inode);
22926 +       h_path->mnt = au_sbr_mnt(sb, bindex);
22927 +       h_sb = h_path->mnt->mnt_sb;
22928 +       if (!force
22929 +           && !au_test_fs_bad_iattr(h_sb)
22930 +           && udba_none)
22931 +               goto out; /* success */
22932 +
22933 +       if (au_dbtop(dentry) == bindex)
22934 +               h_path->dentry = au_h_dptr(dentry, bindex);
22935 +       else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
22936 +               h_path->dentry = au_plink_lkup(inode, bindex);
22937 +               if (IS_ERR(h_path->dentry))
22938 +                       /* pretending success */
22939 +                       h_path->dentry = NULL;
22940 +               else
22941 +                       dput(h_path->dentry);
22942 +       }
22943 +
22944 +out:
22945 +       return err;
22946 +}
22947 +
22948 +static int aufs_getattr(struct user_namespace *userns, const struct path *path,
22949 +                       struct kstat *st, u32 request, unsigned int query)
22950 +{
22951 +       int err;
22952 +       unsigned char positive;
22953 +       struct path h_path;
22954 +       struct dentry *dentry;
22955 +       struct inode *inode;
22956 +       struct super_block *sb;
22957 +
22958 +       dentry = path->dentry;
22959 +       inode = d_inode(dentry);
22960 +       sb = dentry->d_sb;
22961 +       err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22962 +       if (unlikely(err))
22963 +               goto out;
22964 +       err = au_h_path_getattr(dentry, /*inode*/NULL, /*force*/0, &h_path,
22965 +                               /*locked*/0);
22966 +       if (unlikely(err))
22967 +               goto out_si;
22968 +       if (unlikely(!h_path.dentry))
22969 +               /* illegally overlapped or something */
22970 +               goto out_fill; /* pretending success */
22971 +
22972 +       positive = d_is_positive(h_path.dentry);
22973 +       if (positive)
22974 +               /* no vfsub version */
22975 +               err = vfs_getattr(&h_path, st, request, query);
22976 +       if (!err) {
22977 +               if (positive)
22978 +                       au_refresh_iattr(inode, st,
22979 +                                        d_inode(h_path.dentry)->i_nlink);
22980 +               goto out_fill; /* success */
22981 +       }
22982 +       AuTraceErr(err);
22983 +       goto out_di;
22984 +
22985 +out_fill:
22986 +       generic_fillattr(userns, inode, st);
22987 +out_di:
22988 +       di_read_unlock(dentry, AuLock_IR);
22989 +out_si:
22990 +       si_read_unlock(sb);
22991 +out:
22992 +       AuTraceErr(err);
22993 +       return err;
22994 +}
22995 +
22996 +/* ---------------------------------------------------------------------- */
22997 +
22998 +static const char *aufs_get_link(struct dentry *dentry, struct inode *inode,
22999 +                                struct delayed_call *done)
23000 +{
23001 +       const char *ret;
23002 +       struct dentry *h_dentry;
23003 +       struct inode *h_inode;
23004 +       int err;
23005 +       aufs_bindex_t bindex;
23006 +
23007 +       ret = NULL; /* suppress a warning */
23008 +       err = -ECHILD;
23009 +       if (!dentry)
23010 +               goto out;
23011 +
23012 +       err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
23013 +       if (unlikely(err))
23014 +               goto out;
23015 +
23016 +       err = au_d_hashed_positive(dentry);
23017 +       if (unlikely(err))
23018 +               goto out_unlock;
23019 +
23020 +       err = -EINVAL;
23021 +       inode = d_inode(dentry);
23022 +       bindex = au_ibtop(inode);
23023 +       h_inode = au_h_iptr(inode, bindex);
23024 +       if (unlikely(!h_inode->i_op->get_link))
23025 +               goto out_unlock;
23026 +
23027 +       err = -EBUSY;
23028 +       h_dentry = NULL;
23029 +       if (au_dbtop(dentry) <= bindex) {
23030 +               h_dentry = au_h_dptr(dentry, bindex);
23031 +               if (h_dentry)
23032 +                       dget(h_dentry);
23033 +       }
23034 +       if (!h_dentry) {
23035 +               h_dentry = d_find_any_alias(h_inode);
23036 +               if (IS_ERR(h_dentry)) {
23037 +                       err = PTR_ERR(h_dentry);
23038 +                       goto out_unlock;
23039 +               }
23040 +       }
23041 +       if (unlikely(!h_dentry))
23042 +               goto out_unlock;
23043 +
23044 +       err = 0;
23045 +       AuDbg("%ps\n", h_inode->i_op->get_link);
23046 +       AuDbgDentry(h_dentry);
23047 +       ret = vfs_get_link(h_dentry, done);
23048 +       dput(h_dentry);
23049 +       if (IS_ERR(ret))
23050 +               err = PTR_ERR(ret);
23051 +
23052 +out_unlock:
23053 +       aufs_read_unlock(dentry, AuLock_IR);
23054 +out:
23055 +       if (unlikely(err))
23056 +               ret = ERR_PTR(err);
23057 +       AuTraceErrPtr(ret);
23058 +       return ret;
23059 +}
23060 +
23061 +/* ---------------------------------------------------------------------- */
23062 +
23063 +static int au_is_special(struct inode *inode)
23064 +{
23065 +       return (inode->i_mode & (S_IFBLK | S_IFCHR | S_IFIFO | S_IFSOCK));
23066 +}
23067 +
23068 +static int aufs_update_time(struct inode *inode, struct timespec64 *ts,
23069 +                           int flags)
23070 +{
23071 +       int err;
23072 +       aufs_bindex_t bindex;
23073 +       struct super_block *sb;
23074 +       struct inode *h_inode;
23075 +       struct vfsmount *h_mnt;
23076 +
23077 +       sb = inode->i_sb;
23078 +       WARN_ONCE((flags & S_ATIME) && !IS_NOATIME(inode),
23079 +                 "unexpected s_flags 0x%lx", sb->s_flags);
23080 +
23081 +       /* mmap_sem might be acquired already, cf. aufs_mmap() */
23082 +       lockdep_off();
23083 +       si_read_lock(sb, AuLock_FLUSH);
23084 +       ii_write_lock_child(inode);
23085 +
23086 +       err = 0;
23087 +       bindex = au_ibtop(inode);
23088 +       h_inode = au_h_iptr(inode, bindex);
23089 +       if (!au_test_ro(sb, bindex, inode)) {
23090 +               h_mnt = au_sbr_mnt(sb, bindex);
23091 +               err = vfsub_mnt_want_write(h_mnt);
23092 +               if (!err) {
23093 +                       err = vfsub_update_time(h_inode, ts, flags);
23094 +                       vfsub_mnt_drop_write(h_mnt);
23095 +               }
23096 +       } else if (au_is_special(h_inode)) {
23097 +               /*
23098 +                * Never copy-up here.
23099 +                * These special files may already be opened and used for
23100 +                * communicating. If we copied it up, then the communication
23101 +                * would be corrupted.
23102 +                */
23103 +               AuWarn1("timestamps for i%lu are ignored "
23104 +                       "since it is on readonly branch (hi%lu).\n",
23105 +                       inode->i_ino, h_inode->i_ino);
23106 +       } else if (flags & ~S_ATIME) {
23107 +               err = -EIO;
23108 +               AuIOErr1("unexpected flags 0x%x\n", flags);
23109 +               AuDebugOn(1);
23110 +       }
23111 +
23112 +       if (!err)
23113 +               au_cpup_attr_timesizes(inode);
23114 +       ii_write_unlock(inode);
23115 +       si_read_unlock(sb);
23116 +       lockdep_on();
23117 +
23118 +       if (!err && (flags & S_VERSION))
23119 +               inode_inc_iversion(inode);
23120 +
23121 +       return err;
23122 +}
23123 +
23124 +/* ---------------------------------------------------------------------- */
23125 +
23126 +/* no getattr version will be set by module.c:aufs_init() */
23127 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
23128 +       aufs_iop[] = {
23129 +       [AuIop_SYMLINK] = {
23130 +               .permission     = aufs_permission,
23131 +#ifdef CONFIG_FS_POSIX_ACL
23132 +               .get_acl        = aufs_get_acl,
23133 +               .set_acl        = aufs_set_acl, /* unsupport for symlink? */
23134 +#endif
23135 +
23136 +               .setattr        = aufs_setattr,
23137 +               .getattr        = aufs_getattr,
23138 +
23139 +#ifdef CONFIG_AUFS_XATTR
23140 +               .listxattr      = aufs_listxattr,
23141 +#endif
23142 +
23143 +               .get_link       = aufs_get_link
23144 +
23145 +               /* .update_time = aufs_update_time */
23146 +       },
23147 +       [AuIop_DIR] = {
23148 +               .create         = aufs_create,
23149 +               .lookup         = aufs_lookup,
23150 +               .link           = aufs_link,
23151 +               .unlink         = aufs_unlink,
23152 +               .symlink        = aufs_symlink,
23153 +               .mkdir          = aufs_mkdir,
23154 +               .rmdir          = aufs_rmdir,
23155 +               .mknod          = aufs_mknod,
23156 +               .rename         = aufs_rename,
23157 +
23158 +               .permission     = aufs_permission,
23159 +#ifdef CONFIG_FS_POSIX_ACL
23160 +               .get_acl        = aufs_get_acl,
23161 +               .set_acl        = aufs_set_acl,
23162 +#endif
23163 +
23164 +               .setattr        = aufs_setattr,
23165 +               .getattr        = aufs_getattr,
23166 +
23167 +#ifdef CONFIG_AUFS_XATTR
23168 +               .listxattr      = aufs_listxattr,
23169 +#endif
23170 +
23171 +               .update_time    = aufs_update_time,
23172 +               .atomic_open    = aufs_atomic_open,
23173 +               .tmpfile        = aufs_tmpfile
23174 +       },
23175 +       [AuIop_OTHER] = {
23176 +               .permission     = aufs_permission,
23177 +#ifdef CONFIG_FS_POSIX_ACL
23178 +               .get_acl        = aufs_get_acl,
23179 +               .set_acl        = aufs_set_acl,
23180 +#endif
23181 +
23182 +               .setattr        = aufs_setattr,
23183 +               .getattr        = aufs_getattr,
23184 +
23185 +#ifdef CONFIG_AUFS_XATTR
23186 +               .listxattr      = aufs_listxattr,
23187 +#endif
23188 +
23189 +               .update_time    = aufs_update_time
23190 +       }
23191 +};
23192 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
23193 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
23194 +++ linux/fs/aufs/i_op_del.c    2022-11-05 23:02:18.965889284 +0100
23195 @@ -0,0 +1,522 @@
23196 +// SPDX-License-Identifier: GPL-2.0
23197 +/*
23198 + * Copyright (C) 2005-2022 Junjiro R. Okajima
23199 + *
23200 + * This program is free software; you can redistribute it and/or modify
23201 + * it under the terms of the GNU General Public License as published by
23202 + * the Free Software Foundation; either version 2 of the License, or
23203 + * (at your option) any later version.
23204 + *
23205 + * This program is distributed in the hope that it will be useful,
23206 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23207 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
23208 + * GNU General Public License for more details.
23209 + *
23210 + * You should have received a copy of the GNU General Public License
23211 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
23212 + */
23213 +
23214 +/*
23215 + * inode operations (del entry)
23216 + */
23217 +
23218 +#include <linux/iversion.h>
23219 +#include "aufs.h"
23220 +
23221 +/*
23222 + * decide if a new whiteout for @dentry is necessary or not.
23223 + * when it is necessary, prepare the parent dir for the upper branch whose
23224 + * branch index is @bcpup for creation. the actual creation of the whiteout will
23225 + * be done by caller.
23226 + * return value:
23227 + * 0: wh is unnecessary
23228 + * plus: wh is necessary
23229 + * minus: error
23230 + */
23231 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
23232 +{
23233 +       int need_wh, err;
23234 +       aufs_bindex_t btop;
23235 +       struct super_block *sb;
23236 +
23237 +       sb = dentry->d_sb;
23238 +       btop = au_dbtop(dentry);
23239 +       if (*bcpup < 0) {
23240 +               *bcpup = btop;
23241 +               if (au_test_ro(sb, btop, d_inode(dentry))) {
23242 +                       err = AuWbrCopyup(au_sbi(sb), dentry);
23243 +                       *bcpup = err;
23244 +                       if (unlikely(err < 0))
23245 +                               goto out;
23246 +               }
23247 +       } else
23248 +               AuDebugOn(btop < *bcpup
23249 +                         || au_test_ro(sb, *bcpup, d_inode(dentry)));
23250 +       AuDbg("bcpup %d, btop %d\n", *bcpup, btop);
23251 +
23252 +       if (*bcpup != btop) {
23253 +               err = au_cpup_dirs(dentry, *bcpup);
23254 +               if (unlikely(err))
23255 +                       goto out;
23256 +               need_wh = 1;
23257 +       } else {
23258 +               struct au_dinfo *dinfo, *tmp;
23259 +
23260 +               need_wh = -ENOMEM;
23261 +               dinfo = au_di(dentry);
23262 +               tmp = au_di_alloc(sb, AuLsc_DI_TMP);
23263 +               if (tmp) {
23264 +                       au_di_cp(tmp, dinfo);
23265 +                       au_di_swap(tmp, dinfo);
23266 +                       /* returns the number of positive dentries */
23267 +                       need_wh = au_lkup_dentry(dentry, btop + 1,
23268 +                                                /* AuLkup_IGNORE_PERM */ 0);
23269 +                       au_di_swap(tmp, dinfo);
23270 +                       au_rw_write_unlock(&tmp->di_rwsem);
23271 +                       au_di_free(tmp);
23272 +               }
23273 +       }
23274 +       AuDbg("need_wh %d\n", need_wh);
23275 +       err = need_wh;
23276 +
23277 +out:
23278 +       return err;
23279 +}
23280 +
23281 +/*
23282 + * simple tests for the del-entry operations.
23283 + * following the checks in vfs, plus the parent-child relationship.
23284 + */
23285 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
23286 +              struct dentry *h_parent, int isdir)
23287 +{
23288 +       int err;
23289 +       umode_t h_mode;
23290 +       struct dentry *h_dentry, *h_latest;
23291 +       struct inode *h_inode;
23292 +       struct path h_ppath;
23293 +       struct super_block *sb;
23294 +       struct au_branch *br;
23295 +       struct user_namespace *h_userns;
23296 +
23297 +       h_dentry = au_h_dptr(dentry, bindex);
23298 +       if (d_really_is_positive(dentry)) {
23299 +               err = -ENOENT;
23300 +               if (unlikely(d_is_negative(h_dentry)))
23301 +                       goto out;
23302 +               h_inode = d_inode(h_dentry);
23303 +               if (unlikely(!h_inode->i_nlink))
23304 +                       goto out;
23305 +
23306 +               h_mode = h_inode->i_mode;
23307 +               if (!isdir) {
23308 +                       err = -EISDIR;
23309 +                       if (unlikely(S_ISDIR(h_mode)))
23310 +                               goto out;
23311 +               } else if (unlikely(!S_ISDIR(h_mode))) {
23312 +                       err = -ENOTDIR;
23313 +                       goto out;
23314 +               }
23315 +       } else {
23316 +               /* rename(2) case */
23317 +               err = -EIO;
23318 +               if (unlikely(d_is_positive(h_dentry)))
23319 +                       goto out;
23320 +       }
23321 +
23322 +       err = -ENOENT;
23323 +       /* expected parent dir is locked */
23324 +       if (unlikely(h_parent != h_dentry->d_parent))
23325 +               goto out;
23326 +       err = 0;
23327 +
23328 +       /*
23329 +        * rmdir a dir may break the consistency on some filesystem.
23330 +        * let's try heavy test.
23331 +        */
23332 +       err = -EACCES;
23333 +       sb = dentry->d_sb;
23334 +       br = au_sbr(sb, bindex);
23335 +       h_userns = au_br_userns(br);
23336 +       if (unlikely(!au_opt_test(au_mntflags(sb), DIRPERM1)
23337 +                    && au_test_h_perm(h_userns, d_inode(h_parent),
23338 +                                      MAY_EXEC | MAY_WRITE)))
23339 +               goto out;
23340 +
23341 +       h_ppath.dentry = h_parent;
23342 +       h_ppath.mnt = au_br_mnt(br);
23343 +       h_latest = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
23344 +       err = -EIO;
23345 +       if (IS_ERR(h_latest))
23346 +               goto out;
23347 +       if (h_latest == h_dentry)
23348 +               err = 0;
23349 +       dput(h_latest);
23350 +
23351 +out:
23352 +       return err;
23353 +}
23354 +
23355 +/*
23356 + * decide the branch where we operate for @dentry. the branch index will be set
23357 + * @rbcpup. after deciding it, 'pin' it and store the timestamps of the parent
23358 + * dir for reverting.
23359 + * when a new whiteout is necessary, create it.
23360 + */
23361 +static struct dentry*
23362 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
23363 +                   struct au_dtime *dt, struct au_pin *pin)
23364 +{
23365 +       struct dentry *wh_dentry;
23366 +       struct super_block *sb;
23367 +       struct path h_path;
23368 +       int err, need_wh;
23369 +       unsigned int udba;
23370 +       aufs_bindex_t bcpup;
23371 +
23372 +       need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
23373 +       wh_dentry = ERR_PTR(need_wh);
23374 +       if (unlikely(need_wh < 0))
23375 +               goto out;
23376 +
23377 +       sb = dentry->d_sb;
23378 +       udba = au_opt_udba(sb);
23379 +       bcpup = *rbcpup;
23380 +       err = au_pin(pin, dentry, bcpup, udba,
23381 +                    AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23382 +       wh_dentry = ERR_PTR(err);
23383 +       if (unlikely(err))
23384 +               goto out;
23385 +
23386 +       h_path.dentry = au_pinned_h_parent(pin);
23387 +       if (udba != AuOpt_UDBA_NONE
23388 +           && au_dbtop(dentry) == bcpup) {
23389 +               err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
23390 +               wh_dentry = ERR_PTR(err);
23391 +               if (unlikely(err))
23392 +                       goto out_unpin;
23393 +       }
23394 +
23395 +       h_path.mnt = au_sbr_mnt(sb, bcpup);
23396 +       au_dtime_store(dt, au_pinned_parent(pin), &h_path);
23397 +       wh_dentry = NULL;
23398 +       if (!need_wh)
23399 +               goto out; /* success, no need to create whiteout */
23400 +
23401 +       wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
23402 +       if (IS_ERR(wh_dentry))
23403 +               goto out_unpin;
23404 +
23405 +       /* returns with the parent is locked and wh_dentry is dget-ed */
23406 +       goto out; /* success */
23407 +
23408 +out_unpin:
23409 +       au_unpin(pin);
23410 +out:
23411 +       return wh_dentry;
23412 +}
23413 +
23414 +/*
23415 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
23416 + * in order to be revertible and save time for removing many child whiteouts
23417 + * under the dir.
23418 + * returns 1 when there are too many child whiteout and caller should remove
23419 + * them asynchronously. returns 0 when the number of children is enough small to
23420 + * remove now or the branch fs is a remote fs.
23421 + * otherwise return an error.
23422 + */
23423 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
23424 +                          struct au_nhash *whlist, struct inode *dir)
23425 +{
23426 +       int rmdir_later, err, dirwh;
23427 +       struct dentry *h_dentry;
23428 +       struct super_block *sb;
23429 +       struct inode *inode;
23430 +
23431 +       sb = dentry->d_sb;
23432 +       SiMustAnyLock(sb);
23433 +       h_dentry = au_h_dptr(dentry, bindex);
23434 +       err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
23435 +       if (unlikely(err))
23436 +               goto out;
23437 +
23438 +       /* stop monitoring */
23439 +       inode = d_inode(dentry);
23440 +       au_hn_free(au_hi(inode, bindex));
23441 +
23442 +       if (!au_test_fs_remote(h_dentry->d_sb)) {
23443 +               dirwh = au_sbi(sb)->si_dirwh;
23444 +               rmdir_later = (dirwh <= 1);
23445 +               if (!rmdir_later)
23446 +                       rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
23447 +                                                             dirwh);
23448 +               if (rmdir_later)
23449 +                       return rmdir_later;
23450 +       }
23451 +
23452 +       err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
23453 +       if (unlikely(err)) {
23454 +               AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
23455 +                       h_dentry, bindex, err);
23456 +               err = 0;
23457 +       }
23458 +
23459 +out:
23460 +       AuTraceErr(err);
23461 +       return err;
23462 +}
23463 +
23464 +/*
23465 + * final procedure for deleting a entry.
23466 + * maintain dentry and iattr.
23467 + */
23468 +static void epilog(struct inode *dir, struct dentry *dentry,
23469 +                  aufs_bindex_t bindex)
23470 +{
23471 +       struct inode *inode;
23472 +
23473 +       inode = d_inode(dentry);
23474 +       d_drop(dentry);
23475 +       inode->i_ctime = dir->i_ctime;
23476 +
23477 +       au_dir_ts(dir, bindex);
23478 +       inode_inc_iversion(dir);
23479 +}
23480 +
23481 +/*
23482 + * when an error happened, remove the created whiteout and revert everything.
23483 + */
23484 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
23485 +                    aufs_bindex_t bwh, struct dentry *wh_dentry,
23486 +                    struct dentry *dentry, struct au_dtime *dt)
23487 +{
23488 +       int rerr;
23489 +       struct path h_path = {
23490 +               .dentry = wh_dentry,
23491 +               .mnt    = au_sbr_mnt(dir->i_sb, bindex)
23492 +       };
23493 +
23494 +       rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
23495 +       if (!rerr) {
23496 +               au_set_dbwh(dentry, bwh);
23497 +               au_dtime_revert(dt);
23498 +               return 0;
23499 +       }
23500 +
23501 +       AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
23502 +       return -EIO;
23503 +}
23504 +
23505 +/* ---------------------------------------------------------------------- */
23506 +
23507 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
23508 +{
23509 +       int err;
23510 +       aufs_bindex_t bwh, bindex, btop;
23511 +       struct inode *inode, *h_dir, *delegated;
23512 +       struct dentry *parent, *wh_dentry;
23513 +       /* to reduce stack size */
23514 +       struct {
23515 +               struct au_dtime dt;
23516 +               struct au_pin pin;
23517 +               struct path h_path;
23518 +       } *a;
23519 +
23520 +       IMustLock(dir);
23521 +
23522 +       err = -ENOMEM;
23523 +       a = kmalloc(sizeof(*a), GFP_NOFS);
23524 +       if (unlikely(!a))
23525 +               goto out;
23526 +
23527 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
23528 +       if (unlikely(err))
23529 +               goto out_free;
23530 +       err = au_d_hashed_positive(dentry);
23531 +       if (unlikely(err))
23532 +               goto out_unlock;
23533 +       inode = d_inode(dentry);
23534 +       IMustLock(inode);
23535 +       err = -EISDIR;
23536 +       if (unlikely(d_is_dir(dentry)))
23537 +               goto out_unlock; /* possible? */
23538 +
23539 +       btop = au_dbtop(dentry);
23540 +       bwh = au_dbwh(dentry);
23541 +       bindex = -1;
23542 +       parent = dentry->d_parent; /* dir inode is locked */
23543 +       di_write_lock_parent(parent);
23544 +       wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
23545 +                                       &a->pin);
23546 +       err = PTR_ERR(wh_dentry);
23547 +       if (IS_ERR(wh_dentry))
23548 +               goto out_parent;
23549 +
23550 +       a->h_path.mnt = au_sbr_mnt(dentry->d_sb, btop);
23551 +       a->h_path.dentry = au_h_dptr(dentry, btop);
23552 +       dget(a->h_path.dentry);
23553 +       if (bindex == btop) {
23554 +               h_dir = au_pinned_h_dir(&a->pin);
23555 +               delegated = NULL;
23556 +               err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
23557 +               if (unlikely(err == -EWOULDBLOCK)) {
23558 +                       pr_warn("cannot retry for NFSv4 delegation"
23559 +                               " for an internal unlink\n");
23560 +                       iput(delegated);
23561 +               }
23562 +       } else {
23563 +               /* dir inode is locked */
23564 +               h_dir = d_inode(wh_dentry->d_parent);
23565 +               IMustLock(h_dir);
23566 +               err = 0;
23567 +       }
23568 +
23569 +       if (!err) {
23570 +               vfsub_drop_nlink(inode);
23571 +               epilog(dir, dentry, bindex);
23572 +
23573 +               /* update target timestamps */
23574 +               if (bindex == btop) {
23575 +                       vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
23576 +                       /*ignore*/
23577 +                       inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
23578 +               } else
23579 +                       /* todo: this timestamp may be reverted later */
23580 +                       inode->i_ctime = h_dir->i_ctime;
23581 +               goto out_unpin; /* success */
23582 +       }
23583 +
23584 +       /* revert */
23585 +       if (wh_dentry) {
23586 +               int rerr;
23587 +
23588 +               rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23589 +                                &a->dt);
23590 +               if (rerr)
23591 +                       err = rerr;
23592 +       }
23593 +
23594 +out_unpin:
23595 +       au_unpin(&a->pin);
23596 +       dput(wh_dentry);
23597 +       dput(a->h_path.dentry);
23598 +out_parent:
23599 +       di_write_unlock(parent);
23600 +out_unlock:
23601 +       aufs_read_unlock(dentry, AuLock_DW);
23602 +out_free:
23603 +       au_kfree_rcu(a);
23604 +out:
23605 +       return err;
23606 +}
23607 +
23608 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
23609 +{
23610 +       int err, rmdir_later;
23611 +       aufs_bindex_t bwh, bindex, btop;
23612 +       struct inode *inode;
23613 +       struct dentry *parent, *wh_dentry, *h_dentry;
23614 +       struct au_whtmp_rmdir *args;
23615 +       /* to reduce stack size */
23616 +       struct {
23617 +               struct au_dtime dt;
23618 +               struct au_pin pin;
23619 +       } *a;
23620 +
23621 +       IMustLock(dir);
23622 +
23623 +       err = -ENOMEM;
23624 +       a = kmalloc(sizeof(*a), GFP_NOFS);
23625 +       if (unlikely(!a))
23626 +               goto out;
23627 +
23628 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
23629 +       if (unlikely(err))
23630 +               goto out_free;
23631 +       err = au_alive_dir(dentry);
23632 +       if (unlikely(err))
23633 +               goto out_unlock;
23634 +       inode = d_inode(dentry);
23635 +       IMustLock(inode);
23636 +       err = -ENOTDIR;
23637 +       if (unlikely(!d_is_dir(dentry)))
23638 +               goto out_unlock; /* possible? */
23639 +
23640 +       err = -ENOMEM;
23641 +       args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
23642 +       if (unlikely(!args))
23643 +               goto out_unlock;
23644 +
23645 +       parent = dentry->d_parent; /* dir inode is locked */
23646 +       di_write_lock_parent(parent);
23647 +       err = au_test_empty(dentry, &args->whlist);
23648 +       if (unlikely(err))
23649 +               goto out_parent;
23650 +
23651 +       btop = au_dbtop(dentry);
23652 +       bwh = au_dbwh(dentry);
23653 +       bindex = -1;
23654 +       wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
23655 +                                       &a->pin);
23656 +       err = PTR_ERR(wh_dentry);
23657 +       if (IS_ERR(wh_dentry))
23658 +               goto out_parent;
23659 +
23660 +       h_dentry = au_h_dptr(dentry, btop);
23661 +       dget(h_dentry);
23662 +       rmdir_later = 0;
23663 +       if (bindex == btop) {
23664 +               err = renwh_and_rmdir(dentry, btop, &args->whlist, dir);
23665 +               if (err > 0) {
23666 +                       rmdir_later = err;
23667 +                       err = 0;
23668 +               }
23669 +       } else {
23670 +               /* stop monitoring */
23671 +               au_hn_free(au_hi(inode, btop));
23672 +
23673 +               /* dir inode is locked */
23674 +               IMustLock(d_inode(wh_dentry->d_parent));
23675 +               err = 0;
23676 +       }
23677 +
23678 +       if (!err) {
23679 +               vfsub_dead_dir(inode);
23680 +               au_set_dbdiropq(dentry, -1);
23681 +               epilog(dir, dentry, bindex);
23682 +
23683 +               if (rmdir_later) {
23684 +                       au_whtmp_kick_rmdir(dir, btop, h_dentry, args);
23685 +                       args = NULL;
23686 +               }
23687 +
23688 +               goto out_unpin; /* success */
23689 +       }
23690 +
23691 +       /* revert */
23692 +       AuLabel(revert);
23693 +       if (wh_dentry) {
23694 +               int rerr;
23695 +
23696 +               rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23697 +                                &a->dt);
23698 +               if (rerr)
23699 +                       err = rerr;
23700 +       }
23701 +
23702 +out_unpin:
23703 +       au_unpin(&a->pin);
23704 +       dput(wh_dentry);
23705 +       dput(h_dentry);
23706 +out_parent:
23707 +       di_write_unlock(parent);
23708 +       if (args)
23709 +               au_whtmp_rmdir_free(args);
23710 +out_unlock:
23711 +       aufs_read_unlock(dentry, AuLock_DW);
23712 +out_free:
23713 +       au_kfree_rcu(a);
23714 +out:
23715 +       AuTraceErr(err);
23716 +       return err;
23717 +}
23718 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
23719 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
23720 +++ linux/fs/aufs/i_op_ren.c    2022-11-05 23:02:18.965889284 +0100
23721 @@ -0,0 +1,1257 @@
23722 +// SPDX-License-Identifier: GPL-2.0
23723 +/*
23724 + * Copyright (C) 2005-2022 Junjiro R. Okajima
23725 + *
23726 + * This program is free software; you can redistribute it and/or modify
23727 + * it under the terms of the GNU General Public License as published by
23728 + * the Free Software Foundation; either version 2 of the License, or
23729 + * (at your option) any later version.
23730 + *
23731 + * This program is distributed in the hope that it will be useful,
23732 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23733 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
23734 + * GNU General Public License for more details.
23735 + *
23736 + * You should have received a copy of the GNU General Public License
23737 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
23738 + */
23739 +
23740 +/*
23741 + * inode operation (rename entry)
23742 + * todo: this is crazy monster
23743 + */
23744 +
23745 +#include <linux/iversion.h>
23746 +#include "aufs.h"
23747 +
23748 +enum { AuSRC, AuDST, AuSrcDst };
23749 +enum { AuPARENT, AuCHILD, AuParentChild };
23750 +
23751 +#define AuRen_ISDIR_SRC                1
23752 +#define AuRen_ISDIR_DST                (1 << 1)
23753 +#define AuRen_ISSAMEDIR                (1 << 2)
23754 +#define AuRen_WHSRC            (1 << 3)
23755 +#define AuRen_WHDST            (1 << 4)
23756 +#define AuRen_MNT_WRITE                (1 << 5)
23757 +#define AuRen_DT_DSTDIR                (1 << 6)
23758 +#define AuRen_DIROPQ_SRC       (1 << 7)
23759 +#define AuRen_DIROPQ_DST       (1 << 8)
23760 +#define AuRen_DIRREN           (1 << 9)
23761 +#define AuRen_DROPPED_SRC      (1 << 10)
23762 +#define AuRen_DROPPED_DST      (1 << 11)
23763 +#define au_ftest_ren(flags, name)      ((flags) & AuRen_##name)
23764 +#define au_fset_ren(flags, name) \
23765 +       do { (flags) |= AuRen_##name; } while (0)
23766 +#define au_fclr_ren(flags, name) \
23767 +       do { (flags) &= ~AuRen_##name; } while (0)
23768 +
23769 +#ifndef CONFIG_AUFS_DIRREN
23770 +#undef AuRen_DIRREN
23771 +#define AuRen_DIRREN           0
23772 +#endif
23773 +
23774 +struct au_ren_args {
23775 +       struct {
23776 +               struct dentry *dentry, *h_dentry, *parent, *h_parent,
23777 +                       *wh_dentry;
23778 +               struct inode *dir, *inode;
23779 +               struct au_hinode *hdir, *hinode;
23780 +               struct au_dtime dt[AuParentChild];
23781 +               aufs_bindex_t btop, bdiropq;
23782 +       } sd[AuSrcDst];
23783 +
23784 +#define src_dentry     sd[AuSRC].dentry
23785 +#define src_dir                sd[AuSRC].dir
23786 +#define src_inode      sd[AuSRC].inode
23787 +#define src_h_dentry   sd[AuSRC].h_dentry
23788 +#define src_parent     sd[AuSRC].parent
23789 +#define src_h_parent   sd[AuSRC].h_parent
23790 +#define src_wh_dentry  sd[AuSRC].wh_dentry
23791 +#define src_hdir       sd[AuSRC].hdir
23792 +#define src_hinode     sd[AuSRC].hinode
23793 +#define src_h_dir      sd[AuSRC].hdir->hi_inode
23794 +#define src_dt         sd[AuSRC].dt
23795 +#define src_btop       sd[AuSRC].btop
23796 +#define src_bdiropq    sd[AuSRC].bdiropq
23797 +
23798 +#define dst_dentry     sd[AuDST].dentry
23799 +#define dst_dir                sd[AuDST].dir
23800 +#define dst_inode      sd[AuDST].inode
23801 +#define dst_h_dentry   sd[AuDST].h_dentry
23802 +#define dst_parent     sd[AuDST].parent
23803 +#define dst_h_parent   sd[AuDST].h_parent
23804 +#define dst_wh_dentry  sd[AuDST].wh_dentry
23805 +#define dst_hdir       sd[AuDST].hdir
23806 +#define dst_hinode     sd[AuDST].hinode
23807 +#define dst_h_dir      sd[AuDST].hdir->hi_inode
23808 +#define dst_dt         sd[AuDST].dt
23809 +#define dst_btop       sd[AuDST].btop
23810 +#define dst_bdiropq    sd[AuDST].bdiropq
23811 +
23812 +       struct dentry *h_trap;
23813 +       struct au_branch *br;
23814 +       struct path h_path;
23815 +       struct au_nhash whlist;
23816 +       aufs_bindex_t btgt, src_bwh;
23817 +
23818 +       struct {
23819 +               unsigned short auren_flags;
23820 +               unsigned char flags;    /* syscall parameter */
23821 +               unsigned char exchange;
23822 +       } __packed;
23823 +
23824 +       struct au_whtmp_rmdir *thargs;
23825 +       struct dentry *h_dst;
23826 +       struct au_hinode *h_root;
23827 +};
23828 +
23829 +/* ---------------------------------------------------------------------- */
23830 +
23831 +/*
23832 + * functions for reverting.
23833 + * when an error happened in a single rename systemcall, we should revert
23834 + * everything as if nothing happened.
23835 + * we don't need to revert the copied-up/down the parent dir since they are
23836 + * harmless.
23837 + */
23838 +
23839 +#define RevertFailure(fmt, ...) do { \
23840 +       AuIOErr("revert failure: " fmt " (%d, %d)\n", \
23841 +               ##__VA_ARGS__, err, rerr); \
23842 +       err = -EIO; \
23843 +} while (0)
23844 +
23845 +static void au_ren_do_rev_diropq(int err, struct au_ren_args *a, int idx)
23846 +{
23847 +       int rerr;
23848 +       struct dentry *d;
23849 +#define src_or_dst(member) a->sd[idx].member
23850 +
23851 +       d = src_or_dst(dentry); /* {src,dst}_dentry */
23852 +       au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
23853 +       rerr = au_diropq_remove(d, a->btgt);
23854 +       au_hn_inode_unlock(src_or_dst(hinode));
23855 +       au_set_dbdiropq(d, src_or_dst(bdiropq));
23856 +       if (rerr)
23857 +               RevertFailure("remove diropq %pd", d);
23858 +
23859 +#undef src_or_dst_
23860 +}
23861 +
23862 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
23863 +{
23864 +       if (au_ftest_ren(a->auren_flags, DIROPQ_SRC))
23865 +               au_ren_do_rev_diropq(err, a, AuSRC);
23866 +       if (au_ftest_ren(a->auren_flags, DIROPQ_DST))
23867 +               au_ren_do_rev_diropq(err, a, AuDST);
23868 +}
23869 +
23870 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
23871 +{
23872 +       int rerr;
23873 +       struct inode *delegated;
23874 +       struct path h_ppath = {
23875 +               .dentry = a->src_h_parent,
23876 +               .mnt    = a->h_path.mnt
23877 +       };
23878 +
23879 +       a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name, &h_ppath);
23880 +       rerr = PTR_ERR(a->h_path.dentry);
23881 +       if (IS_ERR(a->h_path.dentry)) {
23882 +               RevertFailure("lkup one %pd", a->src_dentry);
23883 +               return;
23884 +       }
23885 +
23886 +       delegated = NULL;
23887 +       rerr = vfsub_rename(a->dst_h_dir,
23888 +                           au_h_dptr(a->src_dentry, a->btgt),
23889 +                           a->src_h_dir, &a->h_path, &delegated, a->flags);
23890 +       if (unlikely(rerr == -EWOULDBLOCK)) {
23891 +               pr_warn("cannot retry for NFSv4 delegation"
23892 +                       " for an internal rename\n");
23893 +               iput(delegated);
23894 +       }
23895 +       d_drop(a->h_path.dentry);
23896 +       dput(a->h_path.dentry);
23897 +       /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
23898 +       if (rerr)
23899 +               RevertFailure("rename %pd", a->src_dentry);
23900 +}
23901 +
23902 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
23903 +{
23904 +       int rerr;
23905 +       struct inode *delegated;
23906 +       struct path h_ppath = {
23907 +               .dentry = a->dst_h_parent,
23908 +               .mnt    = a->h_path.mnt
23909 +       };
23910 +
23911 +       a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name, &h_ppath);
23912 +       rerr = PTR_ERR(a->h_path.dentry);
23913 +       if (IS_ERR(a->h_path.dentry)) {
23914 +               RevertFailure("lkup one %pd", a->dst_dentry);
23915 +               return;
23916 +       }
23917 +       if (d_is_positive(a->h_path.dentry)) {
23918 +               d_drop(a->h_path.dentry);
23919 +               dput(a->h_path.dentry);
23920 +               return;
23921 +       }
23922 +
23923 +       delegated = NULL;
23924 +       rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
23925 +                           &delegated, a->flags);
23926 +       if (unlikely(rerr == -EWOULDBLOCK)) {
23927 +               pr_warn("cannot retry for NFSv4 delegation"
23928 +                       " for an internal rename\n");
23929 +               iput(delegated);
23930 +       }
23931 +       d_drop(a->h_path.dentry);
23932 +       dput(a->h_path.dentry);
23933 +       if (!rerr)
23934 +               au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
23935 +       else
23936 +               RevertFailure("rename %pd", a->h_dst);
23937 +}
23938 +
23939 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
23940 +{
23941 +       int rerr;
23942 +
23943 +       a->h_path.dentry = a->src_wh_dentry;
23944 +       rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
23945 +       au_set_dbwh(a->src_dentry, a->src_bwh);
23946 +       if (rerr)
23947 +               RevertFailure("unlink %pd", a->src_wh_dentry);
23948 +}
23949 +#undef RevertFailure
23950 +
23951 +/* ---------------------------------------------------------------------- */
23952 +
23953 +/*
23954 + * when we have to copyup the renaming entry, do it with the rename-target name
23955 + * in order to minimize the cost (the later actual rename is unnecessary).
23956 + * otherwise rename it on the target branch.
23957 + */
23958 +static int au_ren_or_cpup(struct au_ren_args *a)
23959 +{
23960 +       int err;
23961 +       struct dentry *d;
23962 +       struct inode *delegated;
23963 +
23964 +       d = a->src_dentry;
23965 +       if (au_dbtop(d) == a->btgt) {
23966 +               a->h_path.dentry = a->dst_h_dentry;
23967 +               AuDebugOn(au_dbtop(d) != a->btgt);
23968 +               delegated = NULL;
23969 +               err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
23970 +                                  a->dst_h_dir, &a->h_path, &delegated,
23971 +                                  a->flags);
23972 +               if (unlikely(err == -EWOULDBLOCK)) {
23973 +                       pr_warn("cannot retry for NFSv4 delegation"
23974 +                               " for an internal rename\n");
23975 +                       iput(delegated);
23976 +               }
23977 +       } else
23978 +               BUG();
23979 +
23980 +       if (!err && a->h_dst)
23981 +               /* it will be set to dinfo later */
23982 +               dget(a->h_dst);
23983 +
23984 +       return err;
23985 +}
23986 +
23987 +/* cf. aufs_rmdir() */
23988 +static int au_ren_del_whtmp(struct au_ren_args *a)
23989 +{
23990 +       int err;
23991 +       struct inode *dir;
23992 +
23993 +       dir = a->dst_dir;
23994 +       SiMustAnyLock(dir->i_sb);
23995 +       if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
23996 +                                    au_sbi(dir->i_sb)->si_dirwh)
23997 +           || au_test_fs_remote(a->h_dst->d_sb)) {
23998 +               err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
23999 +               if (unlikely(err))
24000 +                       pr_warn("failed removing whtmp dir %pd (%d), "
24001 +                               "ignored.\n", a->h_dst, err);
24002 +       } else {
24003 +               au_nhash_wh_free(&a->thargs->whlist);
24004 +               a->thargs->whlist = a->whlist;
24005 +               a->whlist.nh_num = 0;
24006 +               au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
24007 +               dput(a->h_dst);
24008 +               a->thargs = NULL;
24009 +       }
24010 +
24011 +       return 0;
24012 +}
24013 +
24014 +/* make it 'opaque' dir. */
24015 +static int au_ren_do_diropq(struct au_ren_args *a, int idx)
24016 +{
24017 +       int err;
24018 +       struct dentry *d, *diropq;
24019 +#define src_or_dst(member) a->sd[idx].member
24020 +
24021 +       err = 0;
24022 +       d = src_or_dst(dentry); /* {src,dst}_dentry */
24023 +       src_or_dst(bdiropq) = au_dbdiropq(d);
24024 +       src_or_dst(hinode) = au_hi(src_or_dst(inode), a->btgt);
24025 +       au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
24026 +       diropq = au_diropq_create(d, a->btgt);
24027 +       au_hn_inode_unlock(src_or_dst(hinode));
24028 +       if (IS_ERR(diropq))
24029 +               err = PTR_ERR(diropq);
24030 +       else
24031 +               dput(diropq);
24032 +
24033 +#undef src_or_dst_
24034 +       return err;
24035 +}
24036 +
24037 +static int au_ren_diropq(struct au_ren_args *a)
24038 +{
24039 +       int err;
24040 +       unsigned char always;
24041 +       struct dentry *d;
24042 +
24043 +       err = 0;
24044 +       d = a->dst_dentry; /* already renamed on the branch */
24045 +       always = !!au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ);
24046 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24047 +           && !au_ftest_ren(a->auren_flags, DIRREN)
24048 +           && a->btgt != au_dbdiropq(a->src_dentry)
24049 +           && (a->dst_wh_dentry
24050 +               || a->btgt <= au_dbdiropq(d)
24051 +               /* hide the lower to keep xino */
24052 +               /* the lowers may not be a dir, but we hide them anyway */
24053 +               || a->btgt < au_dbbot(d)
24054 +               || always)) {
24055 +               AuDbg("here\n");
24056 +               err = au_ren_do_diropq(a, AuSRC);
24057 +               if (unlikely(err))
24058 +                       goto out;
24059 +               au_fset_ren(a->auren_flags, DIROPQ_SRC);
24060 +       }
24061 +       if (!a->exchange)
24062 +               goto out; /* success */
24063 +
24064 +       d = a->src_dentry; /* already renamed on the branch */
24065 +       if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24066 +           && a->btgt != au_dbdiropq(a->dst_dentry)
24067 +           && (a->btgt < au_dbdiropq(d)
24068 +               || a->btgt < au_dbbot(d)
24069 +               || always)) {
24070 +               AuDbgDentry(a->src_dentry);
24071 +               AuDbgDentry(a->dst_dentry);
24072 +               err = au_ren_do_diropq(a, AuDST);
24073 +               if (unlikely(err))
24074 +                       goto out_rev_src;
24075 +               au_fset_ren(a->auren_flags, DIROPQ_DST);
24076 +       }
24077 +       goto out; /* success */
24078 +
24079 +out_rev_src:
24080 +       AuDbg("err %d, reverting src\n", err);
24081 +       au_ren_rev_diropq(err, a);
24082 +out:
24083 +       return err;
24084 +}
24085 +
24086 +static int do_rename(struct au_ren_args *a)
24087 +{
24088 +       int err;
24089 +       struct dentry *d, *h_d;
24090 +
24091 +       if (!a->exchange) {
24092 +               /* prepare workqueue args for asynchronous rmdir */
24093 +               h_d = a->dst_h_dentry;
24094 +               if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24095 +                   /* && !au_ftest_ren(a->auren_flags, DIRREN) */
24096 +                   && d_is_positive(h_d)) {
24097 +                       err = -ENOMEM;
24098 +                       a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb,
24099 +                                                        GFP_NOFS);
24100 +                       if (unlikely(!a->thargs))
24101 +                               goto out;
24102 +                       a->h_dst = dget(h_d);
24103 +               }
24104 +
24105 +               /* create whiteout for src_dentry */
24106 +               if (au_ftest_ren(a->auren_flags, WHSRC)) {
24107 +                       a->src_bwh = au_dbwh(a->src_dentry);
24108 +                       AuDebugOn(a->src_bwh >= 0);
24109 +                       a->src_wh_dentry = au_wh_create(a->src_dentry, a->btgt,
24110 +                                                       a->src_h_parent);
24111 +                       err = PTR_ERR(a->src_wh_dentry);
24112 +                       if (IS_ERR(a->src_wh_dentry))
24113 +                               goto out_thargs;
24114 +               }
24115 +
24116 +               /* lookup whiteout for dentry */
24117 +               if (au_ftest_ren(a->auren_flags, WHDST)) {
24118 +                       h_d = au_wh_lkup(a->dst_h_parent,
24119 +                                        &a->dst_dentry->d_name, a->br);
24120 +                       err = PTR_ERR(h_d);
24121 +                       if (IS_ERR(h_d))
24122 +                               goto out_whsrc;
24123 +                       if (d_is_negative(h_d))
24124 +                               dput(h_d);
24125 +                       else
24126 +                               a->dst_wh_dentry = h_d;
24127 +               }
24128 +
24129 +               /* rename dentry to tmpwh */
24130 +               if (a->thargs) {
24131 +                       err = au_whtmp_ren(a->dst_h_dentry, a->br);
24132 +                       if (unlikely(err))
24133 +                               goto out_whdst;
24134 +
24135 +                       d = a->dst_dentry;
24136 +                       au_set_h_dptr(d, a->btgt, NULL);
24137 +                       err = au_lkup_neg(d, a->btgt, /*wh*/0);
24138 +                       if (unlikely(err))
24139 +                               goto out_whtmp;
24140 +                       a->dst_h_dentry = au_h_dptr(d, a->btgt);
24141 +               }
24142 +       }
24143 +
24144 +       BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_btop != a->btgt);
24145 +#if 0 /* debugging */
24146 +       BUG_ON(!au_ftest_ren(a->auren_flags, DIRREN)
24147 +              && d_is_positive(a->dst_h_dentry)
24148 +              && a->src_btop != a->btgt);
24149 +#endif
24150 +
24151 +       /* rename by vfs_rename or cpup */
24152 +       err = au_ren_or_cpup(a);
24153 +       if (unlikely(err))
24154 +               /* leave the copied-up one */
24155 +               goto out_whtmp;
24156 +
24157 +       /* make dir opaque */
24158 +       err = au_ren_diropq(a);
24159 +       if (unlikely(err))
24160 +               goto out_rename;
24161 +
24162 +       /* update target timestamps */
24163 +       if (a->exchange) {
24164 +               AuDebugOn(au_dbtop(a->dst_dentry) != a->btgt);
24165 +               a->h_path.dentry = au_h_dptr(a->dst_dentry, a->btgt);
24166 +               vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24167 +               a->dst_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24168 +       }
24169 +       AuDebugOn(au_dbtop(a->src_dentry) != a->btgt);
24170 +       a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
24171 +       vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24172 +       a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24173 +
24174 +       if (!a->exchange) {
24175 +               /* remove whiteout for dentry */
24176 +               if (a->dst_wh_dentry) {
24177 +                       a->h_path.dentry = a->dst_wh_dentry;
24178 +                       err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
24179 +                                                 a->dst_dentry);
24180 +                       if (unlikely(err))
24181 +                               goto out_diropq;
24182 +               }
24183 +
24184 +               /* remove whtmp */
24185 +               if (a->thargs)
24186 +                       au_ren_del_whtmp(a); /* ignore this error */
24187 +
24188 +               au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
24189 +       }
24190 +       err = 0;
24191 +       goto out_success;
24192 +
24193 +out_diropq:
24194 +       au_ren_rev_diropq(err, a);
24195 +out_rename:
24196 +       au_ren_rev_rename(err, a);
24197 +       dput(a->h_dst);
24198 +out_whtmp:
24199 +       if (a->thargs)
24200 +               au_ren_rev_whtmp(err, a);
24201 +out_whdst:
24202 +       dput(a->dst_wh_dentry);
24203 +       a->dst_wh_dentry = NULL;
24204 +out_whsrc:
24205 +       if (a->src_wh_dentry)
24206 +               au_ren_rev_whsrc(err, a);
24207 +out_success:
24208 +       dput(a->src_wh_dentry);
24209 +       dput(a->dst_wh_dentry);
24210 +out_thargs:
24211 +       if (a->thargs) {
24212 +               dput(a->h_dst);
24213 +               au_whtmp_rmdir_free(a->thargs);
24214 +               a->thargs = NULL;
24215 +       }
24216 +out:
24217 +       return err;
24218 +}
24219 +
24220 +/* ---------------------------------------------------------------------- */
24221 +
24222 +/*
24223 + * test if @dentry dir can be rename destination or not.
24224 + * success means, it is a logically empty dir.
24225 + */
24226 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
24227 +{
24228 +       return au_test_empty(dentry, whlist);
24229 +}
24230 +
24231 +/*
24232 + * test if @a->src_dentry dir can be rename source or not.
24233 + * if it can, return 0.
24234 + * success means,
24235 + * - it is a logically empty dir.
24236 + * - or, it exists on writable branch and has no children including whiteouts
24237 + *   on the lower branch unless DIRREN is on.
24238 + */
24239 +static int may_rename_srcdir(struct au_ren_args *a)
24240 +{
24241 +       int err;
24242 +       unsigned int rdhash;
24243 +       aufs_bindex_t btop, btgt;
24244 +       struct dentry *dentry;
24245 +       struct super_block *sb;
24246 +       struct au_sbinfo *sbinfo;
24247 +
24248 +       dentry = a->src_dentry;
24249 +       sb = dentry->d_sb;
24250 +       sbinfo = au_sbi(sb);
24251 +       if (au_opt_test(sbinfo->si_mntflags, DIRREN))
24252 +               au_fset_ren(a->auren_flags, DIRREN);
24253 +
24254 +       btgt = a->btgt;
24255 +       btop = au_dbtop(dentry);
24256 +       if (btop != btgt) {
24257 +               struct au_nhash whlist;
24258 +
24259 +               SiMustAnyLock(sb);
24260 +               rdhash = sbinfo->si_rdhash;
24261 +               if (!rdhash)
24262 +                       rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
24263 +                                                          dentry));
24264 +               err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
24265 +               if (unlikely(err))
24266 +                       goto out;
24267 +               err = au_test_empty(dentry, &whlist);
24268 +               au_nhash_wh_free(&whlist);
24269 +               goto out;
24270 +       }
24271 +
24272 +       if (btop == au_dbtaildir(dentry))
24273 +               return 0; /* success */
24274 +
24275 +       err = au_test_empty_lower(dentry);
24276 +
24277 +out:
24278 +       if (err == -ENOTEMPTY) {
24279 +               if (au_ftest_ren(a->auren_flags, DIRREN)) {
24280 +                       err = 0;
24281 +               } else {
24282 +                       AuWarn1("renaming dir who has child(ren) on multiple "
24283 +                               "branches, is not supported\n");
24284 +                       err = -EXDEV;
24285 +               }
24286 +       }
24287 +       return err;
24288 +}
24289 +
24290 +/* side effect: sets whlist and h_dentry */
24291 +static int au_ren_may_dir(struct au_ren_args *a)
24292 +{
24293 +       int err;
24294 +       unsigned int rdhash;
24295 +       struct dentry *d;
24296 +
24297 +       d = a->dst_dentry;
24298 +       SiMustAnyLock(d->d_sb);
24299 +
24300 +       err = 0;
24301 +       if (au_ftest_ren(a->auren_flags, ISDIR_DST) && a->dst_inode) {
24302 +               rdhash = au_sbi(d->d_sb)->si_rdhash;
24303 +               if (!rdhash)
24304 +                       rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
24305 +               err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
24306 +               if (unlikely(err))
24307 +                       goto out;
24308 +
24309 +               if (!a->exchange) {
24310 +                       au_set_dbtop(d, a->dst_btop);
24311 +                       err = may_rename_dstdir(d, &a->whlist);
24312 +                       au_set_dbtop(d, a->btgt);
24313 +               } else
24314 +                       err = may_rename_srcdir(a);
24315 +       }
24316 +       a->dst_h_dentry = au_h_dptr(d, au_dbtop(d));
24317 +       if (unlikely(err))
24318 +               goto out;
24319 +
24320 +       d = a->src_dentry;
24321 +       a->src_h_dentry = au_h_dptr(d, au_dbtop(d));
24322 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24323 +               err = may_rename_srcdir(a);
24324 +               if (unlikely(err)) {
24325 +                       au_nhash_wh_free(&a->whlist);
24326 +                       a->whlist.nh_num = 0;
24327 +               }
24328 +       }
24329 +out:
24330 +       return err;
24331 +}
24332 +
24333 +/* ---------------------------------------------------------------------- */
24334 +
24335 +/*
24336 + * simple tests for rename.
24337 + * following the checks in vfs, plus the parent-child relationship.
24338 + */
24339 +static int au_may_ren(struct au_ren_args *a)
24340 +{
24341 +       int err, isdir;
24342 +       struct inode *h_inode;
24343 +
24344 +       if (a->src_btop == a->btgt) {
24345 +               err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
24346 +                                au_ftest_ren(a->auren_flags, ISDIR_SRC));
24347 +               if (unlikely(err))
24348 +                       goto out;
24349 +               err = -EINVAL;
24350 +               if (unlikely(a->src_h_dentry == a->h_trap))
24351 +                       goto out;
24352 +       }
24353 +
24354 +       err = 0;
24355 +       if (a->dst_btop != a->btgt)
24356 +               goto out;
24357 +
24358 +       err = -ENOTEMPTY;
24359 +       if (unlikely(a->dst_h_dentry == a->h_trap))
24360 +               goto out;
24361 +
24362 +       err = -EIO;
24363 +       isdir = !!au_ftest_ren(a->auren_flags, ISDIR_DST);
24364 +       if (d_really_is_negative(a->dst_dentry)) {
24365 +               if (d_is_negative(a->dst_h_dentry))
24366 +                       err = au_may_add(a->dst_dentry, a->btgt,
24367 +                                        a->dst_h_parent, isdir);
24368 +       } else {
24369 +               if (unlikely(d_is_negative(a->dst_h_dentry)))
24370 +                       goto out;
24371 +               h_inode = d_inode(a->dst_h_dentry);
24372 +               if (h_inode->i_nlink)
24373 +                       err = au_may_del(a->dst_dentry, a->btgt,
24374 +                                        a->dst_h_parent, isdir);
24375 +       }
24376 +
24377 +out:
24378 +       if (unlikely(err == -ENOENT || err == -EEXIST))
24379 +               err = -EIO;
24380 +       AuTraceErr(err);
24381 +       return err;
24382 +}
24383 +
24384 +/* ---------------------------------------------------------------------- */
24385 +
24386 +/*
24387 + * locking order
24388 + * (VFS)
24389 + * - src_dir and dir by lock_rename()
24390 + * - inode if exists
24391 + * (aufs)
24392 + * - lock all
24393 + *   + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
24394 + *     + si_read_lock
24395 + *     + di_write_lock2_child()
24396 + *       + di_write_lock_child()
24397 + *        + ii_write_lock_child()
24398 + *       + di_write_lock_child2()
24399 + *        + ii_write_lock_child2()
24400 + *     + src_parent and parent
24401 + *       + di_write_lock_parent()
24402 + *        + ii_write_lock_parent()
24403 + *       + di_write_lock_parent2()
24404 + *        + ii_write_lock_parent2()
24405 + *   + lower src_dir and dir by vfsub_lock_rename()
24406 + *   + verify the every relationships between child and parent. if any
24407 + *     of them failed, unlock all and return -EBUSY.
24408 + */
24409 +static void au_ren_unlock(struct au_ren_args *a)
24410 +{
24411 +       vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
24412 +                           a->dst_h_parent, a->dst_hdir);
24413 +       if (au_ftest_ren(a->auren_flags, DIRREN)
24414 +           && a->h_root)
24415 +               au_hn_inode_unlock(a->h_root);
24416 +       if (au_ftest_ren(a->auren_flags, MNT_WRITE))
24417 +               vfsub_mnt_drop_write(au_br_mnt(a->br));
24418 +}
24419 +
24420 +static int au_ren_lock(struct au_ren_args *a)
24421 +{
24422 +       int err;
24423 +       unsigned int udba;
24424 +
24425 +       err = 0;
24426 +       a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
24427 +       a->src_hdir = au_hi(a->src_dir, a->btgt);
24428 +       a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
24429 +       a->dst_hdir = au_hi(a->dst_dir, a->btgt);
24430 +
24431 +       err = vfsub_mnt_want_write(au_br_mnt(a->br));
24432 +       if (unlikely(err))
24433 +               goto out;
24434 +       au_fset_ren(a->auren_flags, MNT_WRITE);
24435 +       if (au_ftest_ren(a->auren_flags, DIRREN)) {
24436 +               struct dentry *root;
24437 +               struct inode *dir;
24438 +
24439 +               /*
24440 +                * sbinfo is already locked, so this ii_read_lock is
24441 +                * unnecessary. but our debugging feature checks it.
24442 +                */
24443 +               root = a->src_inode->i_sb->s_root;
24444 +               if (root != a->src_parent && root != a->dst_parent) {
24445 +                       dir = d_inode(root);
24446 +                       ii_read_lock_parent3(dir);
24447 +                       a->h_root = au_hi(dir, a->btgt);
24448 +                       ii_read_unlock(dir);
24449 +                       au_hn_inode_lock_nested(a->h_root, AuLsc_I_PARENT3);
24450 +               }
24451 +       }
24452 +       a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
24453 +                                     a->dst_h_parent, a->dst_hdir);
24454 +       udba = au_opt_udba(a->src_dentry->d_sb);
24455 +       if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
24456 +                    || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
24457 +               err = au_busy_or_stale();
24458 +       if (!err && au_dbtop(a->src_dentry) == a->btgt)
24459 +               err = au_h_verify(a->src_h_dentry, udba,
24460 +                                 d_inode(a->src_h_parent), a->src_h_parent,
24461 +                                 a->br);
24462 +       if (!err && au_dbtop(a->dst_dentry) == a->btgt)
24463 +               err = au_h_verify(a->dst_h_dentry, udba,
24464 +                                 d_inode(a->dst_h_parent), a->dst_h_parent,
24465 +                                 a->br);
24466 +       if (!err)
24467 +               goto out; /* success */
24468 +
24469 +       err = au_busy_or_stale();
24470 +       au_ren_unlock(a);
24471 +
24472 +out:
24473 +       return err;
24474 +}
24475 +
24476 +/* ---------------------------------------------------------------------- */
24477 +
24478 +static void au_ren_refresh_dir(struct au_ren_args *a)
24479 +{
24480 +       struct inode *dir;
24481 +
24482 +       dir = a->dst_dir;
24483 +       inode_inc_iversion(dir);
24484 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24485 +               /* is this updating defined in POSIX? */
24486 +               au_cpup_attr_timesizes(a->src_inode);
24487 +               au_cpup_attr_nlink(dir, /*force*/1);
24488 +       }
24489 +       au_dir_ts(dir, a->btgt);
24490 +
24491 +       if (a->exchange) {
24492 +               dir = a->src_dir;
24493 +               inode_inc_iversion(dir);
24494 +               if (au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24495 +                       /* is this updating defined in POSIX? */
24496 +                       au_cpup_attr_timesizes(a->dst_inode);
24497 +                       au_cpup_attr_nlink(dir, /*force*/1);
24498 +               }
24499 +               au_dir_ts(dir, a->btgt);
24500 +       }
24501 +
24502 +       if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24503 +               return;
24504 +
24505 +       dir = a->src_dir;
24506 +       inode_inc_iversion(dir);
24507 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC))
24508 +               au_cpup_attr_nlink(dir, /*force*/1);
24509 +       au_dir_ts(dir, a->btgt);
24510 +}
24511 +
24512 +static void au_ren_refresh(struct au_ren_args *a)
24513 +{
24514 +       aufs_bindex_t bbot, bindex;
24515 +       struct dentry *d, *h_d;
24516 +       struct inode *i, *h_i;
24517 +       struct super_block *sb;
24518 +
24519 +       d = a->dst_dentry;
24520 +       d_drop(d);
24521 +       if (a->h_dst)
24522 +               /* already dget-ed by au_ren_or_cpup() */
24523 +               au_set_h_dptr(d, a->btgt, a->h_dst);
24524 +
24525 +       i = a->dst_inode;
24526 +       if (i) {
24527 +               if (!a->exchange) {
24528 +                       if (!au_ftest_ren(a->auren_flags, ISDIR_DST))
24529 +                               vfsub_drop_nlink(i);
24530 +                       else {
24531 +                               vfsub_dead_dir(i);
24532 +                               au_cpup_attr_timesizes(i);
24533 +                       }
24534 +                       au_update_dbrange(d, /*do_put_zero*/1);
24535 +               } else
24536 +                       au_cpup_attr_nlink(i, /*force*/1);
24537 +       } else {
24538 +               bbot = a->btgt;
24539 +               for (bindex = au_dbtop(d); bindex < bbot; bindex++)
24540 +                       au_set_h_dptr(d, bindex, NULL);
24541 +               bbot = au_dbbot(d);
24542 +               for (bindex = a->btgt + 1; bindex <= bbot; bindex++)
24543 +                       au_set_h_dptr(d, bindex, NULL);
24544 +               au_update_dbrange(d, /*do_put_zero*/0);
24545 +       }
24546 +
24547 +       if (a->exchange
24548 +           || au_ftest_ren(a->auren_flags, DIRREN)) {
24549 +               d_drop(a->src_dentry);
24550 +               if (au_ftest_ren(a->auren_flags, DIRREN))
24551 +                       au_set_dbwh(a->src_dentry, -1);
24552 +               return;
24553 +       }
24554 +
24555 +       d = a->src_dentry;
24556 +       au_set_dbwh(d, -1);
24557 +       bbot = au_dbbot(d);
24558 +       for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24559 +               h_d = au_h_dptr(d, bindex);
24560 +               if (h_d)
24561 +                       au_set_h_dptr(d, bindex, NULL);
24562 +       }
24563 +       au_set_dbbot(d, a->btgt);
24564 +
24565 +       sb = d->d_sb;
24566 +       i = a->src_inode;
24567 +       if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
24568 +               return; /* success */
24569 +
24570 +       bbot = au_ibbot(i);
24571 +       for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24572 +               h_i = au_h_iptr(i, bindex);
24573 +               if (h_i) {
24574 +                       au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
24575 +                       /* ignore this error */
24576 +                       au_set_h_iptr(i, bindex, NULL, 0);
24577 +               }
24578 +       }
24579 +       au_set_ibbot(i, a->btgt);
24580 +}
24581 +
24582 +/* ---------------------------------------------------------------------- */
24583 +
24584 +/* mainly for link(2) and rename(2) */
24585 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
24586 +{
24587 +       aufs_bindex_t bdiropq, bwh;
24588 +       struct dentry *parent;
24589 +       struct au_branch *br;
24590 +
24591 +       parent = dentry->d_parent;
24592 +       IMustLock(d_inode(parent)); /* dir is locked */
24593 +
24594 +       bdiropq = au_dbdiropq(parent);
24595 +       bwh = au_dbwh(dentry);
24596 +       br = au_sbr(dentry->d_sb, btgt);
24597 +       if (au_br_rdonly(br)
24598 +           || (0 <= bdiropq && bdiropq < btgt)
24599 +           || (0 <= bwh && bwh < btgt))
24600 +               btgt = -1;
24601 +
24602 +       AuDbg("btgt %d\n", btgt);
24603 +       return btgt;
24604 +}
24605 +
24606 +/* sets src_btop, dst_btop and btgt */
24607 +static int au_ren_wbr(struct au_ren_args *a)
24608 +{
24609 +       int err;
24610 +       struct au_wr_dir_args wr_dir_args = {
24611 +               /* .force_btgt  = -1, */
24612 +               .flags          = AuWrDir_ADD_ENTRY
24613 +       };
24614 +
24615 +       a->src_btop = au_dbtop(a->src_dentry);
24616 +       a->dst_btop = au_dbtop(a->dst_dentry);
24617 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24618 +           || au_ftest_ren(a->auren_flags, ISDIR_DST))
24619 +               au_fset_wrdir(wr_dir_args.flags, ISDIR);
24620 +       wr_dir_args.force_btgt = a->src_btop;
24621 +       if (a->dst_inode && a->dst_btop < a->src_btop)
24622 +               wr_dir_args.force_btgt = a->dst_btop;
24623 +       wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
24624 +       err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
24625 +       a->btgt = err;
24626 +       if (a->exchange)
24627 +               au_update_dbtop(a->dst_dentry);
24628 +
24629 +       return err;
24630 +}
24631 +
24632 +static void au_ren_dt(struct au_ren_args *a)
24633 +{
24634 +       a->h_path.dentry = a->src_h_parent;
24635 +       au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
24636 +       if (!au_ftest_ren(a->auren_flags, ISSAMEDIR)) {
24637 +               a->h_path.dentry = a->dst_h_parent;
24638 +               au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
24639 +       }
24640 +
24641 +       au_fclr_ren(a->auren_flags, DT_DSTDIR);
24642 +       if (!au_ftest_ren(a->auren_flags, ISDIR_SRC)
24643 +           && !a->exchange)
24644 +               return;
24645 +
24646 +       a->h_path.dentry = a->src_h_dentry;
24647 +       au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
24648 +       if (d_is_positive(a->dst_h_dentry)) {
24649 +               au_fset_ren(a->auren_flags, DT_DSTDIR);
24650 +               a->h_path.dentry = a->dst_h_dentry;
24651 +               au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
24652 +       }
24653 +}
24654 +
24655 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
24656 +{
24657 +       struct dentry *h_d;
24658 +       struct inode *h_inode;
24659 +
24660 +       au_dtime_revert(a->src_dt + AuPARENT);
24661 +       if (!au_ftest_ren(a->auren_flags, ISSAMEDIR))
24662 +               au_dtime_revert(a->dst_dt + AuPARENT);
24663 +
24664 +       if (au_ftest_ren(a->auren_flags, ISDIR_SRC) && err != -EIO) {
24665 +               h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
24666 +               h_inode = d_inode(h_d);
24667 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
24668 +               au_dtime_revert(a->src_dt + AuCHILD);
24669 +               inode_unlock(h_inode);
24670 +
24671 +               if (au_ftest_ren(a->auren_flags, DT_DSTDIR)) {
24672 +                       h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
24673 +                       h_inode = d_inode(h_d);
24674 +                       inode_lock_nested(h_inode, AuLsc_I_CHILD);
24675 +                       au_dtime_revert(a->dst_dt + AuCHILD);
24676 +                       inode_unlock(h_inode);
24677 +               }
24678 +       }
24679 +}
24680 +
24681 +/* ---------------------------------------------------------------------- */
24682 +
24683 +int aufs_rename(struct user_namespace *userns,
24684 +               struct inode *_src_dir, struct dentry *_src_dentry,
24685 +               struct inode *_dst_dir, struct dentry *_dst_dentry,
24686 +               unsigned int _flags)
24687 +{
24688 +       int err, lock_flags;
24689 +       void *rev;
24690 +       /* reduce stack space */
24691 +       struct au_ren_args *a;
24692 +       struct au_pin pin;
24693 +
24694 +       AuDbg("%pd, %pd, 0x%x\n", _src_dentry, _dst_dentry, _flags);
24695 +       IMustLock(_src_dir);
24696 +       IMustLock(_dst_dir);
24697 +
24698 +       err = -EINVAL;
24699 +       if (unlikely(_flags & RENAME_WHITEOUT))
24700 +               goto out;
24701 +
24702 +       err = -ENOMEM;
24703 +       BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
24704 +       a = kzalloc(sizeof(*a), GFP_NOFS);
24705 +       if (unlikely(!a))
24706 +               goto out;
24707 +
24708 +       a->flags = _flags;
24709 +       BUILD_BUG_ON(sizeof(a->exchange) == sizeof(u8)
24710 +                    && RENAME_EXCHANGE > U8_MAX);
24711 +       a->exchange = _flags & RENAME_EXCHANGE;
24712 +       a->src_dir = _src_dir;
24713 +       a->src_dentry = _src_dentry;
24714 +       a->src_inode = NULL;
24715 +       if (d_really_is_positive(a->src_dentry))
24716 +               a->src_inode = d_inode(a->src_dentry);
24717 +       a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
24718 +       a->dst_dir = _dst_dir;
24719 +       a->dst_dentry = _dst_dentry;
24720 +       a->dst_inode = NULL;
24721 +       if (d_really_is_positive(a->dst_dentry))
24722 +               a->dst_inode = d_inode(a->dst_dentry);
24723 +       a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
24724 +       if (a->dst_inode) {
24725 +               /*
24726 +                * if EXCHANGE && src is non-dir && dst is dir,
24727 +                * dst is not locked.
24728 +                */
24729 +               /* IMustLock(a->dst_inode); */
24730 +               au_igrab(a->dst_inode);
24731 +       }
24732 +
24733 +       err = -ENOTDIR;
24734 +       lock_flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
24735 +       if (d_is_dir(a->src_dentry)) {
24736 +               au_fset_ren(a->auren_flags, ISDIR_SRC);
24737 +               if (unlikely(!a->exchange
24738 +                            && d_really_is_positive(a->dst_dentry)
24739 +                            && !d_is_dir(a->dst_dentry)))
24740 +                       goto out_free;
24741 +               lock_flags |= AuLock_DIRS;
24742 +       }
24743 +       if (a->dst_inode && d_is_dir(a->dst_dentry)) {
24744 +               au_fset_ren(a->auren_flags, ISDIR_DST);
24745 +               if (unlikely(!a->exchange
24746 +                            && d_really_is_positive(a->src_dentry)
24747 +                            && !d_is_dir(a->src_dentry)))
24748 +                       goto out_free;
24749 +               lock_flags |= AuLock_DIRS;
24750 +       }
24751 +       err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
24752 +                                       lock_flags);
24753 +       if (unlikely(err))
24754 +               goto out_free;
24755 +
24756 +       err = au_d_hashed_positive(a->src_dentry);
24757 +       if (unlikely(err))
24758 +               goto out_unlock;
24759 +       err = -ENOENT;
24760 +       if (a->dst_inode) {
24761 +               /*
24762 +                * If it is a dir, VFS unhash it before this
24763 +                * function. It means we cannot rely upon d_unhashed().
24764 +                */
24765 +               if (unlikely(!a->dst_inode->i_nlink))
24766 +                       goto out_unlock;
24767 +               if (!au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24768 +                       err = au_d_hashed_positive(a->dst_dentry);
24769 +                       if (unlikely(err && !a->exchange))
24770 +                               goto out_unlock;
24771 +               } else if (unlikely(IS_DEADDIR(a->dst_inode)))
24772 +                       goto out_unlock;
24773 +       } else if (unlikely(d_unhashed(a->dst_dentry)))
24774 +               goto out_unlock;
24775 +
24776 +       /*
24777 +        * is it possible?
24778 +        * yes, it happened (in linux-3.3-rcN) but I don't know why.
24779 +        * there may exist a problem somewhere else.
24780 +        */
24781 +       err = -EINVAL;
24782 +       if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
24783 +               goto out_unlock;
24784 +
24785 +       au_fset_ren(a->auren_flags, ISSAMEDIR); /* temporary */
24786 +       di_write_lock_parent(a->dst_parent);
24787 +
24788 +       /* which branch we process */
24789 +       err = au_ren_wbr(a);
24790 +       if (unlikely(err < 0))
24791 +               goto out_parent;
24792 +       a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
24793 +       a->h_path.mnt = au_br_mnt(a->br);
24794 +
24795 +       /* are they available to be renamed */
24796 +       err = au_ren_may_dir(a);
24797 +       if (unlikely(err))
24798 +               goto out_children;
24799 +
24800 +       /* prepare the writable parent dir on the same branch */
24801 +       if (a->dst_btop == a->btgt) {
24802 +               au_fset_ren(a->auren_flags, WHDST);
24803 +       } else {
24804 +               err = au_cpup_dirs(a->dst_dentry, a->btgt);
24805 +               if (unlikely(err))
24806 +                       goto out_children;
24807 +       }
24808 +
24809 +       err = 0;
24810 +       if (!a->exchange) {
24811 +               if (a->src_dir != a->dst_dir) {
24812 +                       /*
24813 +                        * this temporary unlock is safe,
24814 +                        * because both dir->i_mutex are locked.
24815 +                        */
24816 +                       di_write_unlock(a->dst_parent);
24817 +                       di_write_lock_parent(a->src_parent);
24818 +                       err = au_wr_dir_need_wh(a->src_dentry,
24819 +                                               au_ftest_ren(a->auren_flags,
24820 +                                                            ISDIR_SRC),
24821 +                                               &a->btgt);
24822 +                       di_write_unlock(a->src_parent);
24823 +                       di_write_lock2_parent(a->src_parent, a->dst_parent,
24824 +                                             /*isdir*/1);
24825 +                       au_fclr_ren(a->auren_flags, ISSAMEDIR);
24826 +               } else
24827 +                       err = au_wr_dir_need_wh(a->src_dentry,
24828 +                                               au_ftest_ren(a->auren_flags,
24829 +                                                            ISDIR_SRC),
24830 +                                               &a->btgt);
24831 +       }
24832 +       if (unlikely(err < 0))
24833 +               goto out_children;
24834 +       if (err)
24835 +               au_fset_ren(a->auren_flags, WHSRC);
24836 +
24837 +       /* cpup src */
24838 +       if (a->src_btop != a->btgt) {
24839 +               err = au_pin(&pin, a->src_dentry, a->btgt,
24840 +                            au_opt_udba(a->src_dentry->d_sb),
24841 +                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24842 +               if (!err) {
24843 +                       struct au_cp_generic cpg = {
24844 +                               .dentry = a->src_dentry,
24845 +                               .bdst   = a->btgt,
24846 +                               .bsrc   = a->src_btop,
24847 +                               .len    = -1,
24848 +                               .pin    = &pin,
24849 +                               .flags  = AuCpup_DTIME | AuCpup_HOPEN
24850 +                       };
24851 +                       AuDebugOn(au_dbtop(a->src_dentry) != a->src_btop);
24852 +                       err = au_sio_cpup_simple(&cpg);
24853 +                       au_unpin(&pin);
24854 +               }
24855 +               if (unlikely(err))
24856 +                       goto out_children;
24857 +               a->src_btop = a->btgt;
24858 +               a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
24859 +               if (!a->exchange)
24860 +                       au_fset_ren(a->auren_flags, WHSRC);
24861 +       }
24862 +
24863 +       /* cpup dst */
24864 +       if (a->exchange && a->dst_inode
24865 +           && a->dst_btop != a->btgt) {
24866 +               err = au_pin(&pin, a->dst_dentry, a->btgt,
24867 +                            au_opt_udba(a->dst_dentry->d_sb),
24868 +                            AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24869 +               if (!err) {
24870 +                       struct au_cp_generic cpg = {
24871 +                               .dentry = a->dst_dentry,
24872 +                               .bdst   = a->btgt,
24873 +                               .bsrc   = a->dst_btop,
24874 +                               .len    = -1,
24875 +                               .pin    = &pin,
24876 +                               .flags  = AuCpup_DTIME | AuCpup_HOPEN
24877 +                       };
24878 +                       err = au_sio_cpup_simple(&cpg);
24879 +                       au_unpin(&pin);
24880 +               }
24881 +               if (unlikely(err))
24882 +                       goto out_children;
24883 +               a->dst_btop = a->btgt;
24884 +               a->dst_h_dentry = au_h_dptr(a->dst_dentry, a->btgt);
24885 +       }
24886 +
24887 +       /* lock them all */
24888 +       err = au_ren_lock(a);
24889 +       if (unlikely(err))
24890 +               /* leave the copied-up one */
24891 +               goto out_children;
24892 +
24893 +       if (!a->exchange) {
24894 +               if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
24895 +                       err = au_may_ren(a);
24896 +               else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
24897 +                       err = -ENAMETOOLONG;
24898 +               if (unlikely(err))
24899 +                       goto out_hdir;
24900 +       }
24901 +
24902 +       /* store timestamps to be revertible */
24903 +       au_ren_dt(a);
24904 +
24905 +       /* store dirren info */
24906 +       if (au_ftest_ren(a->auren_flags, DIRREN)) {
24907 +               err = au_dr_rename(a->src_dentry, a->btgt,
24908 +                                  &a->dst_dentry->d_name, &rev);
24909 +               AuTraceErr(err);
24910 +               if (unlikely(err))
24911 +                       goto out_dt;
24912 +       }
24913 +
24914 +       /* here we go */
24915 +       err = do_rename(a);
24916 +       if (unlikely(err))
24917 +               goto out_dirren;
24918 +
24919 +       if (au_ftest_ren(a->auren_flags, DIRREN))
24920 +               au_dr_rename_fin(a->src_dentry, a->btgt, rev);
24921 +
24922 +       /* update dir attributes */
24923 +       au_ren_refresh_dir(a);
24924 +
24925 +       /* dput/iput all lower dentries */
24926 +       au_ren_refresh(a);
24927 +
24928 +       goto out_hdir; /* success */
24929 +
24930 +out_dirren:
24931 +       if (au_ftest_ren(a->auren_flags, DIRREN))
24932 +               au_dr_rename_rev(a->src_dentry, a->btgt, rev);
24933 +out_dt:
24934 +       au_ren_rev_dt(err, a);
24935 +out_hdir:
24936 +       au_ren_unlock(a);
24937 +out_children:
24938 +       au_nhash_wh_free(&a->whlist);
24939 +       if (err && a->dst_inode && a->dst_btop != a->btgt) {
24940 +               AuDbg("btop %d, btgt %d\n", a->dst_btop, a->btgt);
24941 +               au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
24942 +               au_set_dbtop(a->dst_dentry, a->dst_btop);
24943 +       }
24944 +out_parent:
24945 +       if (!err) {
24946 +               if (d_unhashed(a->src_dentry))
24947 +                       au_fset_ren(a->auren_flags, DROPPED_SRC);
24948 +               if (d_unhashed(a->dst_dentry))
24949 +                       au_fset_ren(a->auren_flags, DROPPED_DST);
24950 +               if (!a->exchange)
24951 +                       d_move(a->src_dentry, a->dst_dentry);
24952 +               else {
24953 +                       d_exchange(a->src_dentry, a->dst_dentry);
24954 +                       if (au_ftest_ren(a->auren_flags, DROPPED_DST))
24955 +                               d_drop(a->dst_dentry);
24956 +               }
24957 +               if (au_ftest_ren(a->auren_flags, DROPPED_SRC))
24958 +                       d_drop(a->src_dentry);
24959 +       } else {
24960 +               au_update_dbtop(a->dst_dentry);
24961 +               if (!a->dst_inode)
24962 +                       d_drop(a->dst_dentry);
24963 +       }
24964 +       if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24965 +               di_write_unlock(a->dst_parent);
24966 +       else
24967 +               di_write_unlock2(a->src_parent, a->dst_parent);
24968 +out_unlock:
24969 +       aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
24970 +out_free:
24971 +       iput(a->dst_inode);
24972 +       if (a->thargs)
24973 +               au_whtmp_rmdir_free(a->thargs);
24974 +       au_kfree_rcu(a);
24975 +out:
24976 +       AuTraceErr(err);
24977 +       return err;
24978 +}
24979 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
24980 --- /usr/share/empty/fs/aufs/Kconfig    1970-01-01 01:00:00.000000000 +0100
24981 +++ linux/fs/aufs/Kconfig       2022-11-05 23:02:18.959222617 +0100
24982 @@ -0,0 +1,199 @@
24983 +# SPDX-License-Identifier: GPL-2.0
24984 +config AUFS_FS
24985 +       tristate "Aufs (Advanced multi layered unification filesystem) support"
24986 +       help
24987 +       Aufs is a stackable unification filesystem such as Unionfs,
24988 +       which unifies several directories and provides a merged single
24989 +       directory.
24990 +       In the early days, aufs was entirely re-designed and
24991 +       re-implemented Unionfs Version 1.x series. Introducing many
24992 +       original ideas, approaches and improvements, it becomes totally
24993 +       different from Unionfs while keeping the basic features.
24994 +
24995 +if AUFS_FS
24996 +choice
24997 +       prompt "Maximum number of branches"
24998 +       default AUFS_BRANCH_MAX_127
24999 +       help
25000 +       Specifies the maximum number of branches (or member directories)
25001 +       in a single aufs. The larger value consumes more system
25002 +       resources and has a minor impact to performance.
25003 +config AUFS_BRANCH_MAX_127
25004 +       bool "127"
25005 +       help
25006 +       Specifies the maximum number of branches (or member directories)
25007 +       in a single aufs. The larger value consumes more system
25008 +       resources and has a minor impact to performance.
25009 +config AUFS_BRANCH_MAX_511
25010 +       bool "511"
25011 +       help
25012 +       Specifies the maximum number of branches (or member directories)
25013 +       in a single aufs. The larger value consumes more system
25014 +       resources and has a minor impact to performance.
25015 +config AUFS_BRANCH_MAX_1023
25016 +       bool "1023"
25017 +       help
25018 +       Specifies the maximum number of branches (or member directories)
25019 +       in a single aufs. The larger value consumes more system
25020 +       resources and has a minor impact to performance.
25021 +config AUFS_BRANCH_MAX_32767
25022 +       bool "32767"
25023 +       help
25024 +       Specifies the maximum number of branches (or member directories)
25025 +       in a single aufs. The larger value consumes more system
25026 +       resources and has a minor impact to performance.
25027 +endchoice
25028 +
25029 +config AUFS_SBILIST
25030 +       bool
25031 +       depends on AUFS_MAGIC_SYSRQ || PROC_FS
25032 +       default y
25033 +       help
25034 +       Automatic configuration for internal use.
25035 +       When aufs supports Magic SysRq or /proc, enabled automatically.
25036 +
25037 +config AUFS_HNOTIFY
25038 +       bool "Detect direct branch access (bypassing aufs)"
25039 +       help
25040 +       If you want to modify files on branches directly, eg. bypassing aufs,
25041 +       and want aufs to detect the changes of them fully, then enable this
25042 +       option and use 'udba=notify' mount option.
25043 +       Currently there is only one available configuration, "fsnotify".
25044 +       It will have a negative impact to the performance.
25045 +       See detail in aufs.5.
25046 +
25047 +choice
25048 +       prompt "method" if AUFS_HNOTIFY
25049 +       default AUFS_HFSNOTIFY
25050 +config AUFS_HFSNOTIFY
25051 +       bool "fsnotify"
25052 +       select FSNOTIFY
25053 +endchoice
25054 +
25055 +config AUFS_EXPORT
25056 +       bool "NFS-exportable aufs"
25057 +       depends on EXPORTFS
25058 +       help
25059 +       If you want to export your mounted aufs via NFS, then enable this
25060 +       option. There are several requirements for this configuration.
25061 +       See detail in aufs.5.
25062 +
25063 +config AUFS_INO_T_64
25064 +       bool
25065 +       depends on AUFS_EXPORT
25066 +       depends on 64BIT && !(ALPHA || S390)
25067 +       default y
25068 +       help
25069 +       Automatic configuration for internal use.
25070 +       /* typedef unsigned long/int __kernel_ino_t */
25071 +       /* alpha and s390x are int */
25072 +
25073 +config AUFS_XATTR
25074 +       bool "support for XATTR/EA (including Security Labels)"
25075 +       help
25076 +       If your branch fs supports XATTR/EA and you want to make them
25077 +       available in aufs too, then enable this opsion and specify the
25078 +       branch attributes for EA.
25079 +       See detail in aufs.5.
25080 +
25081 +config AUFS_FHSM
25082 +       bool "File-based Hierarchical Storage Management"
25083 +       help
25084 +       Hierarchical Storage Management (or HSM) is a well-known feature
25085 +       in the storage world. Aufs provides this feature as file-based.
25086 +       with multiple branches.
25087 +       These multiple branches are prioritized, ie. the topmost one
25088 +       should be the fastest drive and be used heavily.
25089 +
25090 +config AUFS_RDU
25091 +       bool "Readdir in userspace"
25092 +       help
25093 +       Aufs has two methods to provide a merged view for a directory,
25094 +       by a user-space library and by kernel-space natively. The latter
25095 +       is always enabled but sometimes large and slow.
25096 +       If you enable this option, install the library in aufs2-util
25097 +       package, and set some environment variables for your readdir(3),
25098 +       then the work will be handled in user-space which generally
25099 +       shows better performance in most cases.
25100 +       See detail in aufs.5.
25101 +
25102 +config AUFS_DIRREN
25103 +       bool "Workaround for rename(2)-ing a directory"
25104 +       help
25105 +       By default, aufs returns EXDEV error in renameing a dir who has
25106 +       his child on the lower branch, since it is a bad idea to issue
25107 +       rename(2) internally for every lower branch. But user may not
25108 +       accept this behaviour. So here is a workaround to allow such
25109 +       rename(2) and store some extra information on the writable
25110 +       branch. Obviously this costs high (and I don't like it).
25111 +       To use this feature, you need to enable this configuration AND
25112 +       to specify the mount option `dirren.'
25113 +       See details in aufs.5 and the design documents.
25114 +
25115 +config AUFS_SHWH
25116 +       bool "Show whiteouts"
25117 +       help
25118 +       If you want to make the whiteouts in aufs visible, then enable
25119 +       this option and specify 'shwh' mount option. Although it may
25120 +       sounds like philosophy or something, but in technically it
25121 +       simply shows the name of whiteout with keeping its behaviour.
25122 +
25123 +config AUFS_BR_RAMFS
25124 +       bool "Ramfs (initramfs/rootfs) as an aufs branch"
25125 +       help
25126 +       If you want to use ramfs as an aufs branch fs, then enable this
25127 +       option. Generally tmpfs is recommended.
25128 +       Aufs prohibited them to be a branch fs by default, because
25129 +       initramfs becomes unusable after switch_root or something
25130 +       generally. If you sets initramfs as an aufs branch and boot your
25131 +       system by switch_root, you will meet a problem easily since the
25132 +       files in initramfs may be inaccessible.
25133 +       Unless you are going to use ramfs as an aufs branch fs without
25134 +       switch_root or something, leave it N.
25135 +
25136 +config AUFS_BR_FUSE
25137 +       bool "Fuse fs as an aufs branch"
25138 +       depends on FUSE_FS
25139 +       select AUFS_POLL
25140 +       help
25141 +       If you want to use fuse-based userspace filesystem as an aufs
25142 +       branch fs, then enable this option.
25143 +       It implements the internal poll(2) operation which is
25144 +       implemented by fuse only (curretnly).
25145 +
25146 +config AUFS_POLL
25147 +       bool
25148 +       help
25149 +       Automatic configuration for internal use.
25150 +
25151 +config AUFS_BR_HFSPLUS
25152 +       bool "Hfsplus as an aufs branch"
25153 +       depends on HFSPLUS_FS
25154 +       default y
25155 +       help
25156 +       If you want to use hfsplus fs as an aufs branch fs, then enable
25157 +       this option. This option introduces a small overhead at
25158 +       copying-up a file on hfsplus.
25159 +
25160 +config AUFS_BDEV_LOOP
25161 +       bool
25162 +       depends on BLK_DEV_LOOP
25163 +       default y
25164 +       help
25165 +       Automatic configuration for internal use.
25166 +       Convert =[ym] into =y.
25167 +
25168 +config AUFS_DEBUG
25169 +       bool "Debug aufs"
25170 +       help
25171 +       Enable this to compile aufs internal debug code.
25172 +       It will have a negative impact to the performance.
25173 +
25174 +config AUFS_MAGIC_SYSRQ
25175 +       bool
25176 +       depends on AUFS_DEBUG && MAGIC_SYSRQ
25177 +       default y
25178 +       help
25179 +       Automatic configuration for internal use.
25180 +       When aufs supports Magic SysRq, enabled automatically.
25181 +endif
25182 diff -urN /usr/share/empty/fs/aufs/lcnt.h linux/fs/aufs/lcnt.h
25183 --- /usr/share/empty/fs/aufs/lcnt.h     1970-01-01 01:00:00.000000000 +0100
25184 +++ linux/fs/aufs/lcnt.h        2022-11-05 23:02:18.965889284 +0100
25185 @@ -0,0 +1,186 @@
25186 +/* SPDX-License-Identifier: GPL-2.0 */
25187 +/*
25188 + * Copyright (C) 2018-2022 Junjiro R. Okajima
25189 + *
25190 + * This program is free software; you can redistribute it and/or modify
25191 + * it under the terms of the GNU General Public License as published by
25192 + * the Free Software Foundation; either version 2 of the License, or
25193 + * (at your option) any later version.
25194 + *
25195 + * This program is distributed in the hope that it will be useful,
25196 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25197 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25198 + * GNU General Public License for more details.
25199 + *
25200 + * You should have received a copy of the GNU General Public License
25201 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25202 + */
25203 +
25204 +/*
25205 + * simple long counter wrapper
25206 + */
25207 +
25208 +#ifndef __AUFS_LCNT_H__
25209 +#define __AUFS_LCNT_H__
25210 +
25211 +#ifdef __KERNEL__
25212 +
25213 +#include "debug.h"
25214 +
25215 +#define AuLCntATOMIC   1
25216 +#define AuLCntPCPUCNT  2
25217 +/*
25218 + * why does percpu_refcount require extra synchronize_rcu()s in
25219 + * au_br_do_free()
25220 + */
25221 +#define AuLCntPCPUREF  3
25222 +
25223 +/* #define AuLCntChosen        AuLCntATOMIC */
25224 +#define AuLCntChosen   AuLCntPCPUCNT
25225 +/* #define AuLCntChosen        AuLCntPCPUREF */
25226 +
25227 +#if AuLCntChosen == AuLCntATOMIC
25228 +#include <linux/atomic.h>
25229 +
25230 +typedef atomic_long_t au_lcnt_t;
25231 +
25232 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25233 +{
25234 +       atomic_long_set(cnt, 0);
25235 +       return 0;
25236 +}
25237 +
25238 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25239 +{
25240 +       /* empty */
25241 +}
25242 +
25243 +static inline void au_lcnt_fin(au_lcnt_t *cnt __maybe_unused,
25244 +                              int do_sync __maybe_unused)
25245 +{
25246 +       /* empty */
25247 +}
25248 +
25249 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25250 +{
25251 +       atomic_long_inc(cnt);
25252 +}
25253 +
25254 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25255 +{
25256 +       atomic_long_dec(cnt);
25257 +}
25258 +
25259 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25260 +{
25261 +       return atomic_long_read(cnt);
25262 +}
25263 +#endif
25264 +
25265 +#if AuLCntChosen == AuLCntPCPUCNT
25266 +#include <linux/percpu_counter.h>
25267 +
25268 +typedef struct percpu_counter au_lcnt_t;
25269 +
25270 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25271 +{
25272 +       return percpu_counter_init(cnt, 0, GFP_NOFS);
25273 +}
25274 +
25275 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25276 +{
25277 +       /* empty */
25278 +}
25279 +
25280 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync __maybe_unused)
25281 +{
25282 +       percpu_counter_destroy(cnt);
25283 +}
25284 +
25285 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25286 +{
25287 +       percpu_counter_inc(cnt);
25288 +}
25289 +
25290 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25291 +{
25292 +       percpu_counter_dec(cnt);
25293 +}
25294 +
25295 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25296 +{
25297 +       s64 n;
25298 +
25299 +       n = percpu_counter_sum(cnt);
25300 +       BUG_ON(n < 0);
25301 +       if (LONG_MAX != LLONG_MAX
25302 +           && n > LONG_MAX)
25303 +               AuWarn1("%s\n", "wrap-around");
25304 +
25305 +       return n;
25306 +}
25307 +#endif
25308 +
25309 +#if AuLCntChosen == AuLCntPCPUREF
25310 +#include <linux/percpu-refcount.h>
25311 +
25312 +typedef struct percpu_ref au_lcnt_t;
25313 +
25314 +static inline int au_lcnt_init(au_lcnt_t *cnt, percpu_ref_func_t *release)
25315 +{
25316 +       if (!release)
25317 +               release = percpu_ref_exit;
25318 +       return percpu_ref_init(cnt, release, /*percpu mode*/0, GFP_NOFS);
25319 +}
25320 +
25321 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25322 +{
25323 +       synchronize_rcu();
25324 +}
25325 +
25326 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync)
25327 +{
25328 +       percpu_ref_kill(cnt);
25329 +       if (do_sync)
25330 +               au_lcnt_wait_for_fin(cnt);
25331 +}
25332 +
25333 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25334 +{
25335 +       percpu_ref_get(cnt);
25336 +}
25337 +
25338 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25339 +{
25340 +       percpu_ref_put(cnt);
25341 +}
25342 +
25343 +/*
25344 + * avoid calling this func as possible.
25345 + */
25346 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev)
25347 +{
25348 +       long l;
25349 +
25350 +       percpu_ref_switch_to_atomic_sync(cnt);
25351 +       l = atomic_long_read(&cnt->count);
25352 +       if (do_rev)
25353 +               percpu_ref_switch_to_percpu(cnt);
25354 +
25355 +       /* percpu_ref is initialized by 1 instead of 0 */
25356 +       return l - 1;
25357 +}
25358 +#endif
25359 +
25360 +#ifdef CONFIG_AUFS_DEBUG
25361 +#define AuLCntZero(val) do {                   \
25362 +       long l = val;                           \
25363 +       if (l)                                  \
25364 +               AuDbg("%s = %ld\n", #val, l);   \
25365 +} while (0)
25366 +#else
25367 +#define AuLCntZero(val)                do {} while (0)
25368 +#endif
25369 +
25370 +#endif /* __KERNEL__ */
25371 +#endif /* __AUFS_LCNT_H__ */
25372 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
25373 --- /usr/share/empty/fs/aufs/loop.c     1970-01-01 01:00:00.000000000 +0100
25374 +++ linux/fs/aufs/loop.c        2022-11-05 23:02:18.965889284 +0100
25375 @@ -0,0 +1,148 @@
25376 +// SPDX-License-Identifier: GPL-2.0
25377 +/*
25378 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25379 + *
25380 + * This program is free software; you can redistribute it and/or modify
25381 + * it under the terms of the GNU General Public License as published by
25382 + * the Free Software Foundation; either version 2 of the License, or
25383 + * (at your option) any later version.
25384 + *
25385 + * This program is distributed in the hope that it will be useful,
25386 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25387 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25388 + * GNU General Public License for more details.
25389 + *
25390 + * You should have received a copy of the GNU General Public License
25391 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25392 + */
25393 +
25394 +/*
25395 + * support for loopback block device as a branch
25396 + */
25397 +
25398 +#include "aufs.h"
25399 +
25400 +/* added into drivers/block/loop.c */
25401 +static struct file *(*backing_file_func)(struct super_block *sb);
25402 +
25403 +/*
25404 + * test if two lower dentries have overlapping branches.
25405 + */
25406 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
25407 +{
25408 +       struct super_block *h_sb;
25409 +       struct file *backing_file;
25410 +
25411 +       if (unlikely(!backing_file_func)) {
25412 +               /* don't load "loop" module here */
25413 +               backing_file_func = symbol_get(loop_backing_file);
25414 +               if (unlikely(!backing_file_func))
25415 +                       /* "loop" module is not loaded */
25416 +                       return 0;
25417 +       }
25418 +
25419 +       h_sb = h_adding->d_sb;
25420 +       backing_file = backing_file_func(h_sb);
25421 +       if (!backing_file)
25422 +               return 0;
25423 +
25424 +       h_adding = backing_file->f_path.dentry;
25425 +       /*
25426 +        * h_adding can be local NFS.
25427 +        * in this case aufs cannot detect the loop.
25428 +        */
25429 +       if (unlikely(h_adding->d_sb == sb))
25430 +               return 1;
25431 +       return !!au_test_subdir(h_adding, sb->s_root);
25432 +}
25433 +
25434 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
25435 +int au_test_loopback_kthread(void)
25436 +{
25437 +       int ret;
25438 +       struct task_struct *tsk = current;
25439 +       char c, comm[sizeof(tsk->comm)];
25440 +
25441 +       ret = 0;
25442 +       if (tsk->flags & PF_KTHREAD) {
25443 +               get_task_comm(comm, tsk);
25444 +               c = comm[4];
25445 +               ret = ('0' <= c && c <= '9'
25446 +                      && !strncmp(comm, "loop", 4));
25447 +       }
25448 +
25449 +       return ret;
25450 +}
25451 +
25452 +/* ---------------------------------------------------------------------- */
25453 +
25454 +#define au_warn_loopback_step  16
25455 +static int au_warn_loopback_nelem = au_warn_loopback_step;
25456 +static unsigned long *au_warn_loopback_array;
25457 +
25458 +void au_warn_loopback(struct super_block *h_sb)
25459 +{
25460 +       int i, new_nelem;
25461 +       unsigned long *a, magic;
25462 +       static DEFINE_SPINLOCK(spin);
25463 +
25464 +       magic = h_sb->s_magic;
25465 +       spin_lock(&spin);
25466 +       a = au_warn_loopback_array;
25467 +       for (i = 0; i < au_warn_loopback_nelem && *a; i++)
25468 +               if (a[i] == magic) {
25469 +                       spin_unlock(&spin);
25470 +                       return;
25471 +               }
25472 +
25473 +       /* h_sb is new to us, print it */
25474 +       if (i < au_warn_loopback_nelem) {
25475 +               a[i] = magic;
25476 +               goto pr;
25477 +       }
25478 +
25479 +       /* expand the array */
25480 +       new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
25481 +       a = au_kzrealloc(au_warn_loopback_array,
25482 +                        au_warn_loopback_nelem * sizeof(unsigned long),
25483 +                        new_nelem * sizeof(unsigned long), GFP_ATOMIC,
25484 +                        /*may_shrink*/0);
25485 +       if (a) {
25486 +               au_warn_loopback_nelem = new_nelem;
25487 +               au_warn_loopback_array = a;
25488 +               a[i] = magic;
25489 +               goto pr;
25490 +       }
25491 +
25492 +       spin_unlock(&spin);
25493 +       AuWarn1("realloc failed, ignored\n");
25494 +       return;
25495 +
25496 +pr:
25497 +       spin_unlock(&spin);
25498 +       pr_warn("you may want to try another patch for loopback file "
25499 +               "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
25500 +}
25501 +
25502 +int au_loopback_init(void)
25503 +{
25504 +       int err;
25505 +       struct super_block *sb __maybe_unused;
25506 +
25507 +       BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(*au_warn_loopback_array));
25508 +
25509 +       err = 0;
25510 +       au_warn_loopback_array = kcalloc(au_warn_loopback_step,
25511 +                                        sizeof(unsigned long), GFP_NOFS);
25512 +       if (unlikely(!au_warn_loopback_array))
25513 +               err = -ENOMEM;
25514 +
25515 +       return err;
25516 +}
25517 +
25518 +void au_loopback_fin(void)
25519 +{
25520 +       if (backing_file_func)
25521 +               symbol_put(loop_backing_file);
25522 +       au_kfree_try_rcu(au_warn_loopback_array);
25523 +}
25524 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
25525 --- /usr/share/empty/fs/aufs/loop.h     1970-01-01 01:00:00.000000000 +0100
25526 +++ linux/fs/aufs/loop.h        2022-11-05 23:02:18.965889284 +0100
25527 @@ -0,0 +1,55 @@
25528 +/* SPDX-License-Identifier: GPL-2.0 */
25529 +/*
25530 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25531 + *
25532 + * This program is free software; you can redistribute it and/or modify
25533 + * it under the terms of the GNU General Public License as published by
25534 + * the Free Software Foundation; either version 2 of the License, or
25535 + * (at your option) any later version.
25536 + *
25537 + * This program is distributed in the hope that it will be useful,
25538 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25539 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25540 + * GNU General Public License for more details.
25541 + *
25542 + * You should have received a copy of the GNU General Public License
25543 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25544 + */
25545 +
25546 +/*
25547 + * support for loopback mount as a branch
25548 + */
25549 +
25550 +#ifndef __AUFS_LOOP_H__
25551 +#define __AUFS_LOOP_H__
25552 +
25553 +#ifdef __KERNEL__
25554 +
25555 +struct dentry;
25556 +struct super_block;
25557 +
25558 +#ifdef CONFIG_AUFS_BDEV_LOOP
25559 +/* drivers/block/loop.c */
25560 +struct file *loop_backing_file(struct super_block *sb);
25561 +
25562 +/* loop.c */
25563 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
25564 +int au_test_loopback_kthread(void);
25565 +void au_warn_loopback(struct super_block *h_sb);
25566 +
25567 +int au_loopback_init(void);
25568 +void au_loopback_fin(void);
25569 +#else
25570 +AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
25571 +
25572 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
25573 +          struct dentry *h_adding)
25574 +AuStubInt0(au_test_loopback_kthread, void)
25575 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
25576 +
25577 +AuStubInt0(au_loopback_init, void)
25578 +AuStubVoid(au_loopback_fin, void)
25579 +#endif /* BLK_DEV_LOOP */
25580 +
25581 +#endif /* __KERNEL__ */
25582 +#endif /* __AUFS_LOOP_H__ */
25583 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
25584 --- /usr/share/empty/fs/aufs/magic.mk   1970-01-01 01:00:00.000000000 +0100
25585 +++ linux/fs/aufs/magic.mk      2022-11-05 23:02:18.965889284 +0100
25586 @@ -0,0 +1,31 @@
25587 +# SPDX-License-Identifier: GPL-2.0
25588 +
25589 +# defined in ${srctree}/fs/fuse/inode.c
25590 +# tristate
25591 +ifdef CONFIG_FUSE_FS
25592 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
25593 +endif
25594 +
25595 +# defined in ${srctree}/fs/xfs/xfs_sb.h
25596 +# tristate
25597 +ifdef CONFIG_XFS_FS
25598 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
25599 +endif
25600 +
25601 +# defined in ${srctree}/fs/configfs/mount.c
25602 +# tristate
25603 +ifdef CONFIG_CONFIGFS_FS
25604 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
25605 +endif
25606 +
25607 +# defined in ${srctree}/fs/ubifs/ubifs.h
25608 +# tristate
25609 +ifdef CONFIG_UBIFS_FS
25610 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
25611 +endif
25612 +
25613 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
25614 +# tristate
25615 +ifdef CONFIG_HFSPLUS_FS
25616 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
25617 +endif
25618 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
25619 --- /usr/share/empty/fs/aufs/Makefile   1970-01-01 01:00:00.000000000 +0100
25620 +++ linux/fs/aufs/Makefile      2022-11-05 23:02:18.959222617 +0100
25621 @@ -0,0 +1,46 @@
25622 +# SPDX-License-Identifier: GPL-2.0
25623 +
25624 +include ${src}/magic.mk
25625 +ifeq (${CONFIG_AUFS_FS},m)
25626 +include ${src}/conf.mk
25627 +endif
25628 +-include ${src}/priv_def.mk
25629 +
25630 +# cf. include/linux/kernel.h
25631 +# enable pr_debug
25632 +ccflags-y += -DDEBUG
25633 +# sparse requires the full pathname
25634 +ifdef M
25635 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
25636 +else
25637 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
25638 +endif
25639 +
25640 +obj-$(CONFIG_AUFS_FS) += aufs.o
25641 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o fsctx.o \
25642 +       wkq.o vfsub.o dcsub.o \
25643 +       cpup.o whout.o wbr_policy.o \
25644 +       dinfo.o dentry.o \
25645 +       dynop.o \
25646 +       finfo.o file.o f_op.o \
25647 +       dir.o vdir.o \
25648 +       iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
25649 +       mvdown.o ioctl.o
25650 +
25651 +# all are boolean
25652 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
25653 +aufs-$(CONFIG_SYSFS) += sysfs.o
25654 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
25655 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
25656 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
25657 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
25658 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
25659 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
25660 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
25661 +aufs-$(CONFIG_AUFS_DIRREN) += dirren.o
25662 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
25663 +aufs-$(CONFIG_AUFS_POLL) += poll.o
25664 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
25665 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
25666 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
25667 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
25668 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
25669 --- /usr/share/empty/fs/aufs/module.c   1970-01-01 01:00:00.000000000 +0100
25670 +++ linux/fs/aufs/module.c      2022-11-05 23:02:18.965889284 +0100
25671 @@ -0,0 +1,273 @@
25672 +// SPDX-License-Identifier: GPL-2.0
25673 +/*
25674 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25675 + *
25676 + * This program is free software; you can redistribute it and/or modify
25677 + * it under the terms of the GNU General Public License as published by
25678 + * the Free Software Foundation; either version 2 of the License, or
25679 + * (at your option) any later version.
25680 + *
25681 + * This program is distributed in the hope that it will be useful,
25682 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25683 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25684 + * GNU General Public License for more details.
25685 + *
25686 + * You should have received a copy of the GNU General Public License
25687 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25688 + */
25689 +
25690 +/*
25691 + * module global variables and operations
25692 + */
25693 +
25694 +#include <linux/module.h>
25695 +#include <linux/seq_file.h>
25696 +#include "aufs.h"
25697 +
25698 +/* shrinkable realloc */
25699 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink)
25700 +{
25701 +       size_t sz;
25702 +       int diff;
25703 +
25704 +       sz = 0;
25705 +       diff = -1;
25706 +       if (p) {
25707 +#if 0 /* unused */
25708 +               if (!new_sz) {
25709 +                       au_kfree_rcu(p);
25710 +                       p = NULL;
25711 +                       goto out;
25712 +               }
25713 +#else
25714 +               AuDebugOn(!new_sz);
25715 +#endif
25716 +               sz = ksize(p);
25717 +               diff = au_kmidx_sub(sz, new_sz);
25718 +       }
25719 +       if (sz && !diff)
25720 +               goto out;
25721 +
25722 +       if (sz < new_sz)
25723 +               /* expand or SLOB */
25724 +               p = krealloc(p, new_sz, gfp);
25725 +       else if (new_sz < sz && may_shrink) {
25726 +               /* shrink */
25727 +               void *q;
25728 +
25729 +               q = kmalloc(new_sz, gfp);
25730 +               if (q) {
25731 +                       if (p) {
25732 +                               memcpy(q, p, new_sz);
25733 +                               au_kfree_try_rcu(p);
25734 +                       }
25735 +                       p = q;
25736 +               } else
25737 +                       p = NULL;
25738 +       }
25739 +
25740 +out:
25741 +       return p;
25742 +}
25743 +
25744 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
25745 +                  int may_shrink)
25746 +{
25747 +       p = au_krealloc(p, new_sz, gfp, may_shrink);
25748 +       if (p && new_sz > nused)
25749 +               memset(p + nused, 0, new_sz - nused);
25750 +       return p;
25751 +}
25752 +
25753 +/* ---------------------------------------------------------------------- */
25754 +/*
25755 + * aufs caches
25756 + */
25757 +struct kmem_cache *au_cache[AuCache_Last];
25758 +
25759 +static void au_cache_fin(void)
25760 +{
25761 +       int i;
25762 +
25763 +       /*
25764 +        * Make sure all delayed rcu free inodes are flushed before we
25765 +        * destroy cache.
25766 +        */
25767 +       rcu_barrier();
25768 +
25769 +       /* excluding AuCache_HNOTIFY */
25770 +       BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
25771 +       for (i = 0; i < AuCache_HNOTIFY; i++) {
25772 +               kmem_cache_destroy(au_cache[i]);
25773 +               au_cache[i] = NULL;
25774 +       }
25775 +}
25776 +
25777 +static int __init au_cache_init(void)
25778 +{
25779 +       au_cache[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
25780 +       if (au_cache[AuCache_DINFO])
25781 +               /* SLAB_DESTROY_BY_RCU */
25782 +               au_cache[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
25783 +                                                      au_icntnr_init_once);
25784 +       if (au_cache[AuCache_ICNTNR])
25785 +               au_cache[AuCache_FINFO] = AuCacheCtor(au_finfo,
25786 +                                                     au_fi_init_once);
25787 +       if (au_cache[AuCache_FINFO])
25788 +               au_cache[AuCache_VDIR] = AuCache(au_vdir);
25789 +       if (au_cache[AuCache_VDIR])
25790 +               au_cache[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
25791 +       if (au_cache[AuCache_DEHSTR])
25792 +               return 0;
25793 +
25794 +       au_cache_fin();
25795 +       return -ENOMEM;
25796 +}
25797 +
25798 +/* ---------------------------------------------------------------------- */
25799 +
25800 +int au_dir_roflags;
25801 +
25802 +#ifdef CONFIG_AUFS_SBILIST
25803 +/*
25804 + * iterate_supers_type() doesn't protect us from
25805 + * remounting (branch management)
25806 + */
25807 +struct hlist_bl_head au_sbilist;
25808 +#endif
25809 +
25810 +/*
25811 + * functions for module interface.
25812 + */
25813 +MODULE_LICENSE("GPL");
25814 +/* MODULE_LICENSE("GPL v2"); */
25815 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
25816 +MODULE_DESCRIPTION(AUFS_NAME
25817 +       " -- Advanced multi layered unification filesystem");
25818 +MODULE_VERSION(AUFS_VERSION);
25819 +MODULE_ALIAS_FS(AUFS_NAME);
25820 +
25821 +/* this module parameter has no meaning when SYSFS is disabled */
25822 +int sysaufs_brs = 1;
25823 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
25824 +module_param_named(brs, sysaufs_brs, int, 0444);
25825 +
25826 +/* this module parameter has no meaning when USER_NS is disabled */
25827 +bool au_userns;
25828 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
25829 +module_param_named(allow_userns, au_userns, bool, 0444);
25830 +
25831 +/* ---------------------------------------------------------------------- */
25832 +
25833 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
25834 +
25835 +int au_seq_path(struct seq_file *seq, struct path *path)
25836 +{
25837 +       int err;
25838 +
25839 +       err = seq_path(seq, path, au_esc_chars);
25840 +       if (err >= 0)
25841 +               err = 0;
25842 +       else
25843 +               err = -ENOMEM;
25844 +
25845 +       return err;
25846 +}
25847 +
25848 +/* ---------------------------------------------------------------------- */
25849 +
25850 +static int __init aufs_init(void)
25851 +{
25852 +       int err, i;
25853 +       char *p;
25854 +
25855 +       p = au_esc_chars;
25856 +       for (i = 1; i <= ' '; i++)
25857 +               *p++ = i;
25858 +       *p++ = '\\';
25859 +       *p++ = '\x7f';
25860 +       *p = 0;
25861 +
25862 +       au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
25863 +
25864 +       memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
25865 +       for (i = 0; i < AuIop_Last; i++)
25866 +               aufs_iop_nogetattr[i].getattr = NULL;
25867 +
25868 +       memset(au_cache, 0, sizeof(au_cache));  /* including hnotify */
25869 +
25870 +       au_sbilist_init();
25871 +       sysaufs_brs_init();
25872 +       au_debug_init();
25873 +       au_dy_init();
25874 +       err = sysaufs_init();
25875 +       if (unlikely(err))
25876 +               goto out;
25877 +       err = dbgaufs_init();
25878 +       if (unlikely(err))
25879 +               goto out_sysaufs;
25880 +       err = au_procfs_init();
25881 +       if (unlikely(err))
25882 +               goto out_dbgaufs;
25883 +       err = au_wkq_init();
25884 +       if (unlikely(err))
25885 +               goto out_procfs;
25886 +       err = au_loopback_init();
25887 +       if (unlikely(err))
25888 +               goto out_wkq;
25889 +       err = au_hnotify_init();
25890 +       if (unlikely(err))
25891 +               goto out_loopback;
25892 +       err = au_sysrq_init();
25893 +       if (unlikely(err))
25894 +               goto out_hin;
25895 +       err = au_cache_init();
25896 +       if (unlikely(err))
25897 +               goto out_sysrq;
25898 +
25899 +       aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
25900 +       err = register_filesystem(&aufs_fs_type);
25901 +       if (unlikely(err))
25902 +               goto out_cache;
25903 +
25904 +       /* since we define pr_fmt, call printk directly */
25905 +       printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
25906 +       goto out; /* success */
25907 +
25908 +out_cache:
25909 +       au_cache_fin();
25910 +out_sysrq:
25911 +       au_sysrq_fin();
25912 +out_hin:
25913 +       au_hnotify_fin();
25914 +out_loopback:
25915 +       au_loopback_fin();
25916 +out_wkq:
25917 +       au_wkq_fin();
25918 +out_procfs:
25919 +       au_procfs_fin();
25920 +out_dbgaufs:
25921 +       dbgaufs_fin();
25922 +out_sysaufs:
25923 +       sysaufs_fin();
25924 +       au_dy_fin();
25925 +out:
25926 +       return err;
25927 +}
25928 +
25929 +static void __exit aufs_exit(void)
25930 +{
25931 +       unregister_filesystem(&aufs_fs_type);
25932 +       au_cache_fin();
25933 +       au_sysrq_fin();
25934 +       au_hnotify_fin();
25935 +       au_loopback_fin();
25936 +       au_wkq_fin();
25937 +       au_procfs_fin();
25938 +       dbgaufs_fin();
25939 +       sysaufs_fin();
25940 +       au_dy_fin();
25941 +}
25942 +
25943 +module_init(aufs_init);
25944 +module_exit(aufs_exit);
25945 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
25946 --- /usr/share/empty/fs/aufs/module.h   1970-01-01 01:00:00.000000000 +0100
25947 +++ linux/fs/aufs/module.h      2022-11-05 23:02:18.969222617 +0100
25948 @@ -0,0 +1,180 @@
25949 +/* SPDX-License-Identifier: GPL-2.0 */
25950 +/*
25951 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25952 + *
25953 + * This program is free software; you can redistribute it and/or modify
25954 + * it under the terms of the GNU General Public License as published by
25955 + * the Free Software Foundation; either version 2 of the License, or
25956 + * (at your option) any later version.
25957 + *
25958 + * This program is distributed in the hope that it will be useful,
25959 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25960 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
25961 + * GNU General Public License for more details.
25962 + *
25963 + * You should have received a copy of the GNU General Public License
25964 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
25965 + */
25966 +
25967 +/*
25968 + * module initialization and module-global
25969 + */
25970 +
25971 +#ifndef __AUFS_MODULE_H__
25972 +#define __AUFS_MODULE_H__
25973 +
25974 +#ifdef __KERNEL__
25975 +
25976 +#include <linux/slab.h>
25977 +#include "debug.h"
25978 +#include "dentry.h"
25979 +#include "dir.h"
25980 +#include "file.h"
25981 +#include "inode.h"
25982 +
25983 +struct path;
25984 +struct seq_file;
25985 +
25986 +/* module parameters */
25987 +extern int sysaufs_brs;
25988 +extern bool au_userns;
25989 +
25990 +/* ---------------------------------------------------------------------- */
25991 +
25992 +extern int au_dir_roflags;
25993 +
25994 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink);
25995 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
25996 +                  int may_shrink);
25997 +
25998 +/*
25999 + * Comparing the size of the object with sizeof(struct rcu_head)
26000 + * case 1: object is always larger
26001 + *     --> au_kfree_rcu() or au_kfree_do_rcu()
26002 + * case 2: object is always smaller
26003 + *     --> au_kfree_small()
26004 + * case 3: object can be any size
26005 + *     --> au_kfree_try_rcu()
26006 + */
26007 +
26008 +static inline void au_kfree_do_rcu(const void *p)
26009 +{
26010 +       struct {
26011 +               struct rcu_head rcu;
26012 +       } *a = (void *)p;
26013 +
26014 +       kfree_rcu(a, rcu);
26015 +}
26016 +
26017 +#define au_kfree_rcu(_p) do {                                          \
26018 +               typeof(_p) p = (_p);                                    \
26019 +               BUILD_BUG_ON(sizeof(*p) < sizeof(struct rcu_head));     \
26020 +               if (p)                                                  \
26021 +                       au_kfree_do_rcu(p);                             \
26022 +       } while (0)
26023 +
26024 +#define au_kfree_do_sz_test(sz)        (sz >= sizeof(struct rcu_head))
26025 +#define au_kfree_sz_test(p)    (p && au_kfree_do_sz_test(ksize(p)))
26026 +
26027 +static inline void au_kfree_try_rcu(const void *p)
26028 +{
26029 +       if (!p)
26030 +               return;
26031 +       if (au_kfree_sz_test(p))
26032 +               au_kfree_do_rcu(p);
26033 +       else
26034 +               kfree(p);
26035 +}
26036 +
26037 +static inline void au_kfree_small(const void *p)
26038 +{
26039 +       if (!p)
26040 +               return;
26041 +       AuDebugOn(au_kfree_sz_test(p));
26042 +       kfree(p);
26043 +}
26044 +
26045 +static inline int au_kmidx_sub(size_t sz, size_t new_sz)
26046 +{
26047 +#ifndef CONFIG_SLOB
26048 +       return __kmalloc_index(sz, false) - __kmalloc_index(new_sz, false);
26049 +#else
26050 +       return -1; /* SLOB is untested */
26051 +#endif
26052 +}
26053 +
26054 +int au_seq_path(struct seq_file *seq, struct path *path);
26055 +
26056 +#ifdef CONFIG_PROC_FS
26057 +/* procfs.c */
26058 +int __init au_procfs_init(void);
26059 +void au_procfs_fin(void);
26060 +#else
26061 +AuStubInt0(au_procfs_init, void);
26062 +AuStubVoid(au_procfs_fin, void);
26063 +#endif
26064 +
26065 +/* ---------------------------------------------------------------------- */
26066 +
26067 +/* kmem cache */
26068 +enum {
26069 +       AuCache_DINFO,
26070 +       AuCache_ICNTNR,
26071 +       AuCache_FINFO,
26072 +       AuCache_VDIR,
26073 +       AuCache_DEHSTR,
26074 +       AuCache_HNOTIFY, /* must be last */
26075 +       AuCache_Last
26076 +};
26077 +
26078 +extern struct kmem_cache *au_cache[AuCache_Last];
26079 +
26080 +#define AuCacheFlags           (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
26081 +#define AuCache(type)          KMEM_CACHE(type, AuCacheFlags)
26082 +#define AuCacheCtor(type, ctor)        \
26083 +       kmem_cache_create(#type, sizeof(struct type), \
26084 +                         __alignof__(struct type), AuCacheFlags, ctor)
26085 +
26086 +#define AuCacheFuncAlloc(name, index)                                  \
26087 +       static inline struct au_##name *au_cache_alloc_##name(void)     \
26088 +       { return kmem_cache_alloc(au_cache[AuCache_##index], GFP_NOFS); }
26089 +
26090 +#define AuCacheFuncs(name, index)                                      \
26091 +       static inline void au_cache_free_##name##_norcu(struct au_##name *p) \
26092 +       { kmem_cache_free(au_cache[AuCache_##index], p); }              \
26093 +                                                                       \
26094 +       static inline void au_cache_free_##name##_rcu_cb(struct rcu_head *rcu) \
26095 +       { void *p = rcu;                                                \
26096 +               p -= offsetof(struct au_##name, rcu);                   \
26097 +               kmem_cache_free(au_cache[AuCache_##index], p); }        \
26098 +       static inline void au_cache_free_##name##_rcu(struct au_##name *p) \
26099 +       { BUILD_BUG_ON(sizeof(struct au_##name) < sizeof(struct rcu_head)); \
26100 +               call_rcu(&p->rcu, au_cache_free_##name##_rcu_cb); }     \
26101 +                                                                       \
26102 +       static inline void au_cache_free_##name(struct au_##name *p)    \
26103 +       { /* au_cache_free_##name##_norcu(p); */                        \
26104 +               au_cache_free_##name##_rcu(p); }
26105 +
26106 +AuCacheFuncs(dinfo, DINFO);
26107 +AuCacheFuncAlloc(dinfo, DINFO);
26108 +
26109 +AuCacheFuncs(icntnr, ICNTNR);
26110 +static inline struct au_icntnr *au_cache_alloc_icntnr(struct super_block *sb)
26111 +{ return alloc_inode_sb(sb, au_cache[AuCache_ICNTNR], GFP_NOFS); }
26112 +
26113 +AuCacheFuncs(finfo, FINFO);
26114 +AuCacheFuncAlloc(finfo, FINFO);
26115 +
26116 +AuCacheFuncs(vdir, VDIR);
26117 +AuCacheFuncAlloc(vdir, VDIR);
26118 +
26119 +AuCacheFuncs(vdir_dehstr, DEHSTR);
26120 +AuCacheFuncAlloc(vdir_dehstr, DEHSTR);
26121 +
26122 +#ifdef CONFIG_AUFS_HNOTIFY
26123 +AuCacheFuncs(hnotify, HNOTIFY);
26124 +AuCacheFuncAlloc(hnotify, HNOTIFY);
26125 +#endif
26126 +
26127 +#endif /* __KERNEL__ */
26128 +#endif /* __AUFS_MODULE_H__ */
26129 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
26130 --- /usr/share/empty/fs/aufs/mvdown.c   1970-01-01 01:00:00.000000000 +0100
26131 +++ linux/fs/aufs/mvdown.c      2022-11-05 23:02:18.969222617 +0100
26132 @@ -0,0 +1,706 @@
26133 +// SPDX-License-Identifier: GPL-2.0
26134 +/*
26135 + * Copyright (C) 2011-2022 Junjiro R. Okajima
26136 + *
26137 + * This program is free software; you can redistribute it and/or modify
26138 + * it under the terms of the GNU General Public License as published by
26139 + * the Free Software Foundation; either version 2 of the License, or
26140 + * (at your option) any later version.
26141 + *
26142 + * This program is distributed in the hope that it will be useful,
26143 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26144 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
26145 + * GNU General Public License for more details.
26146 + *
26147 + * You should have received a copy of the GNU General Public License
26148 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
26149 + */
26150 +
26151 +/*
26152 + * move-down, opposite of copy-up
26153 + */
26154 +
26155 +#include "aufs.h"
26156 +
26157 +struct au_mvd_args {
26158 +       struct {
26159 +               struct super_block *h_sb;
26160 +               struct dentry *h_parent;
26161 +               struct au_hinode *hdir;
26162 +               struct inode *h_dir, *h_inode;
26163 +               struct au_pin pin;
26164 +       } info[AUFS_MVDOWN_NARRAY];
26165 +
26166 +       struct aufs_mvdown mvdown;
26167 +       struct dentry *dentry, *parent;
26168 +       struct inode *inode, *dir;
26169 +       struct super_block *sb;
26170 +       aufs_bindex_t bopq, bwh, bfound;
26171 +       unsigned char rename_lock;
26172 +};
26173 +
26174 +#define mvd_errno              mvdown.au_errno
26175 +#define mvd_bsrc               mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
26176 +#define mvd_src_brid           mvdown.stbr[AUFS_MVDOWN_UPPER].brid
26177 +#define mvd_bdst               mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
26178 +#define mvd_dst_brid           mvdown.stbr[AUFS_MVDOWN_LOWER].brid
26179 +
26180 +#define mvd_h_src_sb           info[AUFS_MVDOWN_UPPER].h_sb
26181 +#define mvd_h_src_parent       info[AUFS_MVDOWN_UPPER].h_parent
26182 +#define mvd_hdir_src           info[AUFS_MVDOWN_UPPER].hdir
26183 +#define mvd_h_src_dir          info[AUFS_MVDOWN_UPPER].h_dir
26184 +#define mvd_h_src_inode                info[AUFS_MVDOWN_UPPER].h_inode
26185 +#define mvd_pin_src            info[AUFS_MVDOWN_UPPER].pin
26186 +
26187 +#define mvd_h_dst_sb           info[AUFS_MVDOWN_LOWER].h_sb
26188 +#define mvd_h_dst_parent       info[AUFS_MVDOWN_LOWER].h_parent
26189 +#define mvd_hdir_dst           info[AUFS_MVDOWN_LOWER].hdir
26190 +#define mvd_h_dst_dir          info[AUFS_MVDOWN_LOWER].h_dir
26191 +#define mvd_h_dst_inode                info[AUFS_MVDOWN_LOWER].h_inode
26192 +#define mvd_pin_dst            info[AUFS_MVDOWN_LOWER].pin
26193 +
26194 +#define AU_MVD_PR(flag, ...) do {                      \
26195 +               if (flag)                               \
26196 +                       pr_err(__VA_ARGS__);            \
26197 +       } while (0)
26198 +
26199 +static int find_lower_writable(struct au_mvd_args *a)
26200 +{
26201 +       struct super_block *sb;
26202 +       aufs_bindex_t bindex, bbot;
26203 +       struct au_branch *br;
26204 +
26205 +       sb = a->sb;
26206 +       bindex = a->mvd_bsrc;
26207 +       bbot = au_sbbot(sb);
26208 +       if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
26209 +               for (bindex++; bindex <= bbot; bindex++) {
26210 +                       br = au_sbr(sb, bindex);
26211 +                       if (au_br_fhsm(br->br_perm)
26212 +                           && !sb_rdonly(au_br_sb(br)))
26213 +                               return bindex;
26214 +               }
26215 +       else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
26216 +               for (bindex++; bindex <= bbot; bindex++) {
26217 +                       br = au_sbr(sb, bindex);
26218 +                       if (!au_br_rdonly(br))
26219 +                               return bindex;
26220 +               }
26221 +       else
26222 +               for (bindex++; bindex <= bbot; bindex++) {
26223 +                       br = au_sbr(sb, bindex);
26224 +                       if (!sb_rdonly(au_br_sb(br))) {
26225 +                               if (au_br_rdonly(br))
26226 +                                       a->mvdown.flags
26227 +                                               |= AUFS_MVDOWN_ROLOWER_R;
26228 +                               return bindex;
26229 +                       }
26230 +               }
26231 +
26232 +       return -1;
26233 +}
26234 +
26235 +/* make the parent dir on bdst */
26236 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
26237 +{
26238 +       int err;
26239 +
26240 +       err = 0;
26241 +       a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
26242 +       a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
26243 +       a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
26244 +       a->mvd_h_dst_parent = NULL;
26245 +       if (au_dbbot(a->parent) >= a->mvd_bdst)
26246 +               a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26247 +       if (!a->mvd_h_dst_parent) {
26248 +               err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
26249 +               if (unlikely(err)) {
26250 +                       AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
26251 +                       goto out;
26252 +               }
26253 +               a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26254 +       }
26255 +
26256 +out:
26257 +       AuTraceErr(err);
26258 +       return err;
26259 +}
26260 +
26261 +/* lock them all */
26262 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
26263 +{
26264 +       int err;
26265 +       struct dentry *h_trap;
26266 +
26267 +       a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
26268 +       a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
26269 +       err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
26270 +                    au_opt_udba(a->sb),
26271 +                    AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26272 +       AuTraceErr(err);
26273 +       if (unlikely(err)) {
26274 +               AU_MVD_PR(dmsg, "pin_dst failed\n");
26275 +               goto out;
26276 +       }
26277 +
26278 +       if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
26279 +               a->rename_lock = 0;
26280 +               au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26281 +                           AuLsc_DI_PARENT, AuLsc_I_PARENT3,
26282 +                           au_opt_udba(a->sb),
26283 +                           AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26284 +               err = au_do_pin(&a->mvd_pin_src);
26285 +               AuTraceErr(err);
26286 +               a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26287 +               if (unlikely(err)) {
26288 +                       AU_MVD_PR(dmsg, "pin_src failed\n");
26289 +                       goto out_dst;
26290 +               }
26291 +               goto out; /* success */
26292 +       }
26293 +
26294 +       a->rename_lock = 1;
26295 +       au_pin_hdir_unlock(&a->mvd_pin_dst);
26296 +       err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26297 +                    au_opt_udba(a->sb),
26298 +                    AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26299 +       AuTraceErr(err);
26300 +       a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26301 +       if (unlikely(err)) {
26302 +               AU_MVD_PR(dmsg, "pin_src failed\n");
26303 +               au_pin_hdir_lock(&a->mvd_pin_dst);
26304 +               goto out_dst;
26305 +       }
26306 +       au_pin_hdir_unlock(&a->mvd_pin_src);
26307 +       h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26308 +                                  a->mvd_h_dst_parent, a->mvd_hdir_dst);
26309 +       if (h_trap) {
26310 +               err = (h_trap != a->mvd_h_src_parent);
26311 +               if (err)
26312 +                       err = (h_trap != a->mvd_h_dst_parent);
26313 +       }
26314 +       BUG_ON(err); /* it should never happen */
26315 +       if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
26316 +               err = -EBUSY;
26317 +               AuTraceErr(err);
26318 +               vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26319 +                                   a->mvd_h_dst_parent, a->mvd_hdir_dst);
26320 +               au_pin_hdir_lock(&a->mvd_pin_src);
26321 +               au_unpin(&a->mvd_pin_src);
26322 +               au_pin_hdir_lock(&a->mvd_pin_dst);
26323 +               goto out_dst;
26324 +       }
26325 +       goto out; /* success */
26326 +
26327 +out_dst:
26328 +       au_unpin(&a->mvd_pin_dst);
26329 +out:
26330 +       AuTraceErr(err);
26331 +       return err;
26332 +}
26333 +
26334 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
26335 +{
26336 +       if (!a->rename_lock)
26337 +               au_unpin(&a->mvd_pin_src);
26338 +       else {
26339 +               vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26340 +                                   a->mvd_h_dst_parent, a->mvd_hdir_dst);
26341 +               au_pin_hdir_lock(&a->mvd_pin_src);
26342 +               au_unpin(&a->mvd_pin_src);
26343 +               au_pin_hdir_lock(&a->mvd_pin_dst);
26344 +       }
26345 +       au_unpin(&a->mvd_pin_dst);
26346 +}
26347 +
26348 +/* copy-down the file */
26349 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
26350 +{
26351 +       int err;
26352 +       struct au_cp_generic cpg = {
26353 +               .dentry = a->dentry,
26354 +               .bdst   = a->mvd_bdst,
26355 +               .bsrc   = a->mvd_bsrc,
26356 +               .len    = -1,
26357 +               .pin    = &a->mvd_pin_dst,
26358 +               .flags  = AuCpup_DTIME | AuCpup_HOPEN
26359 +       };
26360 +
26361 +       AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
26362 +       if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26363 +               au_fset_cpup(cpg.flags, OVERWRITE);
26364 +       if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
26365 +               au_fset_cpup(cpg.flags, RWDST);
26366 +       err = au_sio_cpdown_simple(&cpg);
26367 +       if (unlikely(err))
26368 +               AU_MVD_PR(dmsg, "cpdown failed\n");
26369 +
26370 +       AuTraceErr(err);
26371 +       return err;
26372 +}
26373 +
26374 +/*
26375 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
26376 + * were sleeping
26377 + */
26378 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
26379 +{
26380 +       int err;
26381 +       struct path h_path;
26382 +       struct au_branch *br;
26383 +       struct inode *delegated;
26384 +
26385 +       br = au_sbr(a->sb, a->mvd_bdst);
26386 +       h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
26387 +       err = PTR_ERR(h_path.dentry);
26388 +       if (IS_ERR(h_path.dentry)) {
26389 +               AU_MVD_PR(dmsg, "wh_lkup failed\n");
26390 +               goto out;
26391 +       }
26392 +
26393 +       err = 0;
26394 +       if (d_is_positive(h_path.dentry)) {
26395 +               h_path.mnt = au_br_mnt(br);
26396 +               delegated = NULL;
26397 +               err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
26398 +                                  &delegated, /*force*/0);
26399 +               if (unlikely(err == -EWOULDBLOCK)) {
26400 +                       pr_warn("cannot retry for NFSv4 delegation"
26401 +                               " for an internal unlink\n");
26402 +                       iput(delegated);
26403 +               }
26404 +               if (unlikely(err))
26405 +                       AU_MVD_PR(dmsg, "wh_unlink failed\n");
26406 +       }
26407 +       dput(h_path.dentry);
26408 +
26409 +out:
26410 +       AuTraceErr(err);
26411 +       return err;
26412 +}
26413 +
26414 +/*
26415 + * unlink the topmost h_dentry
26416 + */
26417 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
26418 +{
26419 +       int err;
26420 +       struct path h_path;
26421 +       struct inode *delegated;
26422 +
26423 +       h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
26424 +       h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
26425 +       delegated = NULL;
26426 +       err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
26427 +       if (unlikely(err == -EWOULDBLOCK)) {
26428 +               pr_warn("cannot retry for NFSv4 delegation"
26429 +                       " for an internal unlink\n");
26430 +               iput(delegated);
26431 +       }
26432 +       if (unlikely(err))
26433 +               AU_MVD_PR(dmsg, "unlink failed\n");
26434 +
26435 +       AuTraceErr(err);
26436 +       return err;
26437 +}
26438 +
26439 +/* Since mvdown succeeded, we ignore an error of this function */
26440 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
26441 +{
26442 +       int err;
26443 +       struct au_branch *br;
26444 +
26445 +       a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
26446 +       br = au_sbr(a->sb, a->mvd_bsrc);
26447 +       err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
26448 +       if (!err) {
26449 +               br = au_sbr(a->sb, a->mvd_bdst);
26450 +               a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
26451 +               err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
26452 +       }
26453 +       if (!err)
26454 +               a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
26455 +       else
26456 +               AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
26457 +}
26458 +
26459 +/*
26460 + * copy-down the file and unlink the bsrc file.
26461 + * - unlink the bdst whout if exist
26462 + * - copy-down the file (with whtmp name and rename)
26463 + * - unlink the bsrc file
26464 + */
26465 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
26466 +{
26467 +       int err;
26468 +
26469 +       err = au_do_mkdir(dmsg, a);
26470 +       if (!err)
26471 +               err = au_do_lock(dmsg, a);
26472 +       if (unlikely(err))
26473 +               goto out;
26474 +
26475 +       /*
26476 +        * do not revert the activities we made on bdst since they should be
26477 +        * harmless in aufs.
26478 +        */
26479 +
26480 +       err = au_do_cpdown(dmsg, a);
26481 +       if (!err)
26482 +               err = au_do_unlink_wh(dmsg, a);
26483 +       if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
26484 +               err = au_do_unlink(dmsg, a);
26485 +       if (unlikely(err))
26486 +               goto out_unlock;
26487 +
26488 +       AuDbg("%pd2, 0x%x, %d --> %d\n",
26489 +             a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
26490 +       if (find_lower_writable(a) < 0)
26491 +               a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
26492 +
26493 +       if (a->mvdown.flags & AUFS_MVDOWN_STFS)
26494 +               au_do_stfs(dmsg, a);
26495 +
26496 +       /* maintain internal array */
26497 +       if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
26498 +               au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
26499 +               au_set_dbtop(a->dentry, a->mvd_bdst);
26500 +               au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
26501 +               au_set_ibtop(a->inode, a->mvd_bdst);
26502 +       } else {
26503 +               /* hide the lower */
26504 +               au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
26505 +               au_set_dbbot(a->dentry, a->mvd_bsrc);
26506 +               au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
26507 +               au_set_ibbot(a->inode, a->mvd_bsrc);
26508 +       }
26509 +       if (au_dbbot(a->dentry) < a->mvd_bdst)
26510 +               au_set_dbbot(a->dentry, a->mvd_bdst);
26511 +       if (au_ibbot(a->inode) < a->mvd_bdst)
26512 +               au_set_ibbot(a->inode, a->mvd_bdst);
26513 +
26514 +out_unlock:
26515 +       au_do_unlock(dmsg, a);
26516 +out:
26517 +       AuTraceErr(err);
26518 +       return err;
26519 +}
26520 +
26521 +/* ---------------------------------------------------------------------- */
26522 +
26523 +/* make sure the file is idle */
26524 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
26525 +{
26526 +       int err, plinked;
26527 +
26528 +       err = 0;
26529 +       plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
26530 +       if (au_dbtop(a->dentry) == a->mvd_bsrc
26531 +           && au_dcount(a->dentry) == 1
26532 +           && atomic_read(&a->inode->i_count) == 1
26533 +           /* && a->mvd_h_src_inode->i_nlink == 1 */
26534 +           && (!plinked || !au_plink_test(a->inode))
26535 +           && a->inode->i_nlink == 1)
26536 +               goto out;
26537 +
26538 +       err = -EBUSY;
26539 +       AU_MVD_PR(dmsg,
26540 +                 "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
26541 +                 a->mvd_bsrc, au_dbtop(a->dentry), au_dcount(a->dentry),
26542 +                 atomic_read(&a->inode->i_count), a->inode->i_nlink,
26543 +                 a->mvd_h_src_inode->i_nlink,
26544 +                 plinked, plinked ? au_plink_test(a->inode) : 0);
26545 +
26546 +out:
26547 +       AuTraceErr(err);
26548 +       return err;
26549 +}
26550 +
26551 +/* make sure the parent dir is fine */
26552 +static int au_mvd_args_parent(const unsigned char dmsg,
26553 +                             struct au_mvd_args *a)
26554 +{
26555 +       int err;
26556 +       aufs_bindex_t bindex;
26557 +
26558 +       err = 0;
26559 +       if (unlikely(au_alive_dir(a->parent))) {
26560 +               err = -ENOENT;
26561 +               AU_MVD_PR(dmsg, "parent dir is dead\n");
26562 +               goto out;
26563 +       }
26564 +
26565 +       a->bopq = au_dbdiropq(a->parent);
26566 +       bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
26567 +       AuDbg("b%d\n", bindex);
26568 +       if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
26569 +                    || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
26570 +               err = -EINVAL;
26571 +               a->mvd_errno = EAU_MVDOWN_OPAQUE;
26572 +               AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
26573 +                         a->bopq, a->mvd_bdst);
26574 +       }
26575 +
26576 +out:
26577 +       AuTraceErr(err);
26578 +       return err;
26579 +}
26580 +
26581 +static int au_mvd_args_intermediate(const unsigned char dmsg,
26582 +                                   struct au_mvd_args *a)
26583 +{
26584 +       int err;
26585 +       struct au_dinfo *dinfo, *tmp;
26586 +
26587 +       /* lookup the next lower positive entry */
26588 +       err = -ENOMEM;
26589 +       tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
26590 +       if (unlikely(!tmp))
26591 +               goto out;
26592 +
26593 +       a->bfound = -1;
26594 +       a->bwh = -1;
26595 +       dinfo = au_di(a->dentry);
26596 +       au_di_cp(tmp, dinfo);
26597 +       au_di_swap(tmp, dinfo);
26598 +
26599 +       /* returns the number of positive dentries */
26600 +       err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1,
26601 +                            /* AuLkup_IGNORE_PERM */ 0);
26602 +       if (!err)
26603 +               a->bwh = au_dbwh(a->dentry);
26604 +       else if (err > 0)
26605 +               a->bfound = au_dbtop(a->dentry);
26606 +
26607 +       au_di_swap(tmp, dinfo);
26608 +       au_rw_write_unlock(&tmp->di_rwsem);
26609 +       au_di_free(tmp);
26610 +       if (unlikely(err < 0))
26611 +               AU_MVD_PR(dmsg, "failed look-up lower\n");
26612 +
26613 +       /*
26614 +        * here, we have these cases.
26615 +        * bfound == -1
26616 +        *      no positive dentry under bsrc. there are more sub-cases.
26617 +        *      bwh < 0
26618 +        *              there no whiteout, we can safely move-down.
26619 +        *      bwh <= bsrc
26620 +        *              impossible
26621 +        *      bsrc < bwh && bwh < bdst
26622 +        *              there is a whiteout on RO branch. cannot proceed.
26623 +        *      bwh == bdst
26624 +        *              there is a whiteout on the RW target branch. it should
26625 +        *              be removed.
26626 +        *      bdst < bwh
26627 +        *              there is a whiteout somewhere unrelated branch.
26628 +        * -1 < bfound && bfound <= bsrc
26629 +        *      impossible.
26630 +        * bfound < bdst
26631 +        *      found, but it is on RO branch between bsrc and bdst. cannot
26632 +        *      proceed.
26633 +        * bfound == bdst
26634 +        *      found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
26635 +        *      error.
26636 +        * bdst < bfound
26637 +        *      found, after we create the file on bdst, it will be hidden.
26638 +        */
26639 +
26640 +       AuDebugOn(a->bfound == -1
26641 +                 && a->bwh != -1
26642 +                 && a->bwh <= a->mvd_bsrc);
26643 +       AuDebugOn(-1 < a->bfound
26644 +                 && a->bfound <= a->mvd_bsrc);
26645 +
26646 +       err = -EINVAL;
26647 +       if (a->bfound == -1
26648 +           && a->mvd_bsrc < a->bwh
26649 +           && a->bwh != -1
26650 +           && a->bwh < a->mvd_bdst) {
26651 +               a->mvd_errno = EAU_MVDOWN_WHITEOUT;
26652 +               AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
26653 +                         a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
26654 +               goto out;
26655 +       } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
26656 +               a->mvd_errno = EAU_MVDOWN_UPPER;
26657 +               AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
26658 +                         a->mvd_bdst, a->bfound);
26659 +               goto out;
26660 +       }
26661 +
26662 +       err = 0; /* success */
26663 +
26664 +out:
26665 +       AuTraceErr(err);
26666 +       return err;
26667 +}
26668 +
26669 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
26670 +{
26671 +       int err;
26672 +
26673 +       err = 0;
26674 +       if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26675 +           && a->bfound == a->mvd_bdst)
26676 +               err = -EEXIST;
26677 +       AuTraceErr(err);
26678 +       return err;
26679 +}
26680 +
26681 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
26682 +{
26683 +       int err;
26684 +       struct au_branch *br;
26685 +
26686 +       err = -EISDIR;
26687 +       if (unlikely(S_ISDIR(a->inode->i_mode)))
26688 +               goto out;
26689 +
26690 +       err = -EINVAL;
26691 +       if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
26692 +               a->mvd_bsrc = au_ibtop(a->inode);
26693 +       else {
26694 +               a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
26695 +               if (unlikely(a->mvd_bsrc < 0
26696 +                            || (a->mvd_bsrc < au_dbtop(a->dentry)
26697 +                                || au_dbbot(a->dentry) < a->mvd_bsrc
26698 +                                || !au_h_dptr(a->dentry, a->mvd_bsrc))
26699 +                            || (a->mvd_bsrc < au_ibtop(a->inode)
26700 +                                || au_ibbot(a->inode) < a->mvd_bsrc
26701 +                                || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
26702 +                       a->mvd_errno = EAU_MVDOWN_NOUPPER;
26703 +                       AU_MVD_PR(dmsg, "no upper\n");
26704 +                       goto out;
26705 +               }
26706 +       }
26707 +       if (unlikely(a->mvd_bsrc == au_sbbot(a->sb))) {
26708 +               a->mvd_errno = EAU_MVDOWN_BOTTOM;
26709 +               AU_MVD_PR(dmsg, "on the bottom\n");
26710 +               goto out;
26711 +       }
26712 +       a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
26713 +       br = au_sbr(a->sb, a->mvd_bsrc);
26714 +       err = au_br_rdonly(br);
26715 +       if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
26716 +               if (unlikely(err))
26717 +                       goto out;
26718 +       } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
26719 +                    || IS_APPEND(a->mvd_h_src_inode))) {
26720 +               if (err)
26721 +                       a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
26722 +               /* go on */
26723 +       } else
26724 +               goto out;
26725 +
26726 +       err = -EINVAL;
26727 +       if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
26728 +               a->mvd_bdst = find_lower_writable(a);
26729 +               if (unlikely(a->mvd_bdst < 0)) {
26730 +                       a->mvd_errno = EAU_MVDOWN_BOTTOM;
26731 +                       AU_MVD_PR(dmsg, "no writable lower branch\n");
26732 +                       goto out;
26733 +               }
26734 +       } else {
26735 +               a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
26736 +               if (unlikely(a->mvd_bdst < 0
26737 +                            || au_sbbot(a->sb) < a->mvd_bdst)) {
26738 +                       a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
26739 +                       AU_MVD_PR(dmsg, "no lower brid\n");
26740 +                       goto out;
26741 +               }
26742 +       }
26743 +
26744 +       err = au_mvd_args_busy(dmsg, a);
26745 +       if (!err)
26746 +               err = au_mvd_args_parent(dmsg, a);
26747 +       if (!err)
26748 +               err = au_mvd_args_intermediate(dmsg, a);
26749 +       if (!err)
26750 +               err = au_mvd_args_exist(dmsg, a);
26751 +       if (!err)
26752 +               AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
26753 +
26754 +out:
26755 +       AuTraceErr(err);
26756 +       return err;
26757 +}
26758 +
26759 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
26760 +{
26761 +       int err, e;
26762 +       unsigned char dmsg;
26763 +       struct au_mvd_args *args;
26764 +       struct inode *inode;
26765 +
26766 +       inode = d_inode(dentry);
26767 +       err = -EPERM;
26768 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
26769 +               goto out;
26770 +
26771 +       err = -ENOMEM;
26772 +       args = kmalloc(sizeof(*args), GFP_NOFS);
26773 +       if (unlikely(!args))
26774 +               goto out;
26775 +
26776 +       err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
26777 +       if (!err)
26778 +               /* VERIFY_WRITE */
26779 +               err = !access_ok(uarg, sizeof(*uarg));
26780 +       if (unlikely(err)) {
26781 +               err = -EFAULT;
26782 +               AuTraceErr(err);
26783 +               goto out_free;
26784 +       }
26785 +       AuDbg("flags 0x%x\n", args->mvdown.flags);
26786 +       args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
26787 +       args->mvdown.au_errno = 0;
26788 +       args->dentry = dentry;
26789 +       args->inode = inode;
26790 +       args->sb = dentry->d_sb;
26791 +
26792 +       err = -ENOENT;
26793 +       dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
26794 +       args->parent = dget_parent(dentry);
26795 +       args->dir = d_inode(args->parent);
26796 +       inode_lock_nested(args->dir, I_MUTEX_PARENT);
26797 +       dput(args->parent);
26798 +       if (unlikely(args->parent != dentry->d_parent)) {
26799 +               AU_MVD_PR(dmsg, "parent dir is moved\n");
26800 +               goto out_dir;
26801 +       }
26802 +
26803 +       inode_lock_nested(inode, I_MUTEX_CHILD);
26804 +       err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
26805 +       if (unlikely(err))
26806 +               goto out_inode;
26807 +
26808 +       di_write_lock_parent(args->parent);
26809 +       err = au_mvd_args(dmsg, args);
26810 +       if (unlikely(err))
26811 +               goto out_parent;
26812 +
26813 +       err = au_do_mvdown(dmsg, args);
26814 +       if (unlikely(err))
26815 +               goto out_parent;
26816 +
26817 +       au_cpup_attr_timesizes(args->dir);
26818 +       au_cpup_attr_timesizes(inode);
26819 +       if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
26820 +               au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
26821 +       /* au_digen_dec(dentry); */
26822 +
26823 +out_parent:
26824 +       di_write_unlock(args->parent);
26825 +       aufs_read_unlock(dentry, AuLock_DW);
26826 +out_inode:
26827 +       inode_unlock(inode);
26828 +out_dir:
26829 +       inode_unlock(args->dir);
26830 +out_free:
26831 +       e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
26832 +       if (unlikely(e))
26833 +               err = -EFAULT;
26834 +       au_kfree_rcu(args);
26835 +out:
26836 +       AuTraceErr(err);
26837 +       return err;
26838 +}
26839 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
26840 --- /usr/share/empty/fs/aufs/opts.c     1970-01-01 01:00:00.000000000 +0100
26841 +++ linux/fs/aufs/opts.c        2022-11-05 23:02:18.969222617 +0100
26842 @@ -0,0 +1,1032 @@
26843 +// SPDX-License-Identifier: GPL-2.0
26844 +/*
26845 + * Copyright (C) 2005-2022 Junjiro R. Okajima
26846 + *
26847 + * This program is free software; you can redistribute it and/or modify
26848 + * it under the terms of the GNU General Public License as published by
26849 + * the Free Software Foundation; either version 2 of the License, or
26850 + * (at your option) any later version.
26851 + *
26852 + * This program is distributed in the hope that it will be useful,
26853 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26854 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
26855 + * GNU General Public License for more details.
26856 + *
26857 + * You should have received a copy of the GNU General Public License
26858 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
26859 + */
26860 +
26861 +/*
26862 + * mount options/flags
26863 + */
26864 +
26865 +#include <linux/types.h> /* a distribution requires */
26866 +#include <linux/parser.h>
26867 +#include "aufs.h"
26868 +
26869 +/* ---------------------------------------------------------------------- */
26870 +
26871 +static const char *au_parser_pattern(int val, match_table_t tbl)
26872 +{
26873 +       struct match_token *p;
26874 +
26875 +       p = tbl;
26876 +       while (p->pattern) {
26877 +               if (p->token == val)
26878 +                       return p->pattern;
26879 +               p++;
26880 +       }
26881 +       BUG();
26882 +       return "??";
26883 +}
26884 +
26885 +static const char *au_optstr(int *val, match_table_t tbl)
26886 +{
26887 +       struct match_token *p;
26888 +       int v;
26889 +
26890 +       v = *val;
26891 +       if (!v)
26892 +               goto out;
26893 +       p = tbl;
26894 +       while (p->pattern) {
26895 +               if (p->token
26896 +                   && (v & p->token) == p->token) {
26897 +                       *val &= ~p->token;
26898 +                       return p->pattern;
26899 +               }
26900 +               p++;
26901 +       }
26902 +
26903 +out:
26904 +       return NULL;
26905 +}
26906 +
26907 +/* ---------------------------------------------------------------------- */
26908 +
26909 +static match_table_t brperm = {
26910 +       {AuBrPerm_RO, AUFS_BRPERM_RO},
26911 +       {AuBrPerm_RR, AUFS_BRPERM_RR},
26912 +       {AuBrPerm_RW, AUFS_BRPERM_RW},
26913 +       {0, NULL}
26914 +};
26915 +
26916 +static match_table_t brattr = {
26917 +       /* general */
26918 +       {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
26919 +       {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
26920 +       /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
26921 +       {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
26922 +#ifdef CONFIG_AUFS_FHSM
26923 +       {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
26924 +#endif
26925 +#ifdef CONFIG_AUFS_XATTR
26926 +       {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
26927 +       {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
26928 +       {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
26929 +       {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
26930 +       {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
26931 +       {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
26932 +#endif
26933 +
26934 +       /* ro/rr branch */
26935 +       {AuBrRAttr_WH, AUFS_BRRATTR_WH},
26936 +
26937 +       /* rw branch */
26938 +       {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
26939 +       {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
26940 +
26941 +       {0, NULL}
26942 +};
26943 +
26944 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
26945 +{
26946 +       int attr, v;
26947 +       char *p;
26948 +
26949 +       attr = 0;
26950 +       do {
26951 +               p = strchr(str, '+');
26952 +               if (p)
26953 +                       *p = 0;
26954 +               v = match_token(str, table, args);
26955 +               if (v) {
26956 +                       if (v & AuBrAttr_CMOO_Mask)
26957 +                               attr &= ~AuBrAttr_CMOO_Mask;
26958 +                       attr |= v;
26959 +               } else {
26960 +                       if (p)
26961 +                               *p = '+';
26962 +                       pr_warn("ignored branch attribute %s\n", str);
26963 +                       break;
26964 +               }
26965 +               if (p)
26966 +                       str = p + 1;
26967 +       } while (p);
26968 +
26969 +       return attr;
26970 +}
26971 +
26972 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
26973 +{
26974 +       int sz;
26975 +       const char *p;
26976 +       char *q;
26977 +
26978 +       q = str->a;
26979 +       *q = 0;
26980 +       p = au_optstr(&perm, brattr);
26981 +       if (p) {
26982 +               sz = strlen(p);
26983 +               memcpy(q, p, sz + 1);
26984 +               q += sz;
26985 +       } else
26986 +               goto out;
26987 +
26988 +       do {
26989 +               p = au_optstr(&perm, brattr);
26990 +               if (p) {
26991 +                       *q++ = '+';
26992 +                       sz = strlen(p);
26993 +                       memcpy(q, p, sz + 1);
26994 +                       q += sz;
26995 +               }
26996 +       } while (p);
26997 +
26998 +out:
26999 +       return q - str->a;
27000 +}
27001 +
27002 +int au_br_perm_val(char *perm)
27003 +{
27004 +       int val, bad, sz;
27005 +       char *p;
27006 +       substring_t args[MAX_OPT_ARGS];
27007 +       au_br_perm_str_t attr;
27008 +
27009 +       p = strchr(perm, '+');
27010 +       if (p)
27011 +               *p = 0;
27012 +       val = match_token(perm, brperm, args);
27013 +       if (!val) {
27014 +               if (p)
27015 +                       *p = '+';
27016 +               pr_warn("ignored branch permission %s\n", perm);
27017 +               val = AuBrPerm_RO;
27018 +               goto out;
27019 +       }
27020 +       if (!p)
27021 +               goto out;
27022 +
27023 +       val |= br_attr_val(p + 1, brattr, args);
27024 +
27025 +       bad = 0;
27026 +       switch (val & AuBrPerm_Mask) {
27027 +       case AuBrPerm_RO:
27028 +       case AuBrPerm_RR:
27029 +               bad = val & AuBrWAttr_Mask;
27030 +               val &= ~AuBrWAttr_Mask;
27031 +               break;
27032 +       case AuBrPerm_RW:
27033 +               bad = val & AuBrRAttr_Mask;
27034 +               val &= ~AuBrRAttr_Mask;
27035 +               break;
27036 +       }
27037 +
27038 +       /*
27039 +        * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
27040 +        * does not treat it as an error, just warning.
27041 +        * this is a tiny guard for the user operation.
27042 +        */
27043 +       if (val & AuBrAttr_UNPIN) {
27044 +               bad |= AuBrAttr_UNPIN;
27045 +               val &= ~AuBrAttr_UNPIN;
27046 +       }
27047 +
27048 +       if (unlikely(bad)) {
27049 +               sz = au_do_optstr_br_attr(&attr, bad);
27050 +               AuDebugOn(!sz);
27051 +               pr_warn("ignored branch attribute %s\n", attr.a);
27052 +       }
27053 +
27054 +out:
27055 +       return val;
27056 +}
27057 +
27058 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
27059 +{
27060 +       au_br_perm_str_t attr;
27061 +       const char *p;
27062 +       char *q;
27063 +       int sz;
27064 +
27065 +       q = str->a;
27066 +       p = au_optstr(&perm, brperm);
27067 +       AuDebugOn(!p || !*p);
27068 +       sz = strlen(p);
27069 +       memcpy(q, p, sz + 1);
27070 +       q += sz;
27071 +
27072 +       sz = au_do_optstr_br_attr(&attr, perm);
27073 +       if (sz) {
27074 +               *q++ = '+';
27075 +               memcpy(q, attr.a, sz + 1);
27076 +       }
27077 +
27078 +       AuDebugOn(strlen(str->a) >= sizeof(str->a));
27079 +}
27080 +
27081 +/* ---------------------------------------------------------------------- */
27082 +
27083 +static match_table_t udbalevel = {
27084 +       {AuOpt_UDBA_REVAL, "reval"},
27085 +       {AuOpt_UDBA_NONE, "none"},
27086 +#ifdef CONFIG_AUFS_HNOTIFY
27087 +       {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
27088 +#ifdef CONFIG_AUFS_HFSNOTIFY
27089 +       {AuOpt_UDBA_HNOTIFY, "fsnotify"},
27090 +#endif
27091 +#endif
27092 +       {-1, NULL}
27093 +};
27094 +
27095 +int au_udba_val(char *str)
27096 +{
27097 +       substring_t args[MAX_OPT_ARGS];
27098 +
27099 +       return match_token(str, udbalevel, args);
27100 +}
27101 +
27102 +const char *au_optstr_udba(int udba)
27103 +{
27104 +       return au_parser_pattern(udba, udbalevel);
27105 +}
27106 +
27107 +/* ---------------------------------------------------------------------- */
27108 +
27109 +static match_table_t au_wbr_create_policy = {
27110 +       {AuWbrCreate_TDP, "tdp"},
27111 +       {AuWbrCreate_TDP, "top-down-parent"},
27112 +       {AuWbrCreate_RR, "rr"},
27113 +       {AuWbrCreate_RR, "round-robin"},
27114 +       {AuWbrCreate_MFS, "mfs"},
27115 +       {AuWbrCreate_MFS, "most-free-space"},
27116 +       {AuWbrCreate_MFSV, "mfs:%d"},
27117 +       {AuWbrCreate_MFSV, "most-free-space:%d"},
27118 +
27119 +       /* top-down regardless the parent, and then mfs */
27120 +       {AuWbrCreate_TDMFS, "tdmfs:%d"},
27121 +       {AuWbrCreate_TDMFSV, "tdmfs:%d:%d"},
27122 +
27123 +       {AuWbrCreate_MFSRR, "mfsrr:%d"},
27124 +       {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
27125 +       {AuWbrCreate_PMFS, "pmfs"},
27126 +       {AuWbrCreate_PMFSV, "pmfs:%d"},
27127 +       {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
27128 +       {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
27129 +
27130 +       {-1, NULL}
27131 +};
27132 +
27133 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
27134 +                           struct au_opt_wbr_create *create)
27135 +{
27136 +       int err;
27137 +       unsigned long long ull;
27138 +
27139 +       err = 0;
27140 +       if (!match_u64(arg, &ull))
27141 +               create->mfsrr_watermark = ull;
27142 +       else {
27143 +               pr_err("bad integer in %s\n", str);
27144 +               err = -EINVAL;
27145 +       }
27146 +
27147 +       return err;
27148 +}
27149 +
27150 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
27151 +                         struct au_opt_wbr_create *create)
27152 +{
27153 +       int n, err;
27154 +
27155 +       err = 0;
27156 +       if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
27157 +               create->mfs_second = n;
27158 +       else {
27159 +               pr_err("bad integer in %s\n", str);
27160 +               err = -EINVAL;
27161 +       }
27162 +
27163 +       return err;
27164 +}
27165 +
27166 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
27167 +{
27168 +       int err, e;
27169 +       substring_t args[MAX_OPT_ARGS];
27170 +
27171 +       err = match_token(str, au_wbr_create_policy, args);
27172 +       create->wbr_create = err;
27173 +       switch (err) {
27174 +       case AuWbrCreate_MFSRRV:
27175 +       case AuWbrCreate_TDMFSV:
27176 +       case AuWbrCreate_PMFSRRV:
27177 +               e = au_wbr_mfs_wmark(&args[0], str, create);
27178 +               if (!e)
27179 +                       e = au_wbr_mfs_sec(&args[1], str, create);
27180 +               if (unlikely(e))
27181 +                       err = e;
27182 +               break;
27183 +       case AuWbrCreate_MFSRR:
27184 +       case AuWbrCreate_TDMFS:
27185 +       case AuWbrCreate_PMFSRR:
27186 +               e = au_wbr_mfs_wmark(&args[0], str, create);
27187 +               if (unlikely(e)) {
27188 +                       err = e;
27189 +                       break;
27190 +               }
27191 +               fallthrough;
27192 +       case AuWbrCreate_MFS:
27193 +       case AuWbrCreate_PMFS:
27194 +               create->mfs_second = AUFS_MFS_DEF_SEC;
27195 +               break;
27196 +       case AuWbrCreate_MFSV:
27197 +       case AuWbrCreate_PMFSV:
27198 +               e = au_wbr_mfs_sec(&args[0], str, create);
27199 +               if (unlikely(e))
27200 +                       err = e;
27201 +               break;
27202 +       }
27203 +
27204 +       return err;
27205 +}
27206 +
27207 +const char *au_optstr_wbr_create(int wbr_create)
27208 +{
27209 +       return au_parser_pattern(wbr_create, au_wbr_create_policy);
27210 +}
27211 +
27212 +static match_table_t au_wbr_copyup_policy = {
27213 +       {AuWbrCopyup_TDP, "tdp"},
27214 +       {AuWbrCopyup_TDP, "top-down-parent"},
27215 +       {AuWbrCopyup_BUP, "bup"},
27216 +       {AuWbrCopyup_BUP, "bottom-up-parent"},
27217 +       {AuWbrCopyup_BU, "bu"},
27218 +       {AuWbrCopyup_BU, "bottom-up"},
27219 +       {-1, NULL}
27220 +};
27221 +
27222 +int au_wbr_copyup_val(char *str)
27223 +{
27224 +       substring_t args[MAX_OPT_ARGS];
27225 +
27226 +       return match_token(str, au_wbr_copyup_policy, args);
27227 +}
27228 +
27229 +const char *au_optstr_wbr_copyup(int wbr_copyup)
27230 +{
27231 +       return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
27232 +}
27233 +
27234 +/* ---------------------------------------------------------------------- */
27235 +
27236 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
27237 +              aufs_bindex_t bindex)
27238 +{
27239 +       int err;
27240 +       struct au_opt_add *add = &opt->add;
27241 +       char *p;
27242 +
27243 +       add->bindex = bindex;
27244 +       add->perm = AuBrPerm_RO;
27245 +       add->pathname = opt_str;
27246 +       p = strchr(opt_str, '=');
27247 +       if (p) {
27248 +               *p++ = 0;
27249 +               if (*p)
27250 +                       add->perm = au_br_perm_val(p);
27251 +       }
27252 +
27253 +       err = vfsub_kern_path(add->pathname, AuOpt_LkupDirFlags, &add->path);
27254 +       if (!err) {
27255 +               if (!p) {
27256 +                       add->perm = AuBrPerm_RO;
27257 +                       if (au_test_fs_rr(add->path.dentry->d_sb))
27258 +                               add->perm = AuBrPerm_RR;
27259 +                       else if (!bindex && !(sb_flags & SB_RDONLY))
27260 +                               add->perm = AuBrPerm_RW;
27261 +               }
27262 +               opt->type = Opt_add;
27263 +               goto out;
27264 +       }
27265 +       pr_err("lookup failed %s (%d)\n", add->pathname, err);
27266 +       err = -EINVAL;
27267 +
27268 +out:
27269 +       return err;
27270 +}
27271 +
27272 +static int au_opt_wbr_create(struct super_block *sb,
27273 +                            struct au_opt_wbr_create *create)
27274 +{
27275 +       int err;
27276 +       struct au_sbinfo *sbinfo;
27277 +
27278 +       SiMustWriteLock(sb);
27279 +
27280 +       err = 1; /* handled */
27281 +       sbinfo = au_sbi(sb);
27282 +       if (sbinfo->si_wbr_create_ops->fin) {
27283 +               err = sbinfo->si_wbr_create_ops->fin(sb);
27284 +               if (!err)
27285 +                       err = 1;
27286 +       }
27287 +
27288 +       sbinfo->si_wbr_create = create->wbr_create;
27289 +       sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
27290 +       switch (create->wbr_create) {
27291 +       case AuWbrCreate_MFSRRV:
27292 +       case AuWbrCreate_MFSRR:
27293 +       case AuWbrCreate_TDMFS:
27294 +       case AuWbrCreate_TDMFSV:
27295 +       case AuWbrCreate_PMFSRR:
27296 +       case AuWbrCreate_PMFSRRV:
27297 +               sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
27298 +               fallthrough;
27299 +       case AuWbrCreate_MFS:
27300 +       case AuWbrCreate_MFSV:
27301 +       case AuWbrCreate_PMFS:
27302 +       case AuWbrCreate_PMFSV:
27303 +               sbinfo->si_wbr_mfs.mfs_expire
27304 +                       = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
27305 +               break;
27306 +       }
27307 +
27308 +       if (sbinfo->si_wbr_create_ops->init)
27309 +               sbinfo->si_wbr_create_ops->init(sb); /* ignore */
27310 +
27311 +       return err;
27312 +}
27313 +
27314 +/*
27315 + * returns,
27316 + * plus: processed without an error
27317 + * zero: unprocessed
27318 + */
27319 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
27320 +                        struct au_opts *opts)
27321 +{
27322 +       int err;
27323 +       struct au_sbinfo *sbinfo;
27324 +
27325 +       SiMustWriteLock(sb);
27326 +
27327 +       err = 1; /* handled */
27328 +       sbinfo = au_sbi(sb);
27329 +       switch (opt->type) {
27330 +       case Opt_udba:
27331 +               sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27332 +               sbinfo->si_mntflags |= opt->udba;
27333 +               opts->given_udba |= opt->udba;
27334 +               break;
27335 +
27336 +       case Opt_plink:
27337 +               if (opt->tf)
27338 +                       au_opt_set(sbinfo->si_mntflags, PLINK);
27339 +               else {
27340 +                       if (au_opt_test(sbinfo->si_mntflags, PLINK))
27341 +                               au_plink_put(sb, /*verbose*/1);
27342 +                       au_opt_clr(sbinfo->si_mntflags, PLINK);
27343 +               }
27344 +               break;
27345 +       case Opt_list_plink:
27346 +               if (au_opt_test(sbinfo->si_mntflags, PLINK))
27347 +                       au_plink_list(sb);
27348 +               break;
27349 +
27350 +       case Opt_dio:
27351 +               if (opt->tf) {
27352 +                       au_opt_set(sbinfo->si_mntflags, DIO);
27353 +                       au_fset_opts(opts->flags, REFRESH_DYAOP);
27354 +               } else {
27355 +                       au_opt_clr(sbinfo->si_mntflags, DIO);
27356 +                       au_fset_opts(opts->flags, REFRESH_DYAOP);
27357 +               }
27358 +               break;
27359 +
27360 +       case Opt_fhsm_sec:
27361 +               au_fhsm_set(sbinfo, opt->fhsm_second);
27362 +               break;
27363 +
27364 +       case Opt_diropq_a:
27365 +               au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27366 +               break;
27367 +       case Opt_diropq_w:
27368 +               au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27369 +               break;
27370 +
27371 +       case Opt_warn_perm:
27372 +               if (opt->tf)
27373 +                       au_opt_set(sbinfo->si_mntflags, WARN_PERM);
27374 +               else
27375 +                       au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
27376 +               break;
27377 +
27378 +       case Opt_verbose:
27379 +               if (opt->tf)
27380 +                       au_opt_set(sbinfo->si_mntflags, VERBOSE);
27381 +               else
27382 +                       au_opt_clr(sbinfo->si_mntflags, VERBOSE);
27383 +               break;
27384 +
27385 +       case Opt_sum:
27386 +               if (opt->tf)
27387 +                       au_opt_set(sbinfo->si_mntflags, SUM);
27388 +               else {
27389 +                       au_opt_clr(sbinfo->si_mntflags, SUM);
27390 +                       au_opt_clr(sbinfo->si_mntflags, SUM_W);
27391 +               }
27392 +               break;
27393 +       case Opt_wsum:
27394 +               au_opt_clr(sbinfo->si_mntflags, SUM);
27395 +               au_opt_set(sbinfo->si_mntflags, SUM_W);
27396 +               break;
27397 +
27398 +       case Opt_wbr_create:
27399 +               err = au_opt_wbr_create(sb, &opt->wbr_create);
27400 +               break;
27401 +       case Opt_wbr_copyup:
27402 +               sbinfo->si_wbr_copyup = opt->wbr_copyup;
27403 +               sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
27404 +               break;
27405 +
27406 +       case Opt_dirwh:
27407 +               sbinfo->si_dirwh = opt->dirwh;
27408 +               break;
27409 +
27410 +       case Opt_rdcache:
27411 +               sbinfo->si_rdcache
27412 +                       = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
27413 +               break;
27414 +       case Opt_rdblk:
27415 +               sbinfo->si_rdblk = opt->rdblk;
27416 +               break;
27417 +       case Opt_rdhash:
27418 +               sbinfo->si_rdhash = opt->rdhash;
27419 +               break;
27420 +
27421 +       case Opt_shwh:
27422 +               if (opt->tf)
27423 +                       au_opt_set(sbinfo->si_mntflags, SHWH);
27424 +               else
27425 +                       au_opt_clr(sbinfo->si_mntflags, SHWH);
27426 +               break;
27427 +
27428 +       case Opt_dirperm1:
27429 +               if (opt->tf)
27430 +                       au_opt_set(sbinfo->si_mntflags, DIRPERM1);
27431 +               else
27432 +                       au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
27433 +               break;
27434 +
27435 +       case Opt_trunc_xino:
27436 +               if (opt->tf)
27437 +                       au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
27438 +               else
27439 +                       au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
27440 +               break;
27441 +
27442 +       case Opt_trunc_xino_path:
27443 +       case Opt_itrunc_xino:
27444 +               err = au_xino_trunc(sb, opt->xino_itrunc.bindex,
27445 +                                   /*idx_begin*/0);
27446 +               if (!err)
27447 +                       err = 1;
27448 +               break;
27449 +
27450 +       case Opt_trunc_xib:
27451 +               if (opt->tf)
27452 +                       au_fset_opts(opts->flags, TRUNC_XIB);
27453 +               else
27454 +                       au_fclr_opts(opts->flags, TRUNC_XIB);
27455 +               break;
27456 +
27457 +       case Opt_dirren:
27458 +               err = 1;
27459 +               if (opt->tf) {
27460 +                       if (!au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27461 +                               err = au_dr_opt_set(sb);
27462 +                               if (!err)
27463 +                                       err = 1;
27464 +                       }
27465 +                       if (err == 1)
27466 +                               au_opt_set(sbinfo->si_mntflags, DIRREN);
27467 +               } else {
27468 +                       if (au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27469 +                               err = au_dr_opt_clr(sb, au_ftest_opts(opts->flags,
27470 +                                                                     DR_FLUSHED));
27471 +                               if (!err)
27472 +                                       err = 1;
27473 +                       }
27474 +                       if (err == 1)
27475 +                               au_opt_clr(sbinfo->si_mntflags, DIRREN);
27476 +               }
27477 +               break;
27478 +
27479 +       case Opt_acl:
27480 +               if (opt->tf)
27481 +                       sb->s_flags |= SB_POSIXACL;
27482 +               else
27483 +                       sb->s_flags &= ~SB_POSIXACL;
27484 +               break;
27485 +
27486 +       default:
27487 +               err = 0;
27488 +               break;
27489 +       }
27490 +
27491 +       return err;
27492 +}
27493 +
27494 +/*
27495 + * returns tri-state.
27496 + * plus: processed without an error
27497 + * zero: unprocessed
27498 + * minus: error
27499 + */
27500 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
27501 +                    struct au_opts *opts)
27502 +{
27503 +       int err, do_refresh;
27504 +
27505 +       err = 0;
27506 +       switch (opt->type) {
27507 +       case Opt_append:
27508 +               opt->add.bindex = au_sbbot(sb) + 1;
27509 +               if (opt->add.bindex < 0)
27510 +                       opt->add.bindex = 0;
27511 +               goto add;
27512 +               /* Always goto add, not fallthrough */
27513 +       case Opt_prepend:
27514 +               opt->add.bindex = 0;
27515 +               fallthrough;
27516 +       add: /* indented label */
27517 +       case Opt_add:
27518 +               err = au_br_add(sb, &opt->add,
27519 +                               au_ftest_opts(opts->flags, REMOUNT));
27520 +               if (!err) {
27521 +                       err = 1;
27522 +                       au_fset_opts(opts->flags, REFRESH);
27523 +               }
27524 +               break;
27525 +
27526 +       case Opt_del:
27527 +       case Opt_idel:
27528 +               err = au_br_del(sb, &opt->del,
27529 +                               au_ftest_opts(opts->flags, REMOUNT));
27530 +               if (!err) {
27531 +                       err = 1;
27532 +                       au_fset_opts(opts->flags, TRUNC_XIB);
27533 +                       au_fset_opts(opts->flags, REFRESH);
27534 +               }
27535 +               break;
27536 +
27537 +       case Opt_mod:
27538 +       case Opt_imod:
27539 +               err = au_br_mod(sb, &opt->mod,
27540 +                               au_ftest_opts(opts->flags, REMOUNT),
27541 +                               &do_refresh);
27542 +               if (!err) {
27543 +                       err = 1;
27544 +                       if (do_refresh)
27545 +                               au_fset_opts(opts->flags, REFRESH);
27546 +               }
27547 +               break;
27548 +       }
27549 +       return err;
27550 +}
27551 +
27552 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
27553 +                      struct au_opt_xino **opt_xino,
27554 +                      struct au_opts *opts)
27555 +{
27556 +       int err;
27557 +
27558 +       err = 0;
27559 +       switch (opt->type) {
27560 +       case Opt_xino:
27561 +               err = au_xino_set(sb, &opt->xino,
27562 +                                 !!au_ftest_opts(opts->flags, REMOUNT));
27563 +               if (!err)
27564 +                       *opt_xino = &opt->xino;
27565 +               break;
27566 +       case Opt_noxino:
27567 +               au_xino_clr(sb);
27568 +               *opt_xino = (void *)-1;
27569 +               break;
27570 +       }
27571 +
27572 +       return err;
27573 +}
27574 +
27575 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
27576 +                  unsigned int pending)
27577 +{
27578 +       int err, fhsm;
27579 +       aufs_bindex_t bindex, bbot;
27580 +       unsigned char do_plink, skip, do_free, can_no_dreval;
27581 +       struct au_branch *br;
27582 +       struct au_wbr *wbr;
27583 +       struct dentry *root, *dentry;
27584 +       struct inode *dir, *h_dir;
27585 +       struct au_sbinfo *sbinfo;
27586 +       struct au_hinode *hdir;
27587 +
27588 +       SiMustAnyLock(sb);
27589 +
27590 +       sbinfo = au_sbi(sb);
27591 +       AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
27592 +
27593 +       if (!(sb_flags & SB_RDONLY)) {
27594 +               if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
27595 +                       pr_warn("first branch should be rw\n");
27596 +               if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
27597 +                       pr_warn_once("shwh should be used with ro\n");
27598 +       }
27599 +
27600 +       if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
27601 +           && !au_opt_test(sbinfo->si_mntflags, XINO))
27602 +               pr_warn_once("udba=*notify requires xino\n");
27603 +
27604 +       if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
27605 +               pr_warn_once("dirperm1 breaks the protection"
27606 +                            " by the permission bits on the lower branch\n");
27607 +
27608 +       err = 0;
27609 +       fhsm = 0;
27610 +       root = sb->s_root;
27611 +       dir = d_inode(root);
27612 +       do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
27613 +       can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
27614 +                                     UDBA_NONE);
27615 +       bbot = au_sbbot(sb);
27616 +       for (bindex = 0; !err && bindex <= bbot; bindex++) {
27617 +               skip = 0;
27618 +               h_dir = au_h_iptr(dir, bindex);
27619 +               br = au_sbr(sb, bindex);
27620 +
27621 +               if ((br->br_perm & AuBrAttr_ICEX)
27622 +                   && !h_dir->i_op->listxattr)
27623 +                       br->br_perm &= ~AuBrAttr_ICEX;
27624 +#if 0 /* untested */
27625 +               if ((br->br_perm & AuBrAttr_ICEX_SEC)
27626 +                   && (au_br_sb(br)->s_flags & SB_NOSEC))
27627 +                       br->br_perm &= ~AuBrAttr_ICEX_SEC;
27628 +#endif
27629 +
27630 +               do_free = 0;
27631 +               wbr = br->br_wbr;
27632 +               if (wbr)
27633 +                       wbr_wh_read_lock(wbr);
27634 +
27635 +               if (!au_br_writable(br->br_perm)) {
27636 +                       do_free = !!wbr;
27637 +                       skip = (!wbr
27638 +                               || (!wbr->wbr_whbase
27639 +                                   && !wbr->wbr_plink
27640 +                                   && !wbr->wbr_orph));
27641 +               } else if (!au_br_wh_linkable(br->br_perm)) {
27642 +                       /* skip = (!br->br_whbase && !br->br_orph); */
27643 +                       skip = (!wbr || !wbr->wbr_whbase);
27644 +                       if (skip && wbr) {
27645 +                               if (do_plink)
27646 +                                       skip = !!wbr->wbr_plink;
27647 +                               else
27648 +                                       skip = !wbr->wbr_plink;
27649 +                       }
27650 +               } else {
27651 +                       /* skip = (br->br_whbase && br->br_ohph); */
27652 +                       skip = (wbr && wbr->wbr_whbase);
27653 +                       if (skip) {
27654 +                               if (do_plink)
27655 +                                       skip = !!wbr->wbr_plink;
27656 +                               else
27657 +                                       skip = !wbr->wbr_plink;
27658 +                       }
27659 +               }
27660 +               if (wbr)
27661 +                       wbr_wh_read_unlock(wbr);
27662 +
27663 +               if (can_no_dreval) {
27664 +                       dentry = br->br_path.dentry;
27665 +                       spin_lock(&dentry->d_lock);
27666 +                       if (dentry->d_flags &
27667 +                           (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
27668 +                               can_no_dreval = 0;
27669 +                       spin_unlock(&dentry->d_lock);
27670 +               }
27671 +
27672 +               if (au_br_fhsm(br->br_perm)) {
27673 +                       fhsm++;
27674 +                       AuDebugOn(!br->br_fhsm);
27675 +               }
27676 +
27677 +               if (skip)
27678 +                       continue;
27679 +
27680 +               hdir = au_hi(dir, bindex);
27681 +               au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
27682 +               if (wbr)
27683 +                       wbr_wh_write_lock(wbr);
27684 +               err = au_wh_init(br, sb);
27685 +               if (wbr)
27686 +                       wbr_wh_write_unlock(wbr);
27687 +               au_hn_inode_unlock(hdir);
27688 +
27689 +               if (!err && do_free) {
27690 +                       au_kfree_rcu(wbr);
27691 +                       br->br_wbr = NULL;
27692 +               }
27693 +       }
27694 +
27695 +       if (can_no_dreval)
27696 +               au_fset_si(sbinfo, NO_DREVAL);
27697 +       else
27698 +               au_fclr_si(sbinfo, NO_DREVAL);
27699 +
27700 +       if (fhsm >= 2) {
27701 +               au_fset_si(sbinfo, FHSM);
27702 +               for (bindex = bbot; bindex >= 0; bindex--) {
27703 +                       br = au_sbr(sb, bindex);
27704 +                       if (au_br_fhsm(br->br_perm)) {
27705 +                               au_fhsm_set_bottom(sb, bindex);
27706 +                               break;
27707 +                       }
27708 +               }
27709 +       } else {
27710 +               au_fclr_si(sbinfo, FHSM);
27711 +               au_fhsm_set_bottom(sb, -1);
27712 +       }
27713 +
27714 +       return err;
27715 +}
27716 +
27717 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
27718 +{
27719 +       int err;
27720 +       unsigned int tmp;
27721 +       aufs_bindex_t bindex, bbot;
27722 +       struct au_opt *opt;
27723 +       struct au_opt_xino *opt_xino, xino;
27724 +       struct au_sbinfo *sbinfo;
27725 +       struct au_branch *br;
27726 +       struct inode *dir;
27727 +
27728 +       SiMustWriteLock(sb);
27729 +
27730 +       err = 0;
27731 +       opt_xino = NULL;
27732 +       opt = opts->opt;
27733 +       while (err >= 0 && opt->type != Opt_tail)
27734 +               err = au_opt_simple(sb, opt++, opts);
27735 +       if (err > 0)
27736 +               err = 0;
27737 +       else if (unlikely(err < 0))
27738 +               goto out;
27739 +
27740 +       /* disable xino and udba temporary */
27741 +       sbinfo = au_sbi(sb);
27742 +       tmp = sbinfo->si_mntflags;
27743 +       au_opt_clr(sbinfo->si_mntflags, XINO);
27744 +       au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
27745 +
27746 +       opt = opts->opt;
27747 +       while (err >= 0 && opt->type != Opt_tail)
27748 +               err = au_opt_br(sb, opt++, opts);
27749 +       if (err > 0)
27750 +               err = 0;
27751 +       else if (unlikely(err < 0))
27752 +               goto out;
27753 +
27754 +       bbot = au_sbbot(sb);
27755 +       if (unlikely(bbot < 0)) {
27756 +               err = -EINVAL;
27757 +               pr_err("no branches\n");
27758 +               goto out;
27759 +       }
27760 +
27761 +       if (au_opt_test(tmp, XINO))
27762 +               au_opt_set(sbinfo->si_mntflags, XINO);
27763 +       opt = opts->opt;
27764 +       while (!err && opt->type != Opt_tail)
27765 +               err = au_opt_xino(sb, opt++, &opt_xino, opts);
27766 +       if (unlikely(err))
27767 +               goto out;
27768 +
27769 +       err = au_opts_verify(sb, sb->s_flags, tmp);
27770 +       if (unlikely(err))
27771 +               goto out;
27772 +
27773 +       /* restore xino */
27774 +       if (au_opt_test(tmp, XINO) && !opt_xino) {
27775 +               xino.file = au_xino_def(sb);
27776 +               err = PTR_ERR(xino.file);
27777 +               if (IS_ERR(xino.file))
27778 +                       goto out;
27779 +
27780 +               err = au_xino_set(sb, &xino, /*remount*/0);
27781 +               fput(xino.file);
27782 +               if (unlikely(err))
27783 +                       goto out;
27784 +       }
27785 +
27786 +       /* restore udba */
27787 +       tmp &= AuOptMask_UDBA;
27788 +       sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27789 +       sbinfo->si_mntflags |= tmp;
27790 +       bbot = au_sbbot(sb);
27791 +       for (bindex = 0; bindex <= bbot; bindex++) {
27792 +               br = au_sbr(sb, bindex);
27793 +               err = au_hnotify_reset_br(tmp, br, br->br_perm);
27794 +               if (unlikely(err))
27795 +                       AuIOErr("hnotify failed on br %d, %d, ignored\n",
27796 +                               bindex, err);
27797 +               /* go on even if err */
27798 +       }
27799 +       if (au_opt_test(tmp, UDBA_HNOTIFY)) {
27800 +               dir = d_inode(sb->s_root);
27801 +               au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
27802 +       }
27803 +
27804 +out:
27805 +       return err;
27806 +}
27807 +
27808 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
27809 +{
27810 +       int err, rerr;
27811 +       unsigned char no_dreval;
27812 +       struct inode *dir;
27813 +       struct au_opt_xino *opt_xino;
27814 +       struct au_opt *opt;
27815 +       struct au_sbinfo *sbinfo;
27816 +
27817 +       SiMustWriteLock(sb);
27818 +
27819 +       err = au_dr_opt_flush(sb);
27820 +       if (unlikely(err))
27821 +               goto out;
27822 +       au_fset_opts(opts->flags, DR_FLUSHED);
27823 +
27824 +       dir = d_inode(sb->s_root);
27825 +       sbinfo = au_sbi(sb);
27826 +       opt_xino = NULL;
27827 +       opt = opts->opt;
27828 +       while (err >= 0 && opt->type != Opt_tail) {
27829 +               err = au_opt_simple(sb, opt, opts);
27830 +               if (!err)
27831 +                       err = au_opt_br(sb, opt, opts);
27832 +               if (!err)
27833 +                       err = au_opt_xino(sb, opt, &opt_xino, opts);
27834 +               opt++;
27835 +       }
27836 +       if (err > 0)
27837 +               err = 0;
27838 +       AuTraceErr(err);
27839 +       /* go on even err */
27840 +
27841 +       no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
27842 +       rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
27843 +       if (unlikely(rerr && !err))
27844 +               err = rerr;
27845 +
27846 +       if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
27847 +               au_fset_opts(opts->flags, REFRESH_IDOP);
27848 +
27849 +       if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
27850 +               rerr = au_xib_trunc(sb);
27851 +               if (unlikely(rerr && !err))
27852 +                       err = rerr;
27853 +       }
27854 +
27855 +       /* will be handled by the caller */
27856 +       if (!au_ftest_opts(opts->flags, REFRESH)
27857 +           && (opts->given_udba
27858 +               || au_opt_test(sbinfo->si_mntflags, XINO)
27859 +               || au_ftest_opts(opts->flags, REFRESH_IDOP)
27860 +                   ))
27861 +               au_fset_opts(opts->flags, REFRESH);
27862 +
27863 +       AuDbg("status 0x%x\n", opts->flags);
27864 +
27865 +out:
27866 +       return err;
27867 +}
27868 +
27869 +/* ---------------------------------------------------------------------- */
27870 +
27871 +unsigned int au_opt_udba(struct super_block *sb)
27872 +{
27873 +       return au_mntflags(sb) & AuOptMask_UDBA;
27874 +}
27875 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
27876 --- /usr/share/empty/fs/aufs/opts.h     1970-01-01 01:00:00.000000000 +0100
27877 +++ linux/fs/aufs/opts.h        2022-11-05 23:02:18.969222617 +0100
27878 @@ -0,0 +1,263 @@
27879 +/* SPDX-License-Identifier: GPL-2.0 */
27880 +/*
27881 + * Copyright (C) 2005-2022 Junjiro R. Okajima
27882 + *
27883 + * This program is free software; you can redistribute it and/or modify
27884 + * it under the terms of the GNU General Public License as published by
27885 + * the Free Software Foundation; either version 2 of the License, or
27886 + * (at your option) any later version.
27887 + *
27888 + * This program is distributed in the hope that it will be useful,
27889 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27890 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
27891 + * GNU General Public License for more details.
27892 + *
27893 + * You should have received a copy of the GNU General Public License
27894 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
27895 + */
27896 +
27897 +/*
27898 + * mount options/flags
27899 + */
27900 +
27901 +#ifndef __AUFS_OPTS_H__
27902 +#define __AUFS_OPTS_H__
27903 +
27904 +#ifdef __KERNEL__
27905 +
27906 +#include <linux/fs_parser.h>
27907 +#include <linux/namei.h>
27908 +#include <linux/path.h>
27909 +
27910 +enum {
27911 +       Opt_br,
27912 +       Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
27913 +       Opt_idel, Opt_imod,
27914 +       Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
27915 +       Opt_xino, Opt_noxino,
27916 +       Opt_trunc_xino, Opt_trunc_xino_v,
27917 +       Opt_trunc_xino_path, Opt_itrunc_xino,
27918 +       Opt_trunc_xib,
27919 +       Opt_shwh,
27920 +       Opt_plink, Opt_list_plink,
27921 +       Opt_udba,
27922 +       Opt_dio,
27923 +       Opt_diropq, Opt_diropq_a, Opt_diropq_w,
27924 +       Opt_warn_perm,
27925 +       Opt_wbr_copyup, Opt_wbr_create,
27926 +       Opt_fhsm_sec,
27927 +       Opt_verbose, Opt_noverbose,
27928 +       Opt_sum, Opt_wsum,
27929 +       Opt_dirperm1,
27930 +       Opt_dirren,
27931 +       Opt_acl,
27932 +       Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
27933 +};
27934 +
27935 +/* ---------------------------------------------------------------------- */
27936 +
27937 +/* mount flags */
27938 +#define AuOpt_XINO             1               /* external inode number bitmap
27939 +                                                  and translation table */
27940 +#define AuOpt_TRUNC_XINO       (1 << 1)        /* truncate xino files */
27941 +#define AuOpt_UDBA_NONE                (1 << 2)        /* users direct branch access */
27942 +#define AuOpt_UDBA_REVAL       (1 << 3)
27943 +#define AuOpt_UDBA_HNOTIFY     (1 << 4)
27944 +#define AuOpt_SHWH             (1 << 5)        /* show whiteout */
27945 +#define AuOpt_PLINK            (1 << 6)        /* pseudo-link */
27946 +#define AuOpt_DIRPERM1         (1 << 7)        /* ignore the lower dir's perm
27947 +                                                  bits */
27948 +#define AuOpt_ALWAYS_DIROPQ    (1 << 9)        /* policy to creating diropq */
27949 +#define AuOpt_SUM              (1 << 10)       /* summation for statfs(2) */
27950 +#define AuOpt_SUM_W            (1 << 11)       /* unimplemented */
27951 +#define AuOpt_WARN_PERM                (1 << 12)       /* warn when add-branch */
27952 +#define AuOpt_VERBOSE          (1 << 13)       /* print the cause of error */
27953 +#define AuOpt_DIO              (1 << 14)       /* direct io */
27954 +#define AuOpt_DIRREN           (1 << 15)       /* directory rename */
27955 +
27956 +#ifndef CONFIG_AUFS_HNOTIFY
27957 +#undef AuOpt_UDBA_HNOTIFY
27958 +#define AuOpt_UDBA_HNOTIFY     0
27959 +#endif
27960 +#ifndef CONFIG_AUFS_DIRREN
27961 +#undef AuOpt_DIRREN
27962 +#define AuOpt_DIRREN           0
27963 +#endif
27964 +#ifndef CONFIG_AUFS_SHWH
27965 +#undef AuOpt_SHWH
27966 +#define AuOpt_SHWH             0
27967 +#endif
27968 +
27969 +#define AuOpt_Def      (AuOpt_XINO \
27970 +                        | AuOpt_UDBA_REVAL \
27971 +                        | AuOpt_PLINK \
27972 +                        /* | AuOpt_DIRPERM1 */ \
27973 +                        | AuOpt_WARN_PERM)
27974 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
27975 +                        | AuOpt_UDBA_REVAL \
27976 +                        | AuOpt_UDBA_HNOTIFY)
27977 +
27978 +#define AuOpt_LkupDirFlags     (LOOKUP_FOLLOW | LOOKUP_DIRECTORY)
27979 +
27980 +#define au_opt_test(flags, name)       (flags & AuOpt_##name)
27981 +#define au_opt_set(flags, name) do { \
27982 +       BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
27983 +       ((flags) |= AuOpt_##name); \
27984 +} while (0)
27985 +#define au_opt_set_udba(flags, name) do { \
27986 +       (flags) &= ~AuOptMask_UDBA; \
27987 +       ((flags) |= AuOpt_##name); \
27988 +} while (0)
27989 +#define au_opt_clr(flags, name) do { \
27990 +       ((flags) &= ~AuOpt_##name); \
27991 +} while (0)
27992 +
27993 +static inline unsigned int au_opts_plink(unsigned int mntflags)
27994 +{
27995 +#ifdef CONFIG_PROC_FS
27996 +       return mntflags;
27997 +#else
27998 +       return mntflags & ~AuOpt_PLINK;
27999 +#endif
28000 +}
28001 +
28002 +/* ---------------------------------------------------------------------- */
28003 +
28004 +/* policies to select one among multiple writable branches */
28005 +enum {
28006 +       AuWbrCreate_TDP,        /* top down parent */
28007 +       AuWbrCreate_RR,         /* round robin */
28008 +       AuWbrCreate_MFS,        /* most free space */
28009 +       AuWbrCreate_MFSV,       /* mfs with seconds */
28010 +       AuWbrCreate_MFSRR,      /* mfs then rr */
28011 +       AuWbrCreate_MFSRRV,     /* mfs then rr with seconds */
28012 +       AuWbrCreate_TDMFS,      /* top down regardless parent and mfs */
28013 +       AuWbrCreate_TDMFSV,     /* top down regardless parent and mfs */
28014 +       AuWbrCreate_PMFS,       /* parent and mfs */
28015 +       AuWbrCreate_PMFSV,      /* parent and mfs with seconds */
28016 +       AuWbrCreate_PMFSRR,     /* parent, mfs and round-robin */
28017 +       AuWbrCreate_PMFSRRV,    /* plus seconds */
28018 +
28019 +       AuWbrCreate_Def = AuWbrCreate_TDP
28020 +};
28021 +
28022 +enum {
28023 +       AuWbrCopyup_TDP,        /* top down parent */
28024 +       AuWbrCopyup_BUP,        /* bottom up parent */
28025 +       AuWbrCopyup_BU,         /* bottom up */
28026 +
28027 +       AuWbrCopyup_Def = AuWbrCopyup_TDP
28028 +};
28029 +
28030 +/* ---------------------------------------------------------------------- */
28031 +
28032 +struct file;
28033 +
28034 +struct au_opt_add {
28035 +       aufs_bindex_t   bindex;
28036 +       char            *pathname;
28037 +       int             perm;
28038 +       struct path     path;
28039 +};
28040 +
28041 +struct au_opt_del {
28042 +       char            *pathname;
28043 +       struct path     h_path;
28044 +};
28045 +
28046 +struct au_opt_mod {
28047 +       char            *path;
28048 +       int             perm;
28049 +       struct dentry   *h_root;
28050 +};
28051 +
28052 +struct au_opt_xino {
28053 +       char            *path;
28054 +       struct file     *file;
28055 +};
28056 +
28057 +struct au_opt_xino_itrunc {
28058 +       aufs_bindex_t   bindex;
28059 +};
28060 +
28061 +struct au_opt_wbr_create {
28062 +       int                     wbr_create;
28063 +       int                     mfs_second;
28064 +       unsigned long long      mfsrr_watermark;
28065 +};
28066 +
28067 +struct au_opt {
28068 +       int type;
28069 +       union {
28070 +               struct au_opt_xino      xino;
28071 +               struct au_opt_xino_itrunc xino_itrunc;
28072 +               struct au_opt_add       add;
28073 +               struct au_opt_del       del;
28074 +               struct au_opt_mod       mod;
28075 +               int                     dirwh;
28076 +               int                     rdcache;
28077 +               unsigned int            rdblk;
28078 +               unsigned int            rdhash;
28079 +               int                     udba;
28080 +               struct au_opt_wbr_create wbr_create;
28081 +               int                     wbr_copyup;
28082 +               unsigned int            fhsm_second;
28083 +               bool                    tf; /* generic flag, true or false */
28084 +       };
28085 +};
28086 +
28087 +/* opts flags */
28088 +#define AuOpts_REMOUNT         1
28089 +#define AuOpts_REFRESH         (1 << 1)
28090 +#define AuOpts_TRUNC_XIB       (1 << 2)
28091 +#define AuOpts_REFRESH_DYAOP   (1 << 3)
28092 +#define AuOpts_REFRESH_IDOP    (1 << 4)
28093 +#define AuOpts_DR_FLUSHED      (1 << 5)
28094 +#define au_ftest_opts(flags, name)     ((flags) & AuOpts_##name)
28095 +#define au_fset_opts(flags, name) \
28096 +       do { (flags) |= AuOpts_##name; } while (0)
28097 +#define au_fclr_opts(flags, name) \
28098 +       do { (flags) &= ~AuOpts_##name; } while (0)
28099 +
28100 +#ifndef CONFIG_AUFS_DIRREN
28101 +#undef AuOpts_DR_FLUSHED
28102 +#define AuOpts_DR_FLUSHED      0
28103 +#endif
28104 +
28105 +struct au_opts {
28106 +       struct au_opt   *opt;
28107 +       int             max_opt;
28108 +
28109 +       unsigned int    given_udba;
28110 +       unsigned int    flags;
28111 +       unsigned long   sb_flags;
28112 +};
28113 +
28114 +/* ---------------------------------------------------------------------- */
28115 +
28116 +/* opts.c */
28117 +int au_br_perm_val(char *perm);
28118 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
28119 +int au_udba_val(char *str);
28120 +const char *au_optstr_udba(int udba);
28121 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create);
28122 +const char *au_optstr_wbr_create(int wbr_create);
28123 +int au_wbr_copyup_val(char *str);
28124 +const char *au_optstr_wbr_copyup(int wbr_copyup);
28125 +
28126 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
28127 +              aufs_bindex_t bindex);
28128 +struct super_block;
28129 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
28130 +                  unsigned int pending);
28131 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
28132 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
28133 +
28134 +unsigned int au_opt_udba(struct super_block *sb);
28135 +
28136 +/* fsctx.c */
28137 +int aufs_fsctx_init(struct fs_context *fc);
28138 +extern const struct fs_parameter_spec aufs_fsctx_paramspec[];
28139 +
28140 +#endif /* __KERNEL__ */
28141 +#endif /* __AUFS_OPTS_H__ */
28142 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
28143 --- /usr/share/empty/fs/aufs/plink.c    1970-01-01 01:00:00.000000000 +0100
28144 +++ linux/fs/aufs/plink.c       2022-11-05 23:02:18.969222617 +0100
28145 @@ -0,0 +1,516 @@
28146 +// SPDX-License-Identifier: GPL-2.0
28147 +/*
28148 + * Copyright (C) 2005-2022 Junjiro R. Okajima
28149 + *
28150 + * This program is free software; you can redistribute it and/or modify
28151 + * it under the terms of the GNU General Public License as published by
28152 + * the Free Software Foundation; either version 2 of the License, or
28153 + * (at your option) any later version.
28154 + *
28155 + * This program is distributed in the hope that it will be useful,
28156 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28157 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28158 + * GNU General Public License for more details.
28159 + *
28160 + * You should have received a copy of the GNU General Public License
28161 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28162 + */
28163 +
28164 +/*
28165 + * pseudo-link
28166 + */
28167 +
28168 +#include "aufs.h"
28169 +
28170 +/*
28171 + * the pseudo-link maintenance mode.
28172 + * during a user process maintains the pseudo-links,
28173 + * prohibit adding a new plink and branch manipulation.
28174 + *
28175 + * Flags
28176 + * NOPLM:
28177 + *     For entry functions which will handle plink, and i_mutex is already held
28178 + *     in VFS.
28179 + *     They cannot wait and should return an error at once.
28180 + *     Callers has to check the error.
28181 + * NOPLMW:
28182 + *     For entry functions which will handle plink, but i_mutex is not held
28183 + *     in VFS.
28184 + *     They can wait the plink maintenance mode to finish.
28185 + *
28186 + * They behave like F_SETLK and F_SETLKW.
28187 + * If the caller never handle plink, then both flags are unnecessary.
28188 + */
28189 +
28190 +int au_plink_maint(struct super_block *sb, int flags)
28191 +{
28192 +       int err;
28193 +       pid_t pid, ppid;
28194 +       struct task_struct *parent, *prev;
28195 +       struct au_sbinfo *sbi;
28196 +
28197 +       SiMustAnyLock(sb);
28198 +
28199 +       err = 0;
28200 +       if (!au_opt_test(au_mntflags(sb), PLINK))
28201 +               goto out;
28202 +
28203 +       sbi = au_sbi(sb);
28204 +       pid = sbi->si_plink_maint_pid;
28205 +       if (!pid || pid == current->pid)
28206 +               goto out;
28207 +
28208 +       /* todo: it highly depends upon /sbin/mount.aufs */
28209 +       prev = NULL;
28210 +       parent = current;
28211 +       ppid = 0;
28212 +       rcu_read_lock();
28213 +       while (1) {
28214 +               parent = rcu_dereference(parent->real_parent);
28215 +               if (parent == prev)
28216 +                       break;
28217 +               ppid = task_pid_vnr(parent);
28218 +               if (pid == ppid) {
28219 +                       rcu_read_unlock();
28220 +                       goto out;
28221 +               }
28222 +               prev = parent;
28223 +       }
28224 +       rcu_read_unlock();
28225 +
28226 +       if (au_ftest_lock(flags, NOPLMW)) {
28227 +               /* if there is no i_mutex lock in VFS, we don't need to wait */
28228 +               /* AuDebugOn(!lockdep_depth(current)); */
28229 +               while (sbi->si_plink_maint_pid) {
28230 +                       si_read_unlock(sb);
28231 +                       /* gave up wake_up_bit() */
28232 +                       wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
28233 +
28234 +                       if (au_ftest_lock(flags, FLUSH))
28235 +                               au_nwt_flush(&sbi->si_nowait);
28236 +                       si_noflush_read_lock(sb);
28237 +               }
28238 +       } else if (au_ftest_lock(flags, NOPLM)) {
28239 +               AuDbg("ppid %d, pid %d\n", ppid, pid);
28240 +               err = -EAGAIN;
28241 +       }
28242 +
28243 +out:
28244 +       return err;
28245 +}
28246 +
28247 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
28248 +{
28249 +       spin_lock(&sbinfo->si_plink_maint_lock);
28250 +       sbinfo->si_plink_maint_pid = 0;
28251 +       spin_unlock(&sbinfo->si_plink_maint_lock);
28252 +       wake_up_all(&sbinfo->si_plink_wq);
28253 +}
28254 +
28255 +int au_plink_maint_enter(struct super_block *sb)
28256 +{
28257 +       int err;
28258 +       struct au_sbinfo *sbinfo;
28259 +
28260 +       err = 0;
28261 +       sbinfo = au_sbi(sb);
28262 +       /* make sure i am the only one in this fs */
28263 +       si_write_lock(sb, AuLock_FLUSH);
28264 +       if (au_opt_test(au_mntflags(sb), PLINK)) {
28265 +               spin_lock(&sbinfo->si_plink_maint_lock);
28266 +               if (!sbinfo->si_plink_maint_pid)
28267 +                       sbinfo->si_plink_maint_pid = current->pid;
28268 +               else
28269 +                       err = -EBUSY;
28270 +               spin_unlock(&sbinfo->si_plink_maint_lock);
28271 +       }
28272 +       si_write_unlock(sb);
28273 +
28274 +       return err;
28275 +}
28276 +
28277 +/* ---------------------------------------------------------------------- */
28278 +
28279 +#ifdef CONFIG_AUFS_DEBUG
28280 +void au_plink_list(struct super_block *sb)
28281 +{
28282 +       int i;
28283 +       struct au_sbinfo *sbinfo;
28284 +       struct hlist_bl_head *hbl;
28285 +       struct hlist_bl_node *pos;
28286 +       struct au_icntnr *icntnr;
28287 +
28288 +       SiMustAnyLock(sb);
28289 +
28290 +       sbinfo = au_sbi(sb);
28291 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28292 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28293 +
28294 +       for (i = 0; i < AuPlink_NHASH; i++) {
28295 +               hbl = sbinfo->si_plink + i;
28296 +               hlist_bl_lock(hbl);
28297 +               hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28298 +                       AuDbg("%lu\n", icntnr->vfs_inode.i_ino);
28299 +               hlist_bl_unlock(hbl);
28300 +       }
28301 +}
28302 +#endif
28303 +
28304 +/* is the inode pseudo-linked? */
28305 +int au_plink_test(struct inode *inode)
28306 +{
28307 +       int found, i;
28308 +       struct au_sbinfo *sbinfo;
28309 +       struct hlist_bl_head *hbl;
28310 +       struct hlist_bl_node *pos;
28311 +       struct au_icntnr *icntnr;
28312 +
28313 +       sbinfo = au_sbi(inode->i_sb);
28314 +       AuRwMustAnyLock(&sbinfo->si_rwsem);
28315 +       AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
28316 +       AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28317 +
28318 +       found = 0;
28319 +       i = au_plink_hash(inode->i_ino);
28320 +       hbl =  sbinfo->si_plink + i;
28321 +       hlist_bl_lock(hbl);
28322 +       hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28323 +               if (&icntnr->vfs_inode == inode) {
28324 +                       found = 1;
28325 +                       break;
28326 +               }
28327 +       hlist_bl_unlock(hbl);
28328 +       return found;
28329 +}
28330 +
28331 +/* ---------------------------------------------------------------------- */
28332 +
28333 +/*
28334 + * generate a name for plink.
28335 + * the file will be stored under AUFS_WH_PLINKDIR.
28336 + */
28337 +/* 20 is max digits length of ulong 64 */
28338 +#define PLINK_NAME_LEN ((20 + 1) * 2)
28339 +
28340 +static int plink_name(char *name, int len, struct inode *inode,
28341 +                     aufs_bindex_t bindex)
28342 +{
28343 +       int rlen;
28344 +       struct inode *h_inode;
28345 +
28346 +       h_inode = au_h_iptr(inode, bindex);
28347 +       rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
28348 +       return rlen;
28349 +}
28350 +
28351 +struct au_do_plink_lkup_args {
28352 +       struct dentry **errp;
28353 +       struct qstr *tgtname;
28354 +       struct path *h_ppath;
28355 +};
28356 +
28357 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
28358 +                                      struct path *h_ppath)
28359 +{
28360 +       struct dentry *h_dentry;
28361 +       struct inode *h_inode;
28362 +
28363 +       h_inode = d_inode(h_ppath->dentry);
28364 +       inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2);
28365 +       h_dentry = vfsub_lkup_one(tgtname, h_ppath);
28366 +       inode_unlock_shared(h_inode);
28367 +
28368 +       return h_dentry;
28369 +}
28370 +
28371 +static void au_call_do_plink_lkup(void *args)
28372 +{
28373 +       struct au_do_plink_lkup_args *a = args;
28374 +       *a->errp = au_do_plink_lkup(a->tgtname, a->h_ppath);
28375 +}
28376 +
28377 +/* lookup the plink-ed @inode under the branch at @bindex */
28378 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
28379 +{
28380 +       struct dentry *h_dentry;
28381 +       struct au_branch *br;
28382 +       struct path h_ppath;
28383 +       int wkq_err;
28384 +       char a[PLINK_NAME_LEN];
28385 +       struct qstr tgtname = QSTR_INIT(a, 0);
28386 +
28387 +       AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28388 +
28389 +       br = au_sbr(inode->i_sb, bindex);
28390 +       h_ppath.dentry = br->br_wbr->wbr_plink;
28391 +       h_ppath.mnt = au_br_mnt(br);
28392 +       tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28393 +
28394 +       if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28395 +               struct au_do_plink_lkup_args args = {
28396 +                       .errp           = &h_dentry,
28397 +                       .tgtname        = &tgtname,
28398 +                       .h_ppath        = &h_ppath
28399 +               };
28400 +
28401 +               wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
28402 +               if (unlikely(wkq_err))
28403 +                       h_dentry = ERR_PTR(wkq_err);
28404 +       } else
28405 +               h_dentry = au_do_plink_lkup(&tgtname, &h_ppath);
28406 +
28407 +       return h_dentry;
28408 +}
28409 +
28410 +/* create a pseudo-link */
28411 +static int do_whplink(struct qstr *tgt, struct path *h_ppath,
28412 +                     struct dentry *h_dentry)
28413 +{
28414 +       int err;
28415 +       struct path h_path;
28416 +       struct inode *h_dir, *delegated;
28417 +
28418 +       h_dir = d_inode(h_ppath->dentry);
28419 +       inode_lock_nested(h_dir, AuLsc_I_CHILD2);
28420 +       h_path.mnt = h_ppath->mnt;
28421 +again:
28422 +       h_path.dentry = vfsub_lkup_one(tgt, h_ppath);
28423 +       err = PTR_ERR(h_path.dentry);
28424 +       if (IS_ERR(h_path.dentry))
28425 +               goto out;
28426 +
28427 +       err = 0;
28428 +       /* wh.plink dir is not monitored */
28429 +       /* todo: is it really safe? */
28430 +       if (d_is_positive(h_path.dentry)
28431 +           && d_inode(h_path.dentry) != d_inode(h_dentry)) {
28432 +               delegated = NULL;
28433 +               err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
28434 +               if (unlikely(err == -EWOULDBLOCK)) {
28435 +                       pr_warn("cannot retry for NFSv4 delegation"
28436 +                               " for an internal unlink\n");
28437 +                       iput(delegated);
28438 +               }
28439 +               dput(h_path.dentry);
28440 +               h_path.dentry = NULL;
28441 +               if (!err)
28442 +                       goto again;
28443 +       }
28444 +       if (!err && d_is_negative(h_path.dentry)) {
28445 +               delegated = NULL;
28446 +               err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
28447 +               if (unlikely(err == -EWOULDBLOCK)) {
28448 +                       pr_warn("cannot retry for NFSv4 delegation"
28449 +                               " for an internal link\n");
28450 +                       iput(delegated);
28451 +               }
28452 +       }
28453 +       dput(h_path.dentry);
28454 +
28455 +out:
28456 +       inode_unlock(h_dir);
28457 +       return err;
28458 +}
28459 +
28460 +struct do_whplink_args {
28461 +       int *errp;
28462 +       struct qstr *tgt;
28463 +       struct path *h_ppath;
28464 +       struct dentry *h_dentry;
28465 +};
28466 +
28467 +static void call_do_whplink(void *args)
28468 +{
28469 +       struct do_whplink_args *a = args;
28470 +       *a->errp = do_whplink(a->tgt, a->h_ppath, a->h_dentry);
28471 +}
28472 +
28473 +static int whplink(struct dentry *h_dentry, struct inode *inode,
28474 +                  aufs_bindex_t bindex)
28475 +{
28476 +       int err, wkq_err;
28477 +       struct au_branch *br;
28478 +       struct au_wbr *wbr;
28479 +       struct path h_ppath;
28480 +       char a[PLINK_NAME_LEN];
28481 +       struct qstr tgtname = QSTR_INIT(a, 0);
28482 +
28483 +       br = au_sbr(inode->i_sb, bindex);
28484 +       wbr = br->br_wbr;
28485 +       h_ppath.dentry = wbr->wbr_plink;
28486 +       h_ppath.mnt = au_br_mnt(br);
28487 +       tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28488 +
28489 +       /* always superio. */
28490 +       if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28491 +               struct do_whplink_args args = {
28492 +                       .errp           = &err,
28493 +                       .tgt            = &tgtname,
28494 +                       .h_ppath        = &h_ppath,
28495 +                       .h_dentry       = h_dentry
28496 +               };
28497 +               wkq_err = au_wkq_wait(call_do_whplink, &args);
28498 +               if (unlikely(wkq_err))
28499 +                       err = wkq_err;
28500 +       } else
28501 +               err = do_whplink(&tgtname, &h_ppath, h_dentry);
28502 +
28503 +       return err;
28504 +}
28505 +
28506 +/*
28507 + * create a new pseudo-link for @h_dentry on @bindex.
28508 + * the linked inode is held in aufs @inode.
28509 + */
28510 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
28511 +                    struct dentry *h_dentry)
28512 +{
28513 +       struct super_block *sb;
28514 +       struct au_sbinfo *sbinfo;
28515 +       struct hlist_bl_head *hbl;
28516 +       struct hlist_bl_node *pos;
28517 +       struct au_icntnr *icntnr;
28518 +       int found, err, cnt, i;
28519 +
28520 +       sb = inode->i_sb;
28521 +       sbinfo = au_sbi(sb);
28522 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28523 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28524 +
28525 +       found = au_plink_test(inode);
28526 +       if (found)
28527 +               return;
28528 +
28529 +       i = au_plink_hash(inode->i_ino);
28530 +       hbl = sbinfo->si_plink + i;
28531 +       au_igrab(inode);
28532 +
28533 +       hlist_bl_lock(hbl);
28534 +       hlist_bl_for_each_entry(icntnr, pos, hbl, plink) {
28535 +               if (&icntnr->vfs_inode == inode) {
28536 +                       found = 1;
28537 +                       break;
28538 +               }
28539 +       }
28540 +       if (!found) {
28541 +               icntnr = container_of(inode, struct au_icntnr, vfs_inode);
28542 +               hlist_bl_add_head(&icntnr->plink, hbl);
28543 +       }
28544 +       hlist_bl_unlock(hbl);
28545 +       if (!found) {
28546 +               cnt = au_hbl_count(hbl);
28547 +#define msg "unexpectedly unbalanced or too many pseudo-links"
28548 +               if (cnt > AUFS_PLINK_WARN)
28549 +                       AuWarn1(msg ", %d\n", cnt);
28550 +#undef msg
28551 +               err = whplink(h_dentry, inode, bindex);
28552 +               if (unlikely(err)) {
28553 +                       pr_warn("err %d, damaged pseudo link.\n", err);
28554 +                       au_hbl_del(&icntnr->plink, hbl);
28555 +                       iput(&icntnr->vfs_inode);
28556 +               }
28557 +       } else
28558 +               iput(&icntnr->vfs_inode);
28559 +}
28560 +
28561 +/* free all plinks */
28562 +void au_plink_put(struct super_block *sb, int verbose)
28563 +{
28564 +       int i, warned;
28565 +       struct au_sbinfo *sbinfo;
28566 +       struct hlist_bl_head *hbl;
28567 +       struct hlist_bl_node *pos, *tmp;
28568 +       struct au_icntnr *icntnr;
28569 +
28570 +       SiMustWriteLock(sb);
28571 +
28572 +       sbinfo = au_sbi(sb);
28573 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28574 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28575 +
28576 +       /* no spin_lock since sbinfo is write-locked */
28577 +       warned = 0;
28578 +       for (i = 0; i < AuPlink_NHASH; i++) {
28579 +               hbl = sbinfo->si_plink + i;
28580 +               if (!warned && verbose && !hlist_bl_empty(hbl)) {
28581 +                       pr_warn("pseudo-link is not flushed");
28582 +                       warned = 1;
28583 +               }
28584 +               hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink)
28585 +                       iput(&icntnr->vfs_inode);
28586 +               INIT_HLIST_BL_HEAD(hbl);
28587 +       }
28588 +}
28589 +
28590 +void au_plink_clean(struct super_block *sb, int verbose)
28591 +{
28592 +       struct dentry *root;
28593 +
28594 +       root = sb->s_root;
28595 +       aufs_write_lock(root);
28596 +       if (au_opt_test(au_mntflags(sb), PLINK))
28597 +               au_plink_put(sb, verbose);
28598 +       aufs_write_unlock(root);
28599 +}
28600 +
28601 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
28602 +{
28603 +       int do_put;
28604 +       aufs_bindex_t btop, bbot, bindex;
28605 +
28606 +       do_put = 0;
28607 +       btop = au_ibtop(inode);
28608 +       bbot = au_ibbot(inode);
28609 +       if (btop >= 0) {
28610 +               for (bindex = btop; bindex <= bbot; bindex++) {
28611 +                       if (!au_h_iptr(inode, bindex)
28612 +                           || au_ii_br_id(inode, bindex) != br_id)
28613 +                               continue;
28614 +                       au_set_h_iptr(inode, bindex, NULL, 0);
28615 +                       do_put = 1;
28616 +                       break;
28617 +               }
28618 +               if (do_put)
28619 +                       for (bindex = btop; bindex <= bbot; bindex++)
28620 +                               if (au_h_iptr(inode, bindex)) {
28621 +                                       do_put = 0;
28622 +                                       break;
28623 +                               }
28624 +       } else
28625 +               do_put = 1;
28626 +
28627 +       return do_put;
28628 +}
28629 +
28630 +/* free the plinks on a branch specified by @br_id */
28631 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
28632 +{
28633 +       struct au_sbinfo *sbinfo;
28634 +       struct hlist_bl_head *hbl;
28635 +       struct hlist_bl_node *pos, *tmp;
28636 +       struct au_icntnr *icntnr;
28637 +       struct inode *inode;
28638 +       int i, do_put;
28639 +
28640 +       SiMustWriteLock(sb);
28641 +
28642 +       sbinfo = au_sbi(sb);
28643 +       AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28644 +       AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28645 +
28646 +       /* no bit_lock since sbinfo is write-locked */
28647 +       for (i = 0; i < AuPlink_NHASH; i++) {
28648 +               hbl = sbinfo->si_plink + i;
28649 +               hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) {
28650 +                       inode = au_igrab(&icntnr->vfs_inode);
28651 +                       ii_write_lock_child(inode);
28652 +                       do_put = au_plink_do_half_refresh(inode, br_id);
28653 +                       if (do_put) {
28654 +                               hlist_bl_del(&icntnr->plink);
28655 +                               iput(inode);
28656 +                       }
28657 +                       ii_write_unlock(inode);
28658 +                       iput(inode);
28659 +               }
28660 +       }
28661 +}
28662 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
28663 --- /usr/share/empty/fs/aufs/poll.c     1970-01-01 01:00:00.000000000 +0100
28664 +++ linux/fs/aufs/poll.c        2022-11-05 23:02:18.969222617 +0100
28665 @@ -0,0 +1,51 @@
28666 +// SPDX-License-Identifier: GPL-2.0
28667 +/*
28668 + * Copyright (C) 2005-2022 Junjiro R. Okajima
28669 + *
28670 + * This program is free software; you can redistribute it and/or modify
28671 + * it under the terms of the GNU General Public License as published by
28672 + * the Free Software Foundation; either version 2 of the License, or
28673 + * (at your option) any later version.
28674 + *
28675 + * This program is distributed in the hope that it will be useful,
28676 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28677 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28678 + * GNU General Public License for more details.
28679 + *
28680 + * You should have received a copy of the GNU General Public License
28681 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28682 + */
28683 +
28684 +/*
28685 + * poll operation
28686 + * There is only one filesystem which implements ->poll operation, currently.
28687 + */
28688 +
28689 +#include "aufs.h"
28690 +
28691 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt)
28692 +{
28693 +       __poll_t mask;
28694 +       struct file *h_file;
28695 +       struct super_block *sb;
28696 +
28697 +       /* We should pretend an error happened. */
28698 +       mask = EPOLLERR /* | EPOLLIN | EPOLLOUT */;
28699 +       sb = file->f_path.dentry->d_sb;
28700 +       si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
28701 +
28702 +       h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
28703 +       if (IS_ERR(h_file)) {
28704 +               AuDbg("h_file %ld\n", PTR_ERR(h_file));
28705 +               goto out;
28706 +       }
28707 +
28708 +       mask = vfs_poll(h_file, pt);
28709 +       fput(h_file); /* instead of au_read_post() */
28710 +
28711 +out:
28712 +       si_read_unlock(sb);
28713 +       if (mask & EPOLLERR)
28714 +               AuDbg("mask 0x%x\n", mask);
28715 +       return mask;
28716 +}
28717 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
28718 --- /usr/share/empty/fs/aufs/posix_acl.c        1970-01-01 01:00:00.000000000 +0100
28719 +++ linux/fs/aufs/posix_acl.c   2022-11-05 23:02:18.969222617 +0100
28720 @@ -0,0 +1,111 @@
28721 +// SPDX-License-Identifier: GPL-2.0
28722 +/*
28723 + * Copyright (C) 2014-2022 Junjiro R. Okajima
28724 + *
28725 + * This program is free software; you can redistribute it and/or modify
28726 + * it under the terms of the GNU General Public License as published by
28727 + * the Free Software Foundation; either version 2 of the License, or
28728 + * (at your option) any later version.
28729 + *
28730 + * This program is distributed in the hope that it will be useful,
28731 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28732 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28733 + * GNU General Public License for more details.
28734 + *
28735 + * You should have received a copy of the GNU General Public License
28736 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28737 + */
28738 +
28739 +/*
28740 + * posix acl operations
28741 + */
28742 +
28743 +#include <linux/fs.h>
28744 +#include "aufs.h"
28745 +
28746 +struct posix_acl *aufs_get_acl(struct inode *inode, int type, bool rcu)
28747 +{
28748 +       struct posix_acl *acl;
28749 +       int err;
28750 +       aufs_bindex_t bindex;
28751 +       struct inode *h_inode;
28752 +       struct super_block *sb;
28753 +
28754 +       acl = ERR_PTR(-ECHILD);
28755 +       if (rcu)
28756 +               goto out;
28757 +
28758 +       acl = NULL;
28759 +       sb = inode->i_sb;
28760 +       si_read_lock(sb, AuLock_FLUSH);
28761 +       ii_read_lock_child(inode);
28762 +       if (!(sb->s_flags & SB_POSIXACL))
28763 +               goto unlock;
28764 +
28765 +       bindex = au_ibtop(inode);
28766 +       h_inode = au_h_iptr(inode, bindex);
28767 +       if (unlikely(!h_inode
28768 +                    || ((h_inode->i_mode & S_IFMT)
28769 +                        != (inode->i_mode & S_IFMT)))) {
28770 +               err = au_busy_or_stale();
28771 +               acl = ERR_PTR(err);
28772 +               goto unlock;
28773 +       }
28774 +
28775 +       /* always topmost only */
28776 +       acl = get_acl(h_inode, type);
28777 +       if (IS_ERR(acl))
28778 +               forget_cached_acl(inode, type);
28779 +       else
28780 +               set_cached_acl(inode, type, acl);
28781 +
28782 +unlock:
28783 +       ii_read_unlock(inode);
28784 +       si_read_unlock(sb);
28785 +
28786 +out:
28787 +       AuTraceErrPtr(acl);
28788 +       return acl;
28789 +}
28790 +
28791 +int aufs_set_acl(struct user_namespace *userns, struct inode *inode,
28792 +                struct posix_acl *acl, int type)
28793 +{
28794 +       int err;
28795 +       ssize_t ssz;
28796 +       struct dentry *dentry;
28797 +       struct au_sxattr arg = {
28798 +               .type = AU_ACL_SET,
28799 +               .u.acl_set = {
28800 +                       .acl    = acl,
28801 +                       .type   = type
28802 +               },
28803 +       };
28804 +
28805 +       IMustLock(inode);
28806 +
28807 +       if (inode->i_ino == AUFS_ROOT_INO)
28808 +               dentry = dget(inode->i_sb->s_root);
28809 +       else {
28810 +               dentry = d_find_alias(inode);
28811 +               if (!dentry)
28812 +                       dentry = d_find_any_alias(inode);
28813 +               if (!dentry) {
28814 +                       pr_warn("cannot handle this inode, "
28815 +                               "please report to aufs-users ML\n");
28816 +                       err = -ENOENT;
28817 +                       goto out;
28818 +               }
28819 +       }
28820 +
28821 +       ssz = au_sxattr(dentry, inode, &arg);
28822 +       /* forget even it if succeeds since the branch might set differently */
28823 +       forget_cached_acl(inode, type);
28824 +       dput(dentry);
28825 +       err = ssz;
28826 +       if (ssz >= 0)
28827 +               err = 0;
28828 +
28829 +out:
28830 +       return err;
28831 +}
28832 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
28833 --- /usr/share/empty/fs/aufs/procfs.c   1970-01-01 01:00:00.000000000 +0100
28834 +++ linux/fs/aufs/procfs.c      2022-11-05 23:02:18.969222617 +0100
28835 @@ -0,0 +1,170 @@
28836 +// SPDX-License-Identifier: GPL-2.0
28837 +/*
28838 + * Copyright (C) 2010-2022 Junjiro R. Okajima
28839 + *
28840 + * This program is free software; you can redistribute it and/or modify
28841 + * it under the terms of the GNU General Public License as published by
28842 + * the Free Software Foundation; either version 2 of the License, or
28843 + * (at your option) any later version.
28844 + *
28845 + * This program is distributed in the hope that it will be useful,
28846 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28847 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
28848 + * GNU General Public License for more details.
28849 + *
28850 + * You should have received a copy of the GNU General Public License
28851 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
28852 + */
28853 +
28854 +/*
28855 + * procfs interfaces
28856 + */
28857 +
28858 +#include <linux/proc_fs.h>
28859 +#include "aufs.h"
28860 +
28861 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
28862 +{
28863 +       struct au_sbinfo *sbinfo;
28864 +
28865 +       sbinfo = file->private_data;
28866 +       if (sbinfo) {
28867 +               au_plink_maint_leave(sbinfo);
28868 +               kobject_put(&sbinfo->si_kobj);
28869 +       }
28870 +
28871 +       return 0;
28872 +}
28873 +
28874 +static void au_procfs_plm_write_clean(struct file *file)
28875 +{
28876 +       struct au_sbinfo *sbinfo;
28877 +
28878 +       sbinfo = file->private_data;
28879 +       if (sbinfo)
28880 +               au_plink_clean(sbinfo->si_sb, /*verbose*/0);
28881 +}
28882 +
28883 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
28884 +{
28885 +       int err;
28886 +       struct super_block *sb;
28887 +       struct au_sbinfo *sbinfo;
28888 +       struct hlist_bl_node *pos;
28889 +
28890 +       err = -EBUSY;
28891 +       if (unlikely(file->private_data))
28892 +               goto out;
28893 +
28894 +       sb = NULL;
28895 +       /* don't use au_sbilist_lock() here */
28896 +       hlist_bl_lock(&au_sbilist);
28897 +       hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
28898 +               if (id == sysaufs_si_id(sbinfo)) {
28899 +                       if (kobject_get_unless_zero(&sbinfo->si_kobj))
28900 +                               sb = sbinfo->si_sb;
28901 +                       break;
28902 +               }
28903 +       hlist_bl_unlock(&au_sbilist);
28904 +
28905 +       err = -EINVAL;
28906 +       if (unlikely(!sb))
28907 +               goto out;
28908 +
28909 +       err = au_plink_maint_enter(sb);
28910 +       if (!err)
28911 +               /* keep kobject_get() */
28912 +               file->private_data = sbinfo;
28913 +       else
28914 +               kobject_put(&sbinfo->si_kobj);
28915 +out:
28916 +       return err;
28917 +}
28918 +
28919 +/*
28920 + * Accept a valid "si=xxxx" only.
28921 + * Once it is accepted successfully, accept "clean" too.
28922 + */
28923 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
28924 +                                  size_t count, loff_t *ppos)
28925 +{
28926 +       ssize_t err;
28927 +       unsigned long id;
28928 +       /* last newline is allowed */
28929 +       char buf[3 + sizeof(unsigned long) * 2 + 1];
28930 +
28931 +       err = -EACCES;
28932 +       if (unlikely(!capable(CAP_SYS_ADMIN)))
28933 +               goto out;
28934 +
28935 +       err = -EINVAL;
28936 +       if (unlikely(count > sizeof(buf)))
28937 +               goto out;
28938 +
28939 +       err = copy_from_user(buf, ubuf, count);
28940 +       if (unlikely(err)) {
28941 +               err = -EFAULT;
28942 +               goto out;
28943 +       }
28944 +       buf[count] = 0;
28945 +
28946 +       err = -EINVAL;
28947 +       if (!strcmp("clean", buf)) {
28948 +               au_procfs_plm_write_clean(file);
28949 +               goto out_success;
28950 +       } else if (unlikely(strncmp("si=", buf, 3)))
28951 +               goto out;
28952 +
28953 +       err = kstrtoul(buf + 3, 16, &id);
28954 +       if (unlikely(err))
28955 +               goto out;
28956 +
28957 +       err = au_procfs_plm_write_si(file, id);
28958 +       if (unlikely(err))
28959 +               goto out;
28960 +
28961 +out_success:
28962 +       err = count; /* success */
28963 +out:
28964 +       return err;
28965 +}
28966 +
28967 +static const struct proc_ops au_procfs_plm_op = {
28968 +       .proc_write     = au_procfs_plm_write,
28969 +       .proc_release   = au_procfs_plm_release
28970 +};
28971 +
28972 +/* ---------------------------------------------------------------------- */
28973 +
28974 +static struct proc_dir_entry *au_procfs_dir;
28975 +
28976 +void au_procfs_fin(void)
28977 +{
28978 +       remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
28979 +       remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
28980 +}
28981 +
28982 +int __init au_procfs_init(void)
28983 +{
28984 +       int err;
28985 +       struct proc_dir_entry *entry;
28986 +
28987 +       err = -ENOMEM;
28988 +       au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
28989 +       if (unlikely(!au_procfs_dir))
28990 +               goto out;
28991 +
28992 +       entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | 0200,
28993 +                           au_procfs_dir, &au_procfs_plm_op);
28994 +       if (unlikely(!entry))
28995 +               goto out_dir;
28996 +
28997 +       err = 0;
28998 +       goto out; /* success */
28999 +
29000 +
29001 +out_dir:
29002 +       remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29003 +out:
29004 +       return err;
29005 +}
29006 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
29007 --- /usr/share/empty/fs/aufs/rdu.c      1970-01-01 01:00:00.000000000 +0100
29008 +++ linux/fs/aufs/rdu.c 2022-11-05 23:02:18.969222617 +0100
29009 @@ -0,0 +1,384 @@
29010 +// SPDX-License-Identifier: GPL-2.0
29011 +/*
29012 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29013 + *
29014 + * This program is free software; you can redistribute it and/or modify
29015 + * it under the terms of the GNU General Public License as published by
29016 + * the Free Software Foundation; either version 2 of the License, or
29017 + * (at your option) any later version.
29018 + *
29019 + * This program is distributed in the hope that it will be useful,
29020 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29021 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29022 + * GNU General Public License for more details.
29023 + *
29024 + * You should have received a copy of the GNU General Public License
29025 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29026 + */
29027 +
29028 +/*
29029 + * readdir in userspace.
29030 + */
29031 +
29032 +#include <linux/compat.h>
29033 +#include <linux/fs_stack.h>
29034 +#include <linux/security.h>
29035 +#include "aufs.h"
29036 +
29037 +/* bits for struct aufs_rdu.flags */
29038 +#define        AuRdu_CALLED    1
29039 +#define        AuRdu_CONT      (1 << 1)
29040 +#define        AuRdu_FULL      (1 << 2)
29041 +#define au_ftest_rdu(flags, name)      ((flags) & AuRdu_##name)
29042 +#define au_fset_rdu(flags, name) \
29043 +       do { (flags) |= AuRdu_##name; } while (0)
29044 +#define au_fclr_rdu(flags, name) \
29045 +       do { (flags) &= ~AuRdu_##name; } while (0)
29046 +
29047 +struct au_rdu_arg {
29048 +       struct dir_context              ctx;
29049 +       struct aufs_rdu                 *rdu;
29050 +       union au_rdu_ent_ul             ent;
29051 +       unsigned long                   end;
29052 +
29053 +       struct super_block              *sb;
29054 +       int                             err;
29055 +};
29056 +
29057 +static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
29058 +                      loff_t offset, u64 h_ino, unsigned int d_type)
29059 +{
29060 +       int err, len;
29061 +       struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
29062 +       struct aufs_rdu *rdu = arg->rdu;
29063 +       struct au_rdu_ent ent;
29064 +
29065 +       err = 0;
29066 +       arg->err = 0;
29067 +       au_fset_rdu(rdu->cookie.flags, CALLED);
29068 +       len = au_rdu_len(nlen);
29069 +       if (arg->ent.ul + len  < arg->end) {
29070 +               ent.ino = h_ino;
29071 +               ent.bindex = rdu->cookie.bindex;
29072 +               ent.type = d_type;
29073 +               ent.nlen = nlen;
29074 +               if (unlikely(nlen > AUFS_MAX_NAMELEN))
29075 +                       ent.type = DT_UNKNOWN;
29076 +
29077 +               /* unnecessary to support mmap_sem since this is a dir */
29078 +               err = -EFAULT;
29079 +               if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
29080 +                       goto out;
29081 +               if (copy_to_user(arg->ent.e->name, name, nlen))
29082 +                       goto out;
29083 +               /* the terminating NULL */
29084 +               if (__put_user(0, arg->ent.e->name + nlen))
29085 +                       goto out;
29086 +               err = 0;
29087 +               /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
29088 +               arg->ent.ul += len;
29089 +               rdu->rent++;
29090 +       } else {
29091 +               err = -EFAULT;
29092 +               au_fset_rdu(rdu->cookie.flags, FULL);
29093 +               rdu->full = 1;
29094 +               rdu->tail = arg->ent;
29095 +       }
29096 +
29097 +out:
29098 +       /* AuTraceErr(err); */
29099 +       return err;
29100 +}
29101 +
29102 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
29103 +{
29104 +       int err;
29105 +       loff_t offset;
29106 +       struct au_rdu_cookie *cookie = &arg->rdu->cookie;
29107 +
29108 +       /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
29109 +       offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
29110 +       err = offset;
29111 +       if (unlikely(offset != cookie->h_pos))
29112 +               goto out;
29113 +
29114 +       err = 0;
29115 +       do {
29116 +               arg->err = 0;
29117 +               au_fclr_rdu(cookie->flags, CALLED);
29118 +               /* smp_mb(); */
29119 +               err = vfsub_iterate_dir(h_file, &arg->ctx);
29120 +               if (err >= 0)
29121 +                       err = arg->err;
29122 +       } while (!err
29123 +                && au_ftest_rdu(cookie->flags, CALLED)
29124 +                && !au_ftest_rdu(cookie->flags, FULL));
29125 +       cookie->h_pos = h_file->f_pos;
29126 +
29127 +out:
29128 +       AuTraceErr(err);
29129 +       return err;
29130 +}
29131 +
29132 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
29133 +{
29134 +       int err;
29135 +       aufs_bindex_t bbot;
29136 +       struct au_rdu_arg arg = {
29137 +               .ctx = {
29138 +                       .actor = au_rdu_fill
29139 +               }
29140 +       };
29141 +       struct dentry *dentry;
29142 +       struct inode *inode;
29143 +       struct file *h_file;
29144 +       struct au_rdu_cookie *cookie = &rdu->cookie;
29145 +
29146 +       /* VERIFY_WRITE */
29147 +       err = !access_ok(rdu->ent.e, rdu->sz);
29148 +       if (unlikely(err)) {
29149 +               err = -EFAULT;
29150 +               AuTraceErr(err);
29151 +               goto out;
29152 +       }
29153 +       rdu->rent = 0;
29154 +       rdu->tail = rdu->ent;
29155 +       rdu->full = 0;
29156 +       arg.rdu = rdu;
29157 +       arg.ent = rdu->ent;
29158 +       arg.end = arg.ent.ul;
29159 +       arg.end += rdu->sz;
29160 +
29161 +       err = -ENOTDIR;
29162 +       if (unlikely(!file->f_op->iterate && !file->f_op->iterate_shared))
29163 +               goto out;
29164 +
29165 +       err = security_file_permission(file, MAY_READ);
29166 +       AuTraceErr(err);
29167 +       if (unlikely(err))
29168 +               goto out;
29169 +
29170 +       dentry = file->f_path.dentry;
29171 +       inode = d_inode(dentry);
29172 +       inode_lock_shared(inode);
29173 +
29174 +       arg.sb = inode->i_sb;
29175 +       err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
29176 +       if (unlikely(err))
29177 +               goto out_mtx;
29178 +       err = au_alive_dir(dentry);
29179 +       if (unlikely(err))
29180 +               goto out_si;
29181 +       /* todo: reval? */
29182 +       fi_read_lock(file);
29183 +
29184 +       err = -EAGAIN;
29185 +       if (unlikely(au_ftest_rdu(cookie->flags, CONT)
29186 +                    && cookie->generation != au_figen(file)))
29187 +               goto out_unlock;
29188 +
29189 +       err = 0;
29190 +       if (!rdu->blk) {
29191 +               rdu->blk = au_sbi(arg.sb)->si_rdblk;
29192 +               if (!rdu->blk)
29193 +                       rdu->blk = au_dir_size(file, /*dentry*/NULL);
29194 +       }
29195 +       bbot = au_fbtop(file);
29196 +       if (cookie->bindex < bbot)
29197 +               cookie->bindex = bbot;
29198 +       bbot = au_fbbot_dir(file);
29199 +       /* AuDbg("b%d, b%d\n", cookie->bindex, bbot); */
29200 +       for (; !err && cookie->bindex <= bbot;
29201 +            cookie->bindex++, cookie->h_pos = 0) {
29202 +               h_file = au_hf_dir(file, cookie->bindex);
29203 +               if (!h_file)
29204 +                       continue;
29205 +
29206 +               au_fclr_rdu(cookie->flags, FULL);
29207 +               err = au_rdu_do(h_file, &arg);
29208 +               AuTraceErr(err);
29209 +               if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
29210 +                       break;
29211 +       }
29212 +       AuDbg("rent %llu\n", rdu->rent);
29213 +
29214 +       if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
29215 +               rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
29216 +               au_fset_rdu(cookie->flags, CONT);
29217 +               cookie->generation = au_figen(file);
29218 +       }
29219 +
29220 +       ii_read_lock_child(inode);
29221 +       fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibtop(inode)));
29222 +       ii_read_unlock(inode);
29223 +
29224 +out_unlock:
29225 +       fi_read_unlock(file);
29226 +out_si:
29227 +       si_read_unlock(arg.sb);
29228 +out_mtx:
29229 +       inode_unlock_shared(inode);
29230 +out:
29231 +       AuTraceErr(err);
29232 +       return err;
29233 +}
29234 +
29235 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
29236 +{
29237 +       int err;
29238 +       ino_t ino;
29239 +       unsigned long long nent;
29240 +       union au_rdu_ent_ul *u;
29241 +       struct au_rdu_ent ent;
29242 +       struct super_block *sb;
29243 +
29244 +       err = 0;
29245 +       nent = rdu->nent;
29246 +       u = &rdu->ent;
29247 +       sb = file->f_path.dentry->d_sb;
29248 +       si_read_lock(sb, AuLock_FLUSH);
29249 +       while (nent-- > 0) {
29250 +               /* unnecessary to support mmap_sem since this is a dir */
29251 +               err = copy_from_user(&ent, u->e, sizeof(ent));
29252 +               if (!err)
29253 +                       /* VERIFY_WRITE */
29254 +                       err = !access_ok(&u->e->ino, sizeof(ino));
29255 +               if (unlikely(err)) {
29256 +                       err = -EFAULT;
29257 +                       AuTraceErr(err);
29258 +                       break;
29259 +               }
29260 +
29261 +               /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
29262 +               if (!ent.wh)
29263 +                       err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
29264 +               else
29265 +                       err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
29266 +                                       &ino);
29267 +               if (unlikely(err)) {
29268 +                       AuTraceErr(err);
29269 +                       break;
29270 +               }
29271 +
29272 +               err = __put_user(ino, &u->e->ino);
29273 +               if (unlikely(err)) {
29274 +                       err = -EFAULT;
29275 +                       AuTraceErr(err);
29276 +                       break;
29277 +               }
29278 +               u->ul += au_rdu_len(ent.nlen);
29279 +       }
29280 +       si_read_unlock(sb);
29281 +
29282 +       return err;
29283 +}
29284 +
29285 +/* ---------------------------------------------------------------------- */
29286 +
29287 +static int au_rdu_verify(struct aufs_rdu *rdu)
29288 +{
29289 +       AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
29290 +             "%llu, b%d, 0x%x, g%u}\n",
29291 +             rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
29292 +             rdu->blk,
29293 +             rdu->rent, rdu->shwh, rdu->full,
29294 +             rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
29295 +             rdu->cookie.generation);
29296 +
29297 +       if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
29298 +               return 0;
29299 +
29300 +       AuDbg("%u:%u\n",
29301 +             rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
29302 +       return -EINVAL;
29303 +}
29304 +
29305 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29306 +{
29307 +       long err, e;
29308 +       struct aufs_rdu rdu;
29309 +       void __user *p = (void __user *)arg;
29310 +
29311 +       err = copy_from_user(&rdu, p, sizeof(rdu));
29312 +       if (unlikely(err)) {
29313 +               err = -EFAULT;
29314 +               AuTraceErr(err);
29315 +               goto out;
29316 +       }
29317 +       err = au_rdu_verify(&rdu);
29318 +       if (unlikely(err))
29319 +               goto out;
29320 +
29321 +       switch (cmd) {
29322 +       case AUFS_CTL_RDU:
29323 +               err = au_rdu(file, &rdu);
29324 +               if (unlikely(err))
29325 +                       break;
29326 +
29327 +               e = copy_to_user(p, &rdu, sizeof(rdu));
29328 +               if (unlikely(e)) {
29329 +                       err = -EFAULT;
29330 +                       AuTraceErr(err);
29331 +               }
29332 +               break;
29333 +       case AUFS_CTL_RDU_INO:
29334 +               err = au_rdu_ino(file, &rdu);
29335 +               break;
29336 +
29337 +       default:
29338 +               /* err = -ENOTTY; */
29339 +               err = -EINVAL;
29340 +       }
29341 +
29342 +out:
29343 +       AuTraceErr(err);
29344 +       return err;
29345 +}
29346 +
29347 +#ifdef CONFIG_COMPAT
29348 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29349 +{
29350 +       long err, e;
29351 +       struct aufs_rdu rdu;
29352 +       void __user *p = compat_ptr(arg);
29353 +
29354 +       /* todo: get_user()? */
29355 +       err = copy_from_user(&rdu, p, sizeof(rdu));
29356 +       if (unlikely(err)) {
29357 +               err = -EFAULT;
29358 +               AuTraceErr(err);
29359 +               goto out;
29360 +       }
29361 +       rdu.ent.e = compat_ptr(rdu.ent.ul);
29362 +       err = au_rdu_verify(&rdu);
29363 +       if (unlikely(err))
29364 +               goto out;
29365 +
29366 +       switch (cmd) {
29367 +       case AUFS_CTL_RDU:
29368 +               err = au_rdu(file, &rdu);
29369 +               if (unlikely(err))
29370 +                       break;
29371 +
29372 +               rdu.ent.ul = ptr_to_compat(rdu.ent.e);
29373 +               rdu.tail.ul = ptr_to_compat(rdu.tail.e);
29374 +               e = copy_to_user(p, &rdu, sizeof(rdu));
29375 +               if (unlikely(e)) {
29376 +                       err = -EFAULT;
29377 +                       AuTraceErr(err);
29378 +               }
29379 +               break;
29380 +       case AUFS_CTL_RDU_INO:
29381 +               err = au_rdu_ino(file, &rdu);
29382 +               break;
29383 +
29384 +       default:
29385 +               /* err = -ENOTTY; */
29386 +               err = -EINVAL;
29387 +       }
29388 +
29389 +out:
29390 +       AuTraceErr(err);
29391 +       return err;
29392 +}
29393 +#endif
29394 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
29395 --- /usr/share/empty/fs/aufs/rwsem.h    1970-01-01 01:00:00.000000000 +0100
29396 +++ linux/fs/aufs/rwsem.h       2022-11-05 23:02:18.969222617 +0100
29397 @@ -0,0 +1,85 @@
29398 +/* SPDX-License-Identifier: GPL-2.0 */
29399 +/*
29400 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29401 + *
29402 + * This program is free software; you can redistribute it and/or modify
29403 + * it under the terms of the GNU General Public License as published by
29404 + * the Free Software Foundation; either version 2 of the License, or
29405 + * (at your option) any later version.
29406 + *
29407 + * This program is distributed in the hope that it will be useful,
29408 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29409 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29410 + * GNU General Public License for more details.
29411 + *
29412 + * You should have received a copy of the GNU General Public License
29413 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29414 + */
29415 +
29416 +/*
29417 + * simple read-write semaphore wrappers
29418 + */
29419 +
29420 +#ifndef __AUFS_RWSEM_H__
29421 +#define __AUFS_RWSEM_H__
29422 +
29423 +#ifdef __KERNEL__
29424 +
29425 +#include "debug.h"
29426 +
29427 +/* in the future, the name 'au_rwsem' will be totally gone */
29428 +#define au_rwsem       rw_semaphore
29429 +
29430 +/* to debug easier, do not make them inlined functions */
29431 +#define AuRwMustNoWaiters(rw)  AuDebugOn(rwsem_is_contended(rw))
29432 +
29433 +#ifdef CONFIG_LOCKDEP
29434 +/* rwsem_is_locked() is unusable */
29435 +#define AuRwMustReadLock(rw)   AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29436 +                                         && !lockdep_recursing(current) \
29437 +                                         && debug_locks                \
29438 +                                         && !lockdep_is_held_type(rw, 1))
29439 +#define AuRwMustWriteLock(rw)  AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29440 +                                         && !lockdep_recursing(current) \
29441 +                                         && debug_locks                \
29442 +                                         && !lockdep_is_held_type(rw, 0))
29443 +#define AuRwMustAnyLock(rw)    AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29444 +                                         && !lockdep_recursing(current) \
29445 +                                         && debug_locks                \
29446 +                                         && !lockdep_is_held(rw))
29447 +#define AuRwDestroy(rw)                AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP)    \
29448 +                                         && !lockdep_recursing(current) \
29449 +                                         && debug_locks                \
29450 +                                         && lockdep_is_held(rw))
29451 +#else
29452 +#define AuRwMustReadLock(rw)   do {} while (0)
29453 +#define AuRwMustWriteLock(rw)  do {} while (0)
29454 +#define AuRwMustAnyLock(rw)    do {} while (0)
29455 +#define AuRwDestroy(rw)                do {} while (0)
29456 +#endif
29457 +
29458 +#define au_rw_init(rw) init_rwsem(rw)
29459 +
29460 +#define au_rw_init_wlock(rw) do {              \
29461 +               au_rw_init(rw);                 \
29462 +               down_write(rw);                 \
29463 +       } while (0)
29464 +
29465 +#define au_rw_init_wlock_nested(rw, lsc) do {  \
29466 +               au_rw_init(rw);                 \
29467 +               down_write_nested(rw, lsc);     \
29468 +       } while (0)
29469 +
29470 +#define au_rw_read_lock(rw)            down_read(rw)
29471 +#define au_rw_read_lock_nested(rw, lsc)        down_read_nested(rw, lsc)
29472 +#define au_rw_read_unlock(rw)          up_read(rw)
29473 +#define au_rw_dgrade_lock(rw)          downgrade_write(rw)
29474 +#define au_rw_write_lock(rw)           down_write(rw)
29475 +#define au_rw_write_lock_nested(rw, lsc) down_write_nested(rw, lsc)
29476 +#define au_rw_write_unlock(rw)         up_write(rw)
29477 +/* why is not _nested version defined? */
29478 +#define au_rw_read_trylock(rw)         down_read_trylock(rw)
29479 +#define au_rw_write_trylock(rw)                down_write_trylock(rw)
29480 +
29481 +#endif /* __KERNEL__ */
29482 +#endif /* __AUFS_RWSEM_H__ */
29483 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
29484 --- /usr/share/empty/fs/aufs/sbinfo.c   1970-01-01 01:00:00.000000000 +0100
29485 +++ linux/fs/aufs/sbinfo.c      2022-11-05 23:02:18.969222617 +0100
29486 @@ -0,0 +1,316 @@
29487 +// SPDX-License-Identifier: GPL-2.0
29488 +/*
29489 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29490 + *
29491 + * This program is free software; you can redistribute it and/or modify
29492 + * it under the terms of the GNU General Public License as published by
29493 + * the Free Software Foundation; either version 2 of the License, or
29494 + * (at your option) any later version.
29495 + *
29496 + * This program is distributed in the hope that it will be useful,
29497 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29498 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29499 + * GNU General Public License for more details.
29500 + *
29501 + * You should have received a copy of the GNU General Public License
29502 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29503 + */
29504 +
29505 +/*
29506 + * superblock private data
29507 + */
29508 +
29509 +#include <linux/iversion.h>
29510 +#include "aufs.h"
29511 +
29512 +/*
29513 + * they are necessary regardless sysfs is disabled.
29514 + */
29515 +void au_si_free(struct kobject *kobj)
29516 +{
29517 +       int i;
29518 +       struct au_sbinfo *sbinfo;
29519 +       char *locked __maybe_unused; /* debug only */
29520 +
29521 +       sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
29522 +       for (i = 0; i < AuPlink_NHASH; i++)
29523 +               AuDebugOn(!hlist_bl_empty(sbinfo->si_plink + i));
29524 +       AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
29525 +
29526 +       AuLCntZero(au_lcnt_read(&sbinfo->si_ninodes, /*do_rev*/0));
29527 +       au_lcnt_fin(&sbinfo->si_ninodes, /*do_sync*/0);
29528 +       AuLCntZero(au_lcnt_read(&sbinfo->si_nfiles, /*do_rev*/0));
29529 +       au_lcnt_fin(&sbinfo->si_nfiles, /*do_sync*/0);
29530 +
29531 +       dbgaufs_si_fin(sbinfo);
29532 +       au_rw_write_lock(&sbinfo->si_rwsem);
29533 +       au_br_free(sbinfo);
29534 +       au_rw_write_unlock(&sbinfo->si_rwsem);
29535 +
29536 +       au_kfree_try_rcu(sbinfo->si_branch);
29537 +       mutex_destroy(&sbinfo->si_xib_mtx);
29538 +       AuRwDestroy(&sbinfo->si_rwsem);
29539 +
29540 +       au_lcnt_wait_for_fin(&sbinfo->si_ninodes);
29541 +       /* si_nfiles is waited too */
29542 +       au_kfree_rcu(sbinfo);
29543 +}
29544 +
29545 +struct au_sbinfo *au_si_alloc(struct super_block *sb)
29546 +{
29547 +       struct au_sbinfo *sbinfo;
29548 +       int err, i;
29549 +
29550 +       err = -ENOMEM;
29551 +       sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
29552 +       if (unlikely(!sbinfo))
29553 +               goto out;
29554 +
29555 +       /* will be reallocated separately */
29556 +       sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
29557 +       if (unlikely(!sbinfo->si_branch))
29558 +               goto out_sbinfo;
29559 +
29560 +       err = sysaufs_si_init(sbinfo);
29561 +       if (!err) {
29562 +               dbgaufs_si_null(sbinfo);
29563 +               err = dbgaufs_si_init(sbinfo);
29564 +               if (unlikely(err))
29565 +                       kobject_put(&sbinfo->si_kobj);
29566 +       }
29567 +       if (unlikely(err))
29568 +               goto out_br;
29569 +
29570 +       au_nwt_init(&sbinfo->si_nowait);
29571 +       au_rw_init_wlock(&sbinfo->si_rwsem);
29572 +
29573 +       au_lcnt_init(&sbinfo->si_ninodes, /*release*/NULL);
29574 +       au_lcnt_init(&sbinfo->si_nfiles, /*release*/NULL);
29575 +
29576 +       sbinfo->si_bbot = -1;
29577 +       sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
29578 +
29579 +       sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
29580 +       sbinfo->si_wbr_create = AuWbrCreate_Def;
29581 +       sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
29582 +       sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
29583 +
29584 +       au_fhsm_init(sbinfo);
29585 +
29586 +       sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
29587 +
29588 +       sbinfo->si_xino_jiffy = jiffies;
29589 +       sbinfo->si_xino_expire
29590 +               = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
29591 +       mutex_init(&sbinfo->si_xib_mtx);
29592 +       /* leave si_xib_last_pindex and si_xib_next_bit */
29593 +
29594 +       INIT_HLIST_BL_HEAD(&sbinfo->si_aopen);
29595 +
29596 +       sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
29597 +       sbinfo->si_rdblk = AUFS_RDBLK_DEF;
29598 +       sbinfo->si_rdhash = AUFS_RDHASH_DEF;
29599 +       sbinfo->si_dirwh = AUFS_DIRWH_DEF;
29600 +
29601 +       for (i = 0; i < AuPlink_NHASH; i++)
29602 +               INIT_HLIST_BL_HEAD(sbinfo->si_plink + i);
29603 +       init_waitqueue_head(&sbinfo->si_plink_wq);
29604 +       spin_lock_init(&sbinfo->si_plink_maint_lock);
29605 +
29606 +       INIT_HLIST_BL_HEAD(&sbinfo->si_files);
29607 +
29608 +       /* with getattr by default */
29609 +       sbinfo->si_iop_array = aufs_iop;
29610 +
29611 +       /* leave other members for sysaufs and si_mnt. */
29612 +       sbinfo->si_sb = sb;
29613 +       if (sb) {
29614 +               sb->s_fs_info = sbinfo;
29615 +               si_pid_set(sb);
29616 +       }
29617 +       return sbinfo; /* success */
29618 +
29619 +out_br:
29620 +       au_kfree_try_rcu(sbinfo->si_branch);
29621 +out_sbinfo:
29622 +       au_kfree_rcu(sbinfo);
29623 +out:
29624 +       return ERR_PTR(err);
29625 +}
29626 +
29627 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink)
29628 +{
29629 +       int err, sz;
29630 +       struct au_branch **brp;
29631 +
29632 +       AuRwMustWriteLock(&sbinfo->si_rwsem);
29633 +
29634 +       err = -ENOMEM;
29635 +       sz = sizeof(*brp) * (sbinfo->si_bbot + 1);
29636 +       if (unlikely(!sz))
29637 +               sz = sizeof(*brp);
29638 +       brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS,
29639 +                          may_shrink);
29640 +       if (brp) {
29641 +               sbinfo->si_branch = brp;
29642 +               err = 0;
29643 +       }
29644 +
29645 +       return err;
29646 +}
29647 +
29648 +/* ---------------------------------------------------------------------- */
29649 +
29650 +unsigned int au_sigen_inc(struct super_block *sb)
29651 +{
29652 +       unsigned int gen;
29653 +       struct inode *inode;
29654 +
29655 +       SiMustWriteLock(sb);
29656 +
29657 +       gen = ++au_sbi(sb)->si_generation;
29658 +       au_update_digen(sb->s_root);
29659 +       inode = d_inode(sb->s_root);
29660 +       au_update_iigen(inode, /*half*/0);
29661 +       inode_inc_iversion(inode);
29662 +       return gen;
29663 +}
29664 +
29665 +aufs_bindex_t au_new_br_id(struct super_block *sb)
29666 +{
29667 +       aufs_bindex_t br_id;
29668 +       int i;
29669 +       struct au_sbinfo *sbinfo;
29670 +
29671 +       SiMustWriteLock(sb);
29672 +
29673 +       sbinfo = au_sbi(sb);
29674 +       for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
29675 +               br_id = ++sbinfo->si_last_br_id;
29676 +               AuDebugOn(br_id < 0);
29677 +               if (br_id && au_br_index(sb, br_id) < 0)
29678 +                       return br_id;
29679 +       }
29680 +
29681 +       return -1;
29682 +}
29683 +
29684 +/* ---------------------------------------------------------------------- */
29685 +
29686 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
29687 +int si_read_lock(struct super_block *sb, int flags)
29688 +{
29689 +       int err;
29690 +
29691 +       err = 0;
29692 +       if (au_ftest_lock(flags, FLUSH))
29693 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
29694 +
29695 +       si_noflush_read_lock(sb);
29696 +       err = au_plink_maint(sb, flags);
29697 +       if (unlikely(err))
29698 +               si_read_unlock(sb);
29699 +
29700 +       return err;
29701 +}
29702 +
29703 +int si_write_lock(struct super_block *sb, int flags)
29704 +{
29705 +       int err;
29706 +
29707 +       if (au_ftest_lock(flags, FLUSH))
29708 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
29709 +
29710 +       si_noflush_write_lock(sb);
29711 +       err = au_plink_maint(sb, flags);
29712 +       if (unlikely(err))
29713 +               si_write_unlock(sb);
29714 +
29715 +       return err;
29716 +}
29717 +
29718 +/* dentry and super_block lock. call at entry point */
29719 +int aufs_read_lock(struct dentry *dentry, int flags)
29720 +{
29721 +       int err;
29722 +       struct super_block *sb;
29723 +
29724 +       sb = dentry->d_sb;
29725 +       err = si_read_lock(sb, flags);
29726 +       if (unlikely(err))
29727 +               goto out;
29728 +
29729 +       if (au_ftest_lock(flags, DW))
29730 +               di_write_lock_child(dentry);
29731 +       else
29732 +               di_read_lock_child(dentry, flags);
29733 +
29734 +       if (au_ftest_lock(flags, GEN)) {
29735 +               err = au_digen_test(dentry, au_sigen(sb));
29736 +               if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
29737 +                       AuDebugOn(!err && au_dbrange_test(dentry));
29738 +               else if (!err)
29739 +                       err = au_dbrange_test(dentry);
29740 +               if (unlikely(err))
29741 +                       aufs_read_unlock(dentry, flags);
29742 +       }
29743 +
29744 +out:
29745 +       return err;
29746 +}
29747 +
29748 +void aufs_read_unlock(struct dentry *dentry, int flags)
29749 +{
29750 +       if (au_ftest_lock(flags, DW))
29751 +               di_write_unlock(dentry);
29752 +       else
29753 +               di_read_unlock(dentry, flags);
29754 +       si_read_unlock(dentry->d_sb);
29755 +}
29756 +
29757 +void aufs_write_lock(struct dentry *dentry)
29758 +{
29759 +       si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
29760 +       di_write_lock_child(dentry);
29761 +}
29762 +
29763 +void aufs_write_unlock(struct dentry *dentry)
29764 +{
29765 +       di_write_unlock(dentry);
29766 +       si_write_unlock(dentry->d_sb);
29767 +}
29768 +
29769 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
29770 +{
29771 +       int err;
29772 +       unsigned int sigen;
29773 +       struct super_block *sb;
29774 +
29775 +       sb = d1->d_sb;
29776 +       err = si_read_lock(sb, flags);
29777 +       if (unlikely(err))
29778 +               goto out;
29779 +
29780 +       di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
29781 +
29782 +       if (au_ftest_lock(flags, GEN)) {
29783 +               sigen = au_sigen(sb);
29784 +               err = au_digen_test(d1, sigen);
29785 +               AuDebugOn(!err && au_dbrange_test(d1));
29786 +               if (!err) {
29787 +                       err = au_digen_test(d2, sigen);
29788 +                       AuDebugOn(!err && au_dbrange_test(d2));
29789 +               }
29790 +               if (unlikely(err))
29791 +                       aufs_read_and_write_unlock2(d1, d2);
29792 +       }
29793 +
29794 +out:
29795 +       return err;
29796 +}
29797 +
29798 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
29799 +{
29800 +       di_write_unlock2(d1, d2);
29801 +       si_read_unlock(d1->d_sb);
29802 +}
29803 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
29804 --- /usr/share/empty/fs/aufs/super.c    1970-01-01 01:00:00.000000000 +0100
29805 +++ linux/fs/aufs/super.c       2022-11-05 23:02:18.969222617 +0100
29806 @@ -0,0 +1,871 @@
29807 +// SPDX-License-Identifier: GPL-2.0
29808 +/*
29809 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29810 + *
29811 + * This program is free software; you can redistribute it and/or modify
29812 + * it under the terms of the GNU General Public License as published by
29813 + * the Free Software Foundation; either version 2 of the License, or
29814 + * (at your option) any later version.
29815 + *
29816 + * This program is distributed in the hope that it will be useful,
29817 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29818 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
29819 + * GNU General Public License for more details.
29820 + *
29821 + * You should have received a copy of the GNU General Public License
29822 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
29823 + */
29824 +
29825 +/*
29826 + * mount and super_block operations
29827 + */
29828 +
29829 +#include <linux/iversion.h>
29830 +#include <linux/mm.h>
29831 +#include <linux/seq_file.h>
29832 +#include <linux/statfs.h>
29833 +#include <linux/vmalloc.h>
29834 +#include "aufs.h"
29835 +
29836 +/*
29837 + * super_operations
29838 + */
29839 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
29840 +{
29841 +       struct au_icntnr *c;
29842 +
29843 +       c = au_cache_alloc_icntnr(sb);
29844 +       if (c) {
29845 +               au_icntnr_init(c);
29846 +               inode_set_iversion(&c->vfs_inode, 1); /* sigen(sb); */
29847 +               c->iinfo.ii_hinode = NULL;
29848 +               return &c->vfs_inode;
29849 +       }
29850 +       return NULL;
29851 +}
29852 +
29853 +static void aufs_destroy_inode(struct inode *inode)
29854 +{
29855 +       if (!au_is_bad_inode(inode))
29856 +               au_iinfo_fin(inode);
29857 +}
29858 +
29859 +static void aufs_free_inode(struct inode *inode)
29860 +{
29861 +       au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
29862 +}
29863 +
29864 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
29865 +{
29866 +       struct inode *inode;
29867 +       int err;
29868 +
29869 +       inode = iget_locked(sb, ino);
29870 +       if (unlikely(!inode)) {
29871 +               inode = ERR_PTR(-ENOMEM);
29872 +               goto out;
29873 +       }
29874 +       if (!(inode->i_state & I_NEW))
29875 +               goto out;
29876 +
29877 +       err = au_xigen_new(inode);
29878 +       if (!err)
29879 +               err = au_iinfo_init(inode);
29880 +       if (!err)
29881 +               inode_inc_iversion(inode);
29882 +       else {
29883 +               iget_failed(inode);
29884 +               inode = ERR_PTR(err);
29885 +       }
29886 +
29887 +out:
29888 +       /* never return NULL */
29889 +       AuDebugOn(!inode);
29890 +       AuTraceErrPtr(inode);
29891 +       return inode;
29892 +}
29893 +
29894 +/* lock free root dinfo */
29895 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
29896 +{
29897 +       int err;
29898 +       aufs_bindex_t bindex, bbot;
29899 +       struct path path;
29900 +       struct au_hdentry *hdp;
29901 +       struct au_branch *br;
29902 +       au_br_perm_str_t perm;
29903 +
29904 +       err = 0;
29905 +       bbot = au_sbbot(sb);
29906 +       bindex = 0;
29907 +       hdp = au_hdentry(au_di(sb->s_root), bindex);
29908 +       for (; !err && bindex <= bbot; bindex++, hdp++) {
29909 +               br = au_sbr(sb, bindex);
29910 +               path.mnt = au_br_mnt(br);
29911 +               path.dentry = hdp->hd_dentry;
29912 +               err = au_seq_path(seq, &path);
29913 +               if (!err) {
29914 +                       au_optstr_br_perm(&perm, br->br_perm);
29915 +                       seq_printf(seq, "=%s", perm.a);
29916 +                       if (bindex != bbot)
29917 +                               seq_putc(seq, ':');
29918 +               }
29919 +       }
29920 +       if (unlikely(err || seq_has_overflowed(seq)))
29921 +               err = -E2BIG;
29922 +
29923 +       return err;
29924 +}
29925 +
29926 +static void au_gen_fmt(char *fmt, int len __maybe_unused, const char *pat,
29927 +                      const char *append)
29928 +{
29929 +       char *p;
29930 +
29931 +       p = fmt;
29932 +       while (*pat != ':')
29933 +               *p++ = *pat++;
29934 +       *p++ = *pat++;
29935 +       strcpy(p, append);
29936 +       AuDebugOn(strlen(fmt) >= len);
29937 +}
29938 +
29939 +static void au_show_wbr_create(struct seq_file *m, int v,
29940 +                              struct au_sbinfo *sbinfo)
29941 +{
29942 +       const char *pat;
29943 +       char fmt[32];
29944 +       struct au_wbr_mfs *mfs;
29945 +
29946 +       AuRwMustAnyLock(&sbinfo->si_rwsem);
29947 +
29948 +       seq_puts(m, ",create=");
29949 +       pat = au_optstr_wbr_create(v);
29950 +       mfs = &sbinfo->si_wbr_mfs;
29951 +       switch (v) {
29952 +       case AuWbrCreate_TDP:
29953 +       case AuWbrCreate_RR:
29954 +       case AuWbrCreate_MFS:
29955 +       case AuWbrCreate_PMFS:
29956 +               seq_puts(m, pat);
29957 +               break;
29958 +       case AuWbrCreate_MFSRR:
29959 +       case AuWbrCreate_TDMFS:
29960 +       case AuWbrCreate_PMFSRR:
29961 +               au_gen_fmt(fmt, sizeof(fmt), pat, "%llu");
29962 +               seq_printf(m, fmt, mfs->mfsrr_watermark);
29963 +               break;
29964 +       case AuWbrCreate_MFSV:
29965 +       case AuWbrCreate_PMFSV:
29966 +               au_gen_fmt(fmt, sizeof(fmt), pat, "%lu");
29967 +               seq_printf(m, fmt,
29968 +                          jiffies_to_msecs(mfs->mfs_expire)
29969 +                          / MSEC_PER_SEC);
29970 +               break;
29971 +       case AuWbrCreate_MFSRRV:
29972 +       case AuWbrCreate_TDMFSV:
29973 +       case AuWbrCreate_PMFSRRV:
29974 +               au_gen_fmt(fmt, sizeof(fmt), pat, "%llu:%lu");
29975 +               seq_printf(m, fmt, mfs->mfsrr_watermark,
29976 +                          jiffies_to_msecs(mfs->mfs_expire) / MSEC_PER_SEC);
29977 +               break;
29978 +       default:
29979 +               BUG();
29980 +       }
29981 +}
29982 +
29983 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
29984 +{
29985 +#ifdef CONFIG_SYSFS
29986 +       return 0;
29987 +#else
29988 +       int err;
29989 +       const int len = sizeof(AUFS_XINO_FNAME) - 1;
29990 +       aufs_bindex_t bindex, brid;
29991 +       struct qstr *name;
29992 +       struct file *f;
29993 +       struct dentry *d, *h_root;
29994 +       struct au_branch *br;
29995 +
29996 +       AuRwMustAnyLock(&sbinfo->si_rwsem);
29997 +
29998 +       err = 0;
29999 +       f = au_sbi(sb)->si_xib;
30000 +       if (!f)
30001 +               goto out;
30002 +
30003 +       /* stop printing the default xino path on the first writable branch */
30004 +       h_root = NULL;
30005 +       bindex = au_xi_root(sb, f->f_path.dentry);
30006 +       if (bindex >= 0) {
30007 +               br = au_sbr_sb(sb, bindex);
30008 +               h_root = au_br_dentry(br);
30009 +       }
30010 +
30011 +       d = f->f_path.dentry;
30012 +       name = &d->d_name;
30013 +       /* safe ->d_parent because the file is unlinked */
30014 +       if (d->d_parent == h_root
30015 +           && name->len == len
30016 +           && !memcmp(name->name, AUFS_XINO_FNAME, len))
30017 +               goto out;
30018 +
30019 +       seq_puts(seq, ",xino=");
30020 +       err = au_xino_path(seq, f);
30021 +
30022 +out:
30023 +       return err;
30024 +#endif
30025 +}
30026 +
30027 +/* seq_file will re-call me in case of too long string */
30028 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
30029 +{
30030 +       int err;
30031 +       unsigned int mnt_flags, v;
30032 +       struct super_block *sb;
30033 +       struct au_sbinfo *sbinfo;
30034 +
30035 +#define AuBool(name, str) do { \
30036 +       v = au_opt_test(mnt_flags, name); \
30037 +       if (v != au_opt_test(AuOpt_Def, name)) \
30038 +               seq_printf(m, ",%s" #str, v ? "" : "no"); \
30039 +} while (0)
30040 +
30041 +#define AuStr(name, str) do { \
30042 +       v = mnt_flags & AuOptMask_##name; \
30043 +       if (v != (AuOpt_Def & AuOptMask_##name)) \
30044 +               seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
30045 +} while (0)
30046 +
30047 +#define AuUInt(name, str, val) do { \
30048 +       if (val != AUFS_##name##_DEF) \
30049 +               seq_printf(m, "," #str "=%u", val); \
30050 +} while (0)
30051 +
30052 +       sb = dentry->d_sb;
30053 +       if (sb->s_flags & SB_POSIXACL)
30054 +               seq_puts(m, ",acl");
30055 +#if 0 /* reserved for future use */
30056 +       if (sb->s_flags & SB_I_VERSION)
30057 +               seq_puts(m, ",i_version");
30058 +#endif
30059 +
30060 +       /* lock free root dinfo */
30061 +       si_noflush_read_lock(sb);
30062 +       sbinfo = au_sbi(sb);
30063 +       seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
30064 +
30065 +       mnt_flags = au_mntflags(sb);
30066 +       if (au_opt_test(mnt_flags, XINO)) {
30067 +               err = au_show_xino(m, sb);
30068 +               if (unlikely(err))
30069 +                       goto out;
30070 +       } else
30071 +               seq_puts(m, ",noxino");
30072 +
30073 +       AuBool(TRUNC_XINO, trunc_xino);
30074 +       AuStr(UDBA, udba);
30075 +       AuBool(SHWH, shwh);
30076 +       AuBool(PLINK, plink);
30077 +       AuBool(DIO, dio);
30078 +       AuBool(DIRPERM1, dirperm1);
30079 +
30080 +       v = sbinfo->si_wbr_create;
30081 +       if (v != AuWbrCreate_Def)
30082 +               au_show_wbr_create(m, v, sbinfo);
30083 +
30084 +       v = sbinfo->si_wbr_copyup;
30085 +       if (v != AuWbrCopyup_Def)
30086 +               seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
30087 +
30088 +       v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
30089 +       if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
30090 +               seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
30091 +
30092 +       AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
30093 +
30094 +       v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
30095 +       AuUInt(RDCACHE, rdcache, v);
30096 +
30097 +       AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
30098 +       AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
30099 +
30100 +       au_fhsm_show(m, sbinfo);
30101 +
30102 +       AuBool(DIRREN, dirren);
30103 +       AuBool(SUM, sum);
30104 +       /* AuBool(SUM_W, wsum); */
30105 +       AuBool(WARN_PERM, warn_perm);
30106 +       AuBool(VERBOSE, verbose);
30107 +
30108 +out:
30109 +       /* be sure to print "br:" last */
30110 +       if (!sysaufs_brs) {
30111 +               seq_puts(m, ",br:");
30112 +               au_show_brs(m, sb);
30113 +       }
30114 +       si_read_unlock(sb);
30115 +       return 0;
30116 +
30117 +#undef AuBool
30118 +#undef AuStr
30119 +#undef AuUInt
30120 +}
30121 +
30122 +/* ---------------------------------------------------------------------- */
30123 +
30124 +/* sum mode which returns the summation for statfs(2) */
30125 +
30126 +static u64 au_add_till_max(u64 a, u64 b)
30127 +{
30128 +       u64 old;
30129 +
30130 +       old = a;
30131 +       a += b;
30132 +       if (old <= a)
30133 +               return a;
30134 +       return ULLONG_MAX;
30135 +}
30136 +
30137 +static u64 au_mul_till_max(u64 a, long mul)
30138 +{
30139 +       u64 old;
30140 +
30141 +       old = a;
30142 +       a *= mul;
30143 +       if (old <= a)
30144 +               return a;
30145 +       return ULLONG_MAX;
30146 +}
30147 +
30148 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
30149 +{
30150 +       int err;
30151 +       long bsize, factor;
30152 +       u64 blocks, bfree, bavail, files, ffree;
30153 +       aufs_bindex_t bbot, bindex, i;
30154 +       unsigned char shared;
30155 +       struct path h_path;
30156 +       struct super_block *h_sb;
30157 +
30158 +       err = 0;
30159 +       bsize = LONG_MAX;
30160 +       files = 0;
30161 +       ffree = 0;
30162 +       blocks = 0;
30163 +       bfree = 0;
30164 +       bavail = 0;
30165 +       bbot = au_sbbot(sb);
30166 +       for (bindex = 0; bindex <= bbot; bindex++) {
30167 +               h_path.mnt = au_sbr_mnt(sb, bindex);
30168 +               h_sb = h_path.mnt->mnt_sb;
30169 +               shared = 0;
30170 +               for (i = 0; !shared && i < bindex; i++)
30171 +                       shared = (au_sbr_sb(sb, i) == h_sb);
30172 +               if (shared)
30173 +                       continue;
30174 +
30175 +               /* sb->s_root for NFS is unreliable */
30176 +               h_path.dentry = h_path.mnt->mnt_root;
30177 +               err = vfs_statfs(&h_path, buf);
30178 +               if (unlikely(err))
30179 +                       goto out;
30180 +
30181 +               if (bsize > buf->f_bsize) {
30182 +                       /*
30183 +                        * we will reduce bsize, so we have to expand blocks
30184 +                        * etc. to match them again
30185 +                        */
30186 +                       factor = (bsize / buf->f_bsize);
30187 +                       blocks = au_mul_till_max(blocks, factor);
30188 +                       bfree = au_mul_till_max(bfree, factor);
30189 +                       bavail = au_mul_till_max(bavail, factor);
30190 +                       bsize = buf->f_bsize;
30191 +               }
30192 +
30193 +               factor = (buf->f_bsize / bsize);
30194 +               blocks = au_add_till_max(blocks,
30195 +                               au_mul_till_max(buf->f_blocks, factor));
30196 +               bfree = au_add_till_max(bfree,
30197 +                               au_mul_till_max(buf->f_bfree, factor));
30198 +               bavail = au_add_till_max(bavail,
30199 +                               au_mul_till_max(buf->f_bavail, factor));
30200 +               files = au_add_till_max(files, buf->f_files);
30201 +               ffree = au_add_till_max(ffree, buf->f_ffree);
30202 +       }
30203 +
30204 +       buf->f_bsize = bsize;
30205 +       buf->f_blocks = blocks;
30206 +       buf->f_bfree = bfree;
30207 +       buf->f_bavail = bavail;
30208 +       buf->f_files = files;
30209 +       buf->f_ffree = ffree;
30210 +       buf->f_frsize = 0;
30211 +
30212 +out:
30213 +       return err;
30214 +}
30215 +
30216 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
30217 +{
30218 +       int err;
30219 +       struct path h_path;
30220 +       struct super_block *sb;
30221 +
30222 +       /* lock free root dinfo */
30223 +       sb = dentry->d_sb;
30224 +       si_noflush_read_lock(sb);
30225 +       if (!au_opt_test(au_mntflags(sb), SUM)) {
30226 +               /* sb->s_root for NFS is unreliable */
30227 +               h_path.mnt = au_sbr_mnt(sb, 0);
30228 +               h_path.dentry = h_path.mnt->mnt_root;
30229 +               err = vfs_statfs(&h_path, buf);
30230 +       } else
30231 +               err = au_statfs_sum(sb, buf);
30232 +       si_read_unlock(sb);
30233 +
30234 +       if (!err) {
30235 +               buf->f_type = AUFS_SUPER_MAGIC;
30236 +               buf->f_namelen = AUFS_MAX_NAMELEN;
30237 +               memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
30238 +       }
30239 +       /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
30240 +
30241 +       return err;
30242 +}
30243 +
30244 +/* ---------------------------------------------------------------------- */
30245 +
30246 +static int aufs_sync_fs(struct super_block *sb, int wait)
30247 +{
30248 +       int err, e;
30249 +       aufs_bindex_t bbot, bindex;
30250 +       struct au_branch *br;
30251 +       struct super_block *h_sb;
30252 +
30253 +       err = 0;
30254 +       si_noflush_read_lock(sb);
30255 +       bbot = au_sbbot(sb);
30256 +       for (bindex = 0; bindex <= bbot; bindex++) {
30257 +               br = au_sbr(sb, bindex);
30258 +               if (!au_br_writable(br->br_perm))
30259 +                       continue;
30260 +
30261 +               h_sb = au_sbr_sb(sb, bindex);
30262 +               e = vfsub_sync_filesystem(h_sb);
30263 +               if (unlikely(e && !err))
30264 +                       err = e;
30265 +               /* go on even if an error happens */
30266 +       }
30267 +       si_read_unlock(sb);
30268 +
30269 +       return err;
30270 +}
30271 +
30272 +/* ---------------------------------------------------------------------- */
30273 +
30274 +/* final actions when unmounting a file system */
30275 +static void aufs_put_super(struct super_block *sb)
30276 +{
30277 +       struct au_sbinfo *sbinfo;
30278 +
30279 +       sbinfo = au_sbi(sb);
30280 +       if (sbinfo)
30281 +               kobject_put(&sbinfo->si_kobj);
30282 +}
30283 +
30284 +/* ---------------------------------------------------------------------- */
30285 +
30286 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30287 +                    struct super_block *sb, void *arg)
30288 +{
30289 +       void *array;
30290 +       unsigned long long n, sz;
30291 +
30292 +       array = NULL;
30293 +       n = 0;
30294 +       if (!*hint)
30295 +               goto out;
30296 +
30297 +       if (*hint > ULLONG_MAX / sizeof(array)) {
30298 +               array = ERR_PTR(-EMFILE);
30299 +               pr_err("hint %llu\n", *hint);
30300 +               goto out;
30301 +       }
30302 +
30303 +       sz = sizeof(array) * *hint;
30304 +       array = kzalloc(sz, GFP_NOFS);
30305 +       if (unlikely(!array))
30306 +               array = vzalloc(sz);
30307 +       if (unlikely(!array)) {
30308 +               array = ERR_PTR(-ENOMEM);
30309 +               goto out;
30310 +       }
30311 +
30312 +       n = cb(sb, array, *hint, arg);
30313 +       AuDebugOn(n > *hint);
30314 +
30315 +out:
30316 +       *hint = n;
30317 +       return array;
30318 +}
30319 +
30320 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
30321 +                                      unsigned long long max __maybe_unused,
30322 +                                      void *arg)
30323 +{
30324 +       unsigned long long n;
30325 +       struct inode **p, *inode;
30326 +       struct list_head *head;
30327 +
30328 +       n = 0;
30329 +       p = a;
30330 +       head = arg;
30331 +       spin_lock(&sb->s_inode_list_lock);
30332 +       list_for_each_entry(inode, head, i_sb_list) {
30333 +               if (!au_is_bad_inode(inode)
30334 +                   && au_ii(inode)->ii_btop >= 0) {
30335 +                       spin_lock(&inode->i_lock);
30336 +                       if (atomic_read(&inode->i_count)) {
30337 +                               au_igrab(inode);
30338 +                               *p++ = inode;
30339 +                               n++;
30340 +                               AuDebugOn(n > max);
30341 +                       }
30342 +                       spin_unlock(&inode->i_lock);
30343 +               }
30344 +       }
30345 +       spin_unlock(&sb->s_inode_list_lock);
30346 +
30347 +       return n;
30348 +}
30349 +
30350 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
30351 +{
30352 +       struct au_sbinfo *sbi;
30353 +
30354 +       sbi = au_sbi(sb);
30355 +       *max = au_lcnt_read(&sbi->si_ninodes, /*do_rev*/1);
30356 +       return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
30357 +}
30358 +
30359 +void au_iarray_free(struct inode **a, unsigned long long max)
30360 +{
30361 +       unsigned long long ull;
30362 +
30363 +       for (ull = 0; ull < max; ull++)
30364 +               iput(a[ull]);
30365 +       kvfree(a);
30366 +}
30367 +
30368 +/* ---------------------------------------------------------------------- */
30369 +
30370 +/*
30371 + * refresh dentry and inode at remount time.
30372 + */
30373 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
30374 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
30375 +                     struct dentry *parent)
30376 +{
30377 +       int err;
30378 +
30379 +       di_write_lock_child(dentry);
30380 +       di_read_lock_parent(parent, AuLock_IR);
30381 +       err = au_refresh_dentry(dentry, parent);
30382 +       if (!err && dir_flags)
30383 +               au_hn_reset(d_inode(dentry), dir_flags);
30384 +       di_read_unlock(parent, AuLock_IR);
30385 +       di_write_unlock(dentry);
30386 +
30387 +       return err;
30388 +}
30389 +
30390 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
30391 +                          struct au_sbinfo *sbinfo,
30392 +                          const unsigned int dir_flags, unsigned int do_idop)
30393 +{
30394 +       int err;
30395 +       struct dentry *parent;
30396 +
30397 +       err = 0;
30398 +       parent = dget_parent(dentry);
30399 +       if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
30400 +               if (d_really_is_positive(dentry)) {
30401 +                       if (!d_is_dir(dentry))
30402 +                               err = au_do_refresh(dentry, /*dir_flags*/0,
30403 +                                                parent);
30404 +                       else {
30405 +                               err = au_do_refresh(dentry, dir_flags, parent);
30406 +                               if (unlikely(err))
30407 +                                       au_fset_si(sbinfo, FAILED_REFRESH_DIR);
30408 +                       }
30409 +               } else
30410 +                       err = au_do_refresh(dentry, /*dir_flags*/0, parent);
30411 +               AuDbgDentry(dentry);
30412 +       }
30413 +       dput(parent);
30414 +
30415 +       if (!err) {
30416 +               if (do_idop)
30417 +                       au_refresh_dop(dentry, /*force_reval*/0);
30418 +       } else
30419 +               au_refresh_dop(dentry, /*force_reval*/1);
30420 +
30421 +       AuTraceErr(err);
30422 +       return err;
30423 +}
30424 +
30425 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
30426 +{
30427 +       int err, i, j, ndentry, e;
30428 +       unsigned int sigen;
30429 +       struct au_dcsub_pages dpages;
30430 +       struct au_dpage *dpage;
30431 +       struct dentry **dentries, *d;
30432 +       struct au_sbinfo *sbinfo;
30433 +       struct dentry *root = sb->s_root;
30434 +       const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
30435 +
30436 +       if (do_idop)
30437 +               au_refresh_dop(root, /*force_reval*/0);
30438 +
30439 +       err = au_dpages_init(&dpages, GFP_NOFS);
30440 +       if (unlikely(err))
30441 +               goto out;
30442 +       err = au_dcsub_pages(&dpages, root, NULL, NULL);
30443 +       if (unlikely(err))
30444 +               goto out_dpages;
30445 +
30446 +       sigen = au_sigen(sb);
30447 +       sbinfo = au_sbi(sb);
30448 +       for (i = 0; i < dpages.ndpage; i++) {
30449 +               dpage = dpages.dpages + i;
30450 +               dentries = dpage->dentries;
30451 +               ndentry = dpage->ndentry;
30452 +               for (j = 0; j < ndentry; j++) {
30453 +                       d = dentries[j];
30454 +                       e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
30455 +                                           do_idop);
30456 +                       if (unlikely(e && !err))
30457 +                               err = e;
30458 +                       /* go on even err */
30459 +               }
30460 +       }
30461 +
30462 +out_dpages:
30463 +       au_dpages_free(&dpages);
30464 +out:
30465 +       return err;
30466 +}
30467 +
30468 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
30469 +{
30470 +       int err, e;
30471 +       unsigned int sigen;
30472 +       unsigned long long max, ull;
30473 +       struct inode *inode, **array;
30474 +
30475 +       array = au_iarray_alloc(sb, &max);
30476 +       err = PTR_ERR(array);
30477 +       if (IS_ERR(array))
30478 +               goto out;
30479 +
30480 +       err = 0;
30481 +       sigen = au_sigen(sb);
30482 +       for (ull = 0; ull < max; ull++) {
30483 +               inode = array[ull];
30484 +               if (unlikely(!inode))
30485 +                       break;
30486 +
30487 +               e = 0;
30488 +               ii_write_lock_child(inode);
30489 +               if (au_iigen(inode, NULL) != sigen) {
30490 +                       e = au_refresh_hinode_self(inode);
30491 +                       if (unlikely(e)) {
30492 +                               au_refresh_iop(inode, /*force_getattr*/1);
30493 +                               pr_err("error %d, i%lu\n", e, inode->i_ino);
30494 +                               if (!err)
30495 +                                       err = e;
30496 +                               /* go on even if err */
30497 +                       }
30498 +               }
30499 +               if (!e && do_idop)
30500 +                       au_refresh_iop(inode, /*force_getattr*/0);
30501 +               ii_write_unlock(inode);
30502 +       }
30503 +
30504 +       au_iarray_free(array, max);
30505 +
30506 +out:
30507 +       return err;
30508 +}
30509 +
30510 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
30511 +{
30512 +       int err, e;
30513 +       unsigned int udba;
30514 +       aufs_bindex_t bindex, bbot;
30515 +       struct dentry *root;
30516 +       struct inode *inode;
30517 +       struct au_branch *br;
30518 +       struct au_sbinfo *sbi;
30519 +
30520 +       au_sigen_inc(sb);
30521 +       sbi = au_sbi(sb);
30522 +       au_fclr_si(sbi, FAILED_REFRESH_DIR);
30523 +
30524 +       root = sb->s_root;
30525 +       DiMustNoWaiters(root);
30526 +       inode = d_inode(root);
30527 +       IiMustNoWaiters(inode);
30528 +
30529 +       udba = au_opt_udba(sb);
30530 +       bbot = au_sbbot(sb);
30531 +       for (bindex = 0; bindex <= bbot; bindex++) {
30532 +               br = au_sbr(sb, bindex);
30533 +               err = au_hnotify_reset_br(udba, br, br->br_perm);
30534 +               if (unlikely(err))
30535 +                       AuIOErr("hnotify failed on br %d, %d, ignored\n",
30536 +                               bindex, err);
30537 +               /* go on even if err */
30538 +       }
30539 +       au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
30540 +
30541 +       if (do_idop) {
30542 +               if (au_ftest_si(sbi, NO_DREVAL)) {
30543 +                       AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
30544 +                       sb->s_d_op = &aufs_dop_noreval;
30545 +                       AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
30546 +                       sbi->si_iop_array = aufs_iop_nogetattr;
30547 +               } else {
30548 +                       AuDebugOn(sb->s_d_op == &aufs_dop);
30549 +                       sb->s_d_op = &aufs_dop;
30550 +                       AuDebugOn(sbi->si_iop_array == aufs_iop);
30551 +                       sbi->si_iop_array = aufs_iop;
30552 +               }
30553 +               pr_info("reset to %ps and %ps\n",
30554 +                       sb->s_d_op, sbi->si_iop_array);
30555 +       }
30556 +
30557 +       di_write_unlock(root);
30558 +       err = au_refresh_d(sb, do_idop);
30559 +       e = au_refresh_i(sb, do_idop);
30560 +       if (unlikely(e && !err))
30561 +               err = e;
30562 +       /* aufs_write_lock() calls ..._child() */
30563 +       di_write_lock_child(root);
30564 +
30565 +       au_cpup_attr_all(inode, /*force*/1);
30566 +
30567 +       if (unlikely(err))
30568 +               AuIOErr("refresh failed, ignored, %d\n", err);
30569 +}
30570 +
30571 +const struct super_operations aufs_sop = {
30572 +       .alloc_inode    = aufs_alloc_inode,
30573 +       .destroy_inode  = aufs_destroy_inode,
30574 +       .free_inode     = aufs_free_inode,
30575 +       /* always deleting, no clearing */
30576 +       .drop_inode     = generic_delete_inode,
30577 +       .show_options   = aufs_show_options,
30578 +       .statfs         = aufs_statfs,
30579 +       .put_super      = aufs_put_super,
30580 +       .sync_fs        = aufs_sync_fs
30581 +};
30582 +
30583 +/* ---------------------------------------------------------------------- */
30584 +
30585 +int au_alloc_root(struct super_block *sb)
30586 +{
30587 +       int err;
30588 +       struct inode *inode;
30589 +       struct dentry *root;
30590 +
30591 +       err = -ENOMEM;
30592 +       inode = au_iget_locked(sb, AUFS_ROOT_INO);
30593 +       err = PTR_ERR(inode);
30594 +       if (IS_ERR(inode))
30595 +               goto out;
30596 +
30597 +       inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
30598 +       inode->i_fop = &aufs_dir_fop;
30599 +       inode->i_mode = S_IFDIR;
30600 +       set_nlink(inode, 2);
30601 +       unlock_new_inode(inode);
30602 +
30603 +       root = d_make_root(inode);
30604 +       if (unlikely(!root))
30605 +               goto out;
30606 +       err = PTR_ERR(root);
30607 +       if (IS_ERR(root))
30608 +               goto out;
30609 +
30610 +       err = au_di_init(root);
30611 +       if (!err) {
30612 +               sb->s_root = root;
30613 +               return 0; /* success */
30614 +       }
30615 +       dput(root);
30616 +
30617 +out:
30618 +       return err;
30619 +}
30620 +
30621 +/* ---------------------------------------------------------------------- */
30622 +
30623 +static void aufs_kill_sb(struct super_block *sb)
30624 +{
30625 +       struct au_sbinfo *sbinfo;
30626 +       struct dentry *root;
30627 +
30628 +       sbinfo = au_sbi(sb);
30629 +       if (!sbinfo)
30630 +               goto out;
30631 +
30632 +       au_sbilist_del(sb);
30633 +
30634 +       root = sb->s_root;
30635 +       if (root)
30636 +               aufs_write_lock(root);
30637 +       else
30638 +               __si_write_lock(sb);
30639 +
30640 +       au_fhsm_fin(sb);
30641 +       if (sbinfo->si_wbr_create_ops->fin)
30642 +               sbinfo->si_wbr_create_ops->fin(sb);
30643 +       if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
30644 +               au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
30645 +               au_remount_refresh(sb, /*do_idop*/0);
30646 +       }
30647 +       if (au_opt_test(sbinfo->si_mntflags, PLINK))
30648 +               au_plink_put(sb, /*verbose*/1);
30649 +       au_xino_clr(sb);
30650 +       if (root)
30651 +               au_dr_opt_flush(sb);
30652 +
30653 +       if (root)
30654 +               aufs_write_unlock(root);
30655 +       else
30656 +               __si_write_unlock(sb);
30657 +
30658 +       sbinfo->si_sb = NULL;
30659 +       au_nwt_flush(&sbinfo->si_nowait);
30660 +
30661 +out:
30662 +       kill_anon_super(sb);
30663 +}
30664 +
30665 +struct file_system_type aufs_fs_type = {
30666 +       .name           = AUFS_FSTYPE,
30667 +       /* a race between rename and others */
30668 +       .fs_flags       = FS_RENAME_DOES_D_MOVE
30669 +                               /* untested */
30670 +                               /*| FS_ALLOW_IDMAP*/
30671 +                               ,
30672 +       .init_fs_context = aufs_fsctx_init,
30673 +       .parameters     = aufs_fsctx_paramspec,
30674 +       .kill_sb        = aufs_kill_sb,
30675 +       /* no need to __module_get() and module_put(). */
30676 +       .owner          = THIS_MODULE,
30677 +};
30678 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
30679 --- /usr/share/empty/fs/aufs/super.h    1970-01-01 01:00:00.000000000 +0100
30680 +++ linux/fs/aufs/super.h       2022-11-05 23:02:18.969222617 +0100
30681 @@ -0,0 +1,592 @@
30682 +/* SPDX-License-Identifier: GPL-2.0 */
30683 +/*
30684 + * Copyright (C) 2005-2022 Junjiro R. Okajima
30685 + *
30686 + * This program is free software; you can redistribute it and/or modify
30687 + * it under the terms of the GNU General Public License as published by
30688 + * the Free Software Foundation; either version 2 of the License, or
30689 + * (at your option) any later version.
30690 + *
30691 + * This program is distributed in the hope that it will be useful,
30692 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30693 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
30694 + * GNU General Public License for more details.
30695 + *
30696 + * You should have received a copy of the GNU General Public License
30697 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
30698 + */
30699 +
30700 +/*
30701 + * super_block operations
30702 + */
30703 +
30704 +#ifndef __AUFS_SUPER_H__
30705 +#define __AUFS_SUPER_H__
30706 +
30707 +#ifdef __KERNEL__
30708 +
30709 +#include <linux/fs.h>
30710 +#include <linux/kobject.h>
30711 +#include "hbl.h"
30712 +#include "lcnt.h"
30713 +#include "rwsem.h"
30714 +#include "wkq.h"
30715 +
30716 +/* policies to select one among multiple writable branches */
30717 +struct au_wbr_copyup_operations {
30718 +       int (*copyup)(struct dentry *dentry);
30719 +};
30720 +
30721 +#define AuWbr_DIR      1               /* target is a dir */
30722 +#define AuWbr_PARENT   (1 << 1)        /* always require a parent */
30723 +
30724 +#define au_ftest_wbr(flags, name)      ((flags) & AuWbr_##name)
30725 +#define au_fset_wbr(flags, name)       { (flags) |= AuWbr_##name; }
30726 +#define au_fclr_wbr(flags, name)       { (flags) &= ~AuWbr_##name; }
30727 +
30728 +struct au_wbr_create_operations {
30729 +       int (*create)(struct dentry *dentry, unsigned int flags);
30730 +       int (*init)(struct super_block *sb);
30731 +       int (*fin)(struct super_block *sb);
30732 +};
30733 +
30734 +struct au_wbr_mfs {
30735 +       struct mutex    mfs_lock; /* protect this structure */
30736 +       unsigned long   mfs_jiffy;
30737 +       unsigned long   mfs_expire;
30738 +       aufs_bindex_t   mfs_bindex;
30739 +
30740 +       unsigned long long      mfsrr_bytes;
30741 +       unsigned long long      mfsrr_watermark;
30742 +};
30743 +
30744 +#define AuPlink_NHASH 100
30745 +static inline int au_plink_hash(ino_t ino)
30746 +{
30747 +       return ino % AuPlink_NHASH;
30748 +}
30749 +
30750 +/* File-based Hierarchical Storage Management */
30751 +struct au_fhsm {
30752 +#ifdef CONFIG_AUFS_FHSM
30753 +       /* allow only one process who can receive the notification */
30754 +       spinlock_t              fhsm_spin;
30755 +       pid_t                   fhsm_pid;
30756 +       wait_queue_head_t       fhsm_wqh;
30757 +       atomic_t                fhsm_readable;
30758 +
30759 +       /* these are protected by si_rwsem */
30760 +       unsigned long           fhsm_expire;
30761 +       aufs_bindex_t           fhsm_bottom;
30762 +#endif
30763 +};
30764 +
30765 +struct au_branch;
30766 +struct au_sbinfo {
30767 +       /* nowait tasks in the system-wide workqueue */
30768 +       struct au_nowait_tasks  si_nowait;
30769 +
30770 +       /*
30771 +        * tried sb->s_umount, but failed due to the dependency between i_mutex.
30772 +        * rwsem for au_sbinfo is necessary.
30773 +        */
30774 +       struct au_rwsem         si_rwsem;
30775 +
30776 +       /*
30777 +        * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
30778 +        * remount.
30779 +        */
30780 +       au_lcnt_t               si_ninodes, si_nfiles;
30781 +
30782 +       /* branch management */
30783 +       unsigned int            si_generation;
30784 +
30785 +       /* see AuSi_ flags */
30786 +       unsigned char           au_si_status;
30787 +
30788 +       aufs_bindex_t           si_bbot;
30789 +
30790 +       /* dirty trick to keep br_id plus */
30791 +       unsigned int            si_last_br_id :
30792 +                               sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
30793 +       struct au_branch        **si_branch;
30794 +
30795 +       /* policy to select a writable branch */
30796 +       unsigned char           si_wbr_copyup;
30797 +       unsigned char           si_wbr_create;
30798 +       struct au_wbr_copyup_operations *si_wbr_copyup_ops;
30799 +       struct au_wbr_create_operations *si_wbr_create_ops;
30800 +
30801 +       /* round robin */
30802 +       atomic_t                si_wbr_rr_next;
30803 +
30804 +       /* most free space */
30805 +       struct au_wbr_mfs       si_wbr_mfs;
30806 +
30807 +       /* File-based Hierarchical Storage Management */
30808 +       struct au_fhsm          si_fhsm;
30809 +
30810 +       /* mount flags */
30811 +       /* include/asm-ia64/siginfo.h defines a macro named si_flags */
30812 +       unsigned int            si_mntflags;
30813 +
30814 +       /* external inode number (bitmap and translation table) */
30815 +       loff_t                  si_ximaxent;    /* max entries in a xino */
30816 +
30817 +       struct file             *si_xib;
30818 +       struct mutex            si_xib_mtx; /* protect xib members */
30819 +       unsigned long           *si_xib_buf;
30820 +       unsigned long           si_xib_last_pindex;
30821 +       int                     si_xib_next_bit;
30822 +
30823 +       unsigned long           si_xino_jiffy;
30824 +       unsigned long           si_xino_expire;
30825 +       /* reserved for future use */
30826 +       /* unsigned long long   si_xib_limit; */        /* Max xib file size */
30827 +
30828 +#ifdef CONFIG_AUFS_EXPORT
30829 +       /* i_generation */
30830 +       /* todo: make xigen file an array to support many inode numbers */
30831 +       struct file             *si_xigen;
30832 +       atomic_t                si_xigen_next;
30833 +#endif
30834 +
30835 +       /* dirty trick to support atomic_open */
30836 +       struct hlist_bl_head    si_aopen;
30837 +
30838 +       /* vdir parameters */
30839 +       unsigned long           si_rdcache;     /* max cache time in jiffies */
30840 +       unsigned int            si_rdblk;       /* deblk size */
30841 +       unsigned int            si_rdhash;      /* hash size */
30842 +
30843 +       /*
30844 +        * If the number of whiteouts are larger than si_dirwh, leave all of
30845 +        * them after au_whtmp_ren to reduce the cost of rmdir(2).
30846 +        * future fsck.aufs or kernel thread will remove them later.
30847 +        * Otherwise, remove all whiteouts and the dir in rmdir(2).
30848 +        */
30849 +       unsigned int            si_dirwh;
30850 +
30851 +       /* pseudo_link list */
30852 +       struct hlist_bl_head    si_plink[AuPlink_NHASH];
30853 +       wait_queue_head_t       si_plink_wq;
30854 +       spinlock_t              si_plink_maint_lock;
30855 +       pid_t                   si_plink_maint_pid;
30856 +
30857 +       /* file list */
30858 +       struct hlist_bl_head    si_files;
30859 +
30860 +       /* with/without getattr, brother of sb->s_d_op */
30861 +       const struct inode_operations *si_iop_array;
30862 +
30863 +       /*
30864 +        * sysfs and lifetime management.
30865 +        * this is not a small structure and it may be a waste of memory in case
30866 +        * of sysfs is disabled, particularly when many aufs-es are mounted.
30867 +        * but using sysfs is majority.
30868 +        */
30869 +       struct kobject          si_kobj;
30870 +#ifdef CONFIG_DEBUG_FS
30871 +       struct dentry            *si_dbgaufs;
30872 +       struct dentry            *si_dbgaufs_plink;
30873 +       struct dentry            *si_dbgaufs_xib;
30874 +#ifdef CONFIG_AUFS_EXPORT
30875 +       struct dentry            *si_dbgaufs_xigen;
30876 +#endif
30877 +#endif
30878 +
30879 +#ifdef CONFIG_AUFS_SBILIST
30880 +       struct hlist_bl_node    si_list;
30881 +#endif
30882 +
30883 +       /* dirty, necessary for unmounting, sysfs and sysrq */
30884 +       struct super_block      *si_sb;
30885 +};
30886 +
30887 +/* sbinfo status flags */
30888 +/*
30889 + * set true when refresh_dirs() failed at remount time.
30890 + * then try refreshing dirs at access time again.
30891 + * if it is false, refreshing dirs at access time is unnecessary
30892 + */
30893 +#define AuSi_FAILED_REFRESH_DIR        1
30894 +#define AuSi_FHSM              (1 << 1)        /* fhsm is active now */
30895 +#define AuSi_NO_DREVAL         (1 << 2)        /* disable all d_revalidate */
30896 +
30897 +#ifndef CONFIG_AUFS_FHSM
30898 +#undef AuSi_FHSM
30899 +#define AuSi_FHSM              0
30900 +#endif
30901 +
30902 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
30903 +                                          unsigned int flag)
30904 +{
30905 +       AuRwMustAnyLock(&sbi->si_rwsem);
30906 +       return sbi->au_si_status & flag;
30907 +}
30908 +#define au_ftest_si(sbinfo, name)      au_do_ftest_si(sbinfo, AuSi_##name)
30909 +#define au_fset_si(sbinfo, name) do { \
30910 +       AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30911 +       (sbinfo)->au_si_status |= AuSi_##name; \
30912 +} while (0)
30913 +#define au_fclr_si(sbinfo, name) do { \
30914 +       AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30915 +       (sbinfo)->au_si_status &= ~AuSi_##name; \
30916 +} while (0)
30917 +
30918 +/* ---------------------------------------------------------------------- */
30919 +
30920 +/* policy to select one among writable branches */
30921 +#define AuWbrCopyup(sbinfo, ...) \
30922 +       ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
30923 +#define AuWbrCreate(sbinfo, ...) \
30924 +       ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
30925 +
30926 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
30927 +#define AuLock_DW              1               /* write-lock dentry */
30928 +#define AuLock_IR              (1 << 1)        /* read-lock inode */
30929 +#define AuLock_IW              (1 << 2)        /* write-lock inode */
30930 +#define AuLock_FLUSH           (1 << 3)        /* wait for 'nowait' tasks */
30931 +#define AuLock_DIRS            (1 << 4)        /* target is a pair of dirs */
30932 +                                               /* except RENAME_EXCHANGE */
30933 +#define AuLock_NOPLM           (1 << 5)        /* return err in plm mode */
30934 +#define AuLock_NOPLMW          (1 << 6)        /* wait for plm mode ends */
30935 +#define AuLock_GEN             (1 << 7)        /* test digen/iigen */
30936 +#define au_ftest_lock(flags, name)     ((flags) & AuLock_##name)
30937 +#define au_fset_lock(flags, name) \
30938 +       do { (flags) |= AuLock_##name; } while (0)
30939 +#define au_fclr_lock(flags, name) \
30940 +       do { (flags) &= ~AuLock_##name; } while (0)
30941 +
30942 +/* ---------------------------------------------------------------------- */
30943 +
30944 +/* super.c */
30945 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
30946 +
30947 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
30948 +                                          unsigned long long max, void *arg);
30949 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30950 +                    struct super_block *sb, void *arg);
30951 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
30952 +void au_iarray_free(struct inode **a, unsigned long long max);
30953 +
30954 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop);
30955 +extern const struct super_operations aufs_sop;
30956 +int au_alloc_root(struct super_block *sb);
30957 +extern struct file_system_type aufs_fs_type;
30958 +
30959 +/* sbinfo.c */
30960 +void au_si_free(struct kobject *kobj);
30961 +struct au_sbinfo *au_si_alloc(struct super_block *sb);
30962 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink);
30963 +
30964 +unsigned int au_sigen_inc(struct super_block *sb);
30965 +aufs_bindex_t au_new_br_id(struct super_block *sb);
30966 +
30967 +int si_read_lock(struct super_block *sb, int flags);
30968 +int si_write_lock(struct super_block *sb, int flags);
30969 +int aufs_read_lock(struct dentry *dentry, int flags);
30970 +void aufs_read_unlock(struct dentry *dentry, int flags);
30971 +void aufs_write_lock(struct dentry *dentry);
30972 +void aufs_write_unlock(struct dentry *dentry);
30973 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
30974 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
30975 +
30976 +/* wbr_policy.c */
30977 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
30978 +extern struct au_wbr_create_operations au_wbr_create_ops[];
30979 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
30980 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
30981 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop);
30982 +
30983 +/* mvdown.c */
30984 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
30985 +
30986 +#ifdef CONFIG_AUFS_FHSM
30987 +/* fhsm.c */
30988 +
30989 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
30990 +{
30991 +       pid_t pid;
30992 +
30993 +       spin_lock(&fhsm->fhsm_spin);
30994 +       pid = fhsm->fhsm_pid;
30995 +       spin_unlock(&fhsm->fhsm_spin);
30996 +
30997 +       return pid;
30998 +}
30999 +
31000 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
31001 +void au_fhsm_wrote_all(struct super_block *sb, int force);
31002 +int au_fhsm_fd(struct super_block *sb, int oflags);
31003 +int au_fhsm_br_alloc(struct au_branch *br);
31004 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
31005 +void au_fhsm_fin(struct super_block *sb);
31006 +void au_fhsm_init(struct au_sbinfo *sbinfo);
31007 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
31008 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
31009 +#else
31010 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
31011 +          int force)
31012 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
31013 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
31014 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
31015 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
31016 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
31017 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
31018 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
31019 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
31020 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
31021 +#endif
31022 +
31023 +/* ---------------------------------------------------------------------- */
31024 +
31025 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
31026 +{
31027 +       return sb->s_fs_info;
31028 +}
31029 +
31030 +/* ---------------------------------------------------------------------- */
31031 +
31032 +#ifdef CONFIG_AUFS_EXPORT
31033 +int au_test_nfsd(void);
31034 +void au_export_init(struct super_block *sb);
31035 +void au_xigen_inc(struct inode *inode);
31036 +int au_xigen_new(struct inode *inode);
31037 +int au_xigen_set(struct super_block *sb, struct path *path);
31038 +void au_xigen_clr(struct super_block *sb);
31039 +
31040 +static inline int au_busy_or_stale(void)
31041 +{
31042 +       if (!au_test_nfsd())
31043 +               return -EBUSY;
31044 +       return -ESTALE;
31045 +}
31046 +#else
31047 +AuStubInt0(au_test_nfsd, void)
31048 +AuStubVoid(au_export_init, struct super_block *sb)
31049 +AuStubVoid(au_xigen_inc, struct inode *inode)
31050 +AuStubInt0(au_xigen_new, struct inode *inode)
31051 +AuStubInt0(au_xigen_set, struct super_block *sb, struct path *path)
31052 +AuStubVoid(au_xigen_clr, struct super_block *sb)
31053 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
31054 +#endif /* CONFIG_AUFS_EXPORT */
31055 +
31056 +/* ---------------------------------------------------------------------- */
31057 +
31058 +#ifdef CONFIG_AUFS_SBILIST
31059 +/* module.c */
31060 +extern struct hlist_bl_head au_sbilist;
31061 +
31062 +static inline void au_sbilist_init(void)
31063 +{
31064 +       INIT_HLIST_BL_HEAD(&au_sbilist);
31065 +}
31066 +
31067 +static inline void au_sbilist_add(struct super_block *sb)
31068 +{
31069 +       au_hbl_add(&au_sbi(sb)->si_list, &au_sbilist);
31070 +}
31071 +
31072 +static inline void au_sbilist_del(struct super_block *sb)
31073 +{
31074 +       au_hbl_del(&au_sbi(sb)->si_list, &au_sbilist);
31075 +}
31076 +
31077 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
31078 +static inline void au_sbilist_lock(void)
31079 +{
31080 +       hlist_bl_lock(&au_sbilist);
31081 +}
31082 +
31083 +static inline void au_sbilist_unlock(void)
31084 +{
31085 +       hlist_bl_unlock(&au_sbilist);
31086 +}
31087 +#define AuGFP_SBILIST  GFP_ATOMIC
31088 +#else
31089 +AuStubVoid(au_sbilist_lock, void)
31090 +AuStubVoid(au_sbilist_unlock, void)
31091 +#define AuGFP_SBILIST  GFP_NOFS
31092 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
31093 +#else
31094 +AuStubVoid(au_sbilist_init, void)
31095 +AuStubVoid(au_sbilist_add, struct super_block *sb)
31096 +AuStubVoid(au_sbilist_del, struct super_block *sb)
31097 +AuStubVoid(au_sbilist_lock, void)
31098 +AuStubVoid(au_sbilist_unlock, void)
31099 +#define AuGFP_SBILIST  GFP_NOFS
31100 +#endif
31101 +
31102 +/* ---------------------------------------------------------------------- */
31103 +
31104 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
31105 +{
31106 +       /*
31107 +        * This function is a dynamic '__init' function actually,
31108 +        * so the tiny check for si_rwsem is unnecessary.
31109 +        */
31110 +       /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
31111 +#ifdef CONFIG_DEBUG_FS
31112 +       sbinfo->si_dbgaufs = NULL;
31113 +       sbinfo->si_dbgaufs_plink = NULL;
31114 +       sbinfo->si_dbgaufs_xib = NULL;
31115 +#ifdef CONFIG_AUFS_EXPORT
31116 +       sbinfo->si_dbgaufs_xigen = NULL;
31117 +#endif
31118 +#endif
31119 +}
31120 +
31121 +/* ---------------------------------------------------------------------- */
31122 +
31123 +/* current->atomic_flags */
31124 +/* this value should never corrupt the ones defined in linux/sched.h */
31125 +#define PFA_AUFS       0x10
31126 +
31127 +TASK_PFA_TEST(AUFS, test_aufs) /* task_test_aufs */
31128 +TASK_PFA_SET(AUFS, aufs)       /* task_set_aufs */
31129 +TASK_PFA_CLEAR(AUFS, aufs)     /* task_clear_aufs */
31130 +
31131 +static inline int si_pid_test(struct super_block *sb)
31132 +{
31133 +       return !!task_test_aufs(current);
31134 +}
31135 +
31136 +static inline void si_pid_clr(struct super_block *sb)
31137 +{
31138 +       AuDebugOn(!task_test_aufs(current));
31139 +       task_clear_aufs(current);
31140 +}
31141 +
31142 +static inline void si_pid_set(struct super_block *sb)
31143 +{
31144 +       AuDebugOn(task_test_aufs(current));
31145 +       task_set_aufs(current);
31146 +}
31147 +
31148 +/* ---------------------------------------------------------------------- */
31149 +
31150 +/* lock superblock. mainly for entry point functions */
31151 +#define __si_read_lock(sb)     au_rw_read_lock(&au_sbi(sb)->si_rwsem)
31152 +#define __si_write_lock(sb)    au_rw_write_lock(&au_sbi(sb)->si_rwsem)
31153 +#define __si_read_trylock(sb)  au_rw_read_trylock(&au_sbi(sb)->si_rwsem)
31154 +#define __si_write_trylock(sb) au_rw_write_trylock(&au_sbi(sb)->si_rwsem)
31155 +/*
31156 +#define __si_read_trylock_nested(sb) \
31157 +       au_rw_read_trylock_nested(&au_sbi(sb)->si_rwsem)
31158 +#define __si_write_trylock_nested(sb) \
31159 +       au_rw_write_trylock_nested(&au_sbi(sb)->si_rwsem)
31160 +*/
31161 +
31162 +#define __si_read_unlock(sb)   au_rw_read_unlock(&au_sbi(sb)->si_rwsem)
31163 +#define __si_write_unlock(sb)  au_rw_write_unlock(&au_sbi(sb)->si_rwsem)
31164 +#define __si_downgrade_lock(sb)        au_rw_dgrade_lock(&au_sbi(sb)->si_rwsem)
31165 +
31166 +#define SiMustNoWaiters(sb)    AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
31167 +#define SiMustAnyLock(sb)      AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
31168 +#define SiMustWriteLock(sb)    AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
31169 +
31170 +static inline void si_noflush_read_lock(struct super_block *sb)
31171 +{
31172 +       __si_read_lock(sb);
31173 +       si_pid_set(sb);
31174 +}
31175 +
31176 +static inline int si_noflush_read_trylock(struct super_block *sb)
31177 +{
31178 +       int locked;
31179 +
31180 +       locked = __si_read_trylock(sb);
31181 +       if (locked)
31182 +               si_pid_set(sb);
31183 +       return locked;
31184 +}
31185 +
31186 +static inline void si_noflush_write_lock(struct super_block *sb)
31187 +{
31188 +       __si_write_lock(sb);
31189 +       si_pid_set(sb);
31190 +}
31191 +
31192 +static inline int si_noflush_write_trylock(struct super_block *sb)
31193 +{
31194 +       int locked;
31195 +
31196 +       locked = __si_write_trylock(sb);
31197 +       if (locked)
31198 +               si_pid_set(sb);
31199 +       return locked;
31200 +}
31201 +
31202 +#if 0 /* reserved */
31203 +static inline int si_read_trylock(struct super_block *sb, int flags)
31204 +{
31205 +       if (au_ftest_lock(flags, FLUSH))
31206 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
31207 +       return si_noflush_read_trylock(sb);
31208 +}
31209 +#endif
31210 +
31211 +static inline void si_read_unlock(struct super_block *sb)
31212 +{
31213 +       si_pid_clr(sb);
31214 +       __si_read_unlock(sb);
31215 +}
31216 +
31217 +#if 0 /* reserved */
31218 +static inline int si_write_trylock(struct super_block *sb, int flags)
31219 +{
31220 +       if (au_ftest_lock(flags, FLUSH))
31221 +               au_nwt_flush(&au_sbi(sb)->si_nowait);
31222 +       return si_noflush_write_trylock(sb);
31223 +}
31224 +#endif
31225 +
31226 +static inline void si_write_unlock(struct super_block *sb)
31227 +{
31228 +       si_pid_clr(sb);
31229 +       __si_write_unlock(sb);
31230 +}
31231 +
31232 +#if 0 /* reserved */
31233 +static inline void si_downgrade_lock(struct super_block *sb)
31234 +{
31235 +       __si_downgrade_lock(sb);
31236 +}
31237 +#endif
31238 +
31239 +/* ---------------------------------------------------------------------- */
31240 +
31241 +static inline aufs_bindex_t au_sbbot(struct super_block *sb)
31242 +{
31243 +       SiMustAnyLock(sb);
31244 +       return au_sbi(sb)->si_bbot;
31245 +}
31246 +
31247 +static inline unsigned int au_mntflags(struct super_block *sb)
31248 +{
31249 +       SiMustAnyLock(sb);
31250 +       return au_sbi(sb)->si_mntflags;
31251 +}
31252 +
31253 +static inline unsigned int au_sigen(struct super_block *sb)
31254 +{
31255 +       SiMustAnyLock(sb);
31256 +       return au_sbi(sb)->si_generation;
31257 +}
31258 +
31259 +static inline struct au_branch *au_sbr(struct super_block *sb,
31260 +                                      aufs_bindex_t bindex)
31261 +{
31262 +       SiMustAnyLock(sb);
31263 +       return au_sbi(sb)->si_branch[0 + bindex];
31264 +}
31265 +
31266 +static inline loff_t au_xi_maxent(struct super_block *sb)
31267 +{
31268 +       SiMustAnyLock(sb);
31269 +       return au_sbi(sb)->si_ximaxent;
31270 +}
31271 +
31272 +#endif /* __KERNEL__ */
31273 +#endif /* __AUFS_SUPER_H__ */
31274 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
31275 --- /usr/share/empty/fs/aufs/sysaufs.c  1970-01-01 01:00:00.000000000 +0100
31276 +++ linux/fs/aufs/sysaufs.c     2022-11-05 23:02:18.969222617 +0100
31277 @@ -0,0 +1,94 @@
31278 +// SPDX-License-Identifier: GPL-2.0
31279 +/*
31280 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31281 + *
31282 + * This program is free software; you can redistribute it and/or modify
31283 + * it under the terms of the GNU General Public License as published by
31284 + * the Free Software Foundation; either version 2 of the License, or
31285 + * (at your option) any later version.
31286 + *
31287 + * This program is distributed in the hope that it will be useful,
31288 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31289 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31290 + * GNU General Public License for more details.
31291 + *
31292 + * You should have received a copy of the GNU General Public License
31293 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31294 + */
31295 +
31296 +/*
31297 + * sysfs interface and lifetime management
31298 + * they are necessary regardless sysfs is disabled.
31299 + */
31300 +
31301 +#include <linux/random.h>
31302 +#include "aufs.h"
31303 +
31304 +unsigned long sysaufs_si_mask;
31305 +struct kset *sysaufs_kset;
31306 +
31307 +#define AuSiAttr(_name) { \
31308 +       .attr   = { .name = __stringify(_name), .mode = 0444 }, \
31309 +       .show   = sysaufs_si_##_name,                           \
31310 +}
31311 +
31312 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
31313 +struct attribute *sysaufs_si_attrs[] = {
31314 +       &sysaufs_si_attr_xi_path.attr,
31315 +       NULL,
31316 +};
31317 +ATTRIBUTE_GROUPS(sysaufs_si);
31318 +
31319 +static const struct sysfs_ops au_sbi_ops = {
31320 +       .show   = sysaufs_si_show
31321 +};
31322 +
31323 +static struct kobj_type au_sbi_ktype = {
31324 +       .release        = au_si_free,
31325 +       .sysfs_ops      = &au_sbi_ops,
31326 +       .default_groups = sysaufs_si_groups
31327 +};
31328 +
31329 +/* ---------------------------------------------------------------------- */
31330 +
31331 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
31332 +{
31333 +       int err;
31334 +
31335 +       sbinfo->si_kobj.kset = sysaufs_kset;
31336 +       /* cf. sysaufs_name() */
31337 +       err = kobject_init_and_add
31338 +               (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
31339 +                SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
31340 +
31341 +       return err;
31342 +}
31343 +
31344 +void sysaufs_fin(void)
31345 +{
31346 +       sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31347 +       kset_unregister(sysaufs_kset);
31348 +}
31349 +
31350 +int __init sysaufs_init(void)
31351 +{
31352 +       int err;
31353 +
31354 +       do {
31355 +               get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
31356 +       } while (!sysaufs_si_mask);
31357 +
31358 +       err = -EINVAL;
31359 +       sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
31360 +       if (unlikely(!sysaufs_kset))
31361 +               goto out;
31362 +       err = PTR_ERR(sysaufs_kset);
31363 +       if (IS_ERR(sysaufs_kset))
31364 +               goto out;
31365 +       err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31366 +       if (unlikely(err))
31367 +               kset_unregister(sysaufs_kset);
31368 +
31369 +out:
31370 +       return err;
31371 +}
31372 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
31373 --- /usr/share/empty/fs/aufs/sysaufs.h  1970-01-01 01:00:00.000000000 +0100
31374 +++ linux/fs/aufs/sysaufs.h     2022-11-05 23:02:18.969222617 +0100
31375 @@ -0,0 +1,102 @@
31376 +/* SPDX-License-Identifier: GPL-2.0 */
31377 +/*
31378 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31379 + *
31380 + * This program is free software; you can redistribute it and/or modify
31381 + * it under the terms of the GNU General Public License as published by
31382 + * the Free Software Foundation; either version 2 of the License, or
31383 + * (at your option) any later version.
31384 + *
31385 + * This program is distributed in the hope that it will be useful,
31386 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31387 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31388 + * GNU General Public License for more details.
31389 + *
31390 + * You should have received a copy of the GNU General Public License
31391 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31392 + */
31393 +
31394 +/*
31395 + * sysfs interface and mount lifetime management
31396 + */
31397 +
31398 +#ifndef __SYSAUFS_H__
31399 +#define __SYSAUFS_H__
31400 +
31401 +#ifdef __KERNEL__
31402 +
31403 +#include <linux/sysfs.h>
31404 +#include "module.h"
31405 +
31406 +struct super_block;
31407 +struct au_sbinfo;
31408 +
31409 +struct sysaufs_si_attr {
31410 +       struct attribute attr;
31411 +       int (*show)(struct seq_file *seq, struct super_block *sb);
31412 +};
31413 +
31414 +/* ---------------------------------------------------------------------- */
31415 +
31416 +/* sysaufs.c */
31417 +extern unsigned long sysaufs_si_mask;
31418 +extern struct kset *sysaufs_kset;
31419 +extern struct attribute *sysaufs_si_attrs[];
31420 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
31421 +int __init sysaufs_init(void);
31422 +void sysaufs_fin(void);
31423 +
31424 +/* ---------------------------------------------------------------------- */
31425 +
31426 +/* some people doesn't like to show a pointer in kernel */
31427 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
31428 +{
31429 +       return sysaufs_si_mask ^ (unsigned long)sbinfo;
31430 +}
31431 +
31432 +#define SysaufsSiNamePrefix    "si_"
31433 +#define SysaufsSiNameLen       (sizeof(SysaufsSiNamePrefix) + 16)
31434 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
31435 +{
31436 +       snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
31437 +                sysaufs_si_id(sbinfo));
31438 +}
31439 +
31440 +struct au_branch;
31441 +#ifdef CONFIG_SYSFS
31442 +/* sysfs.c */
31443 +extern struct attribute_group *sysaufs_attr_group;
31444 +
31445 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
31446 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31447 +                        char *buf);
31448 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
31449 +#ifdef CONFIG_COMPAT
31450 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
31451 +#endif
31452 +
31453 +void sysaufs_br_init(struct au_branch *br);
31454 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
31455 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
31456 +
31457 +#define sysaufs_brs_init()     do {} while (0)
31458 +
31459 +#else
31460 +#define sysaufs_attr_group     NULL
31461 +
31462 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
31463 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
31464 +       struct attribute *attr, char *buf)
31465 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
31466 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
31467 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
31468 +
31469 +static inline void sysaufs_brs_init(void)
31470 +{
31471 +       sysaufs_brs = 0;
31472 +}
31473 +
31474 +#endif /* CONFIG_SYSFS */
31475 +
31476 +#endif /* __KERNEL__ */
31477 +#endif /* __SYSAUFS_H__ */
31478 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
31479 --- /usr/share/empty/fs/aufs/sysfs.c    1970-01-01 01:00:00.000000000 +0100
31480 +++ linux/fs/aufs/sysfs.c       2022-11-05 23:02:18.969222617 +0100
31481 @@ -0,0 +1,374 @@
31482 +// SPDX-License-Identifier: GPL-2.0
31483 +/*
31484 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31485 + *
31486 + * This program is free software; you can redistribute it and/or modify
31487 + * it under the terms of the GNU General Public License as published by
31488 + * the Free Software Foundation; either version 2 of the License, or
31489 + * (at your option) any later version.
31490 + *
31491 + * This program is distributed in the hope that it will be useful,
31492 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31493 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31494 + * GNU General Public License for more details.
31495 + *
31496 + * You should have received a copy of the GNU General Public License
31497 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31498 + */
31499 +
31500 +/*
31501 + * sysfs interface
31502 + */
31503 +
31504 +#include <linux/compat.h>
31505 +#include <linux/seq_file.h>
31506 +#include "aufs.h"
31507 +
31508 +#ifdef CONFIG_AUFS_FS_MODULE
31509 +/* this entry violates the "one line per file" policy of sysfs */
31510 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
31511 +                          char *buf)
31512 +{
31513 +       ssize_t err;
31514 +       static char *conf =
31515 +/* this file is generated at compiling */
31516 +#include "conf.str"
31517 +               ;
31518 +
31519 +       err = snprintf(buf, PAGE_SIZE, conf);
31520 +       if (unlikely(err >= PAGE_SIZE))
31521 +               err = -EFBIG;
31522 +       return err;
31523 +}
31524 +
31525 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
31526 +#endif
31527 +
31528 +static struct attribute *au_attr[] = {
31529 +#ifdef CONFIG_AUFS_FS_MODULE
31530 +       &au_config_attr.attr,
31531 +#endif
31532 +       NULL,   /* need to NULL terminate the list of attributes */
31533 +};
31534 +
31535 +static struct attribute_group sysaufs_attr_group_body = {
31536 +       .attrs = au_attr
31537 +};
31538 +
31539 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
31540 +
31541 +/* ---------------------------------------------------------------------- */
31542 +
31543 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
31544 +{
31545 +       int err;
31546 +
31547 +       SiMustAnyLock(sb);
31548 +
31549 +       err = 0;
31550 +       if (au_opt_test(au_mntflags(sb), XINO)) {
31551 +               err = au_xino_path(seq, au_sbi(sb)->si_xib);
31552 +               seq_putc(seq, '\n');
31553 +       }
31554 +       return err;
31555 +}
31556 +
31557 +/*
31558 + * the lifetime of branch is independent from the entry under sysfs.
31559 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
31560 + * unlinked.
31561 + */
31562 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
31563 +                        aufs_bindex_t bindex, int idx)
31564 +{
31565 +       int err;
31566 +       struct path path;
31567 +       struct dentry *root;
31568 +       struct au_branch *br;
31569 +       au_br_perm_str_t perm;
31570 +
31571 +       AuDbg("b%d\n", bindex);
31572 +
31573 +       err = 0;
31574 +       root = sb->s_root;
31575 +       di_read_lock_parent(root, !AuLock_IR);
31576 +       br = au_sbr(sb, bindex);
31577 +
31578 +       switch (idx) {
31579 +       case AuBrSysfs_BR:
31580 +               path.mnt = au_br_mnt(br);
31581 +               path.dentry = au_h_dptr(root, bindex);
31582 +               err = au_seq_path(seq, &path);
31583 +               if (!err) {
31584 +                       au_optstr_br_perm(&perm, br->br_perm);
31585 +                       seq_printf(seq, "=%s\n", perm.a);
31586 +               }
31587 +               break;
31588 +       case AuBrSysfs_BRID:
31589 +               seq_printf(seq, "%d\n", br->br_id);
31590 +               break;
31591 +       }
31592 +       di_read_unlock(root, !AuLock_IR);
31593 +       if (unlikely(err || seq_has_overflowed(seq)))
31594 +               err = -E2BIG;
31595 +
31596 +       return err;
31597 +}
31598 +
31599 +/* ---------------------------------------------------------------------- */
31600 +
31601 +static struct seq_file *au_seq(char *p, ssize_t len)
31602 +{
31603 +       struct seq_file *seq;
31604 +
31605 +       seq = kzalloc(sizeof(*seq), GFP_NOFS);
31606 +       if (seq) {
31607 +               /* mutex_init(&seq.lock); */
31608 +               seq->buf = p;
31609 +               seq->size = len;
31610 +               return seq; /* success */
31611 +       }
31612 +
31613 +       seq = ERR_PTR(-ENOMEM);
31614 +       return seq;
31615 +}
31616 +
31617 +#define SysaufsBr_PREFIX       "br"
31618 +#define SysaufsBrid_PREFIX     "brid"
31619 +
31620 +/* todo: file size may exceed PAGE_SIZE */
31621 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31622 +                       char *buf)
31623 +{
31624 +       ssize_t err;
31625 +       int idx;
31626 +       long l;
31627 +       aufs_bindex_t bbot;
31628 +       struct au_sbinfo *sbinfo;
31629 +       struct super_block *sb;
31630 +       struct seq_file *seq;
31631 +       char *name;
31632 +       struct attribute **cattr;
31633 +
31634 +       sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
31635 +       sb = sbinfo->si_sb;
31636 +
31637 +       /*
31638 +        * prevent a race condition between sysfs and aufs.
31639 +        * for instance, sysfs_file_read() calls sysfs_get_active_two() which
31640 +        * prohibits maintaining the sysfs entries.
31641 +        * hew we acquire read lock after sysfs_get_active_two().
31642 +        * on the other hand, the remount process may maintain the sysfs/aufs
31643 +        * entries after acquiring write lock.
31644 +        * it can cause a deadlock.
31645 +        * simply we gave up processing read here.
31646 +        */
31647 +       err = -EBUSY;
31648 +       if (unlikely(!si_noflush_read_trylock(sb)))
31649 +               goto out;
31650 +
31651 +       seq = au_seq(buf, PAGE_SIZE);
31652 +       err = PTR_ERR(seq);
31653 +       if (IS_ERR(seq))
31654 +               goto out_unlock;
31655 +
31656 +       name = (void *)attr->name;
31657 +       cattr = sysaufs_si_attrs;
31658 +       while (*cattr) {
31659 +               if (!strcmp(name, (*cattr)->name)) {
31660 +                       err = container_of(*cattr, struct sysaufs_si_attr, attr)
31661 +                               ->show(seq, sb);
31662 +                       goto out_seq;
31663 +               }
31664 +               cattr++;
31665 +       }
31666 +
31667 +       if (!strncmp(name, SysaufsBrid_PREFIX,
31668 +                    sizeof(SysaufsBrid_PREFIX) - 1)) {
31669 +               idx = AuBrSysfs_BRID;
31670 +               name += sizeof(SysaufsBrid_PREFIX) - 1;
31671 +       } else if (!strncmp(name, SysaufsBr_PREFIX,
31672 +                           sizeof(SysaufsBr_PREFIX) - 1)) {
31673 +               idx = AuBrSysfs_BR;
31674 +               name += sizeof(SysaufsBr_PREFIX) - 1;
31675 +       } else
31676 +                 BUG();
31677 +
31678 +       err = kstrtol(name, 10, &l);
31679 +       if (!err) {
31680 +               bbot = au_sbbot(sb);
31681 +               if (l <= bbot)
31682 +                       err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
31683 +               else
31684 +                       err = -ENOENT;
31685 +       }
31686 +
31687 +out_seq:
31688 +       if (!err) {
31689 +               err = seq->count;
31690 +               /* sysfs limit */
31691 +               if (unlikely(err == PAGE_SIZE))
31692 +                       err = -EFBIG;
31693 +       }
31694 +       au_kfree_rcu(seq);
31695 +out_unlock:
31696 +       si_read_unlock(sb);
31697 +out:
31698 +       return err;
31699 +}
31700 +
31701 +/* ---------------------------------------------------------------------- */
31702 +
31703 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
31704 +{
31705 +       int err;
31706 +       int16_t brid;
31707 +       aufs_bindex_t bindex, bbot;
31708 +       size_t sz;
31709 +       char *buf;
31710 +       struct seq_file *seq;
31711 +       struct au_branch *br;
31712 +
31713 +       si_read_lock(sb, AuLock_FLUSH);
31714 +       bbot = au_sbbot(sb);
31715 +       err = bbot + 1;
31716 +       if (!arg)
31717 +               goto out;
31718 +
31719 +       err = -ENOMEM;
31720 +       buf = (void *)__get_free_page(GFP_NOFS);
31721 +       if (unlikely(!buf))
31722 +               goto out;
31723 +
31724 +       seq = au_seq(buf, PAGE_SIZE);
31725 +       err = PTR_ERR(seq);
31726 +       if (IS_ERR(seq))
31727 +               goto out_buf;
31728 +
31729 +       sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
31730 +       for (bindex = 0; bindex <= bbot; bindex++, arg++) {
31731 +               /* VERIFY_WRITE */
31732 +               err = !access_ok(arg, sizeof(*arg));
31733 +               if (unlikely(err))
31734 +                       break;
31735 +
31736 +               br = au_sbr(sb, bindex);
31737 +               brid = br->br_id;
31738 +               BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
31739 +               err = __put_user(brid, &arg->id);
31740 +               if (unlikely(err))
31741 +                       break;
31742 +
31743 +               BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
31744 +               err = __put_user(br->br_perm, &arg->perm);
31745 +               if (unlikely(err))
31746 +                       break;
31747 +
31748 +               err = au_seq_path(seq, &br->br_path);
31749 +               if (unlikely(err))
31750 +                       break;
31751 +               seq_putc(seq, '\0');
31752 +               if (!seq_has_overflowed(seq)) {
31753 +                       err = copy_to_user(arg->path, seq->buf, seq->count);
31754 +                       seq->count = 0;
31755 +                       if (unlikely(err))
31756 +                               break;
31757 +               } else {
31758 +                       err = -E2BIG;
31759 +                       goto out_seq;
31760 +               }
31761 +       }
31762 +       if (unlikely(err))
31763 +               err = -EFAULT;
31764 +
31765 +out_seq:
31766 +       au_kfree_rcu(seq);
31767 +out_buf:
31768 +       free_page((unsigned long)buf);
31769 +out:
31770 +       si_read_unlock(sb);
31771 +       return err;
31772 +}
31773 +
31774 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
31775 +{
31776 +       return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
31777 +}
31778 +
31779 +#ifdef CONFIG_COMPAT
31780 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
31781 +{
31782 +       return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
31783 +}
31784 +#endif
31785 +
31786 +/* ---------------------------------------------------------------------- */
31787 +
31788 +void sysaufs_br_init(struct au_branch *br)
31789 +{
31790 +       int i;
31791 +       struct au_brsysfs *br_sysfs;
31792 +       struct attribute *attr;
31793 +
31794 +       br_sysfs = br->br_sysfs;
31795 +       for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31796 +               attr = &br_sysfs->attr;
31797 +               sysfs_attr_init(attr);
31798 +               attr->name = br_sysfs->name;
31799 +               attr->mode = 0444;
31800 +               br_sysfs++;
31801 +       }
31802 +}
31803 +
31804 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
31805 +{
31806 +       struct au_branch *br;
31807 +       struct kobject *kobj;
31808 +       struct au_brsysfs *br_sysfs;
31809 +       int i;
31810 +       aufs_bindex_t bbot;
31811 +
31812 +       if (!sysaufs_brs)
31813 +               return;
31814 +
31815 +       kobj = &au_sbi(sb)->si_kobj;
31816 +       bbot = au_sbbot(sb);
31817 +       for (; bindex <= bbot; bindex++) {
31818 +               br = au_sbr(sb, bindex);
31819 +               br_sysfs = br->br_sysfs;
31820 +               for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31821 +                       sysfs_remove_file(kobj, &br_sysfs->attr);
31822 +                       br_sysfs++;
31823 +               }
31824 +       }
31825 +}
31826 +
31827 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
31828 +{
31829 +       int err, i;
31830 +       aufs_bindex_t bbot;
31831 +       struct kobject *kobj;
31832 +       struct au_branch *br;
31833 +       struct au_brsysfs *br_sysfs;
31834 +
31835 +       if (!sysaufs_brs)
31836 +               return;
31837 +
31838 +       kobj = &au_sbi(sb)->si_kobj;
31839 +       bbot = au_sbbot(sb);
31840 +       for (; bindex <= bbot; bindex++) {
31841 +               br = au_sbr(sb, bindex);
31842 +               br_sysfs = br->br_sysfs;
31843 +               snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
31844 +                        SysaufsBr_PREFIX "%d", bindex);
31845 +               snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
31846 +                        SysaufsBrid_PREFIX "%d", bindex);
31847 +               for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31848 +                       err = sysfs_create_file(kobj, &br_sysfs->attr);
31849 +                       if (unlikely(err))
31850 +                               pr_warn("failed %s under sysfs(%d)\n",
31851 +                                       br_sysfs->name, err);
31852 +                       br_sysfs++;
31853 +               }
31854 +       }
31855 +}
31856 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
31857 --- /usr/share/empty/fs/aufs/sysrq.c    1970-01-01 01:00:00.000000000 +0100
31858 +++ linux/fs/aufs/sysrq.c       2022-11-05 23:02:18.969222617 +0100
31859 @@ -0,0 +1,149 @@
31860 +// SPDX-License-Identifier: GPL-2.0
31861 +/*
31862 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31863 + *
31864 + * This program is free software; you can redistribute it and/or modify
31865 + * it under the terms of the GNU General Public License as published by
31866 + * the Free Software Foundation; either version 2 of the License, or
31867 + * (at your option) any later version.
31868 + *
31869 + * This program is distributed in the hope that it will be useful,
31870 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31871 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31872 + * GNU General Public License for more details.
31873 + *
31874 + * You should have received a copy of the GNU General Public License
31875 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
31876 + */
31877 +
31878 +/*
31879 + * magic sysrq handler
31880 + */
31881 +
31882 +/* #include <linux/sysrq.h> */
31883 +#include <linux/writeback.h>
31884 +#include "aufs.h"
31885 +
31886 +/* ---------------------------------------------------------------------- */
31887 +
31888 +static void sysrq_sb(struct super_block *sb)
31889 +{
31890 +       char *plevel;
31891 +       struct au_sbinfo *sbinfo;
31892 +       struct file *file;
31893 +       struct hlist_bl_head *files;
31894 +       struct hlist_bl_node *pos;
31895 +       struct au_finfo *finfo;
31896 +       struct inode *i;
31897 +
31898 +       plevel = au_plevel;
31899 +       au_plevel = KERN_WARNING;
31900 +
31901 +       /* since we define pr_fmt, call printk directly */
31902 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
31903 +
31904 +       sbinfo = au_sbi(sb);
31905 +       printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
31906 +       pr("superblock\n");
31907 +       au_dpri_sb(sb);
31908 +
31909 +#if 0 /* reserved */
31910 +       do {
31911 +               int err, i, j, ndentry;
31912 +               struct au_dcsub_pages dpages;
31913 +               struct au_dpage *dpage;
31914 +
31915 +               err = au_dpages_init(&dpages, GFP_ATOMIC);
31916 +               if (unlikely(err))
31917 +                       break;
31918 +               err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
31919 +               if (!err)
31920 +                       for (i = 0; i < dpages.ndpage; i++) {
31921 +                               dpage = dpages.dpages + i;
31922 +                               ndentry = dpage->ndentry;
31923 +                               for (j = 0; j < ndentry; j++)
31924 +                                       au_dpri_dentry(dpage->dentries[j]);
31925 +                       }
31926 +               au_dpages_free(&dpages);
31927 +       } while (0);
31928 +#endif
31929 +
31930 +       pr("isolated inode\n");
31931 +       spin_lock(&sb->s_inode_list_lock);
31932 +       list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
31933 +               spin_lock(&i->i_lock);
31934 +               if (hlist_empty(&i->i_dentry))
31935 +                       au_dpri_inode(i);
31936 +               spin_unlock(&i->i_lock);
31937 +       }
31938 +       spin_unlock(&sb->s_inode_list_lock);
31939 +
31940 +       pr("files\n");
31941 +       files = &au_sbi(sb)->si_files;
31942 +       hlist_bl_lock(files);
31943 +       hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
31944 +               umode_t mode;
31945 +
31946 +               file = finfo->fi_file;
31947 +               mode = file_inode(file)->i_mode;
31948 +               if (!special_file(mode))
31949 +                       au_dpri_file(file);
31950 +       }
31951 +       hlist_bl_unlock(files);
31952 +       pr("done\n");
31953 +
31954 +#undef pr
31955 +       au_plevel = plevel;
31956 +}
31957 +
31958 +/* ---------------------------------------------------------------------- */
31959 +
31960 +/* module parameter */
31961 +static char *aufs_sysrq_key = "a";
31962 +module_param_named(sysrq, aufs_sysrq_key, charp, 0444);
31963 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
31964 +
31965 +static void au_sysrq(int key __maybe_unused)
31966 +{
31967 +       struct au_sbinfo *sbinfo;
31968 +       struct hlist_bl_node *pos;
31969 +
31970 +       lockdep_off();
31971 +       au_sbilist_lock();
31972 +       hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
31973 +               sysrq_sb(sbinfo->si_sb);
31974 +       au_sbilist_unlock();
31975 +       lockdep_on();
31976 +}
31977 +
31978 +static struct sysrq_key_op au_sysrq_op = {
31979 +       .handler        = au_sysrq,
31980 +       .help_msg       = "Aufs",
31981 +       .action_msg     = "Aufs",
31982 +       .enable_mask    = SYSRQ_ENABLE_DUMP
31983 +};
31984 +
31985 +/* ---------------------------------------------------------------------- */
31986 +
31987 +int __init au_sysrq_init(void)
31988 +{
31989 +       int err;
31990 +       char key;
31991 +
31992 +       err = -1;
31993 +       key = *aufs_sysrq_key;
31994 +       if ('a' <= key && key <= 'z')
31995 +               err = register_sysrq_key(key, &au_sysrq_op);
31996 +       if (unlikely(err))
31997 +               pr_err("err %d, sysrq=%c\n", err, key);
31998 +       return err;
31999 +}
32000 +
32001 +void au_sysrq_fin(void)
32002 +{
32003 +       int err;
32004 +
32005 +       err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
32006 +       if (unlikely(err))
32007 +               pr_err("err %d (ignored)\n", err);
32008 +}
32009 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
32010 --- /usr/share/empty/fs/aufs/vdir.c     1970-01-01 01:00:00.000000000 +0100
32011 +++ linux/fs/aufs/vdir.c        2022-11-05 23:02:18.969222617 +0100
32012 @@ -0,0 +1,896 @@
32013 +// SPDX-License-Identifier: GPL-2.0
32014 +/*
32015 + * Copyright (C) 2005-2022 Junjiro R. Okajima
32016 + *
32017 + * This program is free software; you can redistribute it and/or modify
32018 + * it under the terms of the GNU General Public License as published by
32019 + * the Free Software Foundation; either version 2 of the License, or
32020 + * (at your option) any later version.
32021 + *
32022 + * This program is distributed in the hope that it will be useful,
32023 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32024 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
32025 + * GNU General Public License for more details.
32026 + *
32027 + * You should have received a copy of the GNU General Public License
32028 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
32029 + */
32030 +
32031 +/*
32032 + * virtual or vertical directory
32033 + */
32034 +
32035 +#include <linux/iversion.h>
32036 +#include "aufs.h"
32037 +
32038 +static unsigned int calc_size(int nlen)
32039 +{
32040 +       return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
32041 +}
32042 +
32043 +static int set_deblk_end(union au_vdir_deblk_p *p,
32044 +                        union au_vdir_deblk_p *deblk_end)
32045 +{
32046 +       if (calc_size(0) <= deblk_end->deblk - p->deblk) {
32047 +               p->de->de_str.len = 0;
32048 +               /* smp_mb(); */
32049 +               return 0;
32050 +       }
32051 +       return -1; /* error */
32052 +}
32053 +
32054 +/* returns true or false */
32055 +static int is_deblk_end(union au_vdir_deblk_p *p,
32056 +                       union au_vdir_deblk_p *deblk_end)
32057 +{
32058 +       if (calc_size(0) <= deblk_end->deblk - p->deblk)
32059 +               return !p->de->de_str.len;
32060 +       return 1;
32061 +}
32062 +
32063 +static unsigned char *last_deblk(struct au_vdir *vdir)
32064 +{
32065 +       return vdir->vd_deblk[vdir->vd_nblk - 1];
32066 +}
32067 +
32068 +/* ---------------------------------------------------------------------- */
32069 +
32070 +/* estimate the appropriate size for name hash table */
32071 +unsigned int au_rdhash_est(loff_t sz)
32072 +{
32073 +       unsigned int n;
32074 +
32075 +       n = UINT_MAX;
32076 +       sz >>= 10;
32077 +       if (sz < n)
32078 +               n = sz;
32079 +       if (sz < AUFS_RDHASH_DEF)
32080 +               n = AUFS_RDHASH_DEF;
32081 +       /* pr_info("n %u\n", n); */
32082 +       return n;
32083 +}
32084 +
32085 +/*
32086 + * the allocated memory has to be freed by
32087 + * au_nhash_wh_free() or au_nhash_de_free().
32088 + */
32089 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
32090 +{
32091 +       struct hlist_head *head;
32092 +       unsigned int u;
32093 +       size_t sz;
32094 +
32095 +       sz = sizeof(*nhash->nh_head) * num_hash;
32096 +       head = kmalloc(sz, gfp);
32097 +       if (head) {
32098 +               nhash->nh_num = num_hash;
32099 +               nhash->nh_head = head;
32100 +               for (u = 0; u < num_hash; u++)
32101 +                       INIT_HLIST_HEAD(head++);
32102 +               return 0; /* success */
32103 +       }
32104 +
32105 +       return -ENOMEM;
32106 +}
32107 +
32108 +static void nhash_count(struct hlist_head *head)
32109 +{
32110 +#if 0 /* debugging */
32111 +       unsigned long n;
32112 +       struct hlist_node *pos;
32113 +
32114 +       n = 0;
32115 +       hlist_for_each(pos, head)
32116 +               n++;
32117 +       pr_info("%lu\n", n);
32118 +#endif
32119 +}
32120 +
32121 +static void au_nhash_wh_do_free(struct hlist_head *head)
32122 +{
32123 +       struct au_vdir_wh *pos;
32124 +       struct hlist_node *node;
32125 +
32126 +       hlist_for_each_entry_safe(pos, node, head, wh_hash)
32127 +               au_kfree_rcu(pos);
32128 +}
32129 +
32130 +static void au_nhash_de_do_free(struct hlist_head *head)
32131 +{
32132 +       struct au_vdir_dehstr *pos;
32133 +       struct hlist_node *node;
32134 +
32135 +       hlist_for_each_entry_safe(pos, node, head, hash)
32136 +               au_cache_free_vdir_dehstr(pos);
32137 +}
32138 +
32139 +static void au_nhash_do_free(struct au_nhash *nhash,
32140 +                            void (*free)(struct hlist_head *head))
32141 +{
32142 +       unsigned int n;
32143 +       struct hlist_head *head;
32144 +
32145 +       n = nhash->nh_num;
32146 +       if (!n)
32147 +               return;
32148 +
32149 +       head = nhash->nh_head;
32150 +       while (n-- > 0) {
32151 +               nhash_count(head);
32152 +               free(head++);
32153 +       }
32154 +       au_kfree_try_rcu(nhash->nh_head);
32155 +}
32156 +
32157 +void au_nhash_wh_free(struct au_nhash *whlist)
32158 +{
32159 +       au_nhash_do_free(whlist, au_nhash_wh_do_free);
32160 +}
32161 +
32162 +static void au_nhash_de_free(struct au_nhash *delist)
32163 +{
32164 +       au_nhash_do_free(delist, au_nhash_de_do_free);
32165 +}
32166 +
32167 +/* ---------------------------------------------------------------------- */
32168 +
32169 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
32170 +                           int limit)
32171 +{
32172 +       int num;
32173 +       unsigned int u, n;
32174 +       struct hlist_head *head;
32175 +       struct au_vdir_wh *pos;
32176 +
32177 +       num = 0;
32178 +       n = whlist->nh_num;
32179 +       head = whlist->nh_head;
32180 +       for (u = 0; u < n; u++, head++)
32181 +               hlist_for_each_entry(pos, head, wh_hash)
32182 +                       if (pos->wh_bindex == btgt && ++num > limit)
32183 +                               return 1;
32184 +       return 0;
32185 +}
32186 +
32187 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
32188 +                                      unsigned char *name,
32189 +                                      unsigned int len)
32190 +{
32191 +       unsigned int v;
32192 +       /* const unsigned int magic_bit = 12; */
32193 +
32194 +       AuDebugOn(!nhash->nh_num || !nhash->nh_head);
32195 +
32196 +       v = 0;
32197 +       if (len > 8)
32198 +               len = 8;
32199 +       while (len--)
32200 +               v += *name++;
32201 +       /* v = hash_long(v, magic_bit); */
32202 +       v %= nhash->nh_num;
32203 +       return nhash->nh_head + v;
32204 +}
32205 +
32206 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
32207 +                             int nlen)
32208 +{
32209 +       return str->len == nlen && !memcmp(str->name, name, nlen);
32210 +}
32211 +
32212 +/* returns found or not */
32213 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
32214 +{
32215 +       struct hlist_head *head;
32216 +       struct au_vdir_wh *pos;
32217 +       struct au_vdir_destr *str;
32218 +
32219 +       head = au_name_hash(whlist, name, nlen);
32220 +       hlist_for_each_entry(pos, head, wh_hash) {
32221 +               str = &pos->wh_str;
32222 +               AuDbg("%.*s\n", str->len, str->name);
32223 +               if (au_nhash_test_name(str, name, nlen))
32224 +                       return 1;
32225 +       }
32226 +       return 0;
32227 +}
32228 +
32229 +/* returns found(true) or not */
32230 +static int test_known(struct au_nhash *delist, char *name, int nlen)
32231 +{
32232 +       struct hlist_head *head;
32233 +       struct au_vdir_dehstr *pos;
32234 +       struct au_vdir_destr *str;
32235 +
32236 +       head = au_name_hash(delist, name, nlen);
32237 +       hlist_for_each_entry(pos, head, hash) {
32238 +               str = pos->str;
32239 +               AuDbg("%.*s\n", str->len, str->name);
32240 +               if (au_nhash_test_name(str, name, nlen))
32241 +                       return 1;
32242 +       }
32243 +       return 0;
32244 +}
32245 +
32246 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
32247 +                           unsigned char d_type)
32248 +{
32249 +#ifdef CONFIG_AUFS_SHWH
32250 +       wh->wh_ino = ino;
32251 +       wh->wh_type = d_type;
32252 +#endif
32253 +}
32254 +
32255 +/* ---------------------------------------------------------------------- */
32256 +
32257 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
32258 +                      unsigned int d_type, aufs_bindex_t bindex,
32259 +                      unsigned char shwh)
32260 +{
32261 +       int err;
32262 +       struct au_vdir_destr *str;
32263 +       struct au_vdir_wh *wh;
32264 +
32265 +       AuDbg("%.*s\n", nlen, name);
32266 +       AuDebugOn(!whlist->nh_num || !whlist->nh_head);
32267 +
32268 +       err = -ENOMEM;
32269 +       wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
32270 +       if (unlikely(!wh))
32271 +               goto out;
32272 +
32273 +       err = 0;
32274 +       wh->wh_bindex = bindex;
32275 +       if (shwh)
32276 +               au_shwh_init_wh(wh, ino, d_type);
32277 +       str = &wh->wh_str;
32278 +       str->len = nlen;
32279 +       memcpy(str->name, name, nlen);
32280 +       hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
32281 +       /* smp_mb(); */
32282 +
32283 +out:
32284 +       return err;
32285 +}
32286 +
32287 +static int append_deblk(struct au_vdir *vdir)
32288 +{
32289 +       int err;
32290 +       unsigned long ul;
32291 +       const unsigned int deblk_sz = vdir->vd_deblk_sz;
32292 +       union au_vdir_deblk_p p, deblk_end;
32293 +       unsigned char **o;
32294 +
32295 +       err = -ENOMEM;
32296 +       o = au_krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
32297 +                       GFP_NOFS, /*may_shrink*/0);
32298 +       if (unlikely(!o))
32299 +               goto out;
32300 +
32301 +       vdir->vd_deblk = o;
32302 +       p.deblk = kmalloc(deblk_sz, GFP_NOFS);
32303 +       if (p.deblk) {
32304 +               ul = vdir->vd_nblk++;
32305 +               vdir->vd_deblk[ul] = p.deblk;
32306 +               vdir->vd_last.ul = ul;
32307 +               vdir->vd_last.p.deblk = p.deblk;
32308 +               deblk_end.deblk = p.deblk + deblk_sz;
32309 +               err = set_deblk_end(&p, &deblk_end);
32310 +       }
32311 +
32312 +out:
32313 +       return err;
32314 +}
32315 +
32316 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
32317 +                    unsigned int d_type, struct au_nhash *delist)
32318 +{
32319 +       int err;
32320 +       unsigned int sz;
32321 +       const unsigned int deblk_sz = vdir->vd_deblk_sz;
32322 +       union au_vdir_deblk_p p, *room, deblk_end;
32323 +       struct au_vdir_dehstr *dehstr;
32324 +
32325 +       p.deblk = last_deblk(vdir);
32326 +       deblk_end.deblk = p.deblk + deblk_sz;
32327 +       room = &vdir->vd_last.p;
32328 +       AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
32329 +                 || !is_deblk_end(room, &deblk_end));
32330 +
32331 +       sz = calc_size(nlen);
32332 +       if (unlikely(sz > deblk_end.deblk - room->deblk)) {
32333 +               err = append_deblk(vdir);
32334 +               if (unlikely(err))
32335 +                       goto out;
32336 +
32337 +               p.deblk = last_deblk(vdir);
32338 +               deblk_end.deblk = p.deblk + deblk_sz;
32339 +               /* smp_mb(); */
32340 +               AuDebugOn(room->deblk != p.deblk);
32341 +       }
32342 +
32343 +       err = -ENOMEM;
32344 +       dehstr = au_cache_alloc_vdir_dehstr();
32345 +       if (unlikely(!dehstr))
32346 +               goto out;
32347 +
32348 +       dehstr->str = &room->de->de_str;
32349 +       hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
32350 +       room->de->de_ino = ino;
32351 +       room->de->de_type = d_type;
32352 +       room->de->de_str.len = nlen;
32353 +       memcpy(room->de->de_str.name, name, nlen);
32354 +
32355 +       err = 0;
32356 +       room->deblk += sz;
32357 +       if (unlikely(set_deblk_end(room, &deblk_end)))
32358 +               err = append_deblk(vdir);
32359 +       /* smp_mb(); */
32360 +
32361 +out:
32362 +       return err;
32363 +}
32364 +
32365 +/* ---------------------------------------------------------------------- */
32366 +
32367 +void au_vdir_free(struct au_vdir *vdir)
32368 +{
32369 +       unsigned char **deblk;
32370 +
32371 +       deblk = vdir->vd_deblk;
32372 +       while (vdir->vd_nblk--)
32373 +               au_kfree_try_rcu(*deblk++);
32374 +       au_kfree_try_rcu(vdir->vd_deblk);
32375 +       au_cache_free_vdir(vdir);
32376 +}
32377 +
32378 +static struct au_vdir *alloc_vdir(struct file *file)
32379 +{
32380 +       struct au_vdir *vdir;
32381 +       struct super_block *sb;
32382 +       int err;
32383 +
32384 +       sb = file->f_path.dentry->d_sb;
32385 +       SiMustAnyLock(sb);
32386 +
32387 +       err = -ENOMEM;
32388 +       vdir = au_cache_alloc_vdir();
32389 +       if (unlikely(!vdir))
32390 +               goto out;
32391 +
32392 +       vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
32393 +       if (unlikely(!vdir->vd_deblk))
32394 +               goto out_free;
32395 +
32396 +       vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
32397 +       if (!vdir->vd_deblk_sz) {
32398 +               /* estimate the appropriate size for deblk */
32399 +               vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
32400 +               /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
32401 +       }
32402 +       vdir->vd_nblk = 0;
32403 +       vdir->vd_version = 0;
32404 +       vdir->vd_jiffy = 0;
32405 +       err = append_deblk(vdir);
32406 +       if (!err)
32407 +               return vdir; /* success */
32408 +
32409 +       au_kfree_try_rcu(vdir->vd_deblk);
32410 +
32411 +out_free:
32412 +       au_cache_free_vdir(vdir);
32413 +out:
32414 +       vdir = ERR_PTR(err);
32415 +       return vdir;
32416 +}
32417 +
32418 +static int reinit_vdir(struct au_vdir *vdir)
32419 +{
32420 +       int err;
32421 +       union au_vdir_deblk_p p, deblk_end;
32422 +
32423 +       while (vdir->vd_nblk > 1) {
32424 +               au_kfree_try_rcu(vdir->vd_deblk[vdir->vd_nblk - 1]);
32425 +               /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
32426 +               vdir->vd_nblk--;
32427 +       }
32428 +       p.deblk = vdir->vd_deblk[0];
32429 +       deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
32430 +       err = set_deblk_end(&p, &deblk_end);
32431 +       /* keep vd_dblk_sz */
32432 +       vdir->vd_last.ul = 0;
32433 +       vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32434 +       vdir->vd_version = 0;
32435 +       vdir->vd_jiffy = 0;
32436 +       /* smp_mb(); */
32437 +       return err;
32438 +}
32439 +
32440 +/* ---------------------------------------------------------------------- */
32441 +
32442 +#define AuFillVdir_CALLED      1
32443 +#define AuFillVdir_WHABLE      (1 << 1)
32444 +#define AuFillVdir_SHWH                (1 << 2)
32445 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
32446 +#define au_fset_fillvdir(flags, name) \
32447 +       do { (flags) |= AuFillVdir_##name; } while (0)
32448 +#define au_fclr_fillvdir(flags, name) \
32449 +       do { (flags) &= ~AuFillVdir_##name; } while (0)
32450 +
32451 +#ifndef CONFIG_AUFS_SHWH
32452 +#undef AuFillVdir_SHWH
32453 +#define AuFillVdir_SHWH                0
32454 +#endif
32455 +
32456 +struct fillvdir_arg {
32457 +       struct dir_context      ctx;
32458 +       struct file             *file;
32459 +       struct au_vdir          *vdir;
32460 +       struct au_nhash         delist;
32461 +       struct au_nhash         whlist;
32462 +       aufs_bindex_t           bindex;
32463 +       unsigned int            flags;
32464 +       int                     err;
32465 +};
32466 +
32467 +static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
32468 +                   loff_t offset __maybe_unused, u64 h_ino,
32469 +                   unsigned int d_type)
32470 +{
32471 +       struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
32472 +       char *name = (void *)__name;
32473 +       struct super_block *sb;
32474 +       ino_t ino;
32475 +       const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
32476 +
32477 +       arg->err = 0;
32478 +       sb = arg->file->f_path.dentry->d_sb;
32479 +       au_fset_fillvdir(arg->flags, CALLED);
32480 +       /* smp_mb(); */
32481 +       if (nlen <= AUFS_WH_PFX_LEN
32482 +           || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
32483 +               if (test_known(&arg->delist, name, nlen)
32484 +                   || au_nhash_test_known_wh(&arg->whlist, name, nlen))
32485 +                       goto out; /* already exists or whiteouted */
32486 +
32487 +               arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
32488 +               if (!arg->err) {
32489 +                       if (unlikely(nlen > AUFS_MAX_NAMELEN))
32490 +                               d_type = DT_UNKNOWN;
32491 +                       arg->err = append_de(arg->vdir, name, nlen, ino,
32492 +                                            d_type, &arg->delist);
32493 +               }
32494 +       } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
32495 +               name += AUFS_WH_PFX_LEN;
32496 +               nlen -= AUFS_WH_PFX_LEN;
32497 +               if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
32498 +                       goto out; /* already whiteouted */
32499 +
32500 +               ino = 0; /* just to suppress a warning */
32501 +               if (shwh)
32502 +                       arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
32503 +                                            &ino);
32504 +               if (!arg->err) {
32505 +                       if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
32506 +                               d_type = DT_UNKNOWN;
32507 +                       arg->err = au_nhash_append_wh
32508 +                               (&arg->whlist, name, nlen, ino, d_type,
32509 +                                arg->bindex, shwh);
32510 +               }
32511 +       }
32512 +
32513 +out:
32514 +       if (!arg->err)
32515 +               arg->vdir->vd_jiffy = jiffies;
32516 +       /* smp_mb(); */
32517 +       AuTraceErr(arg->err);
32518 +       return arg->err;
32519 +}
32520 +
32521 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
32522 +                         struct au_nhash *whlist, struct au_nhash *delist)
32523 +{
32524 +#ifdef CONFIG_AUFS_SHWH
32525 +       int err;
32526 +       unsigned int nh, u;
32527 +       struct hlist_head *head;
32528 +       struct au_vdir_wh *pos;
32529 +       struct hlist_node *n;
32530 +       char *p, *o;
32531 +       struct au_vdir_destr *destr;
32532 +
32533 +       AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
32534 +
32535 +       err = -ENOMEM;
32536 +       o = p = (void *)__get_free_page(GFP_NOFS);
32537 +       if (unlikely(!p))
32538 +               goto out;
32539 +
32540 +       err = 0;
32541 +       nh = whlist->nh_num;
32542 +       memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32543 +       p += AUFS_WH_PFX_LEN;
32544 +       for (u = 0; u < nh; u++) {
32545 +               head = whlist->nh_head + u;
32546 +               hlist_for_each_entry_safe(pos, n, head, wh_hash) {
32547 +                       destr = &pos->wh_str;
32548 +                       memcpy(p, destr->name, destr->len);
32549 +                       err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
32550 +                                       pos->wh_ino, pos->wh_type, delist);
32551 +                       if (unlikely(err))
32552 +                               break;
32553 +               }
32554 +       }
32555 +
32556 +       free_page((unsigned long)o);
32557 +
32558 +out:
32559 +       AuTraceErr(err);
32560 +       return err;
32561 +#else
32562 +       return 0;
32563 +#endif
32564 +}
32565 +
32566 +static int au_do_read_vdir(struct fillvdir_arg *arg)
32567 +{
32568 +       int err;
32569 +       unsigned int rdhash;
32570 +       loff_t offset;
32571 +       aufs_bindex_t bbot, bindex, btop;
32572 +       unsigned char shwh;
32573 +       struct file *hf, *file;
32574 +       struct super_block *sb;
32575 +
32576 +       file = arg->file;
32577 +       sb = file->f_path.dentry->d_sb;
32578 +       SiMustAnyLock(sb);
32579 +
32580 +       rdhash = au_sbi(sb)->si_rdhash;
32581 +       if (!rdhash)
32582 +               rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
32583 +       err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
32584 +       if (unlikely(err))
32585 +               goto out;
32586 +       err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
32587 +       if (unlikely(err))
32588 +               goto out_delist;
32589 +
32590 +       err = 0;
32591 +       arg->flags = 0;
32592 +       shwh = 0;
32593 +       if (au_opt_test(au_mntflags(sb), SHWH)) {
32594 +               shwh = 1;
32595 +               au_fset_fillvdir(arg->flags, SHWH);
32596 +       }
32597 +       btop = au_fbtop(file);
32598 +       bbot = au_fbbot_dir(file);
32599 +       for (bindex = btop; !err && bindex <= bbot; bindex++) {
32600 +               hf = au_hf_dir(file, bindex);
32601 +               if (!hf)
32602 +                       continue;
32603 +
32604 +               offset = vfsub_llseek(hf, 0, SEEK_SET);
32605 +               err = offset;
32606 +               if (unlikely(offset))
32607 +                       break;
32608 +
32609 +               arg->bindex = bindex;
32610 +               au_fclr_fillvdir(arg->flags, WHABLE);
32611 +               if (shwh
32612 +                   || (bindex != bbot
32613 +                       && au_br_whable(au_sbr_perm(sb, bindex))))
32614 +                       au_fset_fillvdir(arg->flags, WHABLE);
32615 +               do {
32616 +                       arg->err = 0;
32617 +                       au_fclr_fillvdir(arg->flags, CALLED);
32618 +                       /* smp_mb(); */
32619 +                       err = vfsub_iterate_dir(hf, &arg->ctx);
32620 +                       if (err >= 0)
32621 +                               err = arg->err;
32622 +               } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
32623 +
32624 +               /*
32625 +                * dir_relax() may be good for concurrency, but aufs should not
32626 +                * use it since it will cause a lockdep problem.
32627 +                */
32628 +       }
32629 +
32630 +       if (!err && shwh)
32631 +               err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
32632 +
32633 +       au_nhash_wh_free(&arg->whlist);
32634 +
32635 +out_delist:
32636 +       au_nhash_de_free(&arg->delist);
32637 +out:
32638 +       return err;
32639 +}
32640 +
32641 +static int read_vdir(struct file *file, int may_read)
32642 +{
32643 +       int err;
32644 +       unsigned long expire;
32645 +       unsigned char do_read;
32646 +       struct fillvdir_arg arg = {
32647 +               .ctx = {
32648 +                       .actor = fillvdir
32649 +               }
32650 +       };
32651 +       struct inode *inode;
32652 +       struct au_vdir *vdir, *allocated;
32653 +
32654 +       err = 0;
32655 +       inode = file_inode(file);
32656 +       IMustLock(inode);
32657 +       IiMustWriteLock(inode);
32658 +       SiMustAnyLock(inode->i_sb);
32659 +
32660 +       allocated = NULL;
32661 +       do_read = 0;
32662 +       expire = au_sbi(inode->i_sb)->si_rdcache;
32663 +       vdir = au_ivdir(inode);
32664 +       if (!vdir) {
32665 +               do_read = 1;
32666 +               vdir = alloc_vdir(file);
32667 +               err = PTR_ERR(vdir);
32668 +               if (IS_ERR(vdir))
32669 +                       goto out;
32670 +               err = 0;
32671 +               allocated = vdir;
32672 +       } else if (may_read
32673 +                  && (!inode_eq_iversion(inode, vdir->vd_version)
32674 +                      || time_after(jiffies, vdir->vd_jiffy + expire))) {
32675 +               do_read = 1;
32676 +               err = reinit_vdir(vdir);
32677 +               if (unlikely(err))
32678 +                       goto out;
32679 +       }
32680 +
32681 +       if (!do_read)
32682 +               return 0; /* success */
32683 +
32684 +       arg.file = file;
32685 +       arg.vdir = vdir;
32686 +       err = au_do_read_vdir(&arg);
32687 +       if (!err) {
32688 +               /* file->f_pos = 0; */ /* todo: ctx->pos? */
32689 +               vdir->vd_version = inode_query_iversion(inode);
32690 +               vdir->vd_last.ul = 0;
32691 +               vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32692 +               if (allocated)
32693 +                       au_set_ivdir(inode, allocated);
32694 +       } else if (allocated)
32695 +               au_vdir_free(allocated);
32696 +
32697 +out:
32698 +       return err;
32699 +}
32700 +
32701 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
32702 +{
32703 +       int err, rerr;
32704 +       unsigned long ul, n;
32705 +       const unsigned int deblk_sz = src->vd_deblk_sz;
32706 +
32707 +       AuDebugOn(tgt->vd_nblk != 1);
32708 +
32709 +       err = -ENOMEM;
32710 +       if (tgt->vd_nblk < src->vd_nblk) {
32711 +               unsigned char **p;
32712 +
32713 +               p = au_krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
32714 +                               GFP_NOFS, /*may_shrink*/0);
32715 +               if (unlikely(!p))
32716 +                       goto out;
32717 +               tgt->vd_deblk = p;
32718 +       }
32719 +
32720 +       if (tgt->vd_deblk_sz != deblk_sz) {
32721 +               unsigned char *p;
32722 +
32723 +               tgt->vd_deblk_sz = deblk_sz;
32724 +               p = au_krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS,
32725 +                               /*may_shrink*/1);
32726 +               if (unlikely(!p))
32727 +                       goto out;
32728 +               tgt->vd_deblk[0] = p;
32729 +       }
32730 +       memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
32731 +       tgt->vd_version = src->vd_version;
32732 +       tgt->vd_jiffy = src->vd_jiffy;
32733 +
32734 +       n = src->vd_nblk;
32735 +       for (ul = 1; ul < n; ul++) {
32736 +               tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
32737 +                                           GFP_NOFS);
32738 +               if (unlikely(!tgt->vd_deblk[ul]))
32739 +                       goto out;
32740 +               tgt->vd_nblk++;
32741 +       }
32742 +       tgt->vd_nblk = n;
32743 +       tgt->vd_last.ul = tgt->vd_last.ul;
32744 +       tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
32745 +       tgt->vd_last.p.deblk += src->vd_last.p.deblk
32746 +               - src->vd_deblk[src->vd_last.ul];
32747 +       /* smp_mb(); */
32748 +       return 0; /* success */
32749 +
32750 +out:
32751 +       rerr = reinit_vdir(tgt);
32752 +       BUG_ON(rerr);
32753 +       return err;
32754 +}
32755 +
32756 +int au_vdir_init(struct file *file)
32757 +{
32758 +       int err;
32759 +       struct inode *inode;
32760 +       struct au_vdir *vdir_cache, *allocated;
32761 +
32762 +       /* test file->f_pos here instead of ctx->pos */
32763 +       err = read_vdir(file, !file->f_pos);
32764 +       if (unlikely(err))
32765 +               goto out;
32766 +
32767 +       allocated = NULL;
32768 +       vdir_cache = au_fvdir_cache(file);
32769 +       if (!vdir_cache) {
32770 +               vdir_cache = alloc_vdir(file);
32771 +               err = PTR_ERR(vdir_cache);
32772 +               if (IS_ERR(vdir_cache))
32773 +                       goto out;
32774 +               allocated = vdir_cache;
32775 +       } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
32776 +               /* test file->f_pos here instead of ctx->pos */
32777 +               err = reinit_vdir(vdir_cache);
32778 +               if (unlikely(err))
32779 +                       goto out;
32780 +       } else
32781 +               return 0; /* success */
32782 +
32783 +       inode = file_inode(file);
32784 +       err = copy_vdir(vdir_cache, au_ivdir(inode));
32785 +       if (!err) {
32786 +               file->f_version = inode_query_iversion(inode);
32787 +               if (allocated)
32788 +                       au_set_fvdir_cache(file, allocated);
32789 +       } else if (allocated)
32790 +               au_vdir_free(allocated);
32791 +
32792 +out:
32793 +       return err;
32794 +}
32795 +
32796 +static loff_t calc_offset(struct au_vdir *vdir)
32797 +{
32798 +       loff_t offset;
32799 +       union au_vdir_deblk_p p;
32800 +
32801 +       p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
32802 +       offset = vdir->vd_last.p.deblk - p.deblk;
32803 +       offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
32804 +       return offset;
32805 +}
32806 +
32807 +/* returns true or false */
32808 +static int seek_vdir(struct file *file, struct dir_context *ctx)
32809 +{
32810 +       int valid;
32811 +       unsigned int deblk_sz;
32812 +       unsigned long ul, n;
32813 +       loff_t offset;
32814 +       union au_vdir_deblk_p p, deblk_end;
32815 +       struct au_vdir *vdir_cache;
32816 +
32817 +       valid = 1;
32818 +       vdir_cache = au_fvdir_cache(file);
32819 +       offset = calc_offset(vdir_cache);
32820 +       AuDbg("offset %lld\n", offset);
32821 +       if (ctx->pos == offset)
32822 +               goto out;
32823 +
32824 +       vdir_cache->vd_last.ul = 0;
32825 +       vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
32826 +       if (!ctx->pos)
32827 +               goto out;
32828 +
32829 +       valid = 0;
32830 +       deblk_sz = vdir_cache->vd_deblk_sz;
32831 +       ul = div64_u64(ctx->pos, deblk_sz);
32832 +       AuDbg("ul %lu\n", ul);
32833 +       if (ul >= vdir_cache->vd_nblk)
32834 +               goto out;
32835 +
32836 +       n = vdir_cache->vd_nblk;
32837 +       for (; ul < n; ul++) {
32838 +               p.deblk = vdir_cache->vd_deblk[ul];
32839 +               deblk_end.deblk = p.deblk + deblk_sz;
32840 +               offset = ul;
32841 +               offset *= deblk_sz;
32842 +               while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
32843 +                       unsigned int l;
32844 +
32845 +                       l = calc_size(p.de->de_str.len);
32846 +                       offset += l;
32847 +                       p.deblk += l;
32848 +               }
32849 +               if (!is_deblk_end(&p, &deblk_end)) {
32850 +                       valid = 1;
32851 +                       vdir_cache->vd_last.ul = ul;
32852 +                       vdir_cache->vd_last.p = p;
32853 +                       break;
32854 +               }
32855 +       }
32856 +
32857 +out:
32858 +       /* smp_mb(); */
32859 +       if (!valid)
32860 +               AuDbg("valid %d\n", !valid);
32861 +       return valid;
32862 +}
32863 +
32864 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
32865 +{
32866 +       unsigned int l, deblk_sz;
32867 +       union au_vdir_deblk_p deblk_end;
32868 +       struct au_vdir *vdir_cache;
32869 +       struct au_vdir_de *de;
32870 +
32871 +       if (!seek_vdir(file, ctx))
32872 +               return 0;
32873 +
32874 +       vdir_cache = au_fvdir_cache(file);
32875 +       deblk_sz = vdir_cache->vd_deblk_sz;
32876 +       while (1) {
32877 +               deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32878 +               deblk_end.deblk += deblk_sz;
32879 +               while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
32880 +                       de = vdir_cache->vd_last.p.de;
32881 +                       AuDbg("%.*s, off%lld, i%lu, dt%d\n",
32882 +                             de->de_str.len, de->de_str.name, ctx->pos,
32883 +                             (unsigned long)de->de_ino, de->de_type);
32884 +                       if (unlikely(!dir_emit(ctx, de->de_str.name,
32885 +                                              de->de_str.len, de->de_ino,
32886 +                                              de->de_type))) {
32887 +                               /* todo: ignore the error caused by udba? */
32888 +                               /* return err; */
32889 +                               return 0;
32890 +                       }
32891 +
32892 +                       l = calc_size(de->de_str.len);
32893 +                       vdir_cache->vd_last.p.deblk += l;
32894 +                       ctx->pos += l;
32895 +               }
32896 +               if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
32897 +                       vdir_cache->vd_last.ul++;
32898 +                       vdir_cache->vd_last.p.deblk
32899 +                               = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32900 +                       ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
32901 +                       continue;
32902 +               }
32903 +               break;
32904 +       }
32905 +
32906 +       /* smp_mb(); */
32907 +       return 0;
32908 +}
32909 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
32910 --- /usr/share/empty/fs/aufs/vfsub.c    1970-01-01 01:00:00.000000000 +0100
32911 +++ linux/fs/aufs/vfsub.c       2022-11-05 23:02:18.969222617 +0100
32912 @@ -0,0 +1,918 @@
32913 +// SPDX-License-Identifier: GPL-2.0
32914 +/*
32915 + * Copyright (C) 2005-2022 Junjiro R. Okajima
32916 + *
32917 + * This program is free software; you can redistribute it and/or modify
32918 + * it under the terms of the GNU General Public License as published by
32919 + * the Free Software Foundation; either version 2 of the License, or
32920 + * (at your option) any later version.
32921 + *
32922 + * This program is distributed in the hope that it will be useful,
32923 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32924 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
32925 + * GNU General Public License for more details.
32926 + *
32927 + * You should have received a copy of the GNU General Public License
32928 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
32929 + */
32930 +
32931 +/*
32932 + * sub-routines for VFS
32933 + */
32934 +
32935 +#include <linux/mnt_namespace.h>
32936 +#include <linux/nsproxy.h>
32937 +#include <linux/security.h>
32938 +#include <linux/splice.h>
32939 +#include "aufs.h"
32940 +
32941 +#ifdef CONFIG_AUFS_BR_FUSE
32942 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb)
32943 +{
32944 +       if (!au_test_fuse(h_sb) || !au_userns)
32945 +               return 0;
32946 +
32947 +       return is_current_mnt_ns(mnt) ? 0 : -EACCES;
32948 +}
32949 +#endif
32950 +
32951 +int vfsub_sync_filesystem(struct super_block *h_sb)
32952 +{
32953 +       int err;
32954 +
32955 +       lockdep_off();
32956 +       down_read(&h_sb->s_umount);
32957 +       err = sync_filesystem(h_sb);
32958 +       up_read(&h_sb->s_umount);
32959 +       lockdep_on();
32960 +
32961 +       return err;
32962 +}
32963 +
32964 +/* ---------------------------------------------------------------------- */
32965 +
32966 +int vfsub_update_h_iattr(struct path *h_path, int *did)
32967 +{
32968 +       int err;
32969 +       struct kstat st;
32970 +       struct super_block *h_sb;
32971 +
32972 +       /*
32973 +        * Always needs h_path->mnt for LSM or FUSE branch.
32974 +        */
32975 +       AuDebugOn(!h_path->mnt);
32976 +
32977 +       /* for remote fs, leave work for its getattr or d_revalidate */
32978 +       /* for bad i_attr fs, handle them in aufs_getattr() */
32979 +       /* still some fs may acquire i_mutex. we need to skip them */
32980 +       err = 0;
32981 +       if (!did)
32982 +               did = &err;
32983 +       h_sb = h_path->dentry->d_sb;
32984 +       *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
32985 +       if (*did)
32986 +               err = vfsub_getattr(h_path, &st);
32987 +
32988 +       return err;
32989 +}
32990 +
32991 +/* ---------------------------------------------------------------------- */
32992 +
32993 +struct file *vfsub_dentry_open(struct path *path, int flags)
32994 +{
32995 +       return dentry_open(path, flags /* | __FMODE_NONOTIFY */,
32996 +                          current_cred());
32997 +}
32998 +
32999 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
33000 +{
33001 +       struct file *file;
33002 +
33003 +       lockdep_off();
33004 +       file = filp_open(path,
33005 +                        oflags /* | __FMODE_NONOTIFY */,
33006 +                        mode);
33007 +       lockdep_on();
33008 +       if (IS_ERR(file))
33009 +               goto out;
33010 +       vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33011 +
33012 +out:
33013 +       return file;
33014 +}
33015 +
33016 +/*
33017 + * Ideally this function should call VFS:do_last() in order to keep all its
33018 + * checkings. But it is very hard for aufs to regenerate several VFS internal
33019 + * structure such as nameidata. This is a second (or third) best approach.
33020 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
33021 + */
33022 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33023 +                     struct vfsub_aopen_args *args)
33024 +{
33025 +       int err;
33026 +       struct au_branch *br = args->br;
33027 +       struct file *file = args->file;
33028 +       /* copied from linux/fs/namei.c:atomic_open() */
33029 +       struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
33030 +
33031 +       IMustLock(dir);
33032 +       AuDebugOn(!dir->i_op->atomic_open);
33033 +
33034 +       err = au_br_test_oflag(args->open_flag, br);
33035 +       if (unlikely(err))
33036 +               goto out;
33037 +
33038 +       au_lcnt_inc(&br->br_nfiles);
33039 +       file->f_path.dentry = DENTRY_NOT_SET;
33040 +       file->f_path.mnt = au_br_mnt(br);
33041 +       AuDbg("%ps\n", dir->i_op->atomic_open);
33042 +       err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
33043 +                                    args->create_mode);
33044 +       if (unlikely(err < 0)) {
33045 +               au_lcnt_dec(&br->br_nfiles);
33046 +               goto out;
33047 +       }
33048 +
33049 +       /* temporary workaround for nfsv4 branch */
33050 +       if (au_test_nfs(dir->i_sb))
33051 +               nfs_mark_for_revalidate(dir);
33052 +
33053 +       if (file->f_mode & FMODE_CREATED)
33054 +               fsnotify_create(dir, dentry);
33055 +       if (!(file->f_mode & FMODE_OPENED)) {
33056 +               au_lcnt_dec(&br->br_nfiles);
33057 +               goto out;
33058 +       }
33059 +
33060 +       /* todo: call VFS:may_open() here */
33061 +       /* todo: ima_file_check() too? */
33062 +       if (!err && (args->open_flag & __FMODE_EXEC))
33063 +               err = deny_write_access(file);
33064 +       if (!err)
33065 +               fsnotify_open(file);
33066 +       else
33067 +               au_lcnt_dec(&br->br_nfiles);
33068 +       /* note that the file is created and still opened */
33069 +
33070 +out:
33071 +       return err;
33072 +}
33073 +
33074 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
33075 +{
33076 +       int err;
33077 +
33078 +       err = kern_path(name, flags, path);
33079 +       if (!err && d_is_positive(path->dentry))
33080 +               vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33081 +       return err;
33082 +}
33083 +
33084 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33085 +                                            struct path *ppath, int len)
33086 +{
33087 +       struct path path;
33088 +
33089 +       path.dentry = lookup_one_len_unlocked(name, ppath->dentry, len);
33090 +       if (IS_ERR(path.dentry))
33091 +               goto out;
33092 +       if (d_is_positive(path.dentry)) {
33093 +               path.mnt = ppath->mnt;
33094 +               vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33095 +       }
33096 +
33097 +out:
33098 +       AuTraceErrPtr(path.dentry);
33099 +       return path.dentry;
33100 +}
33101 +
33102 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33103 +                                   int len)
33104 +{
33105 +       struct path path;
33106 +
33107 +       /* VFS checks it too, but by WARN_ON_ONCE() */
33108 +       IMustLock(d_inode(ppath->dentry));
33109 +
33110 +       path.dentry = lookup_one_len(name, ppath->dentry, len);
33111 +       if (IS_ERR(path.dentry))
33112 +               goto out;
33113 +       if (d_is_positive(path.dentry)) {
33114 +               path.mnt = ppath->mnt;
33115 +               vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33116 +       }
33117 +
33118 +out:
33119 +       AuTraceErrPtr(path.dentry);
33120 +       return path.dentry;
33121 +}
33122 +
33123 +void vfsub_call_lkup_one(void *args)
33124 +{
33125 +       struct vfsub_lkup_one_args *a = args;
33126 +       *a->errp = vfsub_lkup_one(a->name, a->ppath);
33127 +}
33128 +
33129 +/* ---------------------------------------------------------------------- */
33130 +
33131 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33132 +                                struct dentry *d2, struct au_hinode *hdir2)
33133 +{
33134 +       struct dentry *d;
33135 +
33136 +       lockdep_off();
33137 +       d = lock_rename(d1, d2);
33138 +       lockdep_on();
33139 +       au_hn_suspend(hdir1);
33140 +       if (hdir1 != hdir2)
33141 +               au_hn_suspend(hdir2);
33142 +
33143 +       return d;
33144 +}
33145 +
33146 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33147 +                        struct dentry *d2, struct au_hinode *hdir2)
33148 +{
33149 +       au_hn_resume(hdir1);
33150 +       if (hdir1 != hdir2)
33151 +               au_hn_resume(hdir2);
33152 +       lockdep_off();
33153 +       unlock_rename(d1, d2);
33154 +       lockdep_on();
33155 +}
33156 +
33157 +/* ---------------------------------------------------------------------- */
33158 +
33159 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
33160 +{
33161 +       int err;
33162 +       struct dentry *d;
33163 +       struct user_namespace *userns;
33164 +
33165 +       IMustLock(dir);
33166 +
33167 +       d = path->dentry;
33168 +       path->dentry = d->d_parent;
33169 +       err = security_path_mknod(path, d, mode, 0);
33170 +       path->dentry = d;
33171 +       if (unlikely(err))
33172 +               goto out;
33173 +       userns = mnt_user_ns(path->mnt);
33174 +
33175 +       lockdep_off();
33176 +       err = vfs_create(userns, dir, path->dentry, mode, want_excl);
33177 +       lockdep_on();
33178 +       if (!err) {
33179 +               struct path tmp = *path;
33180 +               int did;
33181 +
33182 +               vfsub_update_h_iattr(&tmp, &did);
33183 +               if (did) {
33184 +                       tmp.dentry = path->dentry->d_parent;
33185 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33186 +               }
33187 +               /*ignore*/
33188 +       }
33189 +
33190 +out:
33191 +       return err;
33192 +}
33193 +
33194 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
33195 +{
33196 +       int err;
33197 +       struct dentry *d;
33198 +       struct user_namespace *userns;
33199 +
33200 +       IMustLock(dir);
33201 +
33202 +       d = path->dentry;
33203 +       path->dentry = d->d_parent;
33204 +       err = security_path_symlink(path, d, symname);
33205 +       path->dentry = d;
33206 +       if (unlikely(err))
33207 +               goto out;
33208 +       userns = mnt_user_ns(path->mnt);
33209 +
33210 +       lockdep_off();
33211 +       err = vfs_symlink(userns, dir, path->dentry, symname);
33212 +       lockdep_on();
33213 +       if (!err) {
33214 +               struct path tmp = *path;
33215 +               int did;
33216 +
33217 +               vfsub_update_h_iattr(&tmp, &did);
33218 +               if (did) {
33219 +                       tmp.dentry = path->dentry->d_parent;
33220 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33221 +               }
33222 +               /*ignore*/
33223 +       }
33224 +
33225 +out:
33226 +       return err;
33227 +}
33228 +
33229 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
33230 +{
33231 +       int err;
33232 +       struct dentry *d;
33233 +       struct user_namespace *userns;
33234 +
33235 +       IMustLock(dir);
33236 +
33237 +       d = path->dentry;
33238 +       path->dentry = d->d_parent;
33239 +       err = security_path_mknod(path, d, mode, new_encode_dev(dev));
33240 +       path->dentry = d;
33241 +       if (unlikely(err))
33242 +               goto out;
33243 +       userns = mnt_user_ns(path->mnt);
33244 +
33245 +       lockdep_off();
33246 +       err = vfs_mknod(userns, dir, path->dentry, mode, dev);
33247 +       lockdep_on();
33248 +       if (!err) {
33249 +               struct path tmp = *path;
33250 +               int did;
33251 +
33252 +               vfsub_update_h_iattr(&tmp, &did);
33253 +               if (did) {
33254 +                       tmp.dentry = path->dentry->d_parent;
33255 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33256 +               }
33257 +               /*ignore*/
33258 +       }
33259 +
33260 +out:
33261 +       return err;
33262 +}
33263 +
33264 +static int au_test_nlink(struct inode *inode)
33265 +{
33266 +       const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
33267 +
33268 +       if (!au_test_fs_no_limit_nlink(inode->i_sb)
33269 +           || inode->i_nlink < link_max)
33270 +               return 0;
33271 +       return -EMLINK;
33272 +}
33273 +
33274 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
33275 +              struct inode **delegated_inode)
33276 +{
33277 +       int err;
33278 +       struct dentry *d;
33279 +       struct user_namespace *userns;
33280 +
33281 +       IMustLock(dir);
33282 +
33283 +       err = au_test_nlink(d_inode(src_dentry));
33284 +       if (unlikely(err))
33285 +               return err;
33286 +
33287 +       /* we don't call may_linkat() */
33288 +       d = path->dentry;
33289 +       path->dentry = d->d_parent;
33290 +       err = security_path_link(src_dentry, path, d);
33291 +       path->dentry = d;
33292 +       if (unlikely(err))
33293 +               goto out;
33294 +       userns = mnt_user_ns(path->mnt);
33295 +
33296 +       lockdep_off();
33297 +       err = vfs_link(src_dentry, userns, dir, path->dentry, delegated_inode);
33298 +       lockdep_on();
33299 +       if (!err) {
33300 +               struct path tmp = *path;
33301 +               int did;
33302 +
33303 +               /* fuse has different memory inode for the same inumber */
33304 +               vfsub_update_h_iattr(&tmp, &did);
33305 +               if (did) {
33306 +                       tmp.dentry = path->dentry->d_parent;
33307 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33308 +                       tmp.dentry = src_dentry;
33309 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33310 +               }
33311 +               /*ignore*/
33312 +       }
33313 +
33314 +out:
33315 +       return err;
33316 +}
33317 +
33318 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
33319 +                struct inode *dir, struct path *path,
33320 +                struct inode **delegated_inode, unsigned int flags)
33321 +{
33322 +       int err;
33323 +       struct renamedata rd;
33324 +       struct path tmp = {
33325 +               .mnt    = path->mnt
33326 +       };
33327 +       struct dentry *d;
33328 +
33329 +       IMustLock(dir);
33330 +       IMustLock(src_dir);
33331 +
33332 +       d = path->dentry;
33333 +       path->dentry = d->d_parent;
33334 +       tmp.dentry = src_dentry->d_parent;
33335 +       err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
33336 +       path->dentry = d;
33337 +       if (unlikely(err))
33338 +               goto out;
33339 +
33340 +       rd.old_mnt_userns = mnt_user_ns(path->mnt);
33341 +       rd.old_dir = src_dir;
33342 +       rd.old_dentry = src_dentry;
33343 +       rd.new_mnt_userns = rd.old_mnt_userns;
33344 +       rd.new_dir = dir;
33345 +       rd.new_dentry = path->dentry;
33346 +       rd.delegated_inode = delegated_inode;
33347 +       rd.flags = flags;
33348 +       lockdep_off();
33349 +       err = vfs_rename(&rd);
33350 +       lockdep_on();
33351 +       if (!err) {
33352 +               int did;
33353 +
33354 +               tmp.dentry = d->d_parent;
33355 +               vfsub_update_h_iattr(&tmp, &did);
33356 +               if (did) {
33357 +                       tmp.dentry = src_dentry;
33358 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33359 +                       tmp.dentry = src_dentry->d_parent;
33360 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33361 +               }
33362 +               /*ignore*/
33363 +       }
33364 +
33365 +out:
33366 +       return err;
33367 +}
33368 +
33369 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
33370 +{
33371 +       int err;
33372 +       struct dentry *d;
33373 +       struct user_namespace *userns;
33374 +
33375 +       IMustLock(dir);
33376 +
33377 +       d = path->dentry;
33378 +       path->dentry = d->d_parent;
33379 +       err = security_path_mkdir(path, d, mode);
33380 +       path->dentry = d;
33381 +       if (unlikely(err))
33382 +               goto out;
33383 +       userns = mnt_user_ns(path->mnt);
33384 +
33385 +       lockdep_off();
33386 +       err = vfs_mkdir(userns, dir, path->dentry, mode);
33387 +       lockdep_on();
33388 +       if (!err) {
33389 +               struct path tmp = *path;
33390 +               int did;
33391 +
33392 +               vfsub_update_h_iattr(&tmp, &did);
33393 +               if (did) {
33394 +                       tmp.dentry = path->dentry->d_parent;
33395 +                       vfsub_update_h_iattr(&tmp, /*did*/NULL);
33396 +               }
33397 +               /*ignore*/
33398 +       }
33399 +
33400 +out:
33401 +       return err;
33402 +}
33403 +
33404 +int vfsub_rmdir(struct inode *dir, struct path *path)
33405 +{
33406 +       int err;
33407 +       struct dentry *d;
33408 +       struct user_namespace *userns;
33409 +
33410 +       IMustLock(dir);
33411 +
33412 +       d = path->dentry;
33413 +       path->dentry = d->d_parent;
33414 +       err = security_path_rmdir(path, d);
33415 +       path->dentry = d;
33416 +       if (unlikely(err))
33417 +               goto out;
33418 +       userns = mnt_user_ns(path->mnt);
33419 +
33420 +       lockdep_off();
33421 +       err = vfs_rmdir(userns, dir, path->dentry);
33422 +       lockdep_on();
33423 +       if (!err) {
33424 +               struct path tmp = {
33425 +                       .dentry = path->dentry->d_parent,
33426 +                       .mnt    = path->mnt
33427 +               };
33428 +
33429 +               vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33430 +       }
33431 +
33432 +out:
33433 +       return err;
33434 +}
33435 +
33436 +/* ---------------------------------------------------------------------- */
33437 +
33438 +/* todo: support mmap_sem? */
33439 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
33440 +                    loff_t *ppos)
33441 +{
33442 +       ssize_t err;
33443 +
33444 +       lockdep_off();
33445 +       err = vfs_read(file, ubuf, count, ppos);
33446 +       lockdep_on();
33447 +       if (err >= 0)
33448 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33449 +       return err;
33450 +}
33451 +
33452 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
33453 +                    loff_t *ppos)
33454 +{
33455 +       ssize_t err;
33456 +
33457 +       lockdep_off();
33458 +       err = kernel_read(file, kbuf, count, ppos);
33459 +       lockdep_on();
33460 +       AuTraceErr(err);
33461 +       if (err >= 0)
33462 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33463 +       return err;
33464 +}
33465 +
33466 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
33467 +                     loff_t *ppos)
33468 +{
33469 +       ssize_t err;
33470 +
33471 +       lockdep_off();
33472 +       err = vfs_write(file, ubuf, count, ppos);
33473 +       lockdep_on();
33474 +       if (err >= 0)
33475 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33476 +       return err;
33477 +}
33478 +
33479 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
33480 +{
33481 +       ssize_t err;
33482 +
33483 +       lockdep_off();
33484 +       err = kernel_write(file, kbuf, count, ppos);
33485 +       lockdep_on();
33486 +       if (err >= 0)
33487 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33488 +       return err;
33489 +}
33490 +
33491 +int vfsub_flush(struct file *file, fl_owner_t id)
33492 +{
33493 +       int err;
33494 +
33495 +       err = 0;
33496 +       if (file->f_op->flush) {
33497 +               if (!au_test_nfs(file->f_path.dentry->d_sb))
33498 +                       err = file->f_op->flush(file, id);
33499 +               else {
33500 +                       lockdep_off();
33501 +                       err = file->f_op->flush(file, id);
33502 +                       lockdep_on();
33503 +               }
33504 +               if (!err)
33505 +                       vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
33506 +               /*ignore*/
33507 +       }
33508 +       return err;
33509 +}
33510 +
33511 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
33512 +{
33513 +       int err;
33514 +
33515 +       AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
33516 +
33517 +       lockdep_off();
33518 +       err = iterate_dir(file, ctx);
33519 +       lockdep_on();
33520 +       if (err >= 0)
33521 +               vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33522 +
33523 +       return err;
33524 +}
33525 +
33526 +long vfsub_splice_to(struct file *in, loff_t *ppos,
33527 +                    struct pipe_inode_info *pipe, size_t len,
33528 +                    unsigned int flags)
33529 +{
33530 +       long err;
33531 +
33532 +       lockdep_off();
33533 +       err = do_splice_to(in, ppos, pipe, len, flags);
33534 +       lockdep_on();
33535 +       file_accessed(in);
33536 +       if (err >= 0)
33537 +               vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
33538 +       return err;
33539 +}
33540 +
33541 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
33542 +                      loff_t *ppos, size_t len, unsigned int flags)
33543 +{
33544 +       long err;
33545 +
33546 +       lockdep_off();
33547 +       err = do_splice_from(pipe, out, ppos, len, flags);
33548 +       lockdep_on();
33549 +       if (err >= 0)
33550 +               vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
33551 +       return err;
33552 +}
33553 +
33554 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
33555 +{
33556 +       int err;
33557 +
33558 +       /* file can be NULL */
33559 +       lockdep_off();
33560 +       err = vfs_fsync(file, datasync);
33561 +       lockdep_on();
33562 +       if (!err) {
33563 +               if (!path) {
33564 +                       AuDebugOn(!file);
33565 +                       path = &file->f_path;
33566 +               }
33567 +               vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33568 +       }
33569 +       return err;
33570 +}
33571 +
33572 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
33573 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33574 +               struct file *h_file)
33575 +{
33576 +       int err;
33577 +       struct inode *h_inode;
33578 +       struct super_block *h_sb;
33579 +       struct user_namespace *h_userns;
33580 +
33581 +       if (!h_file) {
33582 +               err = vfsub_truncate(h_path, length);
33583 +               goto out;
33584 +       }
33585 +
33586 +       h_inode = d_inode(h_path->dentry);
33587 +       h_sb = h_inode->i_sb;
33588 +       lockdep_off();
33589 +       sb_start_write(h_sb);
33590 +       lockdep_on();
33591 +       err = security_path_truncate(h_path);
33592 +       if (!err) {
33593 +               h_userns = mnt_user_ns(h_path->mnt);
33594 +               lockdep_off();
33595 +               err = do_truncate(h_userns, h_path->dentry, length, attr,
33596 +                                 h_file);
33597 +               lockdep_on();
33598 +       }
33599 +       lockdep_off();
33600 +       sb_end_write(h_sb);
33601 +       lockdep_on();
33602 +
33603 +out:
33604 +       return err;
33605 +}
33606 +
33607 +/* ---------------------------------------------------------------------- */
33608 +
33609 +struct au_vfsub_mkdir_args {
33610 +       int *errp;
33611 +       struct inode *dir;
33612 +       struct path *path;
33613 +       int mode;
33614 +};
33615 +
33616 +static void au_call_vfsub_mkdir(void *args)
33617 +{
33618 +       struct au_vfsub_mkdir_args *a = args;
33619 +       *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
33620 +}
33621 +
33622 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
33623 +{
33624 +       int err, do_sio, wkq_err;
33625 +       struct user_namespace *userns;
33626 +
33627 +       userns = mnt_user_ns(path->mnt);
33628 +       do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33629 +       if (!do_sio) {
33630 +               lockdep_off();
33631 +               err = vfsub_mkdir(dir, path, mode);
33632 +               lockdep_on();
33633 +       } else {
33634 +               struct au_vfsub_mkdir_args args = {
33635 +                       .errp   = &err,
33636 +                       .dir    = dir,
33637 +                       .path   = path,
33638 +                       .mode   = mode
33639 +               };
33640 +               wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
33641 +               if (unlikely(wkq_err))
33642 +                       err = wkq_err;
33643 +       }
33644 +
33645 +       return err;
33646 +}
33647 +
33648 +struct au_vfsub_rmdir_args {
33649 +       int *errp;
33650 +       struct inode *dir;
33651 +       struct path *path;
33652 +};
33653 +
33654 +static void au_call_vfsub_rmdir(void *args)
33655 +{
33656 +       struct au_vfsub_rmdir_args *a = args;
33657 +       *a->errp = vfsub_rmdir(a->dir, a->path);
33658 +}
33659 +
33660 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
33661 +{
33662 +       int err, do_sio, wkq_err;
33663 +       struct user_namespace *userns;
33664 +
33665 +       userns = mnt_user_ns(path->mnt);
33666 +       do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33667 +       if (!do_sio) {
33668 +               lockdep_off();
33669 +               err = vfsub_rmdir(dir, path);
33670 +               lockdep_on();
33671 +       } else {
33672 +               struct au_vfsub_rmdir_args args = {
33673 +                       .errp   = &err,
33674 +                       .dir    = dir,
33675 +                       .path   = path
33676 +               };
33677 +               wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
33678 +               if (unlikely(wkq_err))
33679 +                       err = wkq_err;
33680 +       }
33681 +
33682 +       return err;
33683 +}
33684 +
33685 +/* ---------------------------------------------------------------------- */
33686 +
33687 +struct notify_change_args {
33688 +       int *errp;
33689 +       struct path *path;
33690 +       struct iattr *ia;
33691 +       struct inode **delegated_inode;
33692 +};
33693 +
33694 +static void call_notify_change(void *args)
33695 +{
33696 +       struct notify_change_args *a = args;
33697 +       struct inode *h_inode;
33698 +       struct user_namespace *userns;
33699 +
33700 +       h_inode = d_inode(a->path->dentry);
33701 +       IMustLock(h_inode);
33702 +
33703 +       *a->errp = -EPERM;
33704 +       if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
33705 +               userns = mnt_user_ns(a->path->mnt);
33706 +               lockdep_off();
33707 +               *a->errp = notify_change(userns, a->path->dentry, a->ia,
33708 +                                        a->delegated_inode);
33709 +               lockdep_on();
33710 +               if (!*a->errp)
33711 +                       vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
33712 +       }
33713 +       AuTraceErr(*a->errp);
33714 +}
33715 +
33716 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33717 +                       struct inode **delegated_inode)
33718 +{
33719 +       int err;
33720 +       struct notify_change_args args = {
33721 +               .errp                   = &err,
33722 +               .path                   = path,
33723 +               .ia                     = ia,
33724 +               .delegated_inode        = delegated_inode
33725 +       };
33726 +
33727 +       call_notify_change(&args);
33728 +
33729 +       return err;
33730 +}
33731 +
33732 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33733 +                           struct inode **delegated_inode)
33734 +{
33735 +       int err, wkq_err;
33736 +       struct notify_change_args args = {
33737 +               .errp                   = &err,
33738 +               .path                   = path,
33739 +               .ia                     = ia,
33740 +               .delegated_inode        = delegated_inode
33741 +       };
33742 +
33743 +       wkq_err = au_wkq_wait(call_notify_change, &args);
33744 +       if (unlikely(wkq_err))
33745 +               err = wkq_err;
33746 +
33747 +       return err;
33748 +}
33749 +
33750 +/* ---------------------------------------------------------------------- */
33751 +
33752 +struct unlink_args {
33753 +       int *errp;
33754 +       struct inode *dir;
33755 +       struct path *path;
33756 +       struct inode **delegated_inode;
33757 +};
33758 +
33759 +static void call_unlink(void *args)
33760 +{
33761 +       struct unlink_args *a = args;
33762 +       struct dentry *d = a->path->dentry;
33763 +       struct inode *h_inode;
33764 +       struct user_namespace *userns;
33765 +       const int stop_sillyrename = (au_test_nfs(d->d_sb)
33766 +                                     && au_dcount(d) == 1);
33767 +
33768 +       IMustLock(a->dir);
33769 +
33770 +       a->path->dentry = d->d_parent;
33771 +       *a->errp = security_path_unlink(a->path, d);
33772 +       a->path->dentry = d;
33773 +       if (unlikely(*a->errp))
33774 +               return;
33775 +
33776 +       if (!stop_sillyrename)
33777 +               dget(d);
33778 +       h_inode = NULL;
33779 +       if (d_is_positive(d)) {
33780 +               h_inode = d_inode(d);
33781 +               ihold(h_inode);
33782 +       }
33783 +
33784 +       userns = mnt_user_ns(a->path->mnt);
33785 +       lockdep_off();
33786 +       *a->errp = vfs_unlink(userns, a->dir, d, a->delegated_inode);
33787 +       lockdep_on();
33788 +       if (!*a->errp) {
33789 +               struct path tmp = {
33790 +                       .dentry = d->d_parent,
33791 +                       .mnt    = a->path->mnt
33792 +               };
33793 +               vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33794 +       }
33795 +
33796 +       if (!stop_sillyrename)
33797 +               dput(d);
33798 +       if (h_inode)
33799 +               iput(h_inode);
33800 +
33801 +       AuTraceErr(*a->errp);
33802 +}
33803 +
33804 +/*
33805 + * @dir: must be locked.
33806 + * @dentry: target dentry.
33807 + */
33808 +int vfsub_unlink(struct inode *dir, struct path *path,
33809 +                struct inode **delegated_inode, int force)
33810 +{
33811 +       int err;
33812 +       struct unlink_args args = {
33813 +               .errp                   = &err,
33814 +               .dir                    = dir,
33815 +               .path                   = path,
33816 +               .delegated_inode        = delegated_inode
33817 +       };
33818 +
33819 +       if (!force)
33820 +               call_unlink(&args);
33821 +       else {
33822 +               int wkq_err;
33823 +
33824 +               wkq_err = au_wkq_wait(call_unlink, &args);
33825 +               if (unlikely(wkq_err))
33826 +                       err = wkq_err;
33827 +       }
33828 +
33829 +       return err;
33830 +}
33831 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
33832 --- /usr/share/empty/fs/aufs/vfsub.h    1970-01-01 01:00:00.000000000 +0100
33833 +++ linux/fs/aufs/vfsub.h       2022-11-05 23:02:18.969222617 +0100
33834 @@ -0,0 +1,362 @@
33835 +/* SPDX-License-Identifier: GPL-2.0 */
33836 +/*
33837 + * Copyright (C) 2005-2022 Junjiro R. Okajima
33838 + *
33839 + * This program is free software; you can redistribute it and/or modify
33840 + * it under the terms of the GNU General Public License as published by
33841 + * the Free Software Foundation; either version 2 of the License, or
33842 + * (at your option) any later version.
33843 + *
33844 + * This program is distributed in the hope that it will be useful,
33845 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33846 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
33847 + * GNU General Public License for more details.
33848 + *
33849 + * You should have received a copy of the GNU General Public License
33850 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
33851 + */
33852 +
33853 +/*
33854 + * sub-routines for VFS
33855 + */
33856 +
33857 +#ifndef __AUFS_VFSUB_H__
33858 +#define __AUFS_VFSUB_H__
33859 +
33860 +#ifdef __KERNEL__
33861 +
33862 +#include <linux/fs.h>
33863 +#include <linux/mount.h>
33864 +#include <linux/posix_acl.h>
33865 +#include <linux/xattr.h>
33866 +#include "debug.h"
33867 +
33868 +/* copied from linux/fs/internal.h */
33869 +/* todo: BAD approach!! */
33870 +extern void __mnt_drop_write(struct vfsmount *);
33871 +extern struct file *alloc_empty_file(int, const struct cred *);
33872 +
33873 +/* ---------------------------------------------------------------------- */
33874 +
33875 +/* lock subclass for lower inode */
33876 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
33877 +/* reduce? gave up. */
33878 +enum {
33879 +       AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
33880 +       AuLsc_I_PARENT,         /* lower inode, parent first */
33881 +       AuLsc_I_PARENT2,        /* copyup dirs */
33882 +       AuLsc_I_PARENT3,        /* copyup wh */
33883 +       AuLsc_I_CHILD,
33884 +       AuLsc_I_CHILD2,
33885 +       AuLsc_I_End
33886 +};
33887 +
33888 +/* to debug easier, do not make them inlined functions */
33889 +#define MtxMustLock(mtx)       AuDebugOn(!mutex_is_locked(mtx))
33890 +#define IMustLock(i)           AuDebugOn(!inode_is_locked(i))
33891 +
33892 +/* ---------------------------------------------------------------------- */
33893 +
33894 +static inline void vfsub_drop_nlink(struct inode *inode)
33895 +{
33896 +       AuDebugOn(!inode->i_nlink);
33897 +       drop_nlink(inode);
33898 +}
33899 +
33900 +static inline void vfsub_dead_dir(struct inode *inode)
33901 +{
33902 +       AuDebugOn(!S_ISDIR(inode->i_mode));
33903 +       inode->i_flags |= S_DEAD;
33904 +       clear_nlink(inode);
33905 +}
33906 +
33907 +static inline int vfsub_native_ro(struct inode *inode)
33908 +{
33909 +       return sb_rdonly(inode->i_sb)
33910 +               || IS_RDONLY(inode)
33911 +               /* || IS_APPEND(inode) */
33912 +               || IS_IMMUTABLE(inode);
33913 +}
33914 +
33915 +#ifdef CONFIG_AUFS_BR_FUSE
33916 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb);
33917 +#else
33918 +AuStubInt0(vfsub_test_mntns, struct vfsmount *mnt, struct super_block *h_sb);
33919 +#endif
33920 +
33921 +int vfsub_sync_filesystem(struct super_block *h_sb);
33922 +
33923 +/* ---------------------------------------------------------------------- */
33924 +
33925 +int vfsub_update_h_iattr(struct path *h_path, int *did);
33926 +struct file *vfsub_dentry_open(struct path *path, int flags);
33927 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
33928 +struct au_branch;
33929 +struct vfsub_aopen_args {
33930 +       struct file             *file;
33931 +       unsigned int            open_flag;
33932 +       umode_t                 create_mode;
33933 +       struct au_branch        *br;
33934 +};
33935 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33936 +                     struct vfsub_aopen_args *args);
33937 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
33938 +
33939 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33940 +                                            struct path *ppath, int len);
33941 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33942 +                                   int len);
33943 +
33944 +struct vfsub_lkup_one_args {
33945 +       struct dentry **errp;
33946 +       struct qstr *name;
33947 +       struct path *ppath;
33948 +};
33949 +
33950 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
33951 +                                           struct path *ppath)
33952 +{
33953 +       return vfsub_lookup_one_len(name->name, ppath, name->len);
33954 +}
33955 +
33956 +void vfsub_call_lkup_one(void *args);
33957 +
33958 +/* ---------------------------------------------------------------------- */
33959 +
33960 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
33961 +{
33962 +       int err;
33963 +
33964 +       lockdep_off();
33965 +       err = mnt_want_write(mnt);
33966 +       lockdep_on();
33967 +       return err;
33968 +}
33969 +
33970 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
33971 +{
33972 +       lockdep_off();
33973 +       mnt_drop_write(mnt);
33974 +       lockdep_on();
33975 +}
33976 +
33977 +#if 0 /* reserved */
33978 +static inline void vfsub_mnt_drop_write_file(struct file *file)
33979 +{
33980 +       lockdep_off();
33981 +       mnt_drop_write_file(file);
33982 +       lockdep_on();
33983 +}
33984 +#endif
33985 +
33986 +/* ---------------------------------------------------------------------- */
33987 +
33988 +struct au_hinode;
33989 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33990 +                                struct dentry *d2, struct au_hinode *hdir2);
33991 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33992 +                        struct dentry *d2, struct au_hinode *hdir2);
33993 +
33994 +int vfsub_create(struct inode *dir, struct path *path, int mode,
33995 +                bool want_excl);
33996 +int vfsub_symlink(struct inode *dir, struct path *path,
33997 +                 const char *symname);
33998 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
33999 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
34000 +              struct path *path, struct inode **delegated_inode);
34001 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
34002 +                struct inode *hdir, struct path *path,
34003 +                struct inode **delegated_inode, unsigned int flags);
34004 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
34005 +int vfsub_rmdir(struct inode *dir, struct path *path);
34006 +
34007 +/* ---------------------------------------------------------------------- */
34008 +
34009 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
34010 +                    loff_t *ppos);
34011 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
34012 +                       loff_t *ppos);
34013 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
34014 +                     loff_t *ppos);
34015 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
34016 +                     loff_t *ppos);
34017 +int vfsub_flush(struct file *file, fl_owner_t id);
34018 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
34019 +
34020 +static inline loff_t vfsub_f_size_read(struct file *file)
34021 +{
34022 +       return i_size_read(file_inode(file));
34023 +}
34024 +
34025 +static inline unsigned int vfsub_file_flags(struct file *file)
34026 +{
34027 +       unsigned int flags;
34028 +
34029 +       spin_lock(&file->f_lock);
34030 +       flags = file->f_flags;
34031 +       spin_unlock(&file->f_lock);
34032 +
34033 +       return flags;
34034 +}
34035 +
34036 +static inline int vfsub_file_execed(struct file *file)
34037 +{
34038 +       /* todo: direct access f_flags */
34039 +       return !!(vfsub_file_flags(file) & __FMODE_EXEC);
34040 +}
34041 +
34042 +#if 0 /* reserved */
34043 +static inline void vfsub_file_accessed(struct file *h_file)
34044 +{
34045 +       file_accessed(h_file);
34046 +       vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
34047 +}
34048 +#endif
34049 +
34050 +#if 0 /* reserved */
34051 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
34052 +                                    struct dentry *h_dentry)
34053 +{
34054 +       struct path h_path = {
34055 +               .dentry = h_dentry,
34056 +               .mnt    = h_mnt
34057 +       };
34058 +       touch_atime(&h_path);
34059 +       vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
34060 +}
34061 +#endif
34062 +
34063 +static inline int vfsub_update_time(struct inode *h_inode,
34064 +                                   struct timespec64 *ts, int flags)
34065 +{
34066 +       return inode_update_time(h_inode, ts, flags);
34067 +       /* no vfsub_update_h_iattr() since we don't have struct path */
34068 +}
34069 +
34070 +#ifdef CONFIG_FS_POSIX_ACL
34071 +static inline int vfsub_acl_chmod(struct user_namespace *h_userns,
34072 +                                 struct inode *h_inode, umode_t h_mode)
34073 +{
34074 +       int err;
34075 +
34076 +       err = posix_acl_chmod(h_userns, h_inode, h_mode);
34077 +       if (err == -EOPNOTSUPP)
34078 +               err = 0;
34079 +       return err;
34080 +}
34081 +#else
34082 +AuStubInt0(vfsub_acl_chmod, struct user_namespace *h_userns,
34083 +          struct inode *h_inode, umode_t h_mode);
34084 +#endif
34085 +
34086 +long vfsub_splice_to(struct file *in, loff_t *ppos,
34087 +                    struct pipe_inode_info *pipe, size_t len,
34088 +                    unsigned int flags);
34089 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
34090 +                      loff_t *ppos, size_t len, unsigned int flags);
34091 +
34092 +static inline long vfsub_truncate(struct path *path, loff_t length)
34093 +{
34094 +       long err;
34095 +
34096 +       lockdep_off();
34097 +       err = vfs_truncate(path, length);
34098 +       lockdep_on();
34099 +       return err;
34100 +}
34101 +
34102 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
34103 +               struct file *h_file);
34104 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
34105 +
34106 +/*
34107 + * re-use branch fs's ioctl(FICLONE) while aufs itself doesn't support such
34108 + * ioctl.
34109 + */
34110 +static inline loff_t vfsub_clone_file_range(struct file *src, struct file *dst,
34111 +                                           loff_t len)
34112 +{
34113 +       loff_t err;
34114 +
34115 +       lockdep_off();
34116 +       err = vfs_clone_file_range(src, 0, dst, 0, len, /*remap_flags*/0);
34117 +       lockdep_on();
34118 +
34119 +       return err;
34120 +}
34121 +
34122 +/* copy_file_range(2) is a systemcall */
34123 +static inline ssize_t vfsub_copy_file_range(struct file *src, loff_t src_pos,
34124 +                                           struct file *dst, loff_t dst_pos,
34125 +                                           size_t len, unsigned int flags)
34126 +{
34127 +       ssize_t ssz;
34128 +
34129 +       lockdep_off();
34130 +       ssz = vfs_copy_file_range(src, src_pos, dst, dst_pos, len, flags);
34131 +       lockdep_on();
34132 +
34133 +       return ssz;
34134 +}
34135 +
34136 +/* ---------------------------------------------------------------------- */
34137 +
34138 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
34139 +{
34140 +       loff_t err;
34141 +
34142 +       lockdep_off();
34143 +       err = vfs_llseek(file, offset, origin);
34144 +       lockdep_on();
34145 +       return err;
34146 +}
34147 +
34148 +/* ---------------------------------------------------------------------- */
34149 +
34150 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
34151 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
34152 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
34153 +                           struct inode **delegated_inode);
34154 +int vfsub_notify_change(struct path *path, struct iattr *ia,
34155 +                       struct inode **delegated_inode);
34156 +int vfsub_unlink(struct inode *dir, struct path *path,
34157 +                struct inode **delegated_inode, int force);
34158 +
34159 +static inline int vfsub_getattr(const struct path *path, struct kstat *st)
34160 +{
34161 +       return vfs_getattr(path, st, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
34162 +}
34163 +
34164 +/* ---------------------------------------------------------------------- */
34165 +
34166 +/*
34167 + * The parameter 'value' is not const since the id in ACL_USER/ACL_GROUP may be
34168 + * changed by idmapping.
34169 + */
34170 +static inline int vfsub_setxattr(struct user_namespace *userns,
34171 +                                struct dentry *dentry, const char *name,
34172 +                                void *value, size_t size, int flags)
34173 +{
34174 +       int err;
34175 +
34176 +       lockdep_off();
34177 +       err = vfs_setxattr(userns, dentry, name, value, size, flags);
34178 +       lockdep_on();
34179 +
34180 +       return err;
34181 +}
34182 +
34183 +static inline int vfsub_removexattr(struct user_namespace *userns,
34184 +                                   struct dentry *dentry, const char *name)
34185 +{
34186 +       int err;
34187 +
34188 +       lockdep_off();
34189 +       err = vfs_removexattr(userns, dentry, name);
34190 +       lockdep_on();
34191 +
34192 +       return err;
34193 +}
34194 +
34195 +#endif /* __KERNEL__ */
34196 +#endif /* __AUFS_VFSUB_H__ */
34197 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
34198 --- /usr/share/empty/fs/aufs/wbr_policy.c       1970-01-01 01:00:00.000000000 +0100
34199 +++ linux/fs/aufs/wbr_policy.c  2022-11-05 23:02:18.969222617 +0100
34200 @@ -0,0 +1,830 @@
34201 +// SPDX-License-Identifier: GPL-2.0
34202 +/*
34203 + * Copyright (C) 2005-2022 Junjiro R. Okajima
34204 + *
34205 + * This program is free software; you can redistribute it and/or modify
34206 + * it under the terms of the GNU General Public License as published by
34207 + * the Free Software Foundation; either version 2 of the License, or
34208 + * (at your option) any later version.
34209 + *
34210 + * This program is distributed in the hope that it will be useful,
34211 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
34212 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
34213 + * GNU General Public License for more details.
34214 + *
34215 + * You should have received a copy of the GNU General Public License
34216 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
34217 + */
34218 +
34219 +/*
34220 + * policies for selecting one among multiple writable branches
34221 + */
34222 +
34223 +#include <linux/statfs.h>
34224 +#include "aufs.h"
34225 +
34226 +/* subset of cpup_attr() */
34227 +static noinline_for_stack
34228 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
34229 +{
34230 +       int err, sbits;
34231 +       struct iattr ia;
34232 +       struct inode *h_isrc;
34233 +
34234 +       h_isrc = d_inode(h_src);
34235 +       ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
34236 +       ia.ia_mode = h_isrc->i_mode;
34237 +       ia.ia_uid = h_isrc->i_uid;
34238 +       ia.ia_gid = h_isrc->i_gid;
34239 +       sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
34240 +       au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
34241 +       /* no delegation since it is just created */
34242 +       err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34243 +
34244 +       /* is this nfs only? */
34245 +       if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
34246 +               ia.ia_valid = ATTR_FORCE | ATTR_MODE;
34247 +               ia.ia_mode = h_isrc->i_mode;
34248 +               err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34249 +       }
34250 +
34251 +       return err;
34252 +}
34253 +
34254 +#define AuCpdown_PARENT_OPQ    1
34255 +#define AuCpdown_WHED          (1 << 1)
34256 +#define AuCpdown_MADE_DIR      (1 << 2)
34257 +#define AuCpdown_DIROPQ                (1 << 3)
34258 +#define au_ftest_cpdown(flags, name)   ((flags) & AuCpdown_##name)
34259 +#define au_fset_cpdown(flags, name) \
34260 +       do { (flags) |= AuCpdown_##name; } while (0)
34261 +#define au_fclr_cpdown(flags, name) \
34262 +       do { (flags) &= ~AuCpdown_##name; } while (0)
34263 +
34264 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
34265 +                            unsigned int *flags)
34266 +{
34267 +       int err;
34268 +       struct dentry *opq_dentry;
34269 +
34270 +       opq_dentry = au_diropq_create(dentry, bdst);
34271 +       err = PTR_ERR(opq_dentry);
34272 +       if (IS_ERR(opq_dentry))
34273 +               goto out;
34274 +       dput(opq_dentry);
34275 +       au_fset_cpdown(*flags, DIROPQ);
34276 +
34277 +out:
34278 +       return err;
34279 +}
34280 +
34281 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
34282 +                           struct inode *dir, aufs_bindex_t bdst)
34283 +{
34284 +       int err;
34285 +       struct path h_path;
34286 +       struct au_branch *br;
34287 +
34288 +       br = au_sbr(dentry->d_sb, bdst);
34289 +       h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
34290 +       err = PTR_ERR(h_path.dentry);
34291 +       if (IS_ERR(h_path.dentry))
34292 +               goto out;
34293 +
34294 +       err = 0;
34295 +       if (d_is_positive(h_path.dentry)) {
34296 +               h_path.mnt = au_br_mnt(br);
34297 +               err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
34298 +                                         dentry);
34299 +       }
34300 +       dput(h_path.dentry);
34301 +
34302 +out:
34303 +       return err;
34304 +}
34305 +
34306 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
34307 +                        struct au_pin *pin,
34308 +                        struct dentry *h_parent, void *arg)
34309 +{
34310 +       int err, rerr;
34311 +       aufs_bindex_t bopq, btop;
34312 +       struct path h_path;
34313 +       struct dentry *parent;
34314 +       struct inode *h_dir, *h_inode, *inode, *dir;
34315 +       unsigned int *flags = arg;
34316 +
34317 +       btop = au_dbtop(dentry);
34318 +       /* dentry is di-locked */
34319 +       parent = dget_parent(dentry);
34320 +       dir = d_inode(parent);
34321 +       h_dir = d_inode(h_parent);
34322 +       AuDebugOn(h_dir != au_h_iptr(dir, bdst));
34323 +       IMustLock(h_dir);
34324 +
34325 +       err = au_lkup_neg(dentry, bdst, /*wh*/0);
34326 +       if (unlikely(err < 0))
34327 +               goto out;
34328 +       h_path.dentry = au_h_dptr(dentry, bdst);
34329 +       h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
34330 +       err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path, 0755);
34331 +       if (unlikely(err))
34332 +               goto out_put;
34333 +       au_fset_cpdown(*flags, MADE_DIR);
34334 +
34335 +       bopq = au_dbdiropq(dentry);
34336 +       au_fclr_cpdown(*flags, WHED);
34337 +       au_fclr_cpdown(*flags, DIROPQ);
34338 +       if (au_dbwh(dentry) == bdst)
34339 +               au_fset_cpdown(*flags, WHED);
34340 +       if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
34341 +               au_fset_cpdown(*flags, PARENT_OPQ);
34342 +       h_inode = d_inode(h_path.dentry);
34343 +       inode_lock_nested(h_inode, AuLsc_I_CHILD);
34344 +       if (au_ftest_cpdown(*flags, WHED)) {
34345 +               err = au_cpdown_dir_opq(dentry, bdst, flags);
34346 +               if (unlikely(err)) {
34347 +                       inode_unlock(h_inode);
34348 +                       goto out_dir;
34349 +               }
34350 +       }
34351 +
34352 +       err = au_cpdown_attr(&h_path, au_h_dptr(dentry, btop));
34353 +       inode_unlock(h_inode);
34354 +       if (unlikely(err))
34355 +               goto out_opq;
34356 +
34357 +       if (au_ftest_cpdown(*flags, WHED)) {
34358 +               err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
34359 +               if (unlikely(err))
34360 +                       goto out_opq;
34361 +       }
34362 +
34363 +       inode = d_inode(dentry);
34364 +       if (au_ibbot(inode) < bdst)
34365 +               au_set_ibbot(inode, bdst);
34366 +       au_set_h_iptr(inode, bdst, au_igrab(h_inode),
34367 +                     au_hi_flags(inode, /*isdir*/1));
34368 +       au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
34369 +       goto out; /* success */
34370 +
34371 +       /* revert */
34372 +out_opq:
34373 +       if (au_ftest_cpdown(*flags, DIROPQ)) {
34374 +               inode_lock_nested(h_inode, AuLsc_I_CHILD);
34375 +               rerr = au_diropq_remove(dentry, bdst);
34376 +               inode_unlock(h_inode);
34377 +               if (unlikely(rerr)) {
34378 +                       AuIOErr("failed removing diropq for %pd b%d (%d)\n",
34379 +                               dentry, bdst, rerr);
34380 +                       err = -EIO;
34381 +                       goto out;
34382 +               }
34383 +       }
34384 +out_dir:
34385 +       if (au_ftest_cpdown(*flags, MADE_DIR)) {
34386 +               rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
34387 +               if (unlikely(rerr)) {
34388 +                       AuIOErr("failed removing %pd b%d (%d)\n",
34389 +                               dentry, bdst, rerr);
34390 +                       err = -EIO;
34391 +               }
34392 +       }
34393 +out_put:
34394 +       au_set_h_dptr(dentry, bdst, NULL);
34395 +       if (au_dbbot(dentry) == bdst)
34396 +               au_update_dbbot(dentry);
34397 +out:
34398 +       dput(parent);
34399 +       return err;
34400 +}
34401 +
34402 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
34403 +{
34404 +       int err;
34405 +       unsigned int flags;
34406 +
34407 +       flags = 0;
34408 +       err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
34409 +
34410 +       return err;
34411 +}
34412 +
34413 +/* ---------------------------------------------------------------------- */
34414 +
34415 +/* policies for create */
34416 +
34417 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
34418 +{
34419 +       int err, i, j, ndentry;
34420 +       aufs_bindex_t bopq;
34421 +       struct au_dcsub_pages dpages;
34422 +       struct au_dpage *dpage;
34423 +       struct dentry **dentries, *parent, *d;
34424 +
34425 +       err = au_dpages_init(&dpages, GFP_NOFS);
34426 +       if (unlikely(err))
34427 +               goto out;
34428 +       parent = dget_parent(dentry);
34429 +       err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
34430 +       if (unlikely(err))
34431 +               goto out_free;
34432 +
34433 +       err = bindex;
34434 +       for (i = 0; i < dpages.ndpage; i++) {
34435 +               dpage = dpages.dpages + i;
34436 +               dentries = dpage->dentries;
34437 +               ndentry = dpage->ndentry;
34438 +               for (j = 0; j < ndentry; j++) {
34439 +                       d = dentries[j];
34440 +                       di_read_lock_parent2(d, !AuLock_IR);
34441 +                       bopq = au_dbdiropq(d);
34442 +                       di_read_unlock(d, !AuLock_IR);
34443 +                       if (bopq >= 0 && bopq < err)
34444 +                               err = bopq;
34445 +               }
34446 +       }
34447 +
34448 +out_free:
34449 +       dput(parent);
34450 +       au_dpages_free(&dpages);
34451 +out:
34452 +       return err;
34453 +}
34454 +
34455 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
34456 +{
34457 +       for (; bindex >= 0; bindex--)
34458 +               if (!au_br_rdonly(au_sbr(sb, bindex)))
34459 +                       return bindex;
34460 +       return -EROFS;
34461 +}
34462 +
34463 +/* top down parent */
34464 +static int au_wbr_create_tdp(struct dentry *dentry,
34465 +                            unsigned int flags __maybe_unused)
34466 +{
34467 +       int err;
34468 +       aufs_bindex_t btop, bindex;
34469 +       struct super_block *sb;
34470 +       struct dentry *parent, *h_parent;
34471 +
34472 +       sb = dentry->d_sb;
34473 +       btop = au_dbtop(dentry);
34474 +       err = btop;
34475 +       if (!au_br_rdonly(au_sbr(sb, btop)))
34476 +               goto out;
34477 +
34478 +       err = -EROFS;
34479 +       parent = dget_parent(dentry);
34480 +       for (bindex = au_dbtop(parent); bindex < btop; bindex++) {
34481 +               h_parent = au_h_dptr(parent, bindex);
34482 +               if (!h_parent || d_is_negative(h_parent))
34483 +                       continue;
34484 +
34485 +               if (!au_br_rdonly(au_sbr(sb, bindex))) {
34486 +                       err = bindex;
34487 +                       break;
34488 +               }
34489 +       }
34490 +       dput(parent);
34491 +
34492 +       /* bottom up here */
34493 +       if (unlikely(err < 0)) {
34494 +               err = au_wbr_bu(sb, btop - 1);
34495 +               if (err >= 0)
34496 +                       err = au_wbr_nonopq(dentry, err);
34497 +       }
34498 +
34499 +out:
34500 +       AuDbg("b%d\n", err);
34501 +       return err;
34502 +}
34503 +
34504 +/* ---------------------------------------------------------------------- */
34505 +
34506 +/* an exception for the policy other than tdp */
34507 +static int au_wbr_create_exp(struct dentry *dentry)
34508 +{
34509 +       int err;
34510 +       aufs_bindex_t bwh, bdiropq;
34511 +       struct dentry *parent;
34512 +
34513 +       err = -1;
34514 +       bwh = au_dbwh(dentry);
34515 +       parent = dget_parent(dentry);
34516 +       bdiropq = au_dbdiropq(parent);
34517 +       if (bwh >= 0) {
34518 +               if (bdiropq >= 0)
34519 +                       err = min(bdiropq, bwh);
34520 +               else
34521 +                       err = bwh;
34522 +               AuDbg("%d\n", err);
34523 +       } else if (bdiropq >= 0) {
34524 +               err = bdiropq;
34525 +               AuDbg("%d\n", err);
34526 +       }
34527 +       dput(parent);
34528 +
34529 +       if (err >= 0)
34530 +               err = au_wbr_nonopq(dentry, err);
34531 +
34532 +       if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
34533 +               err = -1;
34534 +
34535 +       AuDbg("%d\n", err);
34536 +       return err;
34537 +}
34538 +
34539 +/* ---------------------------------------------------------------------- */
34540 +
34541 +/* round robin */
34542 +static int au_wbr_create_init_rr(struct super_block *sb)
34543 +{
34544 +       int err;
34545 +
34546 +       err = au_wbr_bu(sb, au_sbbot(sb));
34547 +       atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
34548 +       /* smp_mb(); */
34549 +
34550 +       AuDbg("b%d\n", err);
34551 +       return err;
34552 +}
34553 +
34554 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
34555 +{
34556 +       int err, nbr;
34557 +       unsigned int u;
34558 +       aufs_bindex_t bindex, bbot;
34559 +       struct super_block *sb;
34560 +       atomic_t *next;
34561 +
34562 +       err = au_wbr_create_exp(dentry);
34563 +       if (err >= 0)
34564 +               goto out;
34565 +
34566 +       sb = dentry->d_sb;
34567 +       next = &au_sbi(sb)->si_wbr_rr_next;
34568 +       bbot = au_sbbot(sb);
34569 +       nbr = bbot + 1;
34570 +       for (bindex = 0; bindex <= bbot; bindex++) {
34571 +               if (!au_ftest_wbr(flags, DIR)) {
34572 +                       err = atomic_dec_return(next) + 1;
34573 +                       /* modulo for 0 is meaningless */
34574 +                       if (unlikely(!err))
34575 +                               err = atomic_dec_return(next) + 1;
34576 +               } else
34577 +                       err = atomic_read(next);
34578 +               AuDbg("%d\n", err);
34579 +               u = err;
34580 +               err = u % nbr;
34581 +               AuDbg("%d\n", err);
34582 +               if (!au_br_rdonly(au_sbr(sb, err)))
34583 +                       break;
34584 +               err = -EROFS;
34585 +       }
34586 +
34587 +       if (err >= 0)
34588 +               err = au_wbr_nonopq(dentry, err);
34589 +
34590 +out:
34591 +       AuDbg("%d\n", err);
34592 +       return err;
34593 +}
34594 +
34595 +/* ---------------------------------------------------------------------- */
34596 +
34597 +/* most free space */
34598 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
34599 +{
34600 +       struct super_block *sb;
34601 +       struct au_branch *br;
34602 +       struct au_wbr_mfs *mfs;
34603 +       struct dentry *h_parent;
34604 +       aufs_bindex_t bindex, bbot;
34605 +       int err;
34606 +       unsigned long long b, bavail;
34607 +       struct path h_path;
34608 +       /* reduce the stack usage */
34609 +       struct kstatfs *st;
34610 +
34611 +       st = kmalloc(sizeof(*st), GFP_NOFS);
34612 +       if (unlikely(!st)) {
34613 +               AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
34614 +               return;
34615 +       }
34616 +
34617 +       bavail = 0;
34618 +       sb = dentry->d_sb;
34619 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34620 +       MtxMustLock(&mfs->mfs_lock);
34621 +       mfs->mfs_bindex = -EROFS;
34622 +       mfs->mfsrr_bytes = 0;
34623 +       if (!parent) {
34624 +               bindex = 0;
34625 +               bbot = au_sbbot(sb);
34626 +       } else {
34627 +               bindex = au_dbtop(parent);
34628 +               bbot = au_dbtaildir(parent);
34629 +       }
34630 +
34631 +       for (; bindex <= bbot; bindex++) {
34632 +               if (parent) {
34633 +                       h_parent = au_h_dptr(parent, bindex);
34634 +                       if (!h_parent || d_is_negative(h_parent))
34635 +                               continue;
34636 +               }
34637 +               br = au_sbr(sb, bindex);
34638 +               if (au_br_rdonly(br))
34639 +                       continue;
34640 +
34641 +               /* sb->s_root for NFS is unreliable */
34642 +               h_path.mnt = au_br_mnt(br);
34643 +               h_path.dentry = h_path.mnt->mnt_root;
34644 +               err = vfs_statfs(&h_path, st);
34645 +               if (unlikely(err)) {
34646 +                       AuWarn1("failed statfs, b%d, %d\n", bindex, err);
34647 +                       continue;
34648 +               }
34649 +
34650 +               /* when the available size is equal, select the lower one */
34651 +               BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
34652 +                            || sizeof(b) < sizeof(st->f_bsize));
34653 +               b = st->f_bavail * st->f_bsize;
34654 +               br->br_wbr->wbr_bytes = b;
34655 +               if (b >= bavail) {
34656 +                       bavail = b;
34657 +                       mfs->mfs_bindex = bindex;
34658 +                       mfs->mfs_jiffy = jiffies;
34659 +               }
34660 +       }
34661 +
34662 +       mfs->mfsrr_bytes = bavail;
34663 +       AuDbg("b%d\n", mfs->mfs_bindex);
34664 +       au_kfree_rcu(st);
34665 +}
34666 +
34667 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
34668 +{
34669 +       int err;
34670 +       struct dentry *parent;
34671 +       struct super_block *sb;
34672 +       struct au_wbr_mfs *mfs;
34673 +
34674 +       err = au_wbr_create_exp(dentry);
34675 +       if (err >= 0)
34676 +               goto out;
34677 +
34678 +       sb = dentry->d_sb;
34679 +       parent = NULL;
34680 +       if (au_ftest_wbr(flags, PARENT))
34681 +               parent = dget_parent(dentry);
34682 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34683 +       mutex_lock(&mfs->mfs_lock);
34684 +       if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34685 +           || mfs->mfs_bindex < 0
34686 +           || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
34687 +               au_mfs(dentry, parent);
34688 +       mutex_unlock(&mfs->mfs_lock);
34689 +       err = mfs->mfs_bindex;
34690 +       dput(parent);
34691 +
34692 +       if (err >= 0)
34693 +               err = au_wbr_nonopq(dentry, err);
34694 +
34695 +out:
34696 +       AuDbg("b%d\n", err);
34697 +       return err;
34698 +}
34699 +
34700 +static int au_wbr_create_init_mfs(struct super_block *sb)
34701 +{
34702 +       struct au_wbr_mfs *mfs;
34703 +
34704 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34705 +       mutex_init(&mfs->mfs_lock);
34706 +       mfs->mfs_jiffy = 0;
34707 +       mfs->mfs_bindex = -EROFS;
34708 +
34709 +       return 0;
34710 +}
34711 +
34712 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
34713 +{
34714 +       mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
34715 +       return 0;
34716 +}
34717 +
34718 +/* ---------------------------------------------------------------------- */
34719 +
34720 +/* top down regardless parent, and then mfs */
34721 +static int au_wbr_create_tdmfs(struct dentry *dentry,
34722 +                              unsigned int flags __maybe_unused)
34723 +{
34724 +       int err;
34725 +       aufs_bindex_t bwh, btail, bindex, bfound, bmfs;
34726 +       unsigned long long watermark;
34727 +       struct super_block *sb;
34728 +       struct au_wbr_mfs *mfs;
34729 +       struct au_branch *br;
34730 +       struct dentry *parent;
34731 +
34732 +       sb = dentry->d_sb;
34733 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34734 +       mutex_lock(&mfs->mfs_lock);
34735 +       if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34736 +           || mfs->mfs_bindex < 0)
34737 +               au_mfs(dentry, /*parent*/NULL);
34738 +       watermark = mfs->mfsrr_watermark;
34739 +       bmfs = mfs->mfs_bindex;
34740 +       mutex_unlock(&mfs->mfs_lock);
34741 +
34742 +       /* another style of au_wbr_create_exp() */
34743 +       bwh = au_dbwh(dentry);
34744 +       parent = dget_parent(dentry);
34745 +       btail = au_dbtaildir(parent);
34746 +       if (bwh >= 0 && bwh < btail)
34747 +               btail = bwh;
34748 +
34749 +       err = au_wbr_nonopq(dentry, btail);
34750 +       if (unlikely(err < 0))
34751 +               goto out;
34752 +       btail = err;
34753 +       bfound = -1;
34754 +       for (bindex = 0; bindex <= btail; bindex++) {
34755 +               br = au_sbr(sb, bindex);
34756 +               if (au_br_rdonly(br))
34757 +                       continue;
34758 +               if (br->br_wbr->wbr_bytes > watermark) {
34759 +                       bfound = bindex;
34760 +                       break;
34761 +               }
34762 +       }
34763 +       err = bfound;
34764 +       if (err < 0)
34765 +               err = bmfs;
34766 +
34767 +out:
34768 +       dput(parent);
34769 +       AuDbg("b%d\n", err);
34770 +       return err;
34771 +}
34772 +
34773 +/* ---------------------------------------------------------------------- */
34774 +
34775 +/* most free space and then round robin */
34776 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
34777 +{
34778 +       int err;
34779 +       struct au_wbr_mfs *mfs;
34780 +
34781 +       err = au_wbr_create_mfs(dentry, flags);
34782 +       if (err >= 0) {
34783 +               mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
34784 +               mutex_lock(&mfs->mfs_lock);
34785 +               if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
34786 +                       err = au_wbr_create_rr(dentry, flags);
34787 +               mutex_unlock(&mfs->mfs_lock);
34788 +       }
34789 +
34790 +       AuDbg("b%d\n", err);
34791 +       return err;
34792 +}
34793 +
34794 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
34795 +{
34796 +       int err;
34797 +
34798 +       au_wbr_create_init_mfs(sb); /* ignore */
34799 +       err = au_wbr_create_init_rr(sb);
34800 +
34801 +       return err;
34802 +}
34803 +
34804 +/* ---------------------------------------------------------------------- */
34805 +
34806 +/* top down parent and most free space */
34807 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
34808 +{
34809 +       int err, e2;
34810 +       unsigned long long b;
34811 +       aufs_bindex_t bindex, btop, bbot;
34812 +       struct super_block *sb;
34813 +       struct dentry *parent, *h_parent;
34814 +       struct au_branch *br;
34815 +
34816 +       err = au_wbr_create_tdp(dentry, flags);
34817 +       if (unlikely(err < 0))
34818 +               goto out;
34819 +       parent = dget_parent(dentry);
34820 +       btop = au_dbtop(parent);
34821 +       bbot = au_dbtaildir(parent);
34822 +       if (btop == bbot)
34823 +               goto out_parent; /* success */
34824 +
34825 +       e2 = au_wbr_create_mfs(dentry, flags);
34826 +       if (e2 < 0)
34827 +               goto out_parent; /* success */
34828 +
34829 +       /* when the available size is equal, select upper one */
34830 +       sb = dentry->d_sb;
34831 +       br = au_sbr(sb, err);
34832 +       b = br->br_wbr->wbr_bytes;
34833 +       AuDbg("b%d, %llu\n", err, b);
34834 +
34835 +       for (bindex = btop; bindex <= bbot; bindex++) {
34836 +               h_parent = au_h_dptr(parent, bindex);
34837 +               if (!h_parent || d_is_negative(h_parent))
34838 +                       continue;
34839 +
34840 +               br = au_sbr(sb, bindex);
34841 +               if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
34842 +                       b = br->br_wbr->wbr_bytes;
34843 +                       err = bindex;
34844 +                       AuDbg("b%d, %llu\n", err, b);
34845 +               }
34846 +       }
34847 +
34848 +       if (err >= 0)
34849 +               err = au_wbr_nonopq(dentry, err);
34850 +
34851 +out_parent:
34852 +       dput(parent);
34853 +out:
34854 +       AuDbg("b%d\n", err);
34855 +       return err;
34856 +}
34857 +
34858 +/* ---------------------------------------------------------------------- */
34859 +
34860 +/*
34861 + * - top down parent
34862 + * - most free space with parent
34863 + * - most free space round-robin regardless parent
34864 + */
34865 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
34866 +{
34867 +       int err;
34868 +       unsigned long long watermark;
34869 +       struct super_block *sb;
34870 +       struct au_branch *br;
34871 +       struct au_wbr_mfs *mfs;
34872 +
34873 +       err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
34874 +       if (unlikely(err < 0))
34875 +               goto out;
34876 +
34877 +       sb = dentry->d_sb;
34878 +       br = au_sbr(sb, err);
34879 +       mfs = &au_sbi(sb)->si_wbr_mfs;
34880 +       mutex_lock(&mfs->mfs_lock);
34881 +       watermark = mfs->mfsrr_watermark;
34882 +       mutex_unlock(&mfs->mfs_lock);
34883 +       if (br->br_wbr->wbr_bytes < watermark)
34884 +               /* regardless the parent dir */
34885 +               err = au_wbr_create_mfsrr(dentry, flags);
34886 +
34887 +out:
34888 +       AuDbg("b%d\n", err);
34889 +       return err;
34890 +}
34891 +
34892 +/* ---------------------------------------------------------------------- */
34893 +
34894 +/* policies for copyup */
34895 +
34896 +/* top down parent */
34897 +static int au_wbr_copyup_tdp(struct dentry *dentry)
34898 +{
34899 +       return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
34900 +}
34901 +
34902 +/* bottom up parent */
34903 +static int au_wbr_copyup_bup(struct dentry *dentry)
34904 +{
34905 +       int err;
34906 +       aufs_bindex_t bindex, btop;
34907 +       struct dentry *parent, *h_parent;
34908 +       struct super_block *sb;
34909 +
34910 +       err = -EROFS;
34911 +       sb = dentry->d_sb;
34912 +       parent = dget_parent(dentry);
34913 +       btop = au_dbtop(parent);
34914 +       for (bindex = au_dbtop(dentry); bindex >= btop; bindex--) {
34915 +               h_parent = au_h_dptr(parent, bindex);
34916 +               if (!h_parent || d_is_negative(h_parent))
34917 +                       continue;
34918 +
34919 +               if (!au_br_rdonly(au_sbr(sb, bindex))) {
34920 +                       err = bindex;
34921 +                       break;
34922 +               }
34923 +       }
34924 +       dput(parent);
34925 +
34926 +       /* bottom up here */
34927 +       if (unlikely(err < 0))
34928 +               err = au_wbr_bu(sb, btop - 1);
34929 +
34930 +       AuDbg("b%d\n", err);
34931 +       return err;
34932 +}
34933 +
34934 +/* bottom up */
34935 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop)
34936 +{
34937 +       int err;
34938 +
34939 +       err = au_wbr_bu(dentry->d_sb, btop);
34940 +       AuDbg("b%d\n", err);
34941 +       if (err > btop)
34942 +               err = au_wbr_nonopq(dentry, err);
34943 +
34944 +       AuDbg("b%d\n", err);
34945 +       return err;
34946 +}
34947 +
34948 +static int au_wbr_copyup_bu(struct dentry *dentry)
34949 +{
34950 +       int err;
34951 +       aufs_bindex_t btop;
34952 +
34953 +       btop = au_dbtop(dentry);
34954 +       err = au_wbr_do_copyup_bu(dentry, btop);
34955 +       return err;
34956 +}
34957 +
34958 +/* ---------------------------------------------------------------------- */
34959 +
34960 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
34961 +       [AuWbrCopyup_TDP] = {
34962 +               .copyup = au_wbr_copyup_tdp
34963 +       },
34964 +       [AuWbrCopyup_BUP] = {
34965 +               .copyup = au_wbr_copyup_bup
34966 +       },
34967 +       [AuWbrCopyup_BU] = {
34968 +               .copyup = au_wbr_copyup_bu
34969 +       }
34970 +};
34971 +
34972 +struct au_wbr_create_operations au_wbr_create_ops[] = {
34973 +       [AuWbrCreate_TDP] = {
34974 +               .create = au_wbr_create_tdp
34975 +       },
34976 +       [AuWbrCreate_RR] = {
34977 +               .create = au_wbr_create_rr,
34978 +               .init   = au_wbr_create_init_rr
34979 +       },
34980 +       [AuWbrCreate_MFS] = {
34981 +               .create = au_wbr_create_mfs,
34982 +               .init   = au_wbr_create_init_mfs,
34983 +               .fin    = au_wbr_create_fin_mfs
34984 +       },
34985 +       [AuWbrCreate_MFSV] = {
34986 +               .create = au_wbr_create_mfs,
34987 +               .init   = au_wbr_create_init_mfs,
34988 +               .fin    = au_wbr_create_fin_mfs
34989 +       },
34990 +       [AuWbrCreate_MFSRR] = {
34991 +               .create = au_wbr_create_mfsrr,
34992 +               .init   = au_wbr_create_init_mfsrr,
34993 +               .fin    = au_wbr_create_fin_mfs
34994 +       },
34995 +       [AuWbrCreate_MFSRRV] = {
34996 +               .create = au_wbr_create_mfsrr,
34997 +               .init   = au_wbr_create_init_mfsrr,
34998 +               .fin    = au_wbr_create_fin_mfs
34999 +       },
35000 +       [AuWbrCreate_TDMFS] = {
35001 +               .create = au_wbr_create_tdmfs,
35002 +               .init   = au_wbr_create_init_mfs,
35003 +               .fin    = au_wbr_create_fin_mfs
35004 +       },
35005 +       [AuWbrCreate_TDMFSV] = {
35006 +               .create = au_wbr_create_tdmfs,
35007 +               .init   = au_wbr_create_init_mfs,
35008 +               .fin    = au_wbr_create_fin_mfs
35009 +       },
35010 +       [AuWbrCreate_PMFS] = {
35011 +               .create = au_wbr_create_pmfs,
35012 +               .init   = au_wbr_create_init_mfs,
35013 +               .fin    = au_wbr_create_fin_mfs
35014 +       },
35015 +       [AuWbrCreate_PMFSV] = {
35016 +               .create = au_wbr_create_pmfs,
35017 +               .init   = au_wbr_create_init_mfs,
35018 +               .fin    = au_wbr_create_fin_mfs
35019 +       },
35020 +       [AuWbrCreate_PMFSRR] = {
35021 +               .create = au_wbr_create_pmfsrr,
35022 +               .init   = au_wbr_create_init_mfsrr,
35023 +               .fin    = au_wbr_create_fin_mfs
35024 +       },
35025 +       [AuWbrCreate_PMFSRRV] = {
35026 +               .create = au_wbr_create_pmfsrr,
35027 +               .init   = au_wbr_create_init_mfsrr,
35028 +               .fin    = au_wbr_create_fin_mfs
35029 +       }
35030 +};
35031 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
35032 --- /usr/share/empty/fs/aufs/whout.c    1970-01-01 01:00:00.000000000 +0100
35033 +++ linux/fs/aufs/whout.c       2022-11-05 23:02:18.972555950 +0100
35034 @@ -0,0 +1,1072 @@
35035 +// SPDX-License-Identifier: GPL-2.0
35036 +/*
35037 + * Copyright (C) 2005-2022 Junjiro R. Okajima
35038 + *
35039 + * This program is free software; you can redistribute it and/or modify
35040 + * it under the terms of the GNU General Public License as published by
35041 + * the Free Software Foundation; either version 2 of the License, or
35042 + * (at your option) any later version.
35043 + *
35044 + * This program is distributed in the hope that it will be useful,
35045 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35046 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
35047 + * GNU General Public License for more details.
35048 + *
35049 + * You should have received a copy of the GNU General Public License
35050 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
35051 + */
35052 +
35053 +/*
35054 + * whiteout for logical deletion and opaque directory
35055 + */
35056 +
35057 +#include "aufs.h"
35058 +
35059 +#define WH_MASK                        0444
35060 +
35061 +/*
35062 + * If a directory contains this file, then it is opaque.  We start with the
35063 + * .wh. flag so that it is blocked by lookup.
35064 + */
35065 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
35066 +                                          sizeof(AUFS_WH_DIROPQ) - 1);
35067 +
35068 +/*
35069 + * generate whiteout name, which is NOT terminated by NULL.
35070 + * @name: original d_name.name
35071 + * @len: original d_name.len
35072 + * @wh: whiteout qstr
35073 + * returns zero when succeeds, otherwise error.
35074 + * succeeded value as wh->name should be freed by kfree().
35075 + */
35076 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
35077 +{
35078 +       char *p;
35079 +
35080 +       if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
35081 +               return -ENAMETOOLONG;
35082 +
35083 +       wh->len = name->len + AUFS_WH_PFX_LEN;
35084 +       p = kmalloc(wh->len, GFP_NOFS);
35085 +       wh->name = p;
35086 +       if (p) {
35087 +               memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35088 +               memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
35089 +               /* smp_mb(); */
35090 +               return 0;
35091 +       }
35092 +       return -ENOMEM;
35093 +}
35094 +
35095 +/* ---------------------------------------------------------------------- */
35096 +
35097 +/*
35098 + * test if the @wh_name exists under @h_ppath.
35099 + * @try_sio specifies the necessary of super-io.
35100 + */
35101 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
35102 +              struct qstr *wh_name, int try_sio)
35103 +{
35104 +       int err;
35105 +       struct dentry *wh_dentry;
35106 +
35107 +       if (!try_sio)
35108 +               wh_dentry = vfsub_lkup_one(wh_name, h_ppath);
35109 +       else
35110 +               wh_dentry = au_sio_lkup_one(h_userns, wh_name, h_ppath);
35111 +       err = PTR_ERR(wh_dentry);
35112 +       if (IS_ERR(wh_dentry)) {
35113 +               if (err == -ENAMETOOLONG)
35114 +                       err = 0;
35115 +               goto out;
35116 +       }
35117 +
35118 +       err = 0;
35119 +       if (d_is_negative(wh_dentry))
35120 +               goto out_wh; /* success */
35121 +
35122 +       err = 1;
35123 +       if (d_is_reg(wh_dentry))
35124 +               goto out_wh; /* success */
35125 +
35126 +       err = -EIO;
35127 +       AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
35128 +               wh_dentry, d_inode(wh_dentry)->i_mode);
35129 +
35130 +out_wh:
35131 +       dput(wh_dentry);
35132 +out:
35133 +       return err;
35134 +}
35135 +
35136 +/*
35137 + * test if the @h_path->dentry sets opaque or not.
35138 + */
35139 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path)
35140 +{
35141 +       int err;
35142 +       struct inode *h_dir;
35143 +
35144 +       h_dir = d_inode(h_path->dentry);
35145 +       err = au_wh_test(h_userns, h_path, &diropq_name,
35146 +                        au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC));
35147 +       return err;
35148 +}
35149 +
35150 +/*
35151 + * returns a negative dentry whose name is unique and temporary.
35152 + */
35153 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
35154 +                            struct qstr *prefix)
35155 +{
35156 +       struct dentry *dentry;
35157 +       int i;
35158 +       char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
35159 +               *name, *p;
35160 +       /* strict atomic_t is unnecessary here */
35161 +       static unsigned short cnt;
35162 +       struct qstr qs;
35163 +       struct path h_ppath;
35164 +       struct user_namespace *h_userns;
35165 +
35166 +       BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
35167 +
35168 +       name = defname;
35169 +       qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
35170 +       if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
35171 +               dentry = ERR_PTR(-ENAMETOOLONG);
35172 +               if (unlikely(qs.len > NAME_MAX))
35173 +                       goto out;
35174 +               dentry = ERR_PTR(-ENOMEM);
35175 +               name = kmalloc(qs.len + 1, GFP_NOFS);
35176 +               if (unlikely(!name))
35177 +                       goto out;
35178 +       }
35179 +
35180 +       /* doubly whiteout-ed */
35181 +       memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
35182 +       p = name + AUFS_WH_PFX_LEN * 2;
35183 +       memcpy(p, prefix->name, prefix->len);
35184 +       p += prefix->len;
35185 +       *p++ = '.';
35186 +       AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
35187 +
35188 +       h_ppath.dentry = h_parent;
35189 +       h_ppath.mnt = au_br_mnt(br);
35190 +       h_userns = au_br_userns(br);
35191 +       qs.name = name;
35192 +       for (i = 0; i < 3; i++) {
35193 +               sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
35194 +               dentry = au_sio_lkup_one(h_userns, &qs, &h_ppath);
35195 +               if (IS_ERR(dentry) || d_is_negative(dentry))
35196 +                       goto out_name;
35197 +               dput(dentry);
35198 +       }
35199 +       /* pr_warn("could not get random name\n"); */
35200 +       dentry = ERR_PTR(-EEXIST);
35201 +       AuDbg("%.*s\n", AuLNPair(&qs));
35202 +       BUG();
35203 +
35204 +out_name:
35205 +       if (name != defname)
35206 +               au_kfree_try_rcu(name);
35207 +out:
35208 +       AuTraceErrPtr(dentry);
35209 +       return dentry;
35210 +}
35211 +
35212 +/*
35213 + * rename the @h_dentry on @br to the whiteouted temporary name.
35214 + */
35215 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
35216 +{
35217 +       int err;
35218 +       struct path h_path = {
35219 +               .mnt = au_br_mnt(br)
35220 +       };
35221 +       struct inode *h_dir, *delegated;
35222 +       struct dentry *h_parent;
35223 +
35224 +       h_parent = h_dentry->d_parent; /* dir inode is locked */
35225 +       h_dir = d_inode(h_parent);
35226 +       IMustLock(h_dir);
35227 +
35228 +       h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
35229 +       err = PTR_ERR(h_path.dentry);
35230 +       if (IS_ERR(h_path.dentry))
35231 +               goto out;
35232 +
35233 +       /* under the same dir, no need to lock_rename() */
35234 +       delegated = NULL;
35235 +       err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated,
35236 +                          /*flags*/0);
35237 +       AuTraceErr(err);
35238 +       if (unlikely(err == -EWOULDBLOCK)) {
35239 +               pr_warn("cannot retry for NFSv4 delegation"
35240 +                       " for an internal rename\n");
35241 +               iput(delegated);
35242 +       }
35243 +       dput(h_path.dentry);
35244 +
35245 +out:
35246 +       AuTraceErr(err);
35247 +       return err;
35248 +}
35249 +
35250 +/* ---------------------------------------------------------------------- */
35251 +/*
35252 + * functions for removing a whiteout
35253 + */
35254 +
35255 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
35256 +{
35257 +       int err, force;
35258 +       struct inode *delegated;
35259 +
35260 +       /*
35261 +        * forces superio when the dir has a sticky bit.
35262 +        * this may be a violation of unix fs semantics.
35263 +        */
35264 +       force = (h_dir->i_mode & S_ISVTX)
35265 +               && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
35266 +       delegated = NULL;
35267 +       err = vfsub_unlink(h_dir, h_path, &delegated, force);
35268 +       if (unlikely(err == -EWOULDBLOCK)) {
35269 +               pr_warn("cannot retry for NFSv4 delegation"
35270 +                       " for an internal unlink\n");
35271 +               iput(delegated);
35272 +       }
35273 +       return err;
35274 +}
35275 +
35276 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
35277 +                       struct dentry *dentry)
35278 +{
35279 +       int err;
35280 +
35281 +       err = do_unlink_wh(h_dir, h_path);
35282 +       if (!err && dentry)
35283 +               au_set_dbwh(dentry, -1);
35284 +
35285 +       return err;
35286 +}
35287 +
35288 +static int unlink_wh_name(struct path *h_ppath, struct qstr *wh)
35289 +{
35290 +       int err;
35291 +       struct path h_path;
35292 +
35293 +       err = 0;
35294 +       h_path.dentry = vfsub_lkup_one(wh, h_ppath);
35295 +       if (IS_ERR(h_path.dentry))
35296 +               err = PTR_ERR(h_path.dentry);
35297 +       else {
35298 +               if (d_is_reg(h_path.dentry)) {
35299 +                       h_path.mnt = h_ppath->mnt;
35300 +                       err = do_unlink_wh(d_inode(h_ppath->dentry), &h_path);
35301 +               }
35302 +               dput(h_path.dentry);
35303 +       }
35304 +
35305 +       return err;
35306 +}
35307 +
35308 +/* ---------------------------------------------------------------------- */
35309 +/*
35310 + * initialize/clean whiteout for a branch
35311 + */
35312 +
35313 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
35314 +                       const int isdir)
35315 +{
35316 +       int err;
35317 +       struct inode *delegated;
35318 +
35319 +       if (d_is_negative(whpath->dentry))
35320 +               return;
35321 +
35322 +       if (isdir)
35323 +               err = vfsub_rmdir(h_dir, whpath);
35324 +       else {
35325 +               delegated = NULL;
35326 +               err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
35327 +               if (unlikely(err == -EWOULDBLOCK)) {
35328 +                       pr_warn("cannot retry for NFSv4 delegation"
35329 +                               " for an internal unlink\n");
35330 +                       iput(delegated);
35331 +               }
35332 +       }
35333 +       if (unlikely(err))
35334 +               pr_warn("failed removing %pd (%d), ignored.\n",
35335 +                       whpath->dentry, err);
35336 +}
35337 +
35338 +static int test_linkable(struct dentry *h_root)
35339 +{
35340 +       struct inode *h_dir = d_inode(h_root);
35341 +
35342 +       if (h_dir->i_op->link)
35343 +               return 0;
35344 +
35345 +       pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
35346 +              h_root, au_sbtype(h_root->d_sb));
35347 +       return -ENOSYS; /* the branch doesn't have its ->link() */
35348 +}
35349 +
35350 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
35351 +static int au_whdir(struct inode *h_dir, struct path *path)
35352 +{
35353 +       int err;
35354 +
35355 +       err = -EEXIST;
35356 +       if (d_is_negative(path->dentry)) {
35357 +               int mode = 0700;
35358 +
35359 +               if (au_test_nfs(path->dentry->d_sb))
35360 +                       mode |= 0111;
35361 +               err = vfsub_mkdir(h_dir, path, mode);
35362 +       } else if (d_is_dir(path->dentry))
35363 +               err = 0;
35364 +       else
35365 +               pr_err("unknown %pd exists\n", path->dentry);
35366 +
35367 +       return err;
35368 +}
35369 +
35370 +struct au_wh_base {
35371 +       const struct qstr *name;
35372 +       struct dentry *dentry;
35373 +};
35374 +
35375 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
35376 +                         struct path *h_path)
35377 +{
35378 +       h_path->dentry = base[AuBrWh_BASE].dentry;
35379 +       au_wh_clean(h_dir, h_path, /*isdir*/0);
35380 +       h_path->dentry = base[AuBrWh_PLINK].dentry;
35381 +       au_wh_clean(h_dir, h_path, /*isdir*/1);
35382 +       h_path->dentry = base[AuBrWh_ORPH].dentry;
35383 +       au_wh_clean(h_dir, h_path, /*isdir*/1);
35384 +}
35385 +
35386 +/*
35387 + * returns tri-state,
35388 + * minus: error, caller should print the message
35389 + * zero: success
35390 + * plus: error, caller should NOT print the message
35391 + */
35392 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
35393 +                               int do_plink, struct au_wh_base base[],
35394 +                               struct path *h_path)
35395 +{
35396 +       int err;
35397 +       struct inode *h_dir;
35398 +
35399 +       h_dir = d_inode(h_root);
35400 +       h_path->dentry = base[AuBrWh_BASE].dentry;
35401 +       au_wh_clean(h_dir, h_path, /*isdir*/0);
35402 +       h_path->dentry = base[AuBrWh_PLINK].dentry;
35403 +       if (do_plink) {
35404 +               err = test_linkable(h_root);
35405 +               if (unlikely(err)) {
35406 +                       err = 1;
35407 +                       goto out;
35408 +               }
35409 +
35410 +               err = au_whdir(h_dir, h_path);
35411 +               if (unlikely(err))
35412 +                       goto out;
35413 +               wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35414 +       } else
35415 +               au_wh_clean(h_dir, h_path, /*isdir*/1);
35416 +       h_path->dentry = base[AuBrWh_ORPH].dentry;
35417 +       err = au_whdir(h_dir, h_path);
35418 +       if (unlikely(err))
35419 +               goto out;
35420 +       wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35421 +
35422 +out:
35423 +       return err;
35424 +}
35425 +
35426 +/*
35427 + * for the moment, aufs supports the branch filesystem which does not support
35428 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
35429 + * copyup failed. finally, such filesystem will not be used as the writable
35430 + * branch.
35431 + *
35432 + * returns tri-state, see above.
35433 + */
35434 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
35435 +                        int do_plink, struct au_wh_base base[],
35436 +                        struct path *h_path)
35437 +{
35438 +       int err;
35439 +       struct inode *h_dir;
35440 +
35441 +       WbrWhMustWriteLock(wbr);
35442 +
35443 +       err = test_linkable(h_root);
35444 +       if (unlikely(err)) {
35445 +               err = 1;
35446 +               goto out;
35447 +       }
35448 +
35449 +       /*
35450 +        * todo: should this create be done in /sbin/mount.aufs helper?
35451 +        */
35452 +       err = -EEXIST;
35453 +       h_dir = d_inode(h_root);
35454 +       if (d_is_negative(base[AuBrWh_BASE].dentry)) {
35455 +               h_path->dentry = base[AuBrWh_BASE].dentry;
35456 +               err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
35457 +       } else if (d_is_reg(base[AuBrWh_BASE].dentry))
35458 +               err = 0;
35459 +       else
35460 +               pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
35461 +       if (unlikely(err))
35462 +               goto out;
35463 +
35464 +       h_path->dentry = base[AuBrWh_PLINK].dentry;
35465 +       if (do_plink) {
35466 +               err = au_whdir(h_dir, h_path);
35467 +               if (unlikely(err))
35468 +                       goto out;
35469 +               wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35470 +       } else
35471 +               au_wh_clean(h_dir, h_path, /*isdir*/1);
35472 +       wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
35473 +
35474 +       h_path->dentry = base[AuBrWh_ORPH].dentry;
35475 +       err = au_whdir(h_dir, h_path);
35476 +       if (unlikely(err))
35477 +               goto out;
35478 +       wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35479 +
35480 +out:
35481 +       return err;
35482 +}
35483 +
35484 +/*
35485 + * initialize the whiteout base file/dir for @br.
35486 + */
35487 +int au_wh_init(struct au_branch *br, struct super_block *sb)
35488 +{
35489 +       int err, i;
35490 +       const unsigned char do_plink
35491 +               = !!au_opt_test(au_mntflags(sb), PLINK);
35492 +       struct inode *h_dir;
35493 +       struct path path = br->br_path;
35494 +       struct dentry *h_root = path.dentry;
35495 +       struct au_wbr *wbr = br->br_wbr;
35496 +       static const struct qstr base_name[] = {
35497 +               [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
35498 +                                         sizeof(AUFS_BASE_NAME) - 1),
35499 +               [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
35500 +                                          sizeof(AUFS_PLINKDIR_NAME) - 1),
35501 +               [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
35502 +                                         sizeof(AUFS_ORPHDIR_NAME) - 1)
35503 +       };
35504 +       struct au_wh_base base[] = {
35505 +               [AuBrWh_BASE] = {
35506 +                       .name   = base_name + AuBrWh_BASE,
35507 +                       .dentry = NULL
35508 +               },
35509 +               [AuBrWh_PLINK] = {
35510 +                       .name   = base_name + AuBrWh_PLINK,
35511 +                       .dentry = NULL
35512 +               },
35513 +               [AuBrWh_ORPH] = {
35514 +                       .name   = base_name + AuBrWh_ORPH,
35515 +                       .dentry = NULL
35516 +               }
35517 +       };
35518 +
35519 +       if (wbr)
35520 +               WbrWhMustWriteLock(wbr);
35521 +
35522 +       for (i = 0; i < AuBrWh_Last; i++) {
35523 +               /* doubly whiteouted */
35524 +               struct dentry *d;
35525 +
35526 +               d = au_wh_lkup(h_root, (void *)base[i].name, br);
35527 +               err = PTR_ERR(d);
35528 +               if (IS_ERR(d))
35529 +                       goto out;
35530 +
35531 +               base[i].dentry = d;
35532 +               AuDebugOn(wbr
35533 +                         && wbr->wbr_wh[i]
35534 +                         && wbr->wbr_wh[i] != base[i].dentry);
35535 +       }
35536 +
35537 +       if (wbr)
35538 +               for (i = 0; i < AuBrWh_Last; i++) {
35539 +                       dput(wbr->wbr_wh[i]);
35540 +                       wbr->wbr_wh[i] = NULL;
35541 +               }
35542 +
35543 +       err = 0;
35544 +       if (!au_br_writable(br->br_perm)) {
35545 +               h_dir = d_inode(h_root);
35546 +               au_wh_init_ro(h_dir, base, &path);
35547 +       } else if (!au_br_wh_linkable(br->br_perm)) {
35548 +               err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
35549 +               if (err > 0)
35550 +                       goto out;
35551 +               else if (err)
35552 +                       goto out_err;
35553 +       } else {
35554 +               err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
35555 +               if (err > 0)
35556 +                       goto out;
35557 +               else if (err)
35558 +                       goto out_err;
35559 +       }
35560 +       goto out; /* success */
35561 +
35562 +out_err:
35563 +       pr_err("an error(%d) on the writable branch %pd(%s)\n",
35564 +              err, h_root, au_sbtype(h_root->d_sb));
35565 +out:
35566 +       for (i = 0; i < AuBrWh_Last; i++)
35567 +               dput(base[i].dentry);
35568 +       return err;
35569 +}
35570 +
35571 +/* ---------------------------------------------------------------------- */
35572 +/*
35573 + * whiteouts are all hard-linked usually.
35574 + * when its link count reaches a ceiling, we create a new whiteout base
35575 + * asynchronously.
35576 + */
35577 +
35578 +struct reinit_br_wh {
35579 +       struct super_block *sb;
35580 +       struct au_branch *br;
35581 +};
35582 +
35583 +static void reinit_br_wh(void *arg)
35584 +{
35585 +       int err;
35586 +       aufs_bindex_t bindex;
35587 +       struct path h_path;
35588 +       struct reinit_br_wh *a = arg;
35589 +       struct au_wbr *wbr;
35590 +       struct inode *dir, *delegated;
35591 +       struct dentry *h_root;
35592 +       struct au_hinode *hdir;
35593 +
35594 +       err = 0;
35595 +       wbr = a->br->br_wbr;
35596 +       /* big aufs lock */
35597 +       si_noflush_write_lock(a->sb);
35598 +       if (!au_br_writable(a->br->br_perm))
35599 +               goto out;
35600 +       bindex = au_br_index(a->sb, a->br->br_id);
35601 +       if (unlikely(bindex < 0))
35602 +               goto out;
35603 +
35604 +       di_read_lock_parent(a->sb->s_root, AuLock_IR);
35605 +       dir = d_inode(a->sb->s_root);
35606 +       hdir = au_hi(dir, bindex);
35607 +       h_root = au_h_dptr(a->sb->s_root, bindex);
35608 +       AuDebugOn(h_root != au_br_dentry(a->br));
35609 +
35610 +       au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35611 +       wbr_wh_write_lock(wbr);
35612 +       err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
35613 +                         h_root, a->br);
35614 +       if (!err) {
35615 +               h_path.dentry = wbr->wbr_whbase;
35616 +               h_path.mnt = au_br_mnt(a->br);
35617 +               delegated = NULL;
35618 +               err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
35619 +                                  /*force*/0);
35620 +               if (unlikely(err == -EWOULDBLOCK)) {
35621 +                       pr_warn("cannot retry for NFSv4 delegation"
35622 +                               " for an internal unlink\n");
35623 +                       iput(delegated);
35624 +               }
35625 +       } else {
35626 +               pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
35627 +               err = 0;
35628 +       }
35629 +       dput(wbr->wbr_whbase);
35630 +       wbr->wbr_whbase = NULL;
35631 +       if (!err)
35632 +               err = au_wh_init(a->br, a->sb);
35633 +       wbr_wh_write_unlock(wbr);
35634 +       au_hn_inode_unlock(hdir);
35635 +       di_read_unlock(a->sb->s_root, AuLock_IR);
35636 +       if (!err)
35637 +               au_fhsm_wrote(a->sb, bindex, /*force*/0);
35638 +
35639 +out:
35640 +       if (wbr)
35641 +               atomic_dec(&wbr->wbr_wh_running);
35642 +       au_lcnt_dec(&a->br->br_count);
35643 +       si_write_unlock(a->sb);
35644 +       au_nwt_done(&au_sbi(a->sb)->si_nowait);
35645 +       au_kfree_rcu(a);
35646 +       if (unlikely(err))
35647 +               AuIOErr("err %d\n", err);
35648 +}
35649 +
35650 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
35651 +{
35652 +       int do_dec, wkq_err;
35653 +       struct reinit_br_wh *arg;
35654 +
35655 +       do_dec = 1;
35656 +       if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
35657 +               goto out;
35658 +
35659 +       /* ignore ENOMEM */
35660 +       arg = kmalloc(sizeof(*arg), GFP_NOFS);
35661 +       if (arg) {
35662 +               /*
35663 +                * dec(wh_running), kfree(arg) and dec(br_count)
35664 +                * in reinit function
35665 +                */
35666 +               arg->sb = sb;
35667 +               arg->br = br;
35668 +               au_lcnt_inc(&br->br_count);
35669 +               wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
35670 +               if (unlikely(wkq_err)) {
35671 +                       atomic_dec(&br->br_wbr->wbr_wh_running);
35672 +                       au_lcnt_dec(&br->br_count);
35673 +                       au_kfree_rcu(arg);
35674 +               }
35675 +               do_dec = 0;
35676 +       }
35677 +
35678 +out:
35679 +       if (do_dec)
35680 +               atomic_dec(&br->br_wbr->wbr_wh_running);
35681 +}
35682 +
35683 +/* ---------------------------------------------------------------------- */
35684 +
35685 +/*
35686 + * create the whiteout @wh.
35687 + */
35688 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
35689 +                            struct dentry *wh)
35690 +{
35691 +       int err;
35692 +       struct path h_path = {
35693 +               .dentry = wh
35694 +       };
35695 +       struct au_branch *br;
35696 +       struct au_wbr *wbr;
35697 +       struct dentry *h_parent;
35698 +       struct inode *h_dir, *delegated;
35699 +
35700 +       h_parent = wh->d_parent; /* dir inode is locked */
35701 +       h_dir = d_inode(h_parent);
35702 +       IMustLock(h_dir);
35703 +
35704 +       br = au_sbr(sb, bindex);
35705 +       h_path.mnt = au_br_mnt(br);
35706 +       wbr = br->br_wbr;
35707 +       wbr_wh_read_lock(wbr);
35708 +       if (wbr->wbr_whbase) {
35709 +               delegated = NULL;
35710 +               err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
35711 +               if (unlikely(err == -EWOULDBLOCK)) {
35712 +                       pr_warn("cannot retry for NFSv4 delegation"
35713 +                               " for an internal link\n");
35714 +                       iput(delegated);
35715 +               }
35716 +               if (!err || err != -EMLINK)
35717 +                       goto out;
35718 +
35719 +               /* link count full. re-initialize br_whbase. */
35720 +               kick_reinit_br_wh(sb, br);
35721 +       }
35722 +
35723 +       /* return this error in this context */
35724 +       err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
35725 +       if (!err)
35726 +               au_fhsm_wrote(sb, bindex, /*force*/0);
35727 +
35728 +out:
35729 +       wbr_wh_read_unlock(wbr);
35730 +       return err;
35731 +}
35732 +
35733 +/* ---------------------------------------------------------------------- */
35734 +
35735 +/*
35736 + * create or remove the diropq.
35737 + */
35738 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
35739 +                               unsigned int flags)
35740 +{
35741 +       struct dentry *opq_dentry;
35742 +       struct super_block *sb;
35743 +       struct au_branch *br;
35744 +       struct path h_path;
35745 +       int err;
35746 +
35747 +       sb = dentry->d_sb;
35748 +       br = au_sbr(sb, bindex);
35749 +       h_path.dentry = au_h_dptr(dentry, bindex);
35750 +       h_path.mnt = au_br_mnt(br);
35751 +       opq_dentry = vfsub_lkup_one(&diropq_name, &h_path);
35752 +       if (IS_ERR(opq_dentry))
35753 +               goto out;
35754 +
35755 +       if (au_ftest_diropq(flags, CREATE)) {
35756 +               err = link_or_create_wh(sb, bindex, opq_dentry);
35757 +               if (!err) {
35758 +                       au_set_dbdiropq(dentry, bindex);
35759 +                       goto out; /* success */
35760 +               }
35761 +       } else {
35762 +               h_path.dentry = opq_dentry;
35763 +               err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &h_path);
35764 +               if (!err)
35765 +                       au_set_dbdiropq(dentry, -1);
35766 +       }
35767 +       dput(opq_dentry);
35768 +       opq_dentry = ERR_PTR(err);
35769 +
35770 +out:
35771 +       return opq_dentry;
35772 +}
35773 +
35774 +struct do_diropq_args {
35775 +       struct dentry **errp;
35776 +       struct dentry *dentry;
35777 +       aufs_bindex_t bindex;
35778 +       unsigned int flags;
35779 +};
35780 +
35781 +static void call_do_diropq(void *args)
35782 +{
35783 +       struct do_diropq_args *a = args;
35784 +       *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
35785 +}
35786 +
35787 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35788 +                            unsigned int flags)
35789 +{
35790 +       struct dentry *diropq, *h_dentry;
35791 +       struct user_namespace *h_userns;
35792 +
35793 +       h_userns = au_sbr_userns(dentry->d_sb, bindex);
35794 +       h_dentry = au_h_dptr(dentry, bindex);
35795 +       if (!au_test_h_perm_sio(h_userns, d_inode(h_dentry),
35796 +                               MAY_EXEC | MAY_WRITE))
35797 +               diropq = do_diropq(dentry, bindex, flags);
35798 +       else {
35799 +               int wkq_err;
35800 +               struct do_diropq_args args = {
35801 +                       .errp           = &diropq,
35802 +                       .dentry         = dentry,
35803 +                       .bindex         = bindex,
35804 +                       .flags          = flags
35805 +               };
35806 +
35807 +               wkq_err = au_wkq_wait(call_do_diropq, &args);
35808 +               if (unlikely(wkq_err))
35809 +                       diropq = ERR_PTR(wkq_err);
35810 +       }
35811 +
35812 +       return diropq;
35813 +}
35814 +
35815 +/* ---------------------------------------------------------------------- */
35816 +
35817 +/*
35818 + * lookup whiteout dentry.
35819 + * @h_parent: lower parent dentry which must exist and be locked
35820 + * @base_name: name of dentry which will be whiteouted
35821 + * returns dentry for whiteout.
35822 + */
35823 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35824 +                         struct au_branch *br)
35825 +{
35826 +       int err;
35827 +       struct qstr wh_name;
35828 +       struct dentry *wh_dentry;
35829 +       struct path h_path;
35830 +
35831 +       err = au_wh_name_alloc(&wh_name, base_name);
35832 +       wh_dentry = ERR_PTR(err);
35833 +       if (!err) {
35834 +               h_path.dentry = h_parent;
35835 +               h_path.mnt = au_br_mnt(br);
35836 +               wh_dentry = vfsub_lkup_one(&wh_name, &h_path);
35837 +               au_kfree_try_rcu(wh_name.name);
35838 +       }
35839 +       return wh_dentry;
35840 +}
35841 +
35842 +/*
35843 + * link/create a whiteout for @dentry on @bindex.
35844 + */
35845 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35846 +                           struct dentry *h_parent)
35847 +{
35848 +       struct dentry *wh_dentry;
35849 +       struct super_block *sb;
35850 +       int err;
35851 +
35852 +       sb = dentry->d_sb;
35853 +       wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
35854 +       if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
35855 +               err = link_or_create_wh(sb, bindex, wh_dentry);
35856 +               if (!err) {
35857 +                       au_set_dbwh(dentry, bindex);
35858 +                       au_fhsm_wrote(sb, bindex, /*force*/0);
35859 +               } else {
35860 +                       dput(wh_dentry);
35861 +                       wh_dentry = ERR_PTR(err);
35862 +               }
35863 +       }
35864 +
35865 +       return wh_dentry;
35866 +}
35867 +
35868 +/* ---------------------------------------------------------------------- */
35869 +
35870 +/* Delete all whiteouts in this directory on branch bindex. */
35871 +static int del_wh_children(struct path *h_path, struct au_nhash *whlist,
35872 +                          aufs_bindex_t bindex)
35873 +{
35874 +       int err;
35875 +       unsigned long ul, n;
35876 +       struct qstr wh_name;
35877 +       char *p;
35878 +       struct hlist_head *head;
35879 +       struct au_vdir_wh *pos;
35880 +       struct au_vdir_destr *str;
35881 +
35882 +       err = -ENOMEM;
35883 +       p = (void *)__get_free_page(GFP_NOFS);
35884 +       wh_name.name = p;
35885 +       if (unlikely(!wh_name.name))
35886 +               goto out;
35887 +
35888 +       err = 0;
35889 +       memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35890 +       p += AUFS_WH_PFX_LEN;
35891 +       n = whlist->nh_num;
35892 +       head = whlist->nh_head;
35893 +       for (ul = 0; !err && ul < n; ul++, head++) {
35894 +               hlist_for_each_entry(pos, head, wh_hash) {
35895 +                       if (pos->wh_bindex != bindex)
35896 +                               continue;
35897 +
35898 +                       str = &pos->wh_str;
35899 +                       if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
35900 +                               memcpy(p, str->name, str->len);
35901 +                               wh_name.len = AUFS_WH_PFX_LEN + str->len;
35902 +                               err = unlink_wh_name(h_path, &wh_name);
35903 +                               if (!err)
35904 +                                       continue;
35905 +                               break;
35906 +                       }
35907 +                       AuIOErr("whiteout name too long %.*s\n",
35908 +                               str->len, str->name);
35909 +                       err = -EIO;
35910 +                       break;
35911 +               }
35912 +       }
35913 +       free_page((unsigned long)wh_name.name);
35914 +
35915 +out:
35916 +       return err;
35917 +}
35918 +
35919 +struct del_wh_children_args {
35920 +       int *errp;
35921 +       struct path *h_path;
35922 +       struct au_nhash *whlist;
35923 +       aufs_bindex_t bindex;
35924 +};
35925 +
35926 +static void call_del_wh_children(void *args)
35927 +{
35928 +       struct del_wh_children_args *a = args;
35929 +       *a->errp = del_wh_children(a->h_path, a->whlist, a->bindex);
35930 +}
35931 +
35932 +/* ---------------------------------------------------------------------- */
35933 +
35934 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
35935 +{
35936 +       struct au_whtmp_rmdir *whtmp;
35937 +       int err;
35938 +       unsigned int rdhash;
35939 +
35940 +       SiMustAnyLock(sb);
35941 +
35942 +       whtmp = kzalloc(sizeof(*whtmp), gfp);
35943 +       if (unlikely(!whtmp)) {
35944 +               whtmp = ERR_PTR(-ENOMEM);
35945 +               goto out;
35946 +       }
35947 +
35948 +       /* no estimation for dir size */
35949 +       rdhash = au_sbi(sb)->si_rdhash;
35950 +       if (!rdhash)
35951 +               rdhash = AUFS_RDHASH_DEF;
35952 +       err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
35953 +       if (unlikely(err)) {
35954 +               au_kfree_rcu(whtmp);
35955 +               whtmp = ERR_PTR(err);
35956 +       }
35957 +
35958 +out:
35959 +       return whtmp;
35960 +}
35961 +
35962 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
35963 +{
35964 +       if (whtmp->br)
35965 +               au_lcnt_dec(&whtmp->br->br_count);
35966 +       dput(whtmp->wh_dentry);
35967 +       iput(whtmp->dir);
35968 +       au_nhash_wh_free(&whtmp->whlist);
35969 +       au_kfree_rcu(whtmp);
35970 +}
35971 +
35972 +/*
35973 + * rmdir the whiteouted temporary named dir @h_dentry.
35974 + * @whlist: whiteouted children.
35975 + */
35976 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
35977 +                  struct dentry *wh_dentry, struct au_nhash *whlist)
35978 +{
35979 +       int err;
35980 +       unsigned int h_nlink;
35981 +       struct path wh_path;
35982 +       struct inode *wh_inode, *h_dir;
35983 +       struct au_branch *br;
35984 +       struct user_namespace *h_userns;
35985 +
35986 +       h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
35987 +       IMustLock(h_dir);
35988 +
35989 +       br = au_sbr(dir->i_sb, bindex);
35990 +       wh_path.dentry = wh_dentry;
35991 +       wh_path.mnt = au_br_mnt(br);
35992 +       h_userns = au_br_userns(br);
35993 +       wh_inode = d_inode(wh_dentry);
35994 +       inode_lock_nested(wh_inode, AuLsc_I_CHILD);
35995 +
35996 +       /*
35997 +        * someone else might change some whiteouts while we were sleeping.
35998 +        * it means this whlist may have an obsoleted entry.
35999 +        */
36000 +       if (!au_test_h_perm_sio(h_userns, wh_inode, MAY_EXEC | MAY_WRITE))
36001 +               err = del_wh_children(&wh_path, whlist, bindex);
36002 +       else {
36003 +               int wkq_err;
36004 +               struct del_wh_children_args args = {
36005 +                       .errp           = &err,
36006 +                       .h_path         = &wh_path,
36007 +                       .whlist         = whlist,
36008 +                       .bindex         = bindex
36009 +               };
36010 +
36011 +               wkq_err = au_wkq_wait(call_del_wh_children, &args);
36012 +               if (unlikely(wkq_err))
36013 +                       err = wkq_err;
36014 +       }
36015 +       inode_unlock(wh_inode);
36016 +
36017 +       if (!err) {
36018 +               h_nlink = h_dir->i_nlink;
36019 +               err = vfsub_rmdir(h_dir, &wh_path);
36020 +               /* some fs doesn't change the parent nlink in some cases */
36021 +               h_nlink -= h_dir->i_nlink;
36022 +       }
36023 +
36024 +       if (!err) {
36025 +               if (au_ibtop(dir) == bindex) {
36026 +                       /* todo: dir->i_mutex is necessary */
36027 +                       au_cpup_attr_timesizes(dir);
36028 +                       if (h_nlink)
36029 +                               vfsub_drop_nlink(dir);
36030 +               }
36031 +               return 0; /* success */
36032 +       }
36033 +
36034 +       pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
36035 +       return err;
36036 +}
36037 +
36038 +static void call_rmdir_whtmp(void *args)
36039 +{
36040 +       int err;
36041 +       aufs_bindex_t bindex;
36042 +       struct au_whtmp_rmdir *a = args;
36043 +       struct super_block *sb;
36044 +       struct dentry *h_parent;
36045 +       struct inode *h_dir;
36046 +       struct au_hinode *hdir;
36047 +
36048 +       /* rmdir by nfsd may cause deadlock with this i_mutex */
36049 +       /* inode_lock(a->dir); */
36050 +       err = -EROFS;
36051 +       sb = a->dir->i_sb;
36052 +       si_read_lock(sb, !AuLock_FLUSH);
36053 +       if (!au_br_writable(a->br->br_perm))
36054 +               goto out;
36055 +       bindex = au_br_index(sb, a->br->br_id);
36056 +       if (unlikely(bindex < 0))
36057 +               goto out;
36058 +
36059 +       err = -EIO;
36060 +       ii_write_lock_parent(a->dir);
36061 +       h_parent = dget_parent(a->wh_dentry);
36062 +       h_dir = d_inode(h_parent);
36063 +       hdir = au_hi(a->dir, bindex);
36064 +       err = vfsub_mnt_want_write(au_br_mnt(a->br));
36065 +       if (unlikely(err))
36066 +               goto out_mnt;
36067 +       au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
36068 +       err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
36069 +                         a->br);
36070 +       if (!err)
36071 +               err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
36072 +       au_hn_inode_unlock(hdir);
36073 +       vfsub_mnt_drop_write(au_br_mnt(a->br));
36074 +
36075 +out_mnt:
36076 +       dput(h_parent);
36077 +       ii_write_unlock(a->dir);
36078 +out:
36079 +       /* inode_unlock(a->dir); */
36080 +       au_whtmp_rmdir_free(a);
36081 +       si_read_unlock(sb);
36082 +       au_nwt_done(&au_sbi(sb)->si_nowait);
36083 +       if (unlikely(err))
36084 +               AuIOErr("err %d\n", err);
36085 +}
36086 +
36087 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36088 +                        struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
36089 +{
36090 +       int wkq_err;
36091 +       struct super_block *sb;
36092 +
36093 +       IMustLock(dir);
36094 +
36095 +       /* all post-process will be done in do_rmdir_whtmp(). */
36096 +       sb = dir->i_sb;
36097 +       args->dir = au_igrab(dir);
36098 +       args->br = au_sbr(sb, bindex);
36099 +       au_lcnt_inc(&args->br->br_count);
36100 +       args->wh_dentry = dget(wh_dentry);
36101 +       wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
36102 +       if (unlikely(wkq_err)) {
36103 +               pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
36104 +               au_whtmp_rmdir_free(args);
36105 +       }
36106 +}
36107 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
36108 --- /usr/share/empty/fs/aufs/whout.h    1970-01-01 01:00:00.000000000 +0100
36109 +++ linux/fs/aufs/whout.h       2022-11-05 23:02:18.972555950 +0100
36110 @@ -0,0 +1,87 @@
36111 +/* SPDX-License-Identifier: GPL-2.0 */
36112 +/*
36113 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36114 + *
36115 + * This program is free software; you can redistribute it and/or modify
36116 + * it under the terms of the GNU General Public License as published by
36117 + * the Free Software Foundation; either version 2 of the License, or
36118 + * (at your option) any later version.
36119 + *
36120 + * This program is distributed in the hope that it will be useful,
36121 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36122 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36123 + * GNU General Public License for more details.
36124 + *
36125 + * You should have received a copy of the GNU General Public License
36126 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36127 + */
36128 +
36129 +/*
36130 + * whiteout for logical deletion and opaque directory
36131 + */
36132 +
36133 +#ifndef __AUFS_WHOUT_H__
36134 +#define __AUFS_WHOUT_H__
36135 +
36136 +#ifdef __KERNEL__
36137 +
36138 +#include "dir.h"
36139 +
36140 +/* whout.c */
36141 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
36142 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
36143 +              struct qstr *wh_name, int try_sio);
36144 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path);
36145 +struct au_branch;
36146 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
36147 +                            struct qstr *prefix);
36148 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
36149 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
36150 +                       struct dentry *dentry);
36151 +int au_wh_init(struct au_branch *br, struct super_block *sb);
36152 +
36153 +/* diropq flags */
36154 +#define AuDiropq_CREATE        1
36155 +#define au_ftest_diropq(flags, name)   ((flags) & AuDiropq_##name)
36156 +#define au_fset_diropq(flags, name) \
36157 +       do { (flags) |= AuDiropq_##name; } while (0)
36158 +#define au_fclr_diropq(flags, name) \
36159 +       do { (flags) &= ~AuDiropq_##name; } while (0)
36160 +
36161 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
36162 +                            unsigned int flags);
36163 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
36164 +                         struct au_branch *br);
36165 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
36166 +                           struct dentry *h_parent);
36167 +
36168 +/* real rmdir for the whiteout-ed dir */
36169 +struct au_whtmp_rmdir {
36170 +       struct inode *dir;
36171 +       struct au_branch *br;
36172 +       struct dentry *wh_dentry;
36173 +       struct au_nhash whlist;
36174 +};
36175 +
36176 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
36177 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
36178 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36179 +                  struct dentry *wh_dentry, struct au_nhash *whlist);
36180 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36181 +                        struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
36182 +
36183 +/* ---------------------------------------------------------------------- */
36184 +
36185 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
36186 +                                             aufs_bindex_t bindex)
36187 +{
36188 +       return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
36189 +}
36190 +
36191 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
36192 +{
36193 +       return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
36194 +}
36195 +
36196 +#endif /* __KERNEL__ */
36197 +#endif /* __AUFS_WHOUT_H__ */
36198 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
36199 --- /usr/share/empty/fs/aufs/wkq.c      1970-01-01 01:00:00.000000000 +0100
36200 +++ linux/fs/aufs/wkq.c 2022-11-05 23:02:18.972555950 +0100
36201 @@ -0,0 +1,372 @@
36202 +// SPDX-License-Identifier: GPL-2.0
36203 +/*
36204 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36205 + *
36206 + * This program is free software; you can redistribute it and/or modify
36207 + * it under the terms of the GNU General Public License as published by
36208 + * the Free Software Foundation; either version 2 of the License, or
36209 + * (at your option) any later version.
36210 + *
36211 + * This program is distributed in the hope that it will be useful,
36212 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36213 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36214 + * GNU General Public License for more details.
36215 + *
36216 + * You should have received a copy of the GNU General Public License
36217 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36218 + */
36219 +
36220 +/*
36221 + * workqueue for asynchronous/super-io operations
36222 + * todo: try new credential scheme
36223 + */
36224 +
36225 +#include <linux/module.h>
36226 +#include "aufs.h"
36227 +
36228 +/* internal workqueue named AUFS_WKQ_NAME */
36229 +
36230 +static struct workqueue_struct *au_wkq;
36231 +
36232 +struct au_wkinfo {
36233 +       struct work_struct wk;
36234 +       struct kobject *kobj;
36235 +
36236 +       unsigned int flags; /* see wkq.h */
36237 +
36238 +       au_wkq_func_t func;
36239 +       void *args;
36240 +
36241 +#ifdef CONFIG_LOCKDEP
36242 +       int dont_check;
36243 +       struct held_lock **hlock;
36244 +#endif
36245 +
36246 +       struct completion *comp;
36247 +};
36248 +
36249 +/* ---------------------------------------------------------------------- */
36250 +/*
36251 + * Aufs passes some operations to the workqueue such as the internal copyup.
36252 + * This scheme looks rather unnatural for LOCKDEP debugging feature, since the
36253 + * job run by workqueue depends upon the locks acquired in the other task.
36254 + * Delegating a small operation to the workqueue, aufs passes its lockdep
36255 + * information too. And the job in the workqueue restores the info in order to
36256 + * pretend as if it acquired those locks. This is just to make LOCKDEP work
36257 + * correctly and expectedly.
36258 + */
36259 +
36260 +#ifndef CONFIG_LOCKDEP
36261 +AuStubInt0(au_wkq_lockdep_alloc, struct au_wkinfo *wkinfo);
36262 +AuStubVoid(au_wkq_lockdep_free, struct au_wkinfo *wkinfo);
36263 +AuStubVoid(au_wkq_lockdep_pre, struct au_wkinfo *wkinfo);
36264 +AuStubVoid(au_wkq_lockdep_post, struct au_wkinfo *wkinfo);
36265 +AuStubVoid(au_wkq_lockdep_init, struct au_wkinfo *wkinfo);
36266 +#else
36267 +static void au_wkq_lockdep_init(struct au_wkinfo *wkinfo)
36268 +{
36269 +       wkinfo->hlock = NULL;
36270 +       wkinfo->dont_check = 0;
36271 +}
36272 +
36273 +/*
36274 + * 1: matched
36275 + * 0: unmatched
36276 + */
36277 +static int au_wkq_lockdep_test(struct lock_class_key *key, const char *name)
36278 +{
36279 +       static DEFINE_SPINLOCK(spin);
36280 +       static struct {
36281 +               char *name;
36282 +               struct lock_class_key *key;
36283 +       } a[] = {
36284 +               { .name = "&sbinfo->si_rwsem" },
36285 +               { .name = "&finfo->fi_rwsem" },
36286 +               { .name = "&dinfo->di_rwsem" },
36287 +               { .name = "&iinfo->ii_rwsem" }
36288 +       };
36289 +       static int set;
36290 +       int i;
36291 +
36292 +       /* lockless read from 'set.' see below */
36293 +       if (set == ARRAY_SIZE(a)) {
36294 +               for (i = 0; i < ARRAY_SIZE(a); i++)
36295 +                       if (a[i].key == key)
36296 +                               goto match;
36297 +               goto unmatch;
36298 +       }
36299 +
36300 +       spin_lock(&spin);
36301 +       if (set)
36302 +               for (i = 0; i < ARRAY_SIZE(a); i++)
36303 +                       if (a[i].key == key) {
36304 +                               spin_unlock(&spin);
36305 +                               goto match;
36306 +                       }
36307 +       for (i = 0; i < ARRAY_SIZE(a); i++) {
36308 +               if (a[i].key) {
36309 +                       if (unlikely(a[i].key == key)) { /* rare but possible */
36310 +                               spin_unlock(&spin);
36311 +                               goto match;
36312 +                       } else
36313 +                               continue;
36314 +               }
36315 +               if (strstr(a[i].name, name)) {
36316 +                       /*
36317 +                        * the order of these three lines is important for the
36318 +                        * lockless read above.
36319 +                        */
36320 +                       a[i].key = key;
36321 +                       spin_unlock(&spin);
36322 +                       set++;
36323 +                       /* AuDbg("%d, %s\n", set, name); */
36324 +                       goto match;
36325 +               }
36326 +       }
36327 +       spin_unlock(&spin);
36328 +       goto unmatch;
36329 +
36330 +match:
36331 +       return 1;
36332 +unmatch:
36333 +       return 0;
36334 +}
36335 +
36336 +static int au_wkq_lockdep_alloc(struct au_wkinfo *wkinfo)
36337 +{
36338 +       int err, n;
36339 +       struct task_struct *curr;
36340 +       struct held_lock **hl, *held_locks, *p;
36341 +
36342 +       err = 0;
36343 +       curr = current;
36344 +       wkinfo->dont_check = lockdep_recursing(curr);
36345 +       if (wkinfo->dont_check)
36346 +               goto out;
36347 +       n = curr->lockdep_depth;
36348 +       if (!n)
36349 +               goto out;
36350 +
36351 +       err = -ENOMEM;
36352 +       wkinfo->hlock = kmalloc_array(n + 1, sizeof(*wkinfo->hlock), GFP_NOFS);
36353 +       if (unlikely(!wkinfo->hlock))
36354 +               goto out;
36355 +
36356 +       err = 0;
36357 +#if 0 /* left for debugging */
36358 +       if (0 && au_debug_test())
36359 +               lockdep_print_held_locks(curr);
36360 +#endif
36361 +       held_locks = curr->held_locks;
36362 +       hl = wkinfo->hlock;
36363 +       while (n--) {
36364 +               p = held_locks++;
36365 +               if (au_wkq_lockdep_test(p->instance->key, p->instance->name))
36366 +                       *hl++ = p;
36367 +       }
36368 +       *hl = NULL;
36369 +
36370 +out:
36371 +       return err;
36372 +}
36373 +
36374 +static void au_wkq_lockdep_free(struct au_wkinfo *wkinfo)
36375 +{
36376 +       au_kfree_try_rcu(wkinfo->hlock);
36377 +}
36378 +
36379 +static void au_wkq_lockdep_pre(struct au_wkinfo *wkinfo)
36380 +{
36381 +       struct held_lock *p, **hl = wkinfo->hlock;
36382 +       int subclass;
36383 +
36384 +       if (wkinfo->dont_check)
36385 +               lockdep_off();
36386 +       if (!hl)
36387 +               return;
36388 +       while ((p = *hl++)) { /* assignment */
36389 +               subclass = lockdep_hlock_class(p)->subclass;
36390 +               /* AuDbg("%s, %d\n", p->instance->name, subclass); */
36391 +               if (p->read)
36392 +                       rwsem_acquire_read(p->instance, subclass, 0,
36393 +                                          /*p->acquire_ip*/_RET_IP_);
36394 +               else
36395 +                       rwsem_acquire(p->instance, subclass, 0,
36396 +                                     /*p->acquire_ip*/_RET_IP_);
36397 +       }
36398 +}
36399 +
36400 +static void au_wkq_lockdep_post(struct au_wkinfo *wkinfo)
36401 +{
36402 +       struct held_lock *p, **hl = wkinfo->hlock;
36403 +
36404 +       if (wkinfo->dont_check)
36405 +               lockdep_on();
36406 +       if (!hl)
36407 +               return;
36408 +       while ((p = *hl++)) /* assignment */
36409 +               rwsem_release(p->instance, /*p->acquire_ip*/_RET_IP_);
36410 +}
36411 +#endif
36412 +
36413 +static void wkq_func(struct work_struct *wk)
36414 +{
36415 +       struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
36416 +
36417 +       AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
36418 +       AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
36419 +
36420 +       au_wkq_lockdep_pre(wkinfo);
36421 +       wkinfo->func(wkinfo->args);
36422 +       au_wkq_lockdep_post(wkinfo);
36423 +       if (au_ftest_wkq(wkinfo->flags, WAIT))
36424 +               complete(wkinfo->comp);
36425 +       else {
36426 +               kobject_put(wkinfo->kobj);
36427 +               module_put(THIS_MODULE); /* todo: ?? */
36428 +               au_kfree_rcu(wkinfo);
36429 +       }
36430 +}
36431 +
36432 +/*
36433 + * Since struct completion is large, try allocating it dynamically.
36434 + */
36435 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
36436 +
36437 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
36438 +{
36439 +       *comp = kmalloc(sizeof(**comp), GFP_NOFS);
36440 +       if (*comp) {
36441 +               init_completion(*comp);
36442 +               wkinfo->comp = *comp;
36443 +               return 0;
36444 +       }
36445 +       return -ENOMEM;
36446 +}
36447 +
36448 +static void au_wkq_comp_free(struct completion *comp)
36449 +{
36450 +       au_kfree_rcu(comp);
36451 +}
36452 +
36453 +static void au_wkq_run(struct au_wkinfo *wkinfo)
36454 +{
36455 +       if (au_ftest_wkq(wkinfo->flags, NEST)) {
36456 +               if (au_wkq_test()) {
36457 +                       AuWarn1("wkq from wkq, unless silly-rename on NFS,"
36458 +                               " due to a dead dir by UDBA,"
36459 +                               " or async xino write?\n");
36460 +                       AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
36461 +               }
36462 +       } else
36463 +               au_dbg_verify_kthread();
36464 +
36465 +       if (au_ftest_wkq(wkinfo->flags, WAIT)) {
36466 +               INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
36467 +               queue_work(au_wkq, &wkinfo->wk);
36468 +       } else {
36469 +               INIT_WORK(&wkinfo->wk, wkq_func);
36470 +               schedule_work(&wkinfo->wk);
36471 +       }
36472 +}
36473 +
36474 +/*
36475 + * Be careful. It is easy to make deadlock happen.
36476 + * processA: lock, wkq and wait
36477 + * processB: wkq and wait, lock in wkq
36478 + * --> deadlock
36479 + */
36480 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
36481 +{
36482 +       int err;
36483 +       AuWkqCompDeclare(comp);
36484 +       struct au_wkinfo wkinfo = {
36485 +               .flags  = flags,
36486 +               .func   = func,
36487 +               .args   = args
36488 +       };
36489 +
36490 +       err = au_wkq_comp_alloc(&wkinfo, &comp);
36491 +       if (unlikely(err))
36492 +               goto out;
36493 +       err = au_wkq_lockdep_alloc(&wkinfo);
36494 +       if (unlikely(err))
36495 +               goto out_comp;
36496 +       if (!err) {
36497 +               au_wkq_run(&wkinfo);
36498 +               /* no timeout, no interrupt */
36499 +               wait_for_completion(wkinfo.comp);
36500 +       }
36501 +       au_wkq_lockdep_free(&wkinfo);
36502 +
36503 +out_comp:
36504 +       au_wkq_comp_free(comp);
36505 +out:
36506 +       destroy_work_on_stack(&wkinfo.wk);
36507 +       return err;
36508 +}
36509 +
36510 +/*
36511 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
36512 + * problem in a concurrent umounting.
36513 + */
36514 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36515 +                 unsigned int flags)
36516 +{
36517 +       int err;
36518 +       struct au_wkinfo *wkinfo;
36519 +
36520 +       atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
36521 +
36522 +       /*
36523 +        * wkq_func() must free this wkinfo.
36524 +        * it highly depends upon the implementation of workqueue.
36525 +        */
36526 +       err = 0;
36527 +       wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
36528 +       if (wkinfo) {
36529 +               wkinfo->kobj = &au_sbi(sb)->si_kobj;
36530 +               wkinfo->flags = flags & ~AuWkq_WAIT;
36531 +               wkinfo->func = func;
36532 +               wkinfo->args = args;
36533 +               wkinfo->comp = NULL;
36534 +               au_wkq_lockdep_init(wkinfo);
36535 +               kobject_get(wkinfo->kobj);
36536 +               __module_get(THIS_MODULE); /* todo: ?? */
36537 +
36538 +               au_wkq_run(wkinfo);
36539 +       } else {
36540 +               err = -ENOMEM;
36541 +               au_nwt_done(&au_sbi(sb)->si_nowait);
36542 +       }
36543 +
36544 +       return err;
36545 +}
36546 +
36547 +/* ---------------------------------------------------------------------- */
36548 +
36549 +void au_nwt_init(struct au_nowait_tasks *nwt)
36550 +{
36551 +       atomic_set(&nwt->nw_len, 0);
36552 +       /* smp_mb(); */ /* atomic_set */
36553 +       init_waitqueue_head(&nwt->nw_wq);
36554 +}
36555 +
36556 +void au_wkq_fin(void)
36557 +{
36558 +       destroy_workqueue(au_wkq);
36559 +}
36560 +
36561 +int __init au_wkq_init(void)
36562 +{
36563 +       int err;
36564 +
36565 +       err = 0;
36566 +       au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
36567 +       if (IS_ERR(au_wkq))
36568 +               err = PTR_ERR(au_wkq);
36569 +       else if (!au_wkq)
36570 +               err = -ENOMEM;
36571 +
36572 +       return err;
36573 +}
36574 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
36575 --- /usr/share/empty/fs/aufs/wkq.h      1970-01-01 01:00:00.000000000 +0100
36576 +++ linux/fs/aufs/wkq.h 2022-11-05 23:02:18.972555950 +0100
36577 @@ -0,0 +1,89 @@
36578 +/* SPDX-License-Identifier: GPL-2.0 */
36579 +/*
36580 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36581 + *
36582 + * This program is free software; you can redistribute it and/or modify
36583 + * it under the terms of the GNU General Public License as published by
36584 + * the Free Software Foundation; either version 2 of the License, or
36585 + * (at your option) any later version.
36586 + *
36587 + * This program is distributed in the hope that it will be useful,
36588 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36589 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36590 + * GNU General Public License for more details.
36591 + *
36592 + * You should have received a copy of the GNU General Public License
36593 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36594 + */
36595 +
36596 +/*
36597 + * workqueue for asynchronous/super-io operations
36598 + * todo: try new credentials management scheme
36599 + */
36600 +
36601 +#ifndef __AUFS_WKQ_H__
36602 +#define __AUFS_WKQ_H__
36603 +
36604 +#ifdef __KERNEL__
36605 +
36606 +#include <linux/wait.h>
36607 +
36608 +struct super_block;
36609 +
36610 +/* ---------------------------------------------------------------------- */
36611 +
36612 +/*
36613 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
36614 + */
36615 +struct au_nowait_tasks {
36616 +       atomic_t                nw_len;
36617 +       wait_queue_head_t       nw_wq;
36618 +};
36619 +
36620 +/* ---------------------------------------------------------------------- */
36621 +
36622 +typedef void (*au_wkq_func_t)(void *args);
36623 +
36624 +/* wkq flags */
36625 +#define AuWkq_WAIT     1
36626 +#define AuWkq_NEST     (1 << 1)
36627 +#define au_ftest_wkq(flags, name)      ((flags) & AuWkq_##name)
36628 +#define au_fset_wkq(flags, name) \
36629 +       do { (flags) |= AuWkq_##name; } while (0)
36630 +#define au_fclr_wkq(flags, name) \
36631 +       do { (flags) &= ~AuWkq_##name; } while (0)
36632 +
36633 +/* wkq.c */
36634 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
36635 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36636 +                 unsigned int flags);
36637 +void au_nwt_init(struct au_nowait_tasks *nwt);
36638 +int __init au_wkq_init(void);
36639 +void au_wkq_fin(void);
36640 +
36641 +/* ---------------------------------------------------------------------- */
36642 +
36643 +static inline int au_wkq_test(void)
36644 +{
36645 +       return current->flags & PF_WQ_WORKER;
36646 +}
36647 +
36648 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
36649 +{
36650 +       return au_wkq_do_wait(AuWkq_WAIT, func, args);
36651 +}
36652 +
36653 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
36654 +{
36655 +       if (atomic_dec_and_test(&nwt->nw_len))
36656 +               wake_up_all(&nwt->nw_wq);
36657 +}
36658 +
36659 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
36660 +{
36661 +       wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
36662 +       return 0;
36663 +}
36664 +
36665 +#endif /* __KERNEL__ */
36666 +#endif /* __AUFS_WKQ_H__ */
36667 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
36668 --- /usr/share/empty/fs/aufs/xattr.c    1970-01-01 01:00:00.000000000 +0100
36669 +++ linux/fs/aufs/xattr.c       2022-11-05 23:02:18.972555950 +0100
36670 @@ -0,0 +1,372 @@
36671 +// SPDX-License-Identifier: GPL-2.0
36672 +/*
36673 + * Copyright (C) 2014-2022 Junjiro R. Okajima
36674 + *
36675 + * This program is free software; you can redistribute it and/or modify
36676 + * it under the terms of the GNU General Public License as published by
36677 + * the Free Software Foundation; either version 2 of the License, or
36678 + * (at your option) any later version.
36679 + *
36680 + * This program is distributed in the hope that it will be useful,
36681 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36682 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
36683 + * GNU General Public License for more details.
36684 + *
36685 + * You should have received a copy of the GNU General Public License
36686 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
36687 + */
36688 +
36689 +/*
36690 + * handling xattr functions
36691 + */
36692 +
36693 +#include <linux/fs.h>
36694 +#include <linux/posix_acl_xattr.h>
36695 +#include <linux/xattr.h>
36696 +#include "aufs.h"
36697 +
36698 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
36699 +{
36700 +       if (!ignore_flags)
36701 +               goto out;
36702 +       switch (err) {
36703 +       case -ENOMEM:
36704 +       case -EDQUOT:
36705 +               goto out;
36706 +       }
36707 +
36708 +       if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
36709 +               err = 0;
36710 +               goto out;
36711 +       }
36712 +
36713 +#define cmp(brattr, prefix) do {                                       \
36714 +               if (!strncmp(name, XATTR_##prefix##_PREFIX,             \
36715 +                            XATTR_##prefix##_PREFIX_LEN)) {            \
36716 +                       if (ignore_flags & AuBrAttr_ICEX_##brattr)      \
36717 +                               err = 0;                                \
36718 +                       goto out;                                       \
36719 +               }                                                       \
36720 +       } while (0)
36721 +
36722 +       cmp(SEC, SECURITY);
36723 +       cmp(SYS, SYSTEM);
36724 +       cmp(TR, TRUSTED);
36725 +       cmp(USR, USER);
36726 +#undef cmp
36727 +
36728 +       if (ignore_flags & AuBrAttr_ICEX_OTH)
36729 +               err = 0;
36730 +
36731 +out:
36732 +       return err;
36733 +}
36734 +
36735 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
36736 +
36737 +static int au_do_cpup_xattr(struct path *h_dst, struct path *h_src,
36738 +                           char *name, char **buf, unsigned int ignore_flags,
36739 +                           unsigned int verbose)
36740 +{
36741 +       int err;
36742 +       ssize_t ssz;
36743 +       struct inode *h_idst;
36744 +       struct dentry *h_dst_dentry, *h_src_dentry;
36745 +       struct user_namespace *h_dst_userns, *h_src_userns;
36746 +
36747 +       h_src_userns = mnt_user_ns(h_src->mnt);
36748 +       h_src_dentry = h_src->dentry;
36749 +       ssz = vfs_getxattr_alloc(h_src_userns, h_src_dentry, name, buf, 0,
36750 +                                GFP_NOFS);
36751 +       err = ssz;
36752 +       if (unlikely(err <= 0)) {
36753 +               if (err == -ENODATA
36754 +                   || (err == -EOPNOTSUPP
36755 +                       && ((ignore_flags & au_xattr_out_of_list)
36756 +                           || (au_test_nfs_noacl(d_inode(h_src_dentry))
36757 +                               && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
36758 +                                   || !strcmp(name,
36759 +                                              XATTR_NAME_POSIX_ACL_DEFAULT))))
36760 +                           ))
36761 +                       err = 0;
36762 +               if (err && (verbose || au_debug_test()))
36763 +                       pr_err("%s, err %d\n", name, err);
36764 +               goto out;
36765 +       }
36766 +
36767 +       /* unlock it temporary */
36768 +       h_dst_userns = mnt_user_ns(h_dst->mnt);
36769 +       h_dst_dentry = h_dst->dentry;
36770 +       h_idst = d_inode(h_dst_dentry);
36771 +       inode_unlock(h_idst);
36772 +       err = vfsub_setxattr(h_dst_userns, h_dst_dentry, name, *buf, ssz,
36773 +                            /*flags*/0);
36774 +       inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36775 +       if (unlikely(err)) {
36776 +               if (verbose || au_debug_test())
36777 +                       pr_err("%s, err %d\n", name, err);
36778 +               err = au_xattr_ignore(err, name, ignore_flags);
36779 +       }
36780 +
36781 +out:
36782 +       return err;
36783 +}
36784 +
36785 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
36786 +                 unsigned int verbose)
36787 +{
36788 +       int err, unlocked, acl_access, acl_default;
36789 +       ssize_t ssz;
36790 +       struct dentry *h_dst_dentry, *h_src_dentry;
36791 +       struct inode *h_isrc, *h_idst;
36792 +       char *value, *p, *o, *e;
36793 +
36794 +       /* try stopping to update the source inode while we are referencing */
36795 +       /* there should not be the parent-child relationship between them */
36796 +       h_dst_dentry = h_dst->dentry;
36797 +       h_idst = d_inode(h_dst_dentry);
36798 +       h_src_dentry = h_src->dentry;
36799 +       h_isrc = d_inode(h_src_dentry);
36800 +       inode_unlock(h_idst);
36801 +       inode_lock_shared_nested(h_isrc, AuLsc_I_CHILD);
36802 +       inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36803 +       unlocked = 0;
36804 +
36805 +       /* some filesystems don't list POSIX ACL, for example tmpfs */
36806 +       ssz = vfs_listxattr(h_src_dentry, NULL, 0);
36807 +       err = ssz;
36808 +       if (unlikely(err < 0)) {
36809 +               AuTraceErr(err);
36810 +               if (err == -ENODATA
36811 +                   || err == -EOPNOTSUPP)
36812 +                       err = 0;        /* ignore */
36813 +               goto out;
36814 +       }
36815 +
36816 +       err = 0;
36817 +       p = NULL;
36818 +       o = NULL;
36819 +       if (ssz) {
36820 +               err = -ENOMEM;
36821 +               p = kmalloc(ssz, GFP_NOFS);
36822 +               o = p;
36823 +               if (unlikely(!p))
36824 +                       goto out;
36825 +               err = vfs_listxattr(h_src_dentry, p, ssz);
36826 +       }
36827 +       inode_unlock_shared(h_isrc);
36828 +       unlocked = 1;
36829 +       AuDbg("err %d, ssz %zd\n", err, ssz);
36830 +       if (unlikely(err < 0))
36831 +               goto out_free;
36832 +
36833 +       err = 0;
36834 +       e = p + ssz;
36835 +       value = NULL;
36836 +       acl_access = 0;
36837 +       acl_default = 0;
36838 +       while (!err && p < e) {
36839 +               acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
36840 +                                      sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
36841 +               acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
36842 +                                       sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
36843 +                                       - 1);
36844 +               err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
36845 +                                      verbose);
36846 +               p += strlen(p) + 1;
36847 +       }
36848 +       AuTraceErr(err);
36849 +       ignore_flags |= au_xattr_out_of_list;
36850 +       if (!err && !acl_access) {
36851 +               err = au_do_cpup_xattr(h_dst, h_src,
36852 +                                      XATTR_NAME_POSIX_ACL_ACCESS, &value,
36853 +                                      ignore_flags, verbose);
36854 +               AuTraceErr(err);
36855 +       }
36856 +       if (!err && !acl_default) {
36857 +               err = au_do_cpup_xattr(h_dst, h_src,
36858 +                                      XATTR_NAME_POSIX_ACL_DEFAULT, &value,
36859 +                                      ignore_flags, verbose);
36860 +               AuTraceErr(err);
36861 +       }
36862 +
36863 +       au_kfree_try_rcu(value);
36864 +
36865 +out_free:
36866 +       au_kfree_try_rcu(o);
36867 +out:
36868 +       if (!unlocked)
36869 +               inode_unlock_shared(h_isrc);
36870 +       AuTraceErr(err);
36871 +       return err;
36872 +}
36873 +
36874 +/* ---------------------------------------------------------------------- */
36875 +
36876 +static int au_smack_reentering(struct super_block *sb)
36877 +{
36878 +#if IS_ENABLED(CONFIG_SECURITY_SMACK) || IS_ENABLED(CONFIG_SECURITY_SELINUX)
36879 +       /*
36880 +        * as a part of lookup, smack_d_instantiate() is called, and it calls
36881 +        * i_op->getxattr(). ouch.
36882 +        */
36883 +       return si_pid_test(sb);
36884 +#else
36885 +       return 0;
36886 +#endif
36887 +}
36888 +
36889 +enum {
36890 +       AU_XATTR_LIST,
36891 +       AU_XATTR_GET
36892 +};
36893 +
36894 +struct au_lgxattr {
36895 +       int type;
36896 +       union {
36897 +               struct {
36898 +                       char    *list;
36899 +                       size_t  size;
36900 +               } list;
36901 +               struct {
36902 +                       const char      *name;
36903 +                       void            *value;
36904 +                       size_t          size;
36905 +               } get;
36906 +       } u;
36907 +};
36908 +
36909 +static ssize_t au_lgxattr(struct dentry *dentry, struct inode *inode,
36910 +                         struct au_lgxattr *arg)
36911 +{
36912 +       ssize_t err;
36913 +       int reenter;
36914 +       struct path h_path;
36915 +       struct super_block *sb;
36916 +
36917 +       sb = dentry->d_sb;
36918 +       reenter = au_smack_reentering(sb);
36919 +       if (!reenter) {
36920 +               err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
36921 +               if (unlikely(err))
36922 +                       goto out;
36923 +       }
36924 +       err = au_h_path_getattr(dentry, inode, /*force*/1, &h_path, reenter);
36925 +       if (unlikely(err))
36926 +               goto out_si;
36927 +       if (unlikely(!h_path.dentry))
36928 +               /* illegally overlapped or something */
36929 +               goto out_di; /* pretending success */
36930 +
36931 +       /* always topmost entry only */
36932 +       switch (arg->type) {
36933 +       case AU_XATTR_LIST:
36934 +               err = vfs_listxattr(h_path.dentry,
36935 +                                   arg->u.list.list, arg->u.list.size);
36936 +               break;
36937 +       case AU_XATTR_GET:
36938 +               AuDebugOn(d_is_negative(h_path.dentry));
36939 +               err = vfs_getxattr(mnt_user_ns(h_path.mnt), h_path.dentry,
36940 +                                  arg->u.get.name, arg->u.get.value,
36941 +                                  arg->u.get.size);
36942 +               break;
36943 +       }
36944 +
36945 +out_di:
36946 +       if (!reenter)
36947 +               di_read_unlock(dentry, AuLock_IR);
36948 +out_si:
36949 +       if (!reenter)
36950 +               si_read_unlock(sb);
36951 +out:
36952 +       AuTraceErr(err);
36953 +       return err;
36954 +}
36955 +
36956 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
36957 +{
36958 +       struct au_lgxattr arg = {
36959 +               .type = AU_XATTR_LIST,
36960 +               .u.list = {
36961 +                       .list   = list,
36962 +                       .size   = size
36963 +               },
36964 +       };
36965 +
36966 +       return au_lgxattr(dentry, /*inode*/NULL, &arg);
36967 +}
36968 +
36969 +static ssize_t au_getxattr(struct dentry *dentry, struct inode *inode,
36970 +                          const char *name, void *value, size_t size)
36971 +{
36972 +       struct au_lgxattr arg = {
36973 +               .type = AU_XATTR_GET,
36974 +               .u.get = {
36975 +                       .name   = name,
36976 +                       .value  = value,
36977 +                       .size   = size
36978 +               },
36979 +       };
36980 +
36981 +       return au_lgxattr(dentry, inode, &arg);
36982 +}
36983 +
36984 +static int au_setxattr(struct dentry *dentry, struct inode *inode,
36985 +                      const char *name, void *value, size_t size,
36986 +                      int flags)
36987 +{
36988 +       struct au_sxattr arg = {
36989 +               .type = AU_XATTR_SET,
36990 +               .u.set = {
36991 +                       .name   = name,
36992 +                       .value  = value,
36993 +                       .size   = size,
36994 +                       .flags  = flags
36995 +               },
36996 +       };
36997 +
36998 +       return au_sxattr(dentry, inode, &arg);
36999 +}
37000 +
37001 +/* ---------------------------------------------------------------------- */
37002 +
37003 +static int au_xattr_get(const struct xattr_handler *handler,
37004 +                       struct dentry *dentry, struct inode *inode,
37005 +                       const char *name, void *buffer, size_t size)
37006 +{
37007 +       return au_getxattr(dentry, inode, name, buffer, size);
37008 +}
37009 +
37010 +/*
37011 + * The prameter 'void *value' of vfs_setxattr() is NOT const, but
37012 + * xattr_handler->set() requires const. Is it OK? I don't know.
37013 + */
37014 +static int au_xattr_set(const struct xattr_handler *handler,
37015 +                       struct user_namespace *userns,
37016 +                       struct dentry *dentry, struct inode *inode,
37017 +                       const char *name, const void *value, size_t size,
37018 +                       int flags)
37019 +{
37020 +       return au_setxattr(dentry, inode, name, (void *)value, size, flags);
37021 +}
37022 +
37023 +static const struct xattr_handler au_xattr_handler = {
37024 +       .name   = "",
37025 +       .prefix = "",
37026 +       .get    = au_xattr_get,
37027 +       .set    = au_xattr_set
37028 +};
37029 +
37030 +static const struct xattr_handler *au_xattr_handlers[] = {
37031 +#ifdef CONFIG_FS_POSIX_ACL
37032 +       &posix_acl_access_xattr_handler,
37033 +       &posix_acl_default_xattr_handler,
37034 +#endif
37035 +       &au_xattr_handler, /* must be last */
37036 +       NULL
37037 +};
37038 +
37039 +void au_xattr_init(struct super_block *sb)
37040 +{
37041 +       sb->s_xattr = au_xattr_handlers;
37042 +}
37043 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
37044 --- /usr/share/empty/fs/aufs/xino.c     1970-01-01 01:00:00.000000000 +0100
37045 +++ linux/fs/aufs/xino.c        2022-11-05 23:02:18.972555950 +0100
37046 @@ -0,0 +1,1926 @@
37047 +// SPDX-License-Identifier: GPL-2.0
37048 +/*
37049 + * Copyright (C) 2005-2022 Junjiro R. Okajima
37050 + *
37051 + * This program is free software; you can redistribute it and/or modify
37052 + * it under the terms of the GNU General Public License as published by
37053 + * the Free Software Foundation; either version 2 of the License, or
37054 + * (at your option) any later version.
37055 + *
37056 + * This program is distributed in the hope that it will be useful,
37057 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
37058 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
37059 + * GNU General Public License for more details.
37060 + *
37061 + * You should have received a copy of the GNU General Public License
37062 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
37063 + */
37064 +
37065 +/*
37066 + * external inode number translation table and bitmap
37067 + *
37068 + * things to consider
37069 + * - the lifetime
37070 + *   + au_xino object
37071 + *   + XINO files (xino, xib, xigen)
37072 + *   + dynamic debugfs entries (xiN)
37073 + *   + static debugfs entries (xib, xigen)
37074 + *   + static sysfs entry (xi_path)
37075 + * - several entry points to handle them.
37076 + *   + mount(2) without xino option (default)
37077 + *   + mount(2) with xino option
37078 + *   + mount(2) with noxino option
37079 + *   + umount(2)
37080 + *   + remount with add/del branches
37081 + *   + remount with xino/noxino options
37082 + */
37083 +
37084 +#include <linux/seq_file.h>
37085 +#include <linux/statfs.h>
37086 +#include "aufs.h"
37087 +
37088 +static aufs_bindex_t sbr_find_shared(struct super_block *sb, aufs_bindex_t btop,
37089 +                                    aufs_bindex_t bbot,
37090 +                                    struct super_block *h_sb)
37091 +{
37092 +       /* todo: try binary-search if the branches are many */
37093 +       for (; btop <= bbot; btop++)
37094 +               if (h_sb == au_sbr_sb(sb, btop))
37095 +                       return btop;
37096 +       return -1;
37097 +}
37098 +
37099 +/*
37100 + * find another branch who is on the same filesystem of the specified
37101 + * branch{@btgt}. search until @bbot.
37102 + */
37103 +static aufs_bindex_t is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
37104 +                                 aufs_bindex_t bbot)
37105 +{
37106 +       aufs_bindex_t bindex;
37107 +       struct super_block *tgt_sb;
37108 +
37109 +       tgt_sb = au_sbr_sb(sb, btgt);
37110 +       bindex = sbr_find_shared(sb, /*btop*/0, btgt - 1, tgt_sb);
37111 +       if (bindex < 0)
37112 +               bindex = sbr_find_shared(sb, btgt + 1, bbot, tgt_sb);
37113 +
37114 +       return bindex;
37115 +}
37116 +
37117 +/* ---------------------------------------------------------------------- */
37118 +
37119 +/*
37120 + * stop unnecessary notify events at creating xino files
37121 + */
37122 +
37123 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry)
37124 +{
37125 +       aufs_bindex_t bfound, bindex, bbot;
37126 +       struct dentry *parent;
37127 +       struct au_branch *br;
37128 +
37129 +       bfound = -1;
37130 +       parent = dentry->d_parent; /* safe d_parent access */
37131 +       bbot = au_sbbot(sb);
37132 +       for (bindex = 0; bindex <= bbot; bindex++) {
37133 +               br = au_sbr(sb, bindex);
37134 +               if (au_br_dentry(br) == parent) {
37135 +                       bfound = bindex;
37136 +                       break;
37137 +               }
37138 +       }
37139 +
37140 +       AuDbg("bfound b%d\n", bfound);
37141 +       return bfound;
37142 +}
37143 +
37144 +struct au_xino_lock_dir {
37145 +       struct au_hinode *hdir;
37146 +       struct dentry *parent;
37147 +       struct inode *dir;
37148 +};
37149 +
37150 +static struct dentry *au_dget_parent_lock(struct dentry *dentry,
37151 +                                         unsigned int lsc)
37152 +{
37153 +       struct dentry *parent;
37154 +       struct inode *dir;
37155 +
37156 +       parent = dget_parent(dentry);
37157 +       dir = d_inode(parent);
37158 +       inode_lock_nested(dir, lsc);
37159 +#if 0 /* it should not happen */
37160 +       spin_lock(&dentry->d_lock);
37161 +       if (unlikely(dentry->d_parent != parent)) {
37162 +               spin_unlock(&dentry->d_lock);
37163 +               inode_unlock(dir);
37164 +               dput(parent);
37165 +               parent = NULL;
37166 +               goto out;
37167 +       }
37168 +       spin_unlock(&dentry->d_lock);
37169 +
37170 +out:
37171 +#endif
37172 +       return parent;
37173 +}
37174 +
37175 +static void au_xino_lock_dir(struct super_block *sb, struct path *xipath,
37176 +                            struct au_xino_lock_dir *ldir)
37177 +{
37178 +       aufs_bindex_t bindex;
37179 +
37180 +       ldir->hdir = NULL;
37181 +       bindex = au_xi_root(sb, xipath->dentry);
37182 +       if (bindex >= 0) {
37183 +               /* rw branch root */
37184 +               ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
37185 +               au_hn_inode_lock_nested(ldir->hdir, AuLsc_I_PARENT);
37186 +       } else {
37187 +               /* other */
37188 +               ldir->parent = au_dget_parent_lock(xipath->dentry,
37189 +                                                  AuLsc_I_PARENT);
37190 +               ldir->dir = d_inode(ldir->parent);
37191 +       }
37192 +}
37193 +
37194 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
37195 +{
37196 +       if (ldir->hdir)
37197 +               au_hn_inode_unlock(ldir->hdir);
37198 +       else {
37199 +               inode_unlock(ldir->dir);
37200 +               dput(ldir->parent);
37201 +       }
37202 +}
37203 +
37204 +/* ---------------------------------------------------------------------- */
37205 +
37206 +/*
37207 + * create and set a new xino file
37208 + */
37209 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
37210 +                           int wbrtop)
37211 +{
37212 +       struct file *file;
37213 +       struct dentry *h_parent, *d;
37214 +       struct inode *h_dir, *inode;
37215 +       int err;
37216 +       static DEFINE_MUTEX(mtx);
37217 +
37218 +       /*
37219 +        * at mount-time, and the xino file is the default path,
37220 +        * hnotify is disabled so we have no notify events to ignore.
37221 +        * when a user specified the xino, we cannot get au_hdir to be ignored.
37222 +        */
37223 +       if (!wbrtop)
37224 +               mutex_lock(&mtx);
37225 +       file = vfsub_filp_open(fpath, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37226 +                              /* | __FMODE_NONOTIFY */,
37227 +                              0666);
37228 +       if (IS_ERR(file)) {
37229 +               if (!wbrtop)
37230 +                       mutex_unlock(&mtx);
37231 +               if (!silent)
37232 +                       pr_err("open %s(%ld)\n", fpath, PTR_ERR(file));
37233 +               return file;
37234 +       }
37235 +
37236 +       /* keep file count */
37237 +       err = 0;
37238 +       d = file->f_path.dentry;
37239 +       h_parent = au_dget_parent_lock(d, AuLsc_I_PARENT);
37240 +       if (!wbrtop)
37241 +               mutex_unlock(&mtx);
37242 +       /* mnt_want_write() is unnecessary here */
37243 +       h_dir = d_inode(h_parent);
37244 +       inode = file_inode(file);
37245 +       /* no delegation since it is just created */
37246 +       if (inode->i_nlink)
37247 +               err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
37248 +                                  /*force*/0);
37249 +       inode_unlock(h_dir);
37250 +       dput(h_parent);
37251 +       if (unlikely(err)) {
37252 +               if (!silent)
37253 +                       pr_err("unlink %s(%d)\n", fpath, err);
37254 +               goto out;
37255 +       }
37256 +
37257 +       err = -EINVAL;
37258 +       if (unlikely(sb && sb == d->d_sb)) {
37259 +               if (!silent)
37260 +                       pr_err("%s must be outside\n", fpath);
37261 +               goto out;
37262 +       }
37263 +       if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
37264 +               if (!silent)
37265 +                       pr_err("xino doesn't support %s(%s)\n",
37266 +                              fpath, au_sbtype(d->d_sb));
37267 +               goto out;
37268 +       }
37269 +       return file; /* success */
37270 +
37271 +out:
37272 +       fput(file);
37273 +       file = ERR_PTR(err);
37274 +       return file;
37275 +}
37276 +
37277 +/*
37278 + * create a new xinofile at the same place/path as @base.
37279 + */
37280 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
37281 +                            struct file *copy_src)
37282 +{
37283 +       struct file *file;
37284 +       struct dentry *dentry;
37285 +       struct inode *dir, *delegated;
37286 +       struct qstr *name;
37287 +       struct path ppath, path;
37288 +       int err, do_unlock;
37289 +       struct au_xino_lock_dir ldir;
37290 +
37291 +       do_unlock = 1;
37292 +       au_xino_lock_dir(sb, base, &ldir);
37293 +       dentry = base->dentry;
37294 +       ppath.dentry = dentry->d_parent; /* dir inode is locked */
37295 +       ppath.mnt = base->mnt;
37296 +       dir = d_inode(ppath.dentry);
37297 +       IMustLock(dir);
37298 +
37299 +       name = &dentry->d_name;
37300 +       path.dentry = vfsub_lookup_one_len(name->name, &ppath, name->len);
37301 +       if (IS_ERR(path.dentry)) {
37302 +               file = (void *)path.dentry;
37303 +               pr_err("%pd lookup err %ld\n", dentry, PTR_ERR(path.dentry));
37304 +               goto out;
37305 +       }
37306 +
37307 +       /* no need to mnt_want_write() since we call dentry_open() later */
37308 +       err = vfs_create(mnt_user_ns(base->mnt), dir, path.dentry, 0666, NULL);
37309 +       if (unlikely(err)) {
37310 +               file = ERR_PTR(err);
37311 +               pr_err("%pd create err %d\n", dentry, err);
37312 +               goto out_dput;
37313 +       }
37314 +
37315 +       path.mnt = base->mnt;
37316 +       file = vfsub_dentry_open(&path,
37317 +                                O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37318 +                                /* | __FMODE_NONOTIFY */);
37319 +       if (IS_ERR(file)) {
37320 +               pr_err("%pd open err %ld\n", dentry, PTR_ERR(file));
37321 +               goto out_dput;
37322 +       }
37323 +
37324 +       delegated = NULL;
37325 +       err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
37326 +       au_xino_unlock_dir(&ldir);
37327 +       do_unlock = 0;
37328 +       if (unlikely(err == -EWOULDBLOCK)) {
37329 +               pr_warn("cannot retry for NFSv4 delegation"
37330 +                       " for an internal unlink\n");
37331 +               iput(delegated);
37332 +       }
37333 +       if (unlikely(err)) {
37334 +               pr_err("%pd unlink err %d\n", dentry, err);
37335 +               goto out_fput;
37336 +       }
37337 +
37338 +       if (copy_src) {
37339 +               /* no one can touch copy_src xino */
37340 +               err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
37341 +               if (unlikely(err)) {
37342 +                       pr_err("%pd copy err %d\n", dentry, err);
37343 +                       goto out_fput;
37344 +               }
37345 +       }
37346 +       goto out_dput; /* success */
37347 +
37348 +out_fput:
37349 +       fput(file);
37350 +       file = ERR_PTR(err);
37351 +out_dput:
37352 +       dput(path.dentry);
37353 +out:
37354 +       if (do_unlock)
37355 +               au_xino_unlock_dir(&ldir);
37356 +       return file;
37357 +}
37358 +
37359 +struct file *au_xino_file1(struct au_xino *xi)
37360 +{
37361 +       struct file *file;
37362 +       unsigned int u, nfile;
37363 +
37364 +       file = NULL;
37365 +       nfile = xi->xi_nfile;
37366 +       for (u = 0; u < nfile; u++) {
37367 +               file = xi->xi_file[u];
37368 +               if (file)
37369 +                       break;
37370 +       }
37371 +
37372 +       return file;
37373 +}
37374 +
37375 +static int au_xino_file_set(struct au_xino *xi, int idx, struct file *file)
37376 +{
37377 +       int err;
37378 +       struct file *f;
37379 +       void *p;
37380 +
37381 +       if (file)
37382 +               get_file(file);
37383 +
37384 +       err = 0;
37385 +       f = NULL;
37386 +       if (idx < xi->xi_nfile) {
37387 +               f = xi->xi_file[idx];
37388 +               if (f)
37389 +                       fput(f);
37390 +       } else {
37391 +               p = au_kzrealloc(xi->xi_file,
37392 +                                sizeof(*xi->xi_file) * xi->xi_nfile,
37393 +                                sizeof(*xi->xi_file) * (idx + 1),
37394 +                                GFP_NOFS, /*may_shrink*/0);
37395 +               if (p) {
37396 +                       MtxMustLock(&xi->xi_mtx);
37397 +                       xi->xi_file = p;
37398 +                       xi->xi_nfile = idx + 1;
37399 +               } else {
37400 +                       err = -ENOMEM;
37401 +                       if (file)
37402 +                               fput(file);
37403 +                       goto out;
37404 +               }
37405 +       }
37406 +       xi->xi_file[idx] = file;
37407 +
37408 +out:
37409 +       return err;
37410 +}
37411 +
37412 +/*
37413 + * if @xinew->xi is not set, then create new xigen file.
37414 + */
37415 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew)
37416 +{
37417 +       struct file *file;
37418 +       int err;
37419 +
37420 +       SiMustAnyLock(sb);
37421 +
37422 +       file = au_xino_create2(sb, xinew->base, xinew->copy_src);
37423 +       if (IS_ERR(file)) {
37424 +               err = PTR_ERR(file);
37425 +               pr_err("%s[%d], err %d\n",
37426 +                      xinew->xi ? "xino" : "xigen",
37427 +                      xinew->idx, err);
37428 +               goto out;
37429 +       }
37430 +
37431 +       if (xinew->xi)
37432 +               err = au_xino_file_set(xinew->xi, xinew->idx, file);
37433 +       else {
37434 +               BUG();
37435 +               /* todo: make xigen file an array */
37436 +               /* err = au_xigen_file_set(sb, xinew->idx, file); */
37437 +       }
37438 +       fput(file);
37439 +       if (unlikely(err))
37440 +               file = ERR_PTR(err);
37441 +
37442 +out:
37443 +       return file;
37444 +}
37445 +
37446 +/* ---------------------------------------------------------------------- */
37447 +
37448 +/*
37449 + * truncate xino files
37450 + */
37451 +static int au_xino_do_trunc(struct super_block *sb, aufs_bindex_t bindex,
37452 +                           int idx, struct kstatfs *st)
37453 +{
37454 +       int err;
37455 +       blkcnt_t blocks;
37456 +       struct file *file, *new_xino;
37457 +       struct au_xi_new xinew = {
37458 +               .idx = idx
37459 +       };
37460 +
37461 +       err = 0;
37462 +       xinew.xi = au_sbr(sb, bindex)->br_xino;
37463 +       file = au_xino_file(xinew.xi, idx);
37464 +       if (!file)
37465 +               goto out;
37466 +
37467 +       xinew.base = &file->f_path;
37468 +       err = vfs_statfs(xinew.base, st);
37469 +       if (unlikely(err)) {
37470 +               AuErr1("statfs err %d, ignored\n", err);
37471 +               err = 0;
37472 +               goto out;
37473 +       }
37474 +
37475 +       blocks = file_inode(file)->i_blocks;
37476 +       pr_info("begin truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37477 +               bindex, idx, (u64)blocks, st->f_bfree, st->f_blocks);
37478 +
37479 +       xinew.copy_src = file;
37480 +       new_xino = au_xi_new(sb, &xinew);
37481 +       if (IS_ERR(new_xino)) {
37482 +               err = PTR_ERR(new_xino);
37483 +               pr_err("xino(b%d-%d), err %d, ignored\n", bindex, idx, err);
37484 +               goto out;
37485 +       }
37486 +
37487 +       err = vfs_statfs(&new_xino->f_path, st);
37488 +       if (!err)
37489 +               pr_info("end truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37490 +                       bindex, idx, (u64)file_inode(new_xino)->i_blocks,
37491 +                       st->f_bfree, st->f_blocks);
37492 +       else {
37493 +               AuErr1("statfs err %d, ignored\n", err);
37494 +               err = 0;
37495 +       }
37496 +
37497 +out:
37498 +       return err;
37499 +}
37500 +
37501 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin)
37502 +{
37503 +       int err, i;
37504 +       unsigned long jiffy;
37505 +       aufs_bindex_t bbot;
37506 +       struct kstatfs *st;
37507 +       struct au_branch *br;
37508 +       struct au_xino *xi;
37509 +
37510 +       err = -ENOMEM;
37511 +       st = kmalloc(sizeof(*st), GFP_NOFS);
37512 +       if (unlikely(!st))
37513 +               goto out;
37514 +
37515 +       err = -EINVAL;
37516 +       bbot = au_sbbot(sb);
37517 +       if (unlikely(bindex < 0 || bbot < bindex))
37518 +               goto out_st;
37519 +
37520 +       err = 0;
37521 +       jiffy = jiffies;
37522 +       br = au_sbr(sb, bindex);
37523 +       xi = br->br_xino;
37524 +       for (i = idx_begin; !err && i < xi->xi_nfile; i++)
37525 +               err = au_xino_do_trunc(sb, bindex, i, st);
37526 +       if (!err)
37527 +               au_sbi(sb)->si_xino_jiffy = jiffy;
37528 +
37529 +out_st:
37530 +       au_kfree_rcu(st);
37531 +out:
37532 +       return err;
37533 +}
37534 +
37535 +struct xino_do_trunc_args {
37536 +       struct super_block *sb;
37537 +       struct au_branch *br;
37538 +       int idx;
37539 +};
37540 +
37541 +static void xino_do_trunc(void *_args)
37542 +{
37543 +       struct xino_do_trunc_args *args = _args;
37544 +       struct super_block *sb;
37545 +       struct au_branch *br;
37546 +       struct inode *dir;
37547 +       int err, idx;
37548 +       aufs_bindex_t bindex;
37549 +
37550 +       err = 0;
37551 +       sb = args->sb;
37552 +       dir = d_inode(sb->s_root);
37553 +       br = args->br;
37554 +       idx = args->idx;
37555 +
37556 +       si_noflush_write_lock(sb);
37557 +       ii_read_lock_parent(dir);
37558 +       bindex = au_br_index(sb, br->br_id);
37559 +       err = au_xino_trunc(sb, bindex, idx);
37560 +       ii_read_unlock(dir);
37561 +       if (unlikely(err))
37562 +               pr_warn("err b%d, (%d)\n", bindex, err);
37563 +       atomic_dec(&br->br_xino->xi_truncating);
37564 +       au_lcnt_dec(&br->br_count);
37565 +       si_write_unlock(sb);
37566 +       au_nwt_done(&au_sbi(sb)->si_nowait);
37567 +       au_kfree_rcu(args);
37568 +}
37569 +
37570 +/*
37571 + * returns the index in the xi_file array whose corresponding file is necessary
37572 + * to truncate, or -1 which means no need to truncate.
37573 + */
37574 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
37575 +{
37576 +       int err;
37577 +       unsigned int u;
37578 +       struct kstatfs st;
37579 +       struct au_sbinfo *sbinfo;
37580 +       struct au_xino *xi;
37581 +       struct file *file;
37582 +
37583 +       /* todo: si_xino_expire and the ratio should be customizable */
37584 +       sbinfo = au_sbi(sb);
37585 +       if (time_before(jiffies,
37586 +                       sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
37587 +               return -1;
37588 +
37589 +       /* truncation border */
37590 +       xi = br->br_xino;
37591 +       for (u = 0; u < xi->xi_nfile; u++) {
37592 +               file = au_xino_file(xi, u);
37593 +               if (!file)
37594 +                       continue;
37595 +
37596 +               err = vfs_statfs(&file->f_path, &st);
37597 +               if (unlikely(err)) {
37598 +                       AuErr1("statfs err %d, ignored\n", err);
37599 +                       return -1;
37600 +               }
37601 +               if (div64_u64(st.f_bfree * 100, st.f_blocks)
37602 +                   >= AUFS_XINO_DEF_TRUNC)
37603 +                       return u;
37604 +       }
37605 +
37606 +       return -1;
37607 +}
37608 +
37609 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
37610 +{
37611 +       int idx;
37612 +       struct xino_do_trunc_args *args;
37613 +       int wkq_err;
37614 +
37615 +       idx = xino_trunc_test(sb, br);
37616 +       if (idx < 0)
37617 +               return;
37618 +
37619 +       if (atomic_inc_return(&br->br_xino->xi_truncating) > 1)
37620 +               goto out;
37621 +
37622 +       /* lock and kfree() will be called in trunc_xino() */
37623 +       args = kmalloc(sizeof(*args), GFP_NOFS);
37624 +       if (unlikely(!args)) {
37625 +               AuErr1("no memory\n");
37626 +               goto out;
37627 +       }
37628 +
37629 +       au_lcnt_inc(&br->br_count);
37630 +       args->sb = sb;
37631 +       args->br = br;
37632 +       args->idx = idx;
37633 +       wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
37634 +       if (!wkq_err)
37635 +               return; /* success */
37636 +
37637 +       pr_err("wkq %d\n", wkq_err);
37638 +       au_lcnt_dec(&br->br_count);
37639 +       au_kfree_rcu(args);
37640 +
37641 +out:
37642 +       atomic_dec(&br->br_xino->xi_truncating);
37643 +}
37644 +
37645 +/* ---------------------------------------------------------------------- */
37646 +
37647 +struct au_xi_calc {
37648 +       int idx;
37649 +       loff_t pos;
37650 +};
37651 +
37652 +static void au_xi_calc(struct super_block *sb, ino_t h_ino,
37653 +                      struct au_xi_calc *calc)
37654 +{
37655 +       loff_t maxent;
37656 +
37657 +       maxent = au_xi_maxent(sb);
37658 +       calc->idx = div64_u64_rem(h_ino, maxent, &calc->pos);
37659 +       calc->pos *= sizeof(ino_t);
37660 +}
37661 +
37662 +static int au_xino_do_new_async(struct super_block *sb, struct au_branch *br,
37663 +                               struct au_xi_calc *calc)
37664 +{
37665 +       int err;
37666 +       struct file *file;
37667 +       struct au_xino *xi = br->br_xino;
37668 +       struct au_xi_new xinew = {
37669 +               .xi = xi
37670 +       };
37671 +
37672 +       SiMustAnyLock(sb);
37673 +
37674 +       err = 0;
37675 +       if (!xi)
37676 +               goto out;
37677 +
37678 +       mutex_lock(&xi->xi_mtx);
37679 +       file = au_xino_file(xi, calc->idx);
37680 +       if (file)
37681 +               goto out_mtx;
37682 +
37683 +       file = au_xino_file(xi, /*idx*/-1);
37684 +       AuDebugOn(!file);
37685 +       xinew.idx = calc->idx;
37686 +       xinew.base = &file->f_path;
37687 +       /* xinew.copy_src = NULL; */
37688 +       file = au_xi_new(sb, &xinew);
37689 +       if (IS_ERR(file))
37690 +               err = PTR_ERR(file);
37691 +
37692 +out_mtx:
37693 +       mutex_unlock(&xi->xi_mtx);
37694 +out:
37695 +       return err;
37696 +}
37697 +
37698 +struct au_xino_do_new_async_args {
37699 +       struct super_block *sb;
37700 +       struct au_branch *br;
37701 +       struct au_xi_calc calc;
37702 +       ino_t ino;
37703 +};
37704 +
37705 +struct au_xi_writing {
37706 +       struct hlist_bl_node node;
37707 +       ino_t h_ino, ino;
37708 +};
37709 +
37710 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37711 +                           ino_t ino);
37712 +
37713 +static void au_xino_call_do_new_async(void *args)
37714 +{
37715 +       struct au_xino_do_new_async_args *a = args;
37716 +       struct au_branch *br;
37717 +       struct super_block *sb;
37718 +       struct au_sbinfo *sbi;
37719 +       struct inode *root;
37720 +       struct file *file;
37721 +       struct au_xi_writing *del, *p;
37722 +       struct hlist_bl_head *hbl;
37723 +       struct hlist_bl_node *pos;
37724 +       int err;
37725 +
37726 +       br = a->br;
37727 +       sb = a->sb;
37728 +       sbi = au_sbi(sb);
37729 +       si_noflush_read_lock(sb);
37730 +       root = d_inode(sb->s_root);
37731 +       ii_read_lock_child(root);
37732 +       err = au_xino_do_new_async(sb, br, &a->calc);
37733 +       if (unlikely(err)) {
37734 +               AuIOErr("err %d\n", err);
37735 +               goto out;
37736 +       }
37737 +
37738 +       file = au_xino_file(br->br_xino, a->calc.idx);
37739 +       AuDebugOn(!file);
37740 +       err = au_xino_do_write(file, &a->calc, a->ino);
37741 +       if (unlikely(err)) {
37742 +               AuIOErr("err %d\n", err);
37743 +               goto out;
37744 +       }
37745 +
37746 +       del = NULL;
37747 +       hbl = &br->br_xino->xi_writing;
37748 +       hlist_bl_lock(hbl);
37749 +       au_hbl_for_each(pos, hbl) {
37750 +               p = container_of(pos, struct au_xi_writing, node);
37751 +               if (p->ino == a->ino) {
37752 +                       del = p;
37753 +                       hlist_bl_del(&p->node);
37754 +                       break;
37755 +               }
37756 +       }
37757 +       hlist_bl_unlock(hbl);
37758 +       au_kfree_rcu(del);
37759 +
37760 +out:
37761 +       au_lcnt_dec(&br->br_count);
37762 +       ii_read_unlock(root);
37763 +       si_read_unlock(sb);
37764 +       au_nwt_done(&sbi->si_nowait);
37765 +       au_kfree_rcu(a);
37766 +}
37767 +
37768 +/*
37769 + * create a new xino file asynchronously
37770 + */
37771 +static int au_xino_new_async(struct super_block *sb, struct au_branch *br,
37772 +                            struct au_xi_calc *calc, ino_t ino)
37773 +{
37774 +       int err;
37775 +       struct au_xino_do_new_async_args *arg;
37776 +
37777 +       err = -ENOMEM;
37778 +       arg = kmalloc(sizeof(*arg), GFP_NOFS);
37779 +       if (unlikely(!arg))
37780 +               goto out;
37781 +
37782 +       arg->sb = sb;
37783 +       arg->br = br;
37784 +       arg->calc = *calc;
37785 +       arg->ino = ino;
37786 +       au_lcnt_inc(&br->br_count);
37787 +       err = au_wkq_nowait(au_xino_call_do_new_async, arg, sb, AuWkq_NEST);
37788 +       if (unlikely(err)) {
37789 +               pr_err("wkq %d\n", err);
37790 +               au_lcnt_dec(&br->br_count);
37791 +               au_kfree_rcu(arg);
37792 +       }
37793 +
37794 +out:
37795 +       return err;
37796 +}
37797 +
37798 +/*
37799 + * read @ino from xinofile for the specified branch{@sb, @bindex}
37800 + * at the position of @h_ino.
37801 + */
37802 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37803 +                ino_t *ino)
37804 +{
37805 +       int err;
37806 +       ssize_t sz;
37807 +       struct au_xi_calc calc;
37808 +       struct au_sbinfo *sbinfo;
37809 +       struct file *file;
37810 +       struct au_xino *xi;
37811 +       struct hlist_bl_head *hbl;
37812 +       struct hlist_bl_node *pos;
37813 +       struct au_xi_writing *p;
37814 +
37815 +       *ino = 0;
37816 +       if (!au_opt_test(au_mntflags(sb), XINO))
37817 +               return 0; /* no xino */
37818 +
37819 +       err = 0;
37820 +       au_xi_calc(sb, h_ino, &calc);
37821 +       xi = au_sbr(sb, bindex)->br_xino;
37822 +       file = au_xino_file(xi, calc.idx);
37823 +       if (!file) {
37824 +               hbl = &xi->xi_writing;
37825 +               hlist_bl_lock(hbl);
37826 +               au_hbl_for_each(pos, hbl) {
37827 +                       p = container_of(pos, struct au_xi_writing, node);
37828 +                       if (p->h_ino == h_ino) {
37829 +                               AuDbg("hi%llu, i%llu, found\n",
37830 +                                     (u64)p->h_ino, (u64)p->ino);
37831 +                               *ino = p->ino;
37832 +                               break;
37833 +                       }
37834 +               }
37835 +               hlist_bl_unlock(hbl);
37836 +               return 0;
37837 +       } else if (vfsub_f_size_read(file) < calc.pos + sizeof(*ino))
37838 +               return 0; /* no xino */
37839 +
37840 +       sbinfo = au_sbi(sb);
37841 +       sz = xino_fread(file, ino, sizeof(*ino), &calc.pos);
37842 +       if (sz == sizeof(*ino))
37843 +               return 0; /* success */
37844 +
37845 +       err = sz;
37846 +       if (unlikely(sz >= 0)) {
37847 +               err = -EIO;
37848 +               AuIOErr("xino read error (%zd)\n", sz);
37849 +       }
37850 +       return err;
37851 +}
37852 +
37853 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37854 +                           ino_t ino)
37855 +{
37856 +       ssize_t sz;
37857 +
37858 +       sz = xino_fwrite(file, &ino, sizeof(ino), &calc->pos);
37859 +       if (sz == sizeof(ino))
37860 +               return 0; /* success */
37861 +
37862 +       AuIOErr("write failed (%zd)\n", sz);
37863 +       return -EIO;
37864 +}
37865 +
37866 +/*
37867 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
37868 + * at the position of @h_ino.
37869 + * even if @ino is zero, it is written to the xinofile and means no entry.
37870 + * if the size of the xino file on a specific filesystem exceeds the watermark,
37871 + * try truncating it.
37872 + */
37873 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37874 +                 ino_t ino)
37875 +{
37876 +       int err;
37877 +       unsigned int mnt_flags;
37878 +       struct au_xi_calc calc;
37879 +       struct file *file;
37880 +       struct au_branch *br;
37881 +       struct au_xino *xi;
37882 +       struct au_xi_writing *p;
37883 +
37884 +       SiMustAnyLock(sb);
37885 +
37886 +       mnt_flags = au_mntflags(sb);
37887 +       if (!au_opt_test(mnt_flags, XINO))
37888 +               return 0;
37889 +
37890 +       au_xi_calc(sb, h_ino, &calc);
37891 +       br = au_sbr(sb, bindex);
37892 +       xi = br->br_xino;
37893 +       file = au_xino_file(xi, calc.idx);
37894 +       if (!file) {
37895 +               /* store the inum pair into the list */
37896 +               p = kmalloc(sizeof(*p), GFP_NOFS | __GFP_NOFAIL);
37897 +               p->h_ino = h_ino;
37898 +               p->ino = ino;
37899 +               au_hbl_add(&p->node, &xi->xi_writing);
37900 +
37901 +               /* create and write a new xino file asynchronously */
37902 +               err = au_xino_new_async(sb, br, &calc, ino);
37903 +               if (!err)
37904 +                       return 0; /* success */
37905 +               goto out;
37906 +       }
37907 +
37908 +       err = au_xino_do_write(file, &calc, ino);
37909 +       if (!err) {
37910 +               br = au_sbr(sb, bindex);
37911 +               if (au_opt_test(mnt_flags, TRUNC_XINO)
37912 +                   && au_test_fs_trunc_xino(au_br_sb(br)))
37913 +                       xino_try_trunc(sb, br);
37914 +               return 0; /* success */
37915 +       }
37916 +
37917 +out:
37918 +       AuIOErr("write failed (%d)\n", err);
37919 +       return -EIO;
37920 +}
37921 +
37922 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37923 +                             loff_t *pos);
37924 +
37925 +/* todo: unnecessary to support mmap_sem since kernel-space? */
37926 +ssize_t xino_fread(struct file *file, void *kbuf, size_t size, loff_t *pos)
37927 +{
37928 +       ssize_t err;
37929 +       int i;
37930 +       const int prevent_endless = 10;
37931 +
37932 +       i = 0;
37933 +       do {
37934 +               err = vfsub_read_k(file, kbuf, size, pos);
37935 +               if (err == -EINTR
37936 +                   && !au_wkq_test()
37937 +                   && fatal_signal_pending(current)) {
37938 +                       err = xino_fread_wkq(file, kbuf, size, pos);
37939 +                       BUG_ON(err == -EINTR);
37940 +               }
37941 +       } while (i++ < prevent_endless
37942 +                && (err == -EAGAIN || err == -EINTR));
37943 +
37944 +#if 0 /* reserved for future use */
37945 +       if (err > 0)
37946 +               fsnotify_access(file->f_path.dentry);
37947 +#endif
37948 +
37949 +       return err;
37950 +}
37951 +
37952 +struct xino_fread_args {
37953 +       ssize_t *errp;
37954 +       struct file *file;
37955 +       void *buf;
37956 +       size_t size;
37957 +       loff_t *pos;
37958 +};
37959 +
37960 +static void call_xino_fread(void *args)
37961 +{
37962 +       struct xino_fread_args *a = args;
37963 +       *a->errp = xino_fread(a->file, a->buf, a->size, a->pos);
37964 +}
37965 +
37966 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37967 +                             loff_t *pos)
37968 +{
37969 +       ssize_t err;
37970 +       int wkq_err;
37971 +       struct xino_fread_args args = {
37972 +               .errp   = &err,
37973 +               .file   = file,
37974 +               .buf    = buf,
37975 +               .size   = size,
37976 +               .pos    = pos
37977 +       };
37978 +
37979 +       wkq_err = au_wkq_wait(call_xino_fread, &args);
37980 +       if (unlikely(wkq_err))
37981 +               err = wkq_err;
37982 +
37983 +       return err;
37984 +}
37985 +
37986 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
37987 +                              loff_t *pos);
37988 +
37989 +static ssize_t do_xino_fwrite(struct file *file, void *kbuf, size_t size,
37990 +                             loff_t *pos)
37991 +{
37992 +       ssize_t err;
37993 +       int i;
37994 +       const int prevent_endless = 10;
37995 +
37996 +       i = 0;
37997 +       do {
37998 +               err = vfsub_write_k(file, kbuf, size, pos);
37999 +               if (err == -EINTR
38000 +                   && !au_wkq_test()
38001 +                   && fatal_signal_pending(current)) {
38002 +                       err = xino_fwrite_wkq(file, kbuf, size, pos);
38003 +                       BUG_ON(err == -EINTR);
38004 +               }
38005 +       } while (i++ < prevent_endless
38006 +                && (err == -EAGAIN || err == -EINTR));
38007 +
38008 +#if 0 /* reserved for future use */
38009 +       if (err > 0)
38010 +               fsnotify_modify(file->f_path.dentry);
38011 +#endif
38012 +
38013 +       return err;
38014 +}
38015 +
38016 +struct do_xino_fwrite_args {
38017 +       ssize_t *errp;
38018 +       struct file *file;
38019 +       void *buf;
38020 +       size_t size;
38021 +       loff_t *pos;
38022 +};
38023 +
38024 +static void call_do_xino_fwrite(void *args)
38025 +{
38026 +       struct do_xino_fwrite_args *a = args;
38027 +       *a->errp = do_xino_fwrite(a->file, a->buf, a->size, a->pos);
38028 +}
38029 +
38030 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38031 +                              loff_t *pos)
38032 +{
38033 +       ssize_t err;
38034 +       int wkq_err;
38035 +       struct do_xino_fwrite_args args = {
38036 +               .errp   = &err,
38037 +               .file   = file,
38038 +               .buf    = buf,
38039 +               .size   = size,
38040 +               .pos    = pos
38041 +       };
38042 +
38043 +       /*
38044 +        * it breaks RLIMIT_FSIZE and normal user's limit,
38045 +        * users should care about quota and real 'filesystem full.'
38046 +        */
38047 +       wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
38048 +       if (unlikely(wkq_err))
38049 +               err = wkq_err;
38050 +
38051 +       return err;
38052 +}
38053 +
38054 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos)
38055 +{
38056 +       ssize_t err;
38057 +
38058 +       if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
38059 +               lockdep_off();
38060 +               err = do_xino_fwrite(file, buf, size, pos);
38061 +               lockdep_on();
38062 +       } else {
38063 +               lockdep_off();
38064 +               err = xino_fwrite_wkq(file, buf, size, pos);
38065 +               lockdep_on();
38066 +       }
38067 +
38068 +       return err;
38069 +}
38070 +
38071 +/* ---------------------------------------------------------------------- */
38072 +
38073 +/*
38074 + * inode number bitmap
38075 + */
38076 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
38077 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
38078 +{
38079 +       ino_t ino;
38080 +
38081 +       AuDebugOn(bit < 0 || page_bits <= bit);
38082 +       ino = AUFS_FIRST_INO + pindex * page_bits + bit;
38083 +       return ino;
38084 +}
38085 +
38086 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
38087 +{
38088 +       AuDebugOn(ino < AUFS_FIRST_INO);
38089 +       ino -= AUFS_FIRST_INO;
38090 +       *pindex = ino / page_bits;
38091 +       *bit = ino % page_bits;
38092 +}
38093 +
38094 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
38095 +{
38096 +       int err;
38097 +       loff_t pos;
38098 +       ssize_t sz;
38099 +       struct au_sbinfo *sbinfo;
38100 +       struct file *xib;
38101 +       unsigned long *p;
38102 +
38103 +       sbinfo = au_sbi(sb);
38104 +       MtxMustLock(&sbinfo->si_xib_mtx);
38105 +       AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
38106 +                 || !au_opt_test(sbinfo->si_mntflags, XINO));
38107 +
38108 +       if (pindex == sbinfo->si_xib_last_pindex)
38109 +               return 0;
38110 +
38111 +       xib = sbinfo->si_xib;
38112 +       p = sbinfo->si_xib_buf;
38113 +       pos = sbinfo->si_xib_last_pindex;
38114 +       pos *= PAGE_SIZE;
38115 +       sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38116 +       if (unlikely(sz != PAGE_SIZE))
38117 +               goto out;
38118 +
38119 +       pos = pindex;
38120 +       pos *= PAGE_SIZE;
38121 +       if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
38122 +               sz = xino_fread(xib, p, PAGE_SIZE, &pos);
38123 +       else {
38124 +               memset(p, 0, PAGE_SIZE);
38125 +               sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38126 +       }
38127 +       if (sz == PAGE_SIZE) {
38128 +               sbinfo->si_xib_last_pindex = pindex;
38129 +               return 0; /* success */
38130 +       }
38131 +
38132 +out:
38133 +       AuIOErr1("write failed (%zd)\n", sz);
38134 +       err = sz;
38135 +       if (sz >= 0)
38136 +               err = -EIO;
38137 +       return err;
38138 +}
38139 +
38140 +static void au_xib_clear_bit(struct inode *inode)
38141 +{
38142 +       int err, bit;
38143 +       unsigned long pindex;
38144 +       struct super_block *sb;
38145 +       struct au_sbinfo *sbinfo;
38146 +
38147 +       AuDebugOn(inode->i_nlink);
38148 +
38149 +       sb = inode->i_sb;
38150 +       xib_calc_bit(inode->i_ino, &pindex, &bit);
38151 +       AuDebugOn(page_bits <= bit);
38152 +       sbinfo = au_sbi(sb);
38153 +       mutex_lock(&sbinfo->si_xib_mtx);
38154 +       err = xib_pindex(sb, pindex);
38155 +       if (!err) {
38156 +               clear_bit(bit, sbinfo->si_xib_buf);
38157 +               sbinfo->si_xib_next_bit = bit;
38158 +       }
38159 +       mutex_unlock(&sbinfo->si_xib_mtx);
38160 +}
38161 +
38162 +/* ---------------------------------------------------------------------- */
38163 +
38164 +/*
38165 + * truncate a xino bitmap file
38166 + */
38167 +
38168 +/* todo: slow */
38169 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
38170 +{
38171 +       int err, bit;
38172 +       ssize_t sz;
38173 +       unsigned long pindex;
38174 +       loff_t pos, pend;
38175 +       struct au_sbinfo *sbinfo;
38176 +       ino_t *ino;
38177 +       unsigned long *p;
38178 +
38179 +       err = 0;
38180 +       sbinfo = au_sbi(sb);
38181 +       MtxMustLock(&sbinfo->si_xib_mtx);
38182 +       p = sbinfo->si_xib_buf;
38183 +       pend = vfsub_f_size_read(file);
38184 +       pos = 0;
38185 +       while (pos < pend) {
38186 +               sz = xino_fread(file, page, PAGE_SIZE, &pos);
38187 +               err = sz;
38188 +               if (unlikely(sz <= 0))
38189 +                       goto out;
38190 +
38191 +               err = 0;
38192 +               for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
38193 +                       if (unlikely(*ino < AUFS_FIRST_INO))
38194 +                               continue;
38195 +
38196 +                       xib_calc_bit(*ino, &pindex, &bit);
38197 +                       AuDebugOn(page_bits <= bit);
38198 +                       err = xib_pindex(sb, pindex);
38199 +                       if (!err)
38200 +                               set_bit(bit, p);
38201 +                       else
38202 +                               goto out;
38203 +               }
38204 +       }
38205 +
38206 +out:
38207 +       return err;
38208 +}
38209 +
38210 +static int xib_restore(struct super_block *sb)
38211 +{
38212 +       int err, i;
38213 +       unsigned int nfile;
38214 +       aufs_bindex_t bindex, bbot;
38215 +       void *page;
38216 +       struct au_branch *br;
38217 +       struct au_xino *xi;
38218 +       struct file *file;
38219 +
38220 +       err = -ENOMEM;
38221 +       page = (void *)__get_free_page(GFP_NOFS);
38222 +       if (unlikely(!page))
38223 +               goto out;
38224 +
38225 +       err = 0;
38226 +       bbot = au_sbbot(sb);
38227 +       for (bindex = 0; !err && bindex <= bbot; bindex++)
38228 +               if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0) {
38229 +                       br = au_sbr(sb, bindex);
38230 +                       xi = br->br_xino;
38231 +                       nfile = xi->xi_nfile;
38232 +                       for (i = 0; i < nfile; i++) {
38233 +                               file = au_xino_file(xi, i);
38234 +                               if (file)
38235 +                                       err = do_xib_restore(sb, file, page);
38236 +                       }
38237 +               } else
38238 +                       AuDbg("skip shared b%d\n", bindex);
38239 +       free_page((unsigned long)page);
38240 +
38241 +out:
38242 +       return err;
38243 +}
38244 +
38245 +int au_xib_trunc(struct super_block *sb)
38246 +{
38247 +       int err;
38248 +       ssize_t sz;
38249 +       loff_t pos;
38250 +       struct au_sbinfo *sbinfo;
38251 +       unsigned long *p;
38252 +       struct file *file;
38253 +
38254 +       SiMustWriteLock(sb);
38255 +
38256 +       err = 0;
38257 +       sbinfo = au_sbi(sb);
38258 +       if (!au_opt_test(sbinfo->si_mntflags, XINO))
38259 +               goto out;
38260 +
38261 +       file = sbinfo->si_xib;
38262 +       if (vfsub_f_size_read(file) <= PAGE_SIZE)
38263 +               goto out;
38264 +
38265 +       file = au_xino_create2(sb, &sbinfo->si_xib->f_path, NULL);
38266 +       err = PTR_ERR(file);
38267 +       if (IS_ERR(file))
38268 +               goto out;
38269 +       fput(sbinfo->si_xib);
38270 +       sbinfo->si_xib = file;
38271 +
38272 +       p = sbinfo->si_xib_buf;
38273 +       memset(p, 0, PAGE_SIZE);
38274 +       pos = 0;
38275 +       sz = xino_fwrite(sbinfo->si_xib, p, PAGE_SIZE, &pos);
38276 +       if (unlikely(sz != PAGE_SIZE)) {
38277 +               err = sz;
38278 +               AuIOErr("err %d\n", err);
38279 +               if (sz >= 0)
38280 +                       err = -EIO;
38281 +               goto out;
38282 +       }
38283 +
38284 +       mutex_lock(&sbinfo->si_xib_mtx);
38285 +       /* mnt_want_write() is unnecessary here */
38286 +       err = xib_restore(sb);
38287 +       mutex_unlock(&sbinfo->si_xib_mtx);
38288 +
38289 +out:
38290 +       return err;
38291 +}
38292 +
38293 +/* ---------------------------------------------------------------------- */
38294 +
38295 +struct au_xino *au_xino_alloc(unsigned int nfile)
38296 +{
38297 +       struct au_xino *xi;
38298 +
38299 +       xi = kzalloc(sizeof(*xi), GFP_NOFS);
38300 +       if (unlikely(!xi))
38301 +               goto out;
38302 +       xi->xi_nfile = nfile;
38303 +       xi->xi_file = kcalloc(nfile, sizeof(*xi->xi_file), GFP_NOFS);
38304 +       if (unlikely(!xi->xi_file))
38305 +               goto out_free;
38306 +
38307 +       xi->xi_nondir.total = 8; /* initial size */
38308 +       xi->xi_nondir.array = kcalloc(xi->xi_nondir.total, sizeof(ino_t),
38309 +                                     GFP_NOFS);
38310 +       if (unlikely(!xi->xi_nondir.array))
38311 +               goto out_file;
38312 +
38313 +       spin_lock_init(&xi->xi_nondir.spin);
38314 +       init_waitqueue_head(&xi->xi_nondir.wqh);
38315 +       mutex_init(&xi->xi_mtx);
38316 +       INIT_HLIST_BL_HEAD(&xi->xi_writing);
38317 +       atomic_set(&xi->xi_truncating, 0);
38318 +       kref_init(&xi->xi_kref);
38319 +       goto out; /* success */
38320 +
38321 +out_file:
38322 +       au_kfree_try_rcu(xi->xi_file);
38323 +out_free:
38324 +       au_kfree_rcu(xi);
38325 +       xi = NULL;
38326 +out:
38327 +       return xi;
38328 +}
38329 +
38330 +static int au_xino_init(struct au_branch *br, int idx, struct file *file)
38331 +{
38332 +       int err;
38333 +       struct au_xino *xi;
38334 +
38335 +       err = 0;
38336 +       xi = au_xino_alloc(idx + 1);
38337 +       if (unlikely(!xi)) {
38338 +               err = -ENOMEM;
38339 +               goto out;
38340 +       }
38341 +
38342 +       if (file)
38343 +               get_file(file);
38344 +       xi->xi_file[idx] = file;
38345 +       AuDebugOn(br->br_xino);
38346 +       br->br_xino = xi;
38347 +
38348 +out:
38349 +       return err;
38350 +}
38351 +
38352 +static void au_xino_release(struct kref *kref)
38353 +{
38354 +       struct au_xino *xi;
38355 +       int i;
38356 +       unsigned long ul;
38357 +       struct hlist_bl_head *hbl;
38358 +       struct hlist_bl_node *pos, *n;
38359 +       struct au_xi_writing *p;
38360 +
38361 +       xi = container_of(kref, struct au_xino, xi_kref);
38362 +       for (i = 0; i < xi->xi_nfile; i++)
38363 +               if (xi->xi_file[i])
38364 +                       fput(xi->xi_file[i]);
38365 +       for (i = xi->xi_nondir.total - 1; i >= 0; i--)
38366 +               AuDebugOn(xi->xi_nondir.array[i]);
38367 +       mutex_destroy(&xi->xi_mtx);
38368 +       hbl = &xi->xi_writing;
38369 +       ul = au_hbl_count(hbl);
38370 +       if (unlikely(ul)) {
38371 +               pr_warn("xi_writing %lu\n", ul);
38372 +               hlist_bl_lock(hbl);
38373 +               hlist_bl_for_each_entry_safe(p, pos, n, hbl, node) {
38374 +                       hlist_bl_del(&p->node);
38375 +                       /* kmemleak reported au_kfree_rcu() doesn't free it */
38376 +                       kfree(p);
38377 +               }
38378 +               hlist_bl_unlock(hbl);
38379 +       }
38380 +       au_kfree_try_rcu(xi->xi_file);
38381 +       au_kfree_try_rcu(xi->xi_nondir.array);
38382 +       au_kfree_rcu(xi);
38383 +}
38384 +
38385 +int au_xino_put(struct au_branch *br)
38386 +{
38387 +       int ret;
38388 +       struct au_xino *xi;
38389 +
38390 +       ret = 0;
38391 +       xi = br->br_xino;
38392 +       if (xi) {
38393 +               br->br_xino = NULL;
38394 +               ret = kref_put(&xi->xi_kref, au_xino_release);
38395 +       }
38396 +
38397 +       return ret;
38398 +}
38399 +
38400 +/* ---------------------------------------------------------------------- */
38401 +
38402 +/*
38403 + * xino mount option handlers
38404 + */
38405 +
38406 +/* xino bitmap */
38407 +static void xino_clear_xib(struct super_block *sb)
38408 +{
38409 +       struct au_sbinfo *sbinfo;
38410 +
38411 +       SiMustWriteLock(sb);
38412 +
38413 +       sbinfo = au_sbi(sb);
38414 +       if (sbinfo->si_xib)
38415 +               fput(sbinfo->si_xib);
38416 +       sbinfo->si_xib = NULL;
38417 +       if (sbinfo->si_xib_buf)
38418 +               free_page((unsigned long)sbinfo->si_xib_buf);
38419 +       sbinfo->si_xib_buf = NULL;
38420 +}
38421 +
38422 +static int au_xino_set_xib(struct super_block *sb, struct path *path)
38423 +{
38424 +       int err;
38425 +       loff_t pos;
38426 +       struct au_sbinfo *sbinfo;
38427 +       struct file *file;
38428 +       struct super_block *xi_sb;
38429 +
38430 +       SiMustWriteLock(sb);
38431 +
38432 +       sbinfo = au_sbi(sb);
38433 +       file = au_xino_create2(sb, path, sbinfo->si_xib);
38434 +       err = PTR_ERR(file);
38435 +       if (IS_ERR(file))
38436 +               goto out;
38437 +       if (sbinfo->si_xib)
38438 +               fput(sbinfo->si_xib);
38439 +       sbinfo->si_xib = file;
38440 +       xi_sb = file_inode(file)->i_sb;
38441 +       sbinfo->si_ximaxent = xi_sb->s_maxbytes;
38442 +       if (unlikely(sbinfo->si_ximaxent < PAGE_SIZE)) {
38443 +               err = -EIO;
38444 +               pr_err("s_maxbytes(%llu) on %s is too small\n",
38445 +                      (u64)sbinfo->si_ximaxent, au_sbtype(xi_sb));
38446 +               goto out_unset;
38447 +       }
38448 +       sbinfo->si_ximaxent /= sizeof(ino_t);
38449 +
38450 +       err = -ENOMEM;
38451 +       if (!sbinfo->si_xib_buf)
38452 +               sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
38453 +       if (unlikely(!sbinfo->si_xib_buf))
38454 +               goto out_unset;
38455 +
38456 +       sbinfo->si_xib_last_pindex = 0;
38457 +       sbinfo->si_xib_next_bit = 0;
38458 +       if (vfsub_f_size_read(file) < PAGE_SIZE) {
38459 +               pos = 0;
38460 +               err = xino_fwrite(file, sbinfo->si_xib_buf, PAGE_SIZE, &pos);
38461 +               if (unlikely(err != PAGE_SIZE))
38462 +                       goto out_free;
38463 +       }
38464 +       err = 0;
38465 +       goto out; /* success */
38466 +
38467 +out_free:
38468 +       if (sbinfo->si_xib_buf)
38469 +               free_page((unsigned long)sbinfo->si_xib_buf);
38470 +       sbinfo->si_xib_buf = NULL;
38471 +       if (err >= 0)
38472 +               err = -EIO;
38473 +out_unset:
38474 +       fput(sbinfo->si_xib);
38475 +       sbinfo->si_xib = NULL;
38476 +out:
38477 +       AuTraceErr(err);
38478 +       return err;
38479 +}
38480 +
38481 +/* xino for each branch */
38482 +static void xino_clear_br(struct super_block *sb)
38483 +{
38484 +       aufs_bindex_t bindex, bbot;
38485 +       struct au_branch *br;
38486 +
38487 +       bbot = au_sbbot(sb);
38488 +       for (bindex = 0; bindex <= bbot; bindex++) {
38489 +               br = au_sbr(sb, bindex);
38490 +               AuDebugOn(!br);
38491 +               au_xino_put(br);
38492 +       }
38493 +}
38494 +
38495 +static void au_xino_set_br_shared(struct super_block *sb, struct au_branch *br,
38496 +                                 aufs_bindex_t bshared)
38497 +{
38498 +       struct au_branch *brshared;
38499 +
38500 +       brshared = au_sbr(sb, bshared);
38501 +       AuDebugOn(!brshared->br_xino);
38502 +       AuDebugOn(!brshared->br_xino->xi_file);
38503 +       if (br->br_xino != brshared->br_xino) {
38504 +               au_xino_get(brshared);
38505 +               au_xino_put(br);
38506 +               br->br_xino = brshared->br_xino;
38507 +       }
38508 +}
38509 +
38510 +struct au_xino_do_set_br {
38511 +       struct au_branch *br;
38512 +       ino_t h_ino;
38513 +       aufs_bindex_t bshared;
38514 +};
38515 +
38516 +static int au_xino_do_set_br(struct super_block *sb, struct path *path,
38517 +                            struct au_xino_do_set_br *args)
38518 +{
38519 +       int err;
38520 +       struct au_xi_calc calc;
38521 +       struct file *file;
38522 +       struct au_branch *br;
38523 +       struct au_xi_new xinew = {
38524 +               .base = path
38525 +       };
38526 +
38527 +       br = args->br;
38528 +       xinew.xi = br->br_xino;
38529 +       au_xi_calc(sb, args->h_ino, &calc);
38530 +       xinew.copy_src = au_xino_file(xinew.xi, calc.idx);
38531 +       if (args->bshared >= 0)
38532 +               /* shared xino */
38533 +               au_xino_set_br_shared(sb, br, args->bshared);
38534 +       else if (!xinew.xi) {
38535 +               /* new xino */
38536 +               err = au_xino_init(br, calc.idx, xinew.copy_src);
38537 +               if (unlikely(err))
38538 +                       goto out;
38539 +       }
38540 +
38541 +       /* force re-creating */
38542 +       xinew.xi = br->br_xino;
38543 +       xinew.idx = calc.idx;
38544 +       mutex_lock(&xinew.xi->xi_mtx);
38545 +       file = au_xi_new(sb, &xinew);
38546 +       mutex_unlock(&xinew.xi->xi_mtx);
38547 +       err = PTR_ERR(file);
38548 +       if (IS_ERR(file))
38549 +               goto out;
38550 +       AuDebugOn(!file);
38551 +
38552 +       err = au_xino_do_write(file, &calc, AUFS_ROOT_INO);
38553 +       if (unlikely(err))
38554 +               au_xino_put(br);
38555 +
38556 +out:
38557 +       AuTraceErr(err);
38558 +       return err;
38559 +}
38560 +
38561 +static int au_xino_set_br(struct super_block *sb, struct path *path)
38562 +{
38563 +       int err;
38564 +       aufs_bindex_t bindex, bbot;
38565 +       struct au_xino_do_set_br args;
38566 +       struct inode *inode;
38567 +
38568 +       SiMustWriteLock(sb);
38569 +
38570 +       bbot = au_sbbot(sb);
38571 +       inode = d_inode(sb->s_root);
38572 +       for (bindex = 0; bindex <= bbot; bindex++) {
38573 +               args.h_ino = au_h_iptr(inode, bindex)->i_ino;
38574 +               args.br = au_sbr(sb, bindex);
38575 +               args.bshared = is_sb_shared(sb, bindex, bindex - 1);
38576 +               err = au_xino_do_set_br(sb, path, &args);
38577 +               if (unlikely(err))
38578 +                       break;
38579 +       }
38580 +
38581 +       AuTraceErr(err);
38582 +       return err;
38583 +}
38584 +
38585 +void au_xino_clr(struct super_block *sb)
38586 +{
38587 +       struct au_sbinfo *sbinfo;
38588 +
38589 +       au_xigen_clr(sb);
38590 +       xino_clear_xib(sb);
38591 +       xino_clear_br(sb);
38592 +       dbgaufs_brs_del(sb, 0);
38593 +       sbinfo = au_sbi(sb);
38594 +       /* lvalue, do not call au_mntflags() */
38595 +       au_opt_clr(sbinfo->si_mntflags, XINO);
38596 +}
38597 +
38598 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount)
38599 +{
38600 +       int err, skip;
38601 +       struct dentry *dentry, *parent, *cur_dentry, *cur_parent;
38602 +       struct qstr *dname, *cur_name;
38603 +       struct file *cur_xino;
38604 +       struct au_sbinfo *sbinfo;
38605 +       struct path *path, *cur_path;
38606 +
38607 +       SiMustWriteLock(sb);
38608 +
38609 +       err = 0;
38610 +       sbinfo = au_sbi(sb);
38611 +       path = &xiopt->file->f_path;
38612 +       dentry = path->dentry;
38613 +       parent = dget_parent(dentry);
38614 +       if (remount) {
38615 +               skip = 0;
38616 +               cur_xino = sbinfo->si_xib;
38617 +               if (cur_xino) {
38618 +                       cur_path = &cur_xino->f_path;
38619 +                       cur_dentry = cur_path->dentry;
38620 +                       cur_parent = dget_parent(cur_dentry);
38621 +                       cur_name = &cur_dentry->d_name;
38622 +                       dname = &dentry->d_name;
38623 +                       skip = (cur_parent == parent
38624 +                               && au_qstreq(dname, cur_name));
38625 +                       dput(cur_parent);
38626 +               }
38627 +               if (skip)
38628 +                       goto out;
38629 +       }
38630 +
38631 +       au_opt_set(sbinfo->si_mntflags, XINO);
38632 +       err = au_xino_set_xib(sb, path);
38633 +       /* si_x{read,write} are set */
38634 +       if (!err)
38635 +               err = au_xigen_set(sb, path);
38636 +       if (!err)
38637 +               err = au_xino_set_br(sb, path);
38638 +       if (!err) {
38639 +               dbgaufs_brs_add(sb, 0, /*topdown*/1);
38640 +               goto out; /* success */
38641 +       }
38642 +
38643 +       /* reset all */
38644 +       AuIOErr("failed setting xino(%d).\n", err);
38645 +       au_xino_clr(sb);
38646 +
38647 +out:
38648 +       dput(parent);
38649 +       return err;
38650 +}
38651 +
38652 +/*
38653 + * create a xinofile at the default place/path.
38654 + */
38655 +struct file *au_xino_def(struct super_block *sb)
38656 +{
38657 +       struct file *file;
38658 +       char *page, *p;
38659 +       struct au_branch *br;
38660 +       struct super_block *h_sb;
38661 +       struct path path;
38662 +       aufs_bindex_t bbot, bindex, bwr;
38663 +
38664 +       br = NULL;
38665 +       bbot = au_sbbot(sb);
38666 +       bwr = -1;
38667 +       for (bindex = 0; bindex <= bbot; bindex++) {
38668 +               br = au_sbr(sb, bindex);
38669 +               if (au_br_writable(br->br_perm)
38670 +                   && !au_test_fs_bad_xino(au_br_sb(br))) {
38671 +                       bwr = bindex;
38672 +                       break;
38673 +               }
38674 +       }
38675 +
38676 +       if (bwr >= 0) {
38677 +               file = ERR_PTR(-ENOMEM);
38678 +               page = (void *)__get_free_page(GFP_NOFS);
38679 +               if (unlikely(!page))
38680 +                       goto out;
38681 +               path.mnt = au_br_mnt(br);
38682 +               path.dentry = au_h_dptr(sb->s_root, bwr);
38683 +               p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
38684 +               file = (void *)p;
38685 +               if (!IS_ERR(p)) {
38686 +                       strcat(p, "/" AUFS_XINO_FNAME);
38687 +                       AuDbg("%s\n", p);
38688 +                       file = au_xino_create(sb, p, /*silent*/0, /*wbrtop*/1);
38689 +               }
38690 +               free_page((unsigned long)page);
38691 +       } else {
38692 +               file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0,
38693 +                                     /*wbrtop*/0);
38694 +               if (IS_ERR(file))
38695 +                       goto out;
38696 +               h_sb = file->f_path.dentry->d_sb;
38697 +               if (unlikely(au_test_fs_bad_xino(h_sb))) {
38698 +                       pr_err("xino doesn't support %s(%s)\n",
38699 +                              AUFS_XINO_DEFPATH, au_sbtype(h_sb));
38700 +                       fput(file);
38701 +                       file = ERR_PTR(-EINVAL);
38702 +               }
38703 +       }
38704 +
38705 +out:
38706 +       return file;
38707 +}
38708 +
38709 +/* ---------------------------------------------------------------------- */
38710 +
38711 +/*
38712 + * initialize the xinofile for the specified branch @br
38713 + * at the place/path where @base_file indicates.
38714 + * test whether another branch is on the same filesystem or not,
38715 + * if found then share the xinofile with another branch.
38716 + */
38717 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
38718 +                   struct path *base)
38719 +{
38720 +       int err;
38721 +       struct au_xino_do_set_br args = {
38722 +               .h_ino  = h_ino,
38723 +               .br     = br
38724 +       };
38725 +
38726 +       args.bshared = sbr_find_shared(sb, /*btop*/0, au_sbbot(sb),
38727 +                                      au_br_sb(br));
38728 +       err = au_xino_do_set_br(sb, base, &args);
38729 +       if (unlikely(err))
38730 +               au_xino_put(br);
38731 +
38732 +       return err;
38733 +}
38734 +
38735 +/* ---------------------------------------------------------------------- */
38736 +
38737 +/*
38738 + * get an unused inode number from bitmap
38739 + */
38740 +ino_t au_xino_new_ino(struct super_block *sb)
38741 +{
38742 +       ino_t ino;
38743 +       unsigned long *p, pindex, ul, pend;
38744 +       struct au_sbinfo *sbinfo;
38745 +       struct file *file;
38746 +       int free_bit, err;
38747 +
38748 +       if (!au_opt_test(au_mntflags(sb), XINO))
38749 +               return iunique(sb, AUFS_FIRST_INO);
38750 +
38751 +       sbinfo = au_sbi(sb);
38752 +       mutex_lock(&sbinfo->si_xib_mtx);
38753 +       p = sbinfo->si_xib_buf;
38754 +       free_bit = sbinfo->si_xib_next_bit;
38755 +       if (free_bit < page_bits && !test_bit(free_bit, p))
38756 +               goto out; /* success */
38757 +       free_bit = find_first_zero_bit(p, page_bits);
38758 +       if (free_bit < page_bits)
38759 +               goto out; /* success */
38760 +
38761 +       pindex = sbinfo->si_xib_last_pindex;
38762 +       for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
38763 +               err = xib_pindex(sb, ul);
38764 +               if (unlikely(err))
38765 +                       goto out_err;
38766 +               free_bit = find_first_zero_bit(p, page_bits);
38767 +               if (free_bit < page_bits)
38768 +                       goto out; /* success */
38769 +       }
38770 +
38771 +       file = sbinfo->si_xib;
38772 +       pend = vfsub_f_size_read(file) / PAGE_SIZE;
38773 +       for (ul = pindex + 1; ul <= pend; ul++) {
38774 +               err = xib_pindex(sb, ul);
38775 +               if (unlikely(err))
38776 +                       goto out_err;
38777 +               free_bit = find_first_zero_bit(p, page_bits);
38778 +               if (free_bit < page_bits)
38779 +                       goto out; /* success */
38780 +       }
38781 +       BUG();
38782 +
38783 +out:
38784 +       set_bit(free_bit, p);
38785 +       sbinfo->si_xib_next_bit = free_bit + 1;
38786 +       pindex = sbinfo->si_xib_last_pindex;
38787 +       mutex_unlock(&sbinfo->si_xib_mtx);
38788 +       ino = xib_calc_ino(pindex, free_bit);
38789 +       AuDbg("i%lu\n", (unsigned long)ino);
38790 +       return ino;
38791 +out_err:
38792 +       mutex_unlock(&sbinfo->si_xib_mtx);
38793 +       AuDbg("i0\n");
38794 +       return 0;
38795 +}
38796 +
38797 +/* for s_op->delete_inode() */
38798 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
38799 +{
38800 +       int err;
38801 +       unsigned int mnt_flags;
38802 +       aufs_bindex_t bindex, bbot, bi;
38803 +       unsigned char try_trunc;
38804 +       struct au_iinfo *iinfo;
38805 +       struct super_block *sb;
38806 +       struct au_hinode *hi;
38807 +       struct inode *h_inode;
38808 +       struct au_branch *br;
38809 +       struct au_xi_calc calc;
38810 +       struct file *file;
38811 +
38812 +       AuDebugOn(au_is_bad_inode(inode));
38813 +
38814 +       sb = inode->i_sb;
38815 +       mnt_flags = au_mntflags(sb);
38816 +       if (!au_opt_test(mnt_flags, XINO)
38817 +           || inode->i_ino == AUFS_ROOT_INO)
38818 +               return;
38819 +
38820 +       if (unlinked) {
38821 +               au_xigen_inc(inode);
38822 +               au_xib_clear_bit(inode);
38823 +       }
38824 +
38825 +       iinfo = au_ii(inode);
38826 +       bindex = iinfo->ii_btop;
38827 +       if (bindex < 0)
38828 +               return;
38829 +
38830 +       try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
38831 +       hi = au_hinode(iinfo, bindex);
38832 +       bbot = iinfo->ii_bbot;
38833 +       for (; bindex <= bbot; bindex++, hi++) {
38834 +               h_inode = hi->hi_inode;
38835 +               if (!h_inode
38836 +                   || (!unlinked && h_inode->i_nlink))
38837 +                       continue;
38838 +
38839 +               /* inode may not be revalidated */
38840 +               bi = au_br_index(sb, hi->hi_id);
38841 +               if (bi < 0)
38842 +                       continue;
38843 +
38844 +               br = au_sbr(sb, bi);
38845 +               au_xi_calc(sb, h_inode->i_ino, &calc);
38846 +               file = au_xino_file(br->br_xino, calc.idx);
38847 +               if (IS_ERR_OR_NULL(file))
38848 +                       continue;
38849 +
38850 +               err = au_xino_do_write(file, &calc, /*ino*/0);
38851 +               if (!err && try_trunc
38852 +                   && au_test_fs_trunc_xino(au_br_sb(br)))
38853 +                       xino_try_trunc(sb, br);
38854 +       }
38855 +}
38856 +
38857 +/* ---------------------------------------------------------------------- */
38858 +
38859 +static int au_xinondir_find(struct au_xino *xi, ino_t h_ino)
38860 +{
38861 +       int found, total, i;
38862 +
38863 +       found = -1;
38864 +       total = xi->xi_nondir.total;
38865 +       for (i = 0; i < total; i++) {
38866 +               if (xi->xi_nondir.array[i] != h_ino)
38867 +                       continue;
38868 +               found = i;
38869 +               break;
38870 +       }
38871 +
38872 +       return found;
38873 +}
38874 +
38875 +static int au_xinondir_expand(struct au_xino *xi)
38876 +{
38877 +       int err, sz;
38878 +       ino_t *p;
38879 +
38880 +       BUILD_BUG_ON(KMALLOC_MAX_SIZE > INT_MAX);
38881 +
38882 +       err = -ENOMEM;
38883 +       sz = xi->xi_nondir.total * sizeof(ino_t);
38884 +       if (unlikely(sz > KMALLOC_MAX_SIZE / 2))
38885 +               goto out;
38886 +       p = au_kzrealloc(xi->xi_nondir.array, sz, sz << 1, GFP_ATOMIC,
38887 +                        /*may_shrink*/0);
38888 +       if (p) {
38889 +               xi->xi_nondir.array = p;
38890 +               xi->xi_nondir.total <<= 1;
38891 +               AuDbg("xi_nondir.total %d\n", xi->xi_nondir.total);
38892 +               err = 0;
38893 +       }
38894 +
38895 +out:
38896 +       return err;
38897 +}
38898 +
38899 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
38900 +                      ino_t h_ino, int idx)
38901 +{
38902 +       struct au_xino *xi;
38903 +
38904 +       AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
38905 +       xi = au_sbr(sb, bindex)->br_xino;
38906 +       AuDebugOn(idx < 0 || xi->xi_nondir.total <= idx);
38907 +
38908 +       spin_lock(&xi->xi_nondir.spin);
38909 +       AuDebugOn(xi->xi_nondir.array[idx] != h_ino);
38910 +       xi->xi_nondir.array[idx] = 0;
38911 +       spin_unlock(&xi->xi_nondir.spin);
38912 +       wake_up_all(&xi->xi_nondir.wqh);
38913 +}
38914 +
38915 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
38916 +                     int *idx)
38917 +{
38918 +       int err, found, empty;
38919 +       struct au_xino *xi;
38920 +
38921 +       err = 0;
38922 +       *idx = -1;
38923 +       if (!au_opt_test(au_mntflags(sb), XINO))
38924 +               goto out; /* no xino */
38925 +
38926 +       xi = au_sbr(sb, bindex)->br_xino;
38927 +
38928 +again:
38929 +       spin_lock(&xi->xi_nondir.spin);
38930 +       found = au_xinondir_find(xi, h_ino);
38931 +       if (found == -1) {
38932 +               empty = au_xinondir_find(xi, /*h_ino*/0);
38933 +               if (empty == -1) {
38934 +                       empty = xi->xi_nondir.total;
38935 +                       err = au_xinondir_expand(xi);
38936 +                       if (unlikely(err))
38937 +                               goto out_unlock;
38938 +               }
38939 +               xi->xi_nondir.array[empty] = h_ino;
38940 +               *idx = empty;
38941 +       } else {
38942 +               spin_unlock(&xi->xi_nondir.spin);
38943 +               wait_event(xi->xi_nondir.wqh,
38944 +                          xi->xi_nondir.array[found] != h_ino);
38945 +               goto again;
38946 +       }
38947 +
38948 +out_unlock:
38949 +       spin_unlock(&xi->xi_nondir.spin);
38950 +out:
38951 +       return err;
38952 +}
38953 +
38954 +/* ---------------------------------------------------------------------- */
38955 +
38956 +int au_xino_path(struct seq_file *seq, struct file *file)
38957 +{
38958 +       int err;
38959 +
38960 +       err = au_seq_path(seq, &file->f_path);
38961 +       if (unlikely(err))
38962 +               goto out;
38963 +
38964 +#define Deleted "\\040(deleted)"
38965 +       seq->count -= sizeof(Deleted) - 1;
38966 +       AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
38967 +                        sizeof(Deleted) - 1));
38968 +#undef Deleted
38969 +
38970 +out:
38971 +       return err;
38972 +}
38973 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
38974 --- /usr/share/empty/include/uapi/linux/aufs_type.h     1970-01-01 01:00:00.000000000 +0100
38975 +++ linux/include/uapi/linux/aufs_type.h        2022-11-05 23:02:18.972555950 +0100
38976 @@ -0,0 +1,452 @@
38977 +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
38978 +/*
38979 + * Copyright (C) 2005-2022 Junjiro R. Okajima
38980 + *
38981 + * This program is free software; you can redistribute it and/or modify
38982 + * it under the terms of the GNU General Public License as published by
38983 + * the Free Software Foundation; either version 2 of the License, or
38984 + * (at your option) any later version.
38985 + *
38986 + * This program is distributed in the hope that it will be useful,
38987 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
38988 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
38989 + * GNU General Public License for more details.
38990 + *
38991 + * You should have received a copy of the GNU General Public License
38992 + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
38993 + */
38994 +
38995 +#ifndef __AUFS_TYPE_H__
38996 +#define __AUFS_TYPE_H__
38997 +
38998 +#define AUFS_NAME      "aufs"
38999 +
39000 +#ifdef __KERNEL__
39001 +/*
39002 + * define it before including all other headers.
39003 + * sched.h may use pr_* macros before defining "current", so define the
39004 + * no-current version first, and re-define later.
39005 + */
39006 +#define pr_fmt(fmt)    AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
39007 +#include <linux/sched.h>
39008 +#undef pr_fmt
39009 +#define pr_fmt(fmt) \
39010 +               AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
39011 +               (int)sizeof(current->comm), current->comm, current->pid
39012 +#include <linux/limits.h>
39013 +#else
39014 +#include <stdint.h>
39015 +#include <sys/types.h>
39016 +#include <limits.h>
39017 +#endif /* __KERNEL__ */
39018 +
39019 +#define AUFS_VERSION   "6.0-20221031"
39020 +
39021 +/* todo? move this to linux-2.6.19/include/magic.h */
39022 +#define AUFS_SUPER_MAGIC       ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
39023 +
39024 +/* ---------------------------------------------------------------------- */
39025 +
39026 +#ifdef __KERNEL__
39027 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
39028 +typedef int8_t aufs_bindex_t;
39029 +#define AUFS_BRANCH_MAX 127
39030 +#else
39031 +typedef int16_t aufs_bindex_t;
39032 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
39033 +#define AUFS_BRANCH_MAX 511
39034 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
39035 +#define AUFS_BRANCH_MAX 1023
39036 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
39037 +#define AUFS_BRANCH_MAX 32767
39038 +#endif
39039 +#endif
39040 +
39041 +#ifndef AUFS_BRANCH_MAX
39042 +#error unknown CONFIG_AUFS_BRANCH_MAX value
39043 +#endif
39044 +#endif /* __KERNEL__ */
39045 +
39046 +/* ---------------------------------------------------------------------- */
39047 +
39048 +#define AUFS_FSTYPE            AUFS_NAME
39049 +
39050 +#define AUFS_ROOT_INO          2
39051 +#define AUFS_FIRST_INO         11
39052 +
39053 +#define AUFS_WH_PFX            ".wh."
39054 +#define AUFS_WH_PFX_LEN                ((int)sizeof(AUFS_WH_PFX) - 1)
39055 +#define AUFS_WH_TMP_LEN                4
39056 +/* a limit for rmdir/rename a dir and copyup */
39057 +#define AUFS_MAX_NAMELEN       (NAME_MAX \
39058 +                               - AUFS_WH_PFX_LEN * 2   /* doubly whiteouted */\
39059 +                               - 1                     /* dot */\
39060 +                               - AUFS_WH_TMP_LEN)      /* hex */
39061 +#define AUFS_XINO_FNAME                "." AUFS_NAME ".xino"
39062 +#define AUFS_XINO_DEFPATH      "/tmp/" AUFS_XINO_FNAME
39063 +#define AUFS_XINO_DEF_SEC      30 /* seconds */
39064 +#define AUFS_XINO_DEF_TRUNC    45 /* percentage */
39065 +#define AUFS_DIRWH_DEF         3
39066 +#define AUFS_RDCACHE_DEF       10 /* seconds */
39067 +#define AUFS_RDCACHE_MAX       3600 /* seconds */
39068 +#define AUFS_RDBLK_DEF         512 /* bytes */
39069 +#define AUFS_RDHASH_DEF                32
39070 +#define AUFS_WKQ_NAME          AUFS_NAME "d"
39071 +#define AUFS_MFS_DEF_SEC       30 /* seconds */
39072 +#define AUFS_MFS_MAX_SEC       3600 /* seconds */
39073 +#define AUFS_FHSM_CACHE_DEF_SEC        30 /* seconds */
39074 +#define AUFS_PLINK_WARN                50 /* number of plinks in a single bucket */
39075 +
39076 +/* pseudo-link maintenace under /proc */
39077 +#define AUFS_PLINK_MAINT_NAME  "plink_maint"
39078 +#define AUFS_PLINK_MAINT_DIR   "fs/" AUFS_NAME
39079 +#define AUFS_PLINK_MAINT_PATH  AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
39080 +
39081 +/* dirren, renamed dir */
39082 +#define AUFS_DR_INFO_PFX       AUFS_WH_PFX ".dr."
39083 +#define AUFS_DR_BRHINO_NAME    AUFS_WH_PFX "hino"
39084 +/* whiteouted doubly */
39085 +#define AUFS_WH_DR_INFO_PFX    AUFS_WH_PFX AUFS_DR_INFO_PFX
39086 +#define AUFS_WH_DR_BRHINO      AUFS_WH_PFX AUFS_DR_BRHINO_NAME
39087 +
39088 +#define AUFS_DIROPQ_NAME       AUFS_WH_PFX ".opq" /* whiteouted doubly */
39089 +#define AUFS_WH_DIROPQ         AUFS_WH_PFX AUFS_DIROPQ_NAME
39090 +
39091 +#define AUFS_BASE_NAME         AUFS_WH_PFX AUFS_NAME
39092 +#define AUFS_PLINKDIR_NAME     AUFS_WH_PFX "plnk"
39093 +#define AUFS_ORPHDIR_NAME      AUFS_WH_PFX "orph"
39094 +
39095 +/* doubly whiteouted */
39096 +#define AUFS_WH_BASE           AUFS_WH_PFX AUFS_BASE_NAME
39097 +#define AUFS_WH_PLINKDIR       AUFS_WH_PFX AUFS_PLINKDIR_NAME
39098 +#define AUFS_WH_ORPHDIR                AUFS_WH_PFX AUFS_ORPHDIR_NAME
39099 +
39100 +/* branch permissions and attributes */
39101 +#define AUFS_BRPERM_RW         "rw"
39102 +#define AUFS_BRPERM_RO         "ro"
39103 +#define AUFS_BRPERM_RR         "rr"
39104 +#define AUFS_BRATTR_COO_REG    "coo_reg"
39105 +#define AUFS_BRATTR_COO_ALL    "coo_all"
39106 +#define AUFS_BRATTR_FHSM       "fhsm"
39107 +#define AUFS_BRATTR_UNPIN      "unpin"
39108 +#define AUFS_BRATTR_ICEX       "icex"
39109 +#define AUFS_BRATTR_ICEX_SEC   "icexsec"
39110 +#define AUFS_BRATTR_ICEX_SYS   "icexsys"
39111 +#define AUFS_BRATTR_ICEX_TR    "icextr"
39112 +#define AUFS_BRATTR_ICEX_USR   "icexusr"
39113 +#define AUFS_BRATTR_ICEX_OTH   "icexoth"
39114 +#define AUFS_BRRATTR_WH                "wh"
39115 +#define AUFS_BRWATTR_NLWH      "nolwh"
39116 +#define AUFS_BRWATTR_MOO       "moo"
39117 +
39118 +#define AuBrPerm_RW            1               /* writable, hardlinkable wh */
39119 +#define AuBrPerm_RO            (1 << 1)        /* readonly */
39120 +#define AuBrPerm_RR            (1 << 2)        /* natively readonly */
39121 +#define AuBrPerm_Mask          (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
39122 +
39123 +#define AuBrAttr_COO_REG       (1 << 3)        /* copy-up on open */
39124 +#define AuBrAttr_COO_ALL       (1 << 4)
39125 +#define AuBrAttr_COO_Mask      (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
39126 +
39127 +#define AuBrAttr_FHSM          (1 << 5)        /* file-based hsm */
39128 +#define AuBrAttr_UNPIN         (1 << 6)        /* rename-able top dir of
39129 +                                                  branch. meaningless since
39130 +                                                  linux-3.18-rc1 */
39131 +
39132 +/* ignore error in copying XATTR */
39133 +#define AuBrAttr_ICEX_SEC      (1 << 7)
39134 +#define AuBrAttr_ICEX_SYS      (1 << 8)
39135 +#define AuBrAttr_ICEX_TR       (1 << 9)
39136 +#define AuBrAttr_ICEX_USR      (1 << 10)
39137 +#define AuBrAttr_ICEX_OTH      (1 << 11)
39138 +#define AuBrAttr_ICEX          (AuBrAttr_ICEX_SEC      \
39139 +                                | AuBrAttr_ICEX_SYS    \
39140 +                                | AuBrAttr_ICEX_TR     \
39141 +                                | AuBrAttr_ICEX_USR    \
39142 +                                | AuBrAttr_ICEX_OTH)
39143 +
39144 +#define AuBrRAttr_WH           (1 << 12)       /* whiteout-able */
39145 +#define AuBrRAttr_Mask         AuBrRAttr_WH
39146 +
39147 +#define AuBrWAttr_NoLinkWH     (1 << 13)       /* un-hardlinkable whiteouts */
39148 +#define AuBrWAttr_MOO          (1 << 14)       /* move-up on open */
39149 +#define AuBrWAttr_Mask         (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
39150 +
39151 +#define AuBrAttr_CMOO_Mask     (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
39152 +
39153 +/* #warning test userspace */
39154 +#ifdef __KERNEL__
39155 +#ifndef CONFIG_AUFS_FHSM
39156 +#undef AuBrAttr_FHSM
39157 +#define AuBrAttr_FHSM          0
39158 +#endif
39159 +#ifndef CONFIG_AUFS_XATTR
39160 +#undef AuBrAttr_ICEX
39161 +#define AuBrAttr_ICEX          0
39162 +#undef AuBrAttr_ICEX_SEC
39163 +#define AuBrAttr_ICEX_SEC      0
39164 +#undef AuBrAttr_ICEX_SYS
39165 +#define AuBrAttr_ICEX_SYS      0
39166 +#undef AuBrAttr_ICEX_TR
39167 +#define AuBrAttr_ICEX_TR       0
39168 +#undef AuBrAttr_ICEX_USR
39169 +#define AuBrAttr_ICEX_USR      0
39170 +#undef AuBrAttr_ICEX_OTH
39171 +#define AuBrAttr_ICEX_OTH      0
39172 +#endif
39173 +#endif
39174 +
39175 +/* the longest combination */
39176 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
39177 +#define AuBrPermStrSz  sizeof(AUFS_BRPERM_RW                   \
39178 +                              "+" AUFS_BRATTR_COO_REG          \
39179 +                              "+" AUFS_BRATTR_FHSM             \
39180 +                              "+" AUFS_BRATTR_UNPIN            \
39181 +                              "+" AUFS_BRATTR_ICEX_SEC         \
39182 +                              "+" AUFS_BRATTR_ICEX_SYS         \
39183 +                              "+" AUFS_BRATTR_ICEX_USR         \
39184 +                              "+" AUFS_BRATTR_ICEX_OTH         \
39185 +                              "+" AUFS_BRWATTR_NLWH)
39186 +
39187 +typedef struct {
39188 +       char a[AuBrPermStrSz];
39189 +} au_br_perm_str_t;
39190 +
39191 +static inline int au_br_writable(int brperm)
39192 +{
39193 +       return brperm & AuBrPerm_RW;
39194 +}
39195 +
39196 +static inline int au_br_whable(int brperm)
39197 +{
39198 +       return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
39199 +}
39200 +
39201 +static inline int au_br_wh_linkable(int brperm)
39202 +{
39203 +       return !(brperm & AuBrWAttr_NoLinkWH);
39204 +}
39205 +
39206 +static inline int au_br_cmoo(int brperm)
39207 +{
39208 +       return brperm & AuBrAttr_CMOO_Mask;
39209 +}
39210 +
39211 +static inline int au_br_fhsm(int brperm)
39212 +{
39213 +       return brperm & AuBrAttr_FHSM;
39214 +}
39215 +
39216 +/* ---------------------------------------------------------------------- */
39217 +
39218 +/* ioctl */
39219 +enum {
39220 +       /* readdir in userspace */
39221 +       AuCtl_RDU,
39222 +       AuCtl_RDU_INO,
39223 +
39224 +       AuCtl_WBR_FD,   /* pathconf wrapper */
39225 +       AuCtl_IBUSY,    /* busy inode */
39226 +       AuCtl_MVDOWN,   /* move-down */
39227 +       AuCtl_BR,       /* info about branches */
39228 +       AuCtl_FHSM_FD   /* connection for fhsm */
39229 +};
39230 +
39231 +/* borrowed from linux/include/linux/kernel.h */
39232 +#ifndef ALIGN
39233 +#ifdef _GNU_SOURCE
39234 +#define ALIGN(x, a)            __ALIGN_MASK(x, (typeof(x))(a)-1)
39235 +#else
39236 +#define ALIGN(x, a)            (((x) + (a) - 1) & ~((a) - 1))
39237 +#endif
39238 +#define __ALIGN_MASK(x, mask)  (((x)+(mask))&~(mask))
39239 +#endif
39240 +
39241 +/* borrowed from linux/include/linux/compiler-gcc3.h */
39242 +#ifndef __aligned
39243 +#define __aligned(x)                   __attribute__((aligned(x)))
39244 +#endif
39245 +
39246 +#ifdef __KERNEL__
39247 +#ifndef __packed
39248 +#define __packed                       __attribute__((packed))
39249 +#endif
39250 +#endif
39251 +
39252 +struct au_rdu_cookie {
39253 +       uint64_t        h_pos;
39254 +       int16_t         bindex;
39255 +       uint8_t         flags;
39256 +       uint8_t         pad;
39257 +       uint32_t        generation;
39258 +} __aligned(8);
39259 +
39260 +struct au_rdu_ent {
39261 +       uint64_t        ino;
39262 +       int16_t         bindex;
39263 +       uint8_t         type;
39264 +       uint8_t         nlen;
39265 +       uint8_t         wh;
39266 +       char            name[];
39267 +} __aligned(8);
39268 +
39269 +static inline int au_rdu_len(int nlen)
39270 +{
39271 +       /* include the terminating NULL */
39272 +       return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
39273 +                    sizeof(uint64_t));
39274 +}
39275 +
39276 +union au_rdu_ent_ul {
39277 +       struct au_rdu_ent __user        *e;
39278 +       uint64_t                        ul;
39279 +};
39280 +
39281 +enum {
39282 +       AufsCtlRduV_SZ,
39283 +       AufsCtlRduV_End
39284 +};
39285 +
39286 +struct aufs_rdu {
39287 +       /* input */
39288 +       union {
39289 +               uint64_t        sz;     /* AuCtl_RDU */
39290 +               uint64_t        nent;   /* AuCtl_RDU_INO */
39291 +       };
39292 +       union au_rdu_ent_ul     ent;
39293 +       uint16_t                verify[AufsCtlRduV_End];
39294 +
39295 +       /* input/output */
39296 +       uint32_t                blk;
39297 +
39298 +       /* output */
39299 +       union au_rdu_ent_ul     tail;
39300 +       /* number of entries which were added in a single call */
39301 +       uint64_t                rent;
39302 +       uint8_t                 full;
39303 +       uint8_t                 shwh;
39304 +
39305 +       struct au_rdu_cookie    cookie;
39306 +} __aligned(8);
39307 +
39308 +/* ---------------------------------------------------------------------- */
39309 +
39310 +/* dirren. the branch is identified by the filename who contains this */
39311 +struct au_drinfo {
39312 +       uint64_t ino;
39313 +       union {
39314 +               uint8_t oldnamelen;
39315 +               uint64_t _padding;
39316 +       };
39317 +       uint8_t oldname[];
39318 +} __aligned(8);
39319 +
39320 +struct au_drinfo_fdata {
39321 +       uint32_t magic;
39322 +       struct au_drinfo drinfo;
39323 +} __aligned(8);
39324 +
39325 +#define AUFS_DRINFO_MAGIC_V1   ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x01)
39326 +/* future */
39327 +#define AUFS_DRINFO_MAGIC_V2   ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x02)
39328 +
39329 +/* ---------------------------------------------------------------------- */
39330 +
39331 +struct aufs_wbr_fd {
39332 +       uint32_t        oflags;
39333 +       int16_t         brid;
39334 +} __aligned(8);
39335 +
39336 +/* ---------------------------------------------------------------------- */
39337 +
39338 +struct aufs_ibusy {
39339 +       uint64_t        ino, h_ino;
39340 +       int16_t         bindex;
39341 +} __aligned(8);
39342 +
39343 +/* ---------------------------------------------------------------------- */
39344 +
39345 +/* error code for move-down */
39346 +/* the actual message strings are implemented in aufs-util.git */
39347 +enum {
39348 +       EAU_MVDOWN_OPAQUE = 1,
39349 +       EAU_MVDOWN_WHITEOUT,
39350 +       EAU_MVDOWN_UPPER,
39351 +       EAU_MVDOWN_BOTTOM,
39352 +       EAU_MVDOWN_NOUPPER,
39353 +       EAU_MVDOWN_NOLOWERBR,
39354 +       EAU_Last
39355 +};
39356 +
39357 +/* flags for move-down */
39358 +#define AUFS_MVDOWN_DMSG       1
39359 +#define AUFS_MVDOWN_OWLOWER    (1 << 1)        /* overwrite lower */
39360 +#define AUFS_MVDOWN_KUPPER     (1 << 2)        /* keep upper */
39361 +#define AUFS_MVDOWN_ROLOWER    (1 << 3)        /* do even if lower is RO */
39362 +#define AUFS_MVDOWN_ROLOWER_R  (1 << 4)        /* did on lower RO */
39363 +#define AUFS_MVDOWN_ROUPPER    (1 << 5)        /* do even if upper is RO */
39364 +#define AUFS_MVDOWN_ROUPPER_R  (1 << 6)        /* did on upper RO */
39365 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7)        /* upper brid */
39366 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8)        /* lower brid */
39367 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9)        /* find fhsm attr for lower */
39368 +#define AUFS_MVDOWN_STFS       (1 << 10)       /* req. stfs */
39369 +#define AUFS_MVDOWN_STFS_FAILED        (1 << 11)       /* output: stfs is unusable */
39370 +#define AUFS_MVDOWN_BOTTOM     (1 << 12)       /* output: no more lowers */
39371 +
39372 +/* index for move-down */
39373 +enum {
39374 +       AUFS_MVDOWN_UPPER,
39375 +       AUFS_MVDOWN_LOWER,
39376 +       AUFS_MVDOWN_NARRAY
39377 +};
39378 +
39379 +/*
39380 + * additional info of move-down
39381 + * number of free blocks and inodes.
39382 + * subset of struct kstatfs, but smaller and always 64bit.
39383 + */
39384 +struct aufs_stfs {
39385 +       uint64_t        f_blocks;
39386 +       uint64_t        f_bavail;
39387 +       uint64_t        f_files;
39388 +       uint64_t        f_ffree;
39389 +};
39390 +
39391 +struct aufs_stbr {
39392 +       int16_t                 brid;   /* optional input */
39393 +       int16_t                 bindex; /* output */
39394 +       struct aufs_stfs        stfs;   /* output when AUFS_MVDOWN_STFS set */
39395 +} __aligned(8);
39396 +
39397 +struct aufs_mvdown {
39398 +       uint32_t                flags;                  /* input/output */
39399 +       struct aufs_stbr        stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
39400 +       int8_t                  au_errno;               /* output */
39401 +} __aligned(8);
39402 +
39403 +/* ---------------------------------------------------------------------- */
39404 +
39405 +union aufs_brinfo {
39406 +       /* PATH_MAX may differ between kernel-space and user-space */
39407 +       char    _spacer[4096];
39408 +       struct {
39409 +               int16_t id;
39410 +               int     perm;
39411 +               char    path[];
39412 +       };
39413 +} __aligned(8);
39414 +
39415 +/* ---------------------------------------------------------------------- */
39416 +
39417 +#define AuCtlType              'A'
39418 +#define AUFS_CTL_RDU           _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
39419 +#define AUFS_CTL_RDU_INO       _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
39420 +#define AUFS_CTL_WBR_FD                _IOW(AuCtlType, AuCtl_WBR_FD, \
39421 +                                    struct aufs_wbr_fd)
39422 +#define AUFS_CTL_IBUSY         _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
39423 +#define AUFS_CTL_MVDOWN                _IOWR(AuCtlType, AuCtl_MVDOWN, \
39424 +                                     struct aufs_mvdown)
39425 +#define AUFS_CTL_BRINFO                _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
39426 +#define AUFS_CTL_FHSM_FD       _IOW(AuCtlType, AuCtl_FHSM_FD, int)
39427 +
39428 +#endif /* __AUFS_TYPE_H__ */
39429 SPDX-License-Identifier: GPL-2.0
39430 aufs6.0 loopback patch
39431
39432 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
39433 index 8ca56da3ca69..9ac6f67edf09 100644
39434 --- a/drivers/block/loop.c
39435 +++ b/drivers/block/loop.c
39436 @@ -54,7 +54,7 @@ struct loop_device {
39437         int             lo_flags;
39438         char            lo_file_name[LO_NAME_SIZE];
39439
39440 -       struct file *   lo_backing_file;
39441 +       struct file     *lo_backing_file, *lo_backing_virt_file;
39442         struct block_device *lo_device;
39443
39444         gfp_t           old_gfp_mask;
39445 @@ -510,6 +510,15 @@ static inline void loop_update_dio(struct loop_device *lo)
39446                                 lo->use_dio);
39447  }
39448
39449 +static struct file *loop_real_file(struct file *file)
39450 +{
39451 +       struct file *f = NULL;
39452 +
39453 +       if (file->f_path.dentry->d_sb->s_op->real_loop)
39454 +               f = file->f_path.dentry->d_sb->s_op->real_loop(file);
39455 +       return f;
39456 +}
39457 +
39458  static void loop_reread_partitions(struct loop_device *lo)
39459  {
39460         int rc;
39461 @@ -567,6 +576,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39462  {
39463         struct file *file = fget(arg);
39464         struct file *old_file;
39465 +       struct file *f, *virt_file = NULL, *old_virt_file;
39466         int error;
39467         bool partscan;
39468         bool is_loop;
39469 @@ -590,11 +600,19 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39470         if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
39471                 goto out_err;
39472
39473 +       f = loop_real_file(file);
39474 +       if (f) {
39475 +               virt_file = file;
39476 +               file = f;
39477 +               get_file(file);
39478 +       }
39479 +
39480         error = loop_validate_file(file, bdev);
39481         if (error)
39482                 goto out_err;
39483
39484         old_file = lo->lo_backing_file;
39485 +       old_virt_file = lo->lo_backing_virt_file;
39486
39487         error = -EINVAL;
39488
39489 @@ -607,6 +625,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39490         blk_mq_freeze_queue(lo->lo_queue);
39491         mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
39492         lo->lo_backing_file = file;
39493 +       lo->lo_backing_virt_file = virt_file;
39494         lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
39495         mapping_set_gfp_mask(file->f_mapping,
39496                              lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39497 @@ -629,6 +648,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39498          * dependency.
39499          */
39500         fput(old_file);
39501 +       if (old_virt_file)
39502 +               fput(old_virt_file);
39503         if (partscan)
39504                 loop_reread_partitions(lo);
39505
39506 @@ -642,6 +663,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39507         loop_global_unlock(lo, is_loop);
39508  out_putf:
39509         fput(file);
39510 +       if (virt_file)
39511 +               fput(virt_file);
39512         goto done;
39513  }
39514
39515 @@ -1013,6 +1036,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39516                           const struct loop_config *config)
39517  {
39518         struct file *file = fget(config->fd);
39519 +       struct file *f, *virt_file = NULL;
39520         struct inode *inode;
39521         struct address_space *mapping;
39522         int error;
39523 @@ -1031,6 +1055,13 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39524         /* suppress uevents while reconfiguring the device */
39525         dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 1);
39526
39527 +       f = loop_real_file(file);
39528 +       if (f) {
39529 +               virt_file = file;
39530 +               file = f;
39531 +               get_file(file);
39532 +       }
39533 +
39534         /*
39535          * If we don't hold exclusive handle for the device, upgrade to it
39536          * here to avoid changing device under exclusive owner.
39537 @@ -1091,6 +1122,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39538         lo->use_dio = lo->lo_flags & LO_FLAGS_DIRECT_IO;
39539         lo->lo_device = bdev;
39540         lo->lo_backing_file = file;
39541 +       lo->lo_backing_virt_file = virt_file;
39542         lo->old_gfp_mask = mapping_gfp_mask(mapping);
39543         mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39544
39545 @@ -1146,6 +1178,8 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39546                 bd_abort_claiming(bdev, loop_configure);
39547  out_putf:
39548         fput(file);
39549 +       if (virt_file)
39550 +               fput(virt_file);
39551         /* This is safe: open() is still holding a reference. */
39552         module_put(THIS_MODULE);
39553         goto done;
39554 @@ -1154,6 +1188,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39555  static void __loop_clr_fd(struct loop_device *lo, bool release)
39556  {
39557         struct file *filp;
39558 +       struct file *virt_filp = lo->lo_backing_virt_file;
39559         gfp_t gfp = lo->old_gfp_mask;
39560
39561         if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags))
39562 @@ -1170,6 +1205,7 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39563         spin_lock_irq(&lo->lo_lock);
39564         filp = lo->lo_backing_file;
39565         lo->lo_backing_file = NULL;
39566 +       lo->lo_backing_virt_file = NULL;
39567         spin_unlock_irq(&lo->lo_lock);
39568
39569         lo->lo_device = NULL;
39570 @@ -1232,6 +1268,8 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39571          * fput can take open_mutex which is usually taken before lo_mutex.
39572          */
39573         fput(filp);
39574 +       if (virt_filp)
39575 +               fput(virt_filp);
39576  }
39577
39578  static int loop_clr_fd(struct loop_device *lo)
39579 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
39580 index 5c5641222d67..79504ea71dad 100644
39581 --- a/fs/aufs/f_op.c
39582 +++ b/fs/aufs/f_op.c
39583 @@ -304,7 +304,7 @@ static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
39584         if (IS_ERR(h_file))
39585                 goto out;
39586
39587 -       if (au_test_loopback_kthread()) {
39588 +       if (0 && au_test_loopback_kthread()) {
39589                 au_warn_loopback(h_file->f_path.dentry->d_sb);
39590                 if (file->f_mapping != h_file->f_mapping) {
39591                         file->f_mapping = h_file->f_mapping;
39592 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
39593 index 58043e31e5f3..e2bfae6f9d59 100644
39594 --- a/fs/aufs/loop.c
39595 +++ b/fs/aufs/loop.c
39596 @@ -133,3 +133,19 @@ void au_loopback_fin(void)
39597                 symbol_put(loop_backing_file);
39598         au_kfree_try_rcu(au_warn_loopback_array);
39599  }
39600 +
39601 +/* ---------------------------------------------------------------------- */
39602 +
39603 +/* support the loopback block device insude aufs */
39604 +
39605 +struct file *aufs_real_loop(struct file *file)
39606 +{
39607 +       struct file *f;
39608 +
39609 +       BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
39610 +       fi_read_lock(file);
39611 +       f = au_hf_top(file);
39612 +       fi_read_unlock(file);
39613 +       AuDebugOn(!f);
39614 +       return f;
39615 +}
39616 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
39617 index 03d4908a6c03..34d356e181d5 100644
39618 --- a/fs/aufs/loop.h
39619 +++ b/fs/aufs/loop.h
39620 @@ -26,6 +26,8 @@ void au_warn_loopback(struct super_block *h_sb);
39621
39622  int au_loopback_init(void);
39623  void au_loopback_fin(void);
39624 +
39625 +struct file *aufs_real_loop(struct file *file);
39626  #else
39627  AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
39628
39629 @@ -36,6 +38,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
39630
39631  AuStubInt0(au_loopback_init, void)
39632  AuStubVoid(au_loopback_fin, void)
39633 +
39634 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
39635  #endif /* BLK_DEV_LOOP */
39636
39637  #endif /* __KERNEL__ */
39638 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
39639 index 81922d4faf54..c8a62c267d72 100644
39640 --- a/fs/aufs/super.c
39641 +++ b/fs/aufs/super.c
39642 @@ -758,7 +758,10 @@ const struct super_operations aufs_sop = {
39643         .show_options   = aufs_show_options,
39644         .statfs         = aufs_statfs,
39645         .put_super      = aufs_put_super,
39646 -       .sync_fs        = aufs_sync_fs
39647 +       .sync_fs        = aufs_sync_fs,
39648 +#ifdef CONFIG_AUFS_BDEV_LOOP
39649 +       .real_loop      = aufs_real_loop
39650 +#endif
39651  };
39652
39653  /* ---------------------------------------------------------------------- */
39654 diff --git a/include/linux/fs.h b/include/linux/fs.h
39655 index 9c5a33fce3a8..a3c40bc66852 100644
39656 --- a/include/linux/fs.h
39657 +++ b/include/linux/fs.h
39658 @@ -2253,6 +2253,10 @@ struct super_operations {
39659                                   struct shrink_control *);
39660         long (*free_cached_objects)(struct super_block *,
39661                                     struct shrink_control *);
39662 +#if IS_ENABLED(CONFIG_BLK_DEV_LOOP) || IS_ENABLED(CONFIG_BLK_DEV_LOOP_MODULE)
39663 +       /* and aufs */
39664 +       struct file *(*real_loop)(struct file *);
39665 +#endif
39666  };
39667
39668  /*