1 SPDX-License-Identifier: GPL-2.0
2 aufs5.x-rcN kbuild patch
4 diff --git a/fs/Kconfig b/fs/Kconfig
5 index a547307c1ae8..919ca5404804 100644
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"
14 endif # MISC_FILESYSTEMS
16 diff --git a/fs/Makefile b/fs/Makefile
17 index 93b80529f8e8..a26c58d09435 100644
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 aufs5.x-rcN base patch
28 diff --git a/MAINTAINERS b/MAINTAINERS
29 index d30f26e07cd3..2331a15f3550 100644
32 @@ -3426,6 +3426,19 @@ F: include/uapi/linux/audit.h
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
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/
47 +F: include/uapi/linux/aufs_type.h
49 AUXILIARY DISPLAY DRIVERS
50 M: Miguel Ojeda <ojeda@kernel.org>
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,
62 + * no get/put for file.
64 +struct file *loop_backing_file(struct super_block *sb)
67 + struct loop_device *l;
70 + if (MAJOR(sb->s_dev) == LOOP_MAJOR) {
71 + l = sb->s_bdev->bd_disk->private_data;
72 + ret = l->lo_backing_file;
76 +EXPORT_SYMBOL_GPL(loop_backing_file);
78 /* loop sysfs attributes */
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
85 @@ -1345,7 +1345,7 @@ enum d_walk_ret {
87 * The @enter() callbacks are called with d_lock held.
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 *))
93 struct dentry *this_parent;
94 diff --git a/fs/fcntl.c b/fs/fcntl.c
95 index 146c9ab0cd4b..c980a0524118 100644
100 #define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
102 -static int setfl(int fd, struct file * filp, unsigned long arg)
103 +int setfl(int fd, struct file *filp, unsigned long arg)
105 struct inode * inode = file_inode(filp);
107 @@ -63,6 +63,8 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
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);
116 diff --git a/fs/namespace.c b/fs/namespace.c
117 index df137ba19d37..faae7f8644ac 100644
120 @@ -852,6 +852,12 @@ static inline int check_mnt(struct mount *mnt)
121 return mnt->mnt_ns == current->nsproxy->mnt_ns;
124 +/* for aufs, CONFIG_AUFS_BR_FUSE */
125 +int is_current_mnt_ns(struct vfsmount *mnt)
127 + return check_mnt(real_mount(mnt));
131 * vfsmount lock must be held for write
133 diff --git a/fs/splice.c b/fs/splice.c
134 index 0878b852b355..d83f9e5cddbc 100644
137 @@ -756,8 +756,8 @@ static int warn_unsupported(struct file *file, const char *op)
139 * Attempt to initiate a splice from pipe to file.
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)
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,
150 * Attempt to initiate a splice from a file to a pipe.
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)
159 unsigned int p_space;
161 diff --git a/include/linux/fs.h b/include/linux/fs.h
162 index 9eced4cc286e..773ebd5696e4 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);
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 @@ -2615,6 +2617,7 @@ extern int current_umask(void);
182 extern void ihold(struct inode * inode);
183 extern void iput(struct inode *);
184 extern int generic_update_time(struct inode *, struct timespec64 *, int);
185 +extern int update_time(struct inode *, struct timespec64 *, int);
188 extern struct kobject *fs_kobj;
189 @@ -2795,6 +2798,7 @@ static inline bool sb_is_blkdev_sb(struct super_block *sb)
192 void emergency_thaw_all(void);
193 +extern int __sync_filesystem(struct super_block *, int);
194 extern int sync_filesystem(struct super_block *);
195 extern const struct file_operations def_blk_fops;
196 extern const struct file_operations def_chr_fops;
197 diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
198 index 1f1099dac3f0..81d10b8e9dc0 100644
199 --- a/include/linux/lockdep.h
200 +++ b/include/linux/lockdep.h
201 @@ -248,6 +248,8 @@ static inline int lockdep_match_key(struct lockdep_map *lock,
202 return lock->key == key;
205 +struct lock_class *lockdep_hlock_class(struct held_lock *hlock);
210 diff --git a/include/linux/mnt_namespace.h b/include/linux/mnt_namespace.h
211 index 8f882f5881e8..6b9808f09843 100644
212 --- a/include/linux/mnt_namespace.h
213 +++ b/include/linux/mnt_namespace.h
214 @@ -7,12 +7,15 @@ struct mnt_namespace;
216 struct user_namespace;
220 extern struct mnt_namespace *copy_mnt_ns(unsigned long, struct mnt_namespace *,
221 struct user_namespace *, struct fs_struct *);
222 extern void put_mnt_ns(struct mnt_namespace *ns);
223 extern struct ns_common *from_mnt_ns(struct mnt_namespace *);
225 +extern int is_current_mnt_ns(struct vfsmount *mnt);
227 extern const struct file_operations proc_mounts_operations;
228 extern const struct file_operations proc_mountinfo_operations;
229 extern const struct file_operations proc_mountstats_operations;
230 diff --git a/include/linux/splice.h b/include/linux/splice.h
231 index a55179fd60fc..8e21c53cf883 100644
232 --- a/include/linux/splice.h
233 +++ b/include/linux/splice.h
234 @@ -93,4 +93,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
236 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
237 extern const struct pipe_buf_operations default_pipe_buf_ops;
239 +extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
240 + loff_t *ppos, size_t len, unsigned int flags);
241 +extern long do_splice_to(struct file *in, loff_t *ppos,
242 + struct pipe_inode_info *pipe, size_t len,
243 + unsigned int flags);
245 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
246 index 64a13eb56078..c9f30e12aa32 100644
247 --- a/kernel/locking/lockdep.c
248 +++ b/kernel/locking/lockdep.c
249 @@ -217,7 +217,7 @@ unsigned long max_lock_class_idx;
250 struct lock_class lock_classes[MAX_LOCKDEP_KEYS];
251 DECLARE_BITMAP(lock_classes_in_use, MAX_LOCKDEP_KEYS);
253 -static inline struct lock_class *hlock_class(struct held_lock *hlock)
254 +inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
256 unsigned int class_idx = hlock->class_idx;
258 @@ -238,6 +238,7 @@ static inline struct lock_class *hlock_class(struct held_lock *hlock)
260 return lock_classes + class_idx;
262 +#define hlock_class(hlock) lockdep_hlock_class(hlock)
264 #ifdef CONFIG_LOCK_STAT
265 static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
266 SPDX-License-Identifier: GPL-2.0
267 aufs5.x-rcN mmap patch
269 diff --git a/fs/proc/base.c b/fs/proc/base.c
270 index 93f7e3d971e4..23aa1626020a 100644
273 @@ -2218,7 +2218,7 @@ static int map_files_get_link(struct dentry *dentry, struct path *path)
275 vma = find_exact_vma(mm, vm_start, vm_end);
276 if (vma && vma->vm_file) {
277 - *path = vma->vm_file->f_path;
278 + *path = vma_pr_or_file(vma)->f_path;
282 diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
283 index 4d3493579458..42edd9a42c78 100644
284 --- a/fs/proc/nommu.c
285 +++ b/fs/proc/nommu.c
286 @@ -39,7 +39,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
287 file = region->vm_file;
290 - struct inode *inode = file_inode(region->vm_file);
291 + struct inode *inode;
293 + file = vmr_pr_or_file(region);
294 + inode = file_inode(file);
295 dev = inode->i_sb->s_dev;
298 diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
299 index 4e0023643f8b..09394cf6e349 100644
300 --- a/fs/proc/task_mmu.c
301 +++ b/fs/proc/task_mmu.c
302 @@ -281,7 +281,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma)
303 const char *name = NULL;
306 - struct inode *inode = file_inode(vma->vm_file);
307 + struct inode *inode;
309 + file = vma_pr_or_file(vma);
310 + inode = file_inode(file);
311 dev = inode->i_sb->s_dev;
313 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
314 @@ -1912,7 +1915,7 @@ static int show_numa_map(struct seq_file *m, void *v)
315 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
316 struct vm_area_struct *vma = v;
317 struct numa_maps *md = &numa_priv->md;
318 - struct file *file = vma->vm_file;
319 + struct file *file = vma_pr_or_file(vma);
320 struct mm_struct *mm = vma->vm_mm;
321 struct mempolicy *pol;
323 diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
324 index a6d21fc0033c..02c2de31196e 100644
325 --- a/fs/proc/task_nommu.c
326 +++ b/fs/proc/task_nommu.c
327 @@ -155,7 +155,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
331 - struct inode *inode = file_inode(vma->vm_file);
332 + struct inode *inode;
334 + file = vma_pr_or_file(vma);
335 + inode = file_inode(file);
336 dev = inode->i_sb->s_dev;
338 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
339 diff --git a/include/linux/mm.h b/include/linux/mm.h
340 index 21f8b27bd9fd..de9572fe212b 100644
341 --- a/include/linux/mm.h
342 +++ b/include/linux/mm.h
343 @@ -1858,6 +1858,43 @@ static inline void unmap_shared_mapping_range(struct address_space *mapping,
344 unmap_mapping_range(mapping, holebegin, holelen, 0);
347 +#if IS_ENABLED(CONFIG_AUFS_FS)
348 +extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
349 +extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
351 +extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
352 +extern void vma_do_fput(struct vm_area_struct *, const char[], int);
354 +#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
356 +#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
358 +#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
359 +#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
362 +extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
363 +extern void vmr_do_fput(struct vm_region *, const char[], int);
365 +#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
367 +#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
368 +#endif /* !CONFIG_MMU */
372 +#define vma_file_update_time(vma) file_update_time((vma)->vm_file)
373 +#define vma_pr_or_file(vma) (vma)->vm_file
374 +#define vma_get_file(vma) get_file((vma)->vm_file)
375 +#define vma_fput(vma) fput((vma)->vm_file)
378 +#define vmr_pr_or_file(region) (region)->vm_file
379 +#define vmr_fput(region) fput((region)->vm_file)
380 +#endif /* !CONFIG_MMU */
382 +#endif /* CONFIG_AUFS_FS */
384 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
385 void *buf, int len, unsigned int gup_flags);
386 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
387 diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
388 index cf97f3884fda..a783130a669d 100644
389 --- a/include/linux/mm_types.h
390 +++ b/include/linux/mm_types.h
391 @@ -372,6 +372,9 @@ struct vm_region {
392 unsigned long vm_top; /* region allocated to here */
393 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
394 struct file *vm_file; /* the backing file or NULL */
395 +#if IS_ENABLED(CONFIG_AUFS_FS)
396 + struct file *vm_prfile; /* the virtual backing file or NULL */
399 int vm_usage; /* region usage count (access under nommu_region_sem) */
400 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
401 @@ -468,6 +471,9 @@ struct vm_area_struct {
402 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
404 struct file * vm_file; /* File we map to (can be NULL). */
405 +#if IS_ENABLED(CONFIG_AUFS_FS)
406 + struct file *vm_prfile; /* shadow of vm_file */
408 void * vm_private_data; /* was vm_pte (shared mem) */
411 diff --git a/kernel/fork.c b/kernel/fork.c
412 index 90c85b17bf69..55406414ade4 100644
415 @@ -662,7 +662,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
417 struct address_space *mapping = file->f_mapping;
421 i_mmap_lock_write(mapping);
422 if (tmp->vm_flags & VM_SHARED)
423 mapping_allow_writable(mapping);
424 diff --git a/mm/Makefile b/mm/Makefile
425 index 9a564f836403..86a4be7f34a8 100644
428 @@ -134,3 +134,4 @@ obj-$(CONFIG_IO_MAPPING) += io-mapping.o
429 obj-$(CONFIG_HAVE_BOOTMEM_INFO_NODE) += bootmem_info.o
430 obj-$(CONFIG_GENERIC_IOREMAP) += ioremap.o
431 obj-$(CONFIG_SHRINKER_DEBUG) += shrinker_debug.o
432 +obj-$(CONFIG_AUFS_FS:m=y) += prfile.o
433 diff --git a/mm/filemap.c b/mm/filemap.c
434 index 15800334147b..2c22abb4a14d 100644
437 @@ -3403,7 +3403,7 @@ vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf)
438 vm_fault_t ret = VM_FAULT_LOCKED;
440 sb_start_pagefault(mapping->host->i_sb);
441 - file_update_time(vmf->vma->vm_file);
442 + vma_file_update_time(vmf->vma);
444 if (folio->mapping != mapping) {
446 diff --git a/mm/mmap.c b/mm/mmap.c
447 index 9d780f415be3..e744ba276454 100644
450 @@ -142,7 +142,7 @@ static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
451 if (vma->vm_ops && vma->vm_ops->close)
452 vma->vm_ops->close(vma);
454 - fput(vma->vm_file);
456 mpol_put(vma_policy(vma));
459 @@ -911,7 +911,7 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
462 uprobe_munmap(next, next->vm_start, next->vm_end);
467 anon_vma_merge(vma, next);
468 @@ -1845,7 +1845,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
469 if (vma->vm_ops && vma->vm_ops->close)
470 vma->vm_ops->close(vma);
472 - fput(vma->vm_file);
476 /* Undo any partial mapping done by a device driver. */
477 @@ -2705,7 +2705,7 @@ int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
481 - get_file(new->vm_file);
484 if (new->vm_ops && new->vm_ops->open)
485 new->vm_ops->open(new);
486 @@ -2724,7 +2724,7 @@ int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
487 if (new->vm_ops && new->vm_ops->close)
488 new->vm_ops->close(new);
490 - fput(new->vm_file);
492 unlink_anon_vmas(new);
494 mpol_put(vma_policy(new));
495 @@ -2897,6 +2897,9 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
496 unsigned long populate = 0;
497 unsigned long ret = -EINVAL;
499 +#if IS_ENABLED(CONFIG_AUFS_FS)
500 + struct file *prfile;
503 pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/mm/remap_file_pages.rst.\n",
504 current->comm, current->pid);
505 @@ -2952,10 +2955,34 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
506 if (vma->vm_flags & VM_LOCKED)
509 +#if IS_ENABLED(CONFIG_AUFS_FS)
511 + file = vma->vm_file;
512 + prfile = vma->vm_prfile;
513 + ret = do_mmap(vma->vm_file, start, size,
514 + prot, flags, pgoff, &populate, NULL);
515 + if (!IS_ERR_VALUE(ret) && file && prfile) {
516 + struct vm_area_struct *new_vma;
518 + new_vma = find_vma(mm, ret);
519 + if (!new_vma->vm_prfile)
520 + new_vma->vm_prfile = prfile;
521 + if (new_vma != vma)
525 + * two fput()s instead of vma_fput(vma),
526 + * coz vma may not be available anymore.
532 file = get_file(vma->vm_file);
533 ret = do_mmap(vma->vm_file, start, size,
534 prot, flags, pgoff, &populate, NULL);
536 +#endif /* CONFIG_AUFS_FS */
538 mmap_write_unlock(mm);
540 @@ -3229,7 +3256,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
541 if (anon_vma_clone(new_vma, vma))
542 goto out_free_mempol;
543 if (new_vma->vm_file)
544 - get_file(new_vma->vm_file);
545 + vma_get_file(new_vma);
546 if (new_vma->vm_ops && new_vma->vm_ops->open)
547 new_vma->vm_ops->open(new_vma);
548 vma_link(mm, new_vma, prev, rb_link, rb_parent);
549 diff --git a/mm/nommu.c b/mm/nommu.c
550 index e819cbc21b39..dc7d40fb197f 100644
553 @@ -524,7 +524,7 @@ static void __put_nommu_region(struct vm_region *region)
554 up_write(&nommu_region_sem);
557 - fput(region->vm_file);
560 /* IO memory and memory shared directly out of the pagecache
561 * from ramfs/tmpfs mustn't be released here */
562 @@ -656,7 +656,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
563 if (vma->vm_ops && vma->vm_ops->close)
564 vma->vm_ops->close(vma);
566 - fput(vma->vm_file);
568 put_nommu_region(vma->vm_region);
571 @@ -1176,7 +1176,7 @@ unsigned long do_mmap(struct file *file,
572 goto error_just_free;
575 - fput(region->vm_file);
577 kmem_cache_free(vm_region_jar, region);
580 @@ -1253,10 +1253,10 @@ unsigned long do_mmap(struct file *file,
581 up_write(&nommu_region_sem);
584 - fput(region->vm_file);
586 kmem_cache_free(vm_region_jar, region);
588 - fput(vma->vm_file);
593 diff --git a/mm/prfile.c b/mm/prfile.c
595 index 000000000000..511543ab1b41
599 +// SPDX-License-Identifier: GPL-2.0
601 + * Mainly for aufs which mmap(2) different file and wants to print different
602 + * path in /proc/PID/maps.
603 + * Call these functions via macros defined in linux/mm.h.
605 + * See Documentation/filesystems/aufs/design/06mmap.txt
607 + * Copyright (c) 2014-2021 Junjro R. Okajima
608 + * Copyright (c) 2014 Ian Campbell
611 +#include <linux/mm.h>
612 +#include <linux/file.h>
613 +#include <linux/fs.h>
615 +/* #define PRFILE_TRACE */
616 +static inline void prfile_trace(struct file *f, struct file *pr,
617 + const char func[], int line, const char func2[])
621 + pr_info("%s:%d: %s, %pD2\n", func, line, func2, f);
625 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
628 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
630 + prfile_trace(f, pr, func, line, __func__);
631 + file_update_time(f);
633 + file_update_time(pr);
636 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
639 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
641 + prfile_trace(f, pr, func, line, __func__);
642 + return (f && pr) ? pr : f;
645 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
647 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
649 + prfile_trace(f, pr, func, line, __func__);
655 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
657 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
659 + prfile_trace(f, pr, func, line, __func__);
666 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
669 + struct file *f = region->vm_file, *pr = region->vm_prfile;
671 + prfile_trace(f, pr, func, line, __func__);
672 + return (f && pr) ? pr : f;
675 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
677 + struct file *f = region->vm_file, *pr = region->vm_prfile;
679 + prfile_trace(f, pr, func, line, __func__);
684 +#endif /* !CONFIG_MMU */
685 SPDX-License-Identifier: GPL-2.0
686 aufs5.x-rcN standalone patch
688 diff --git a/fs/dcache.c b/fs/dcache.c
689 index d753ad7d77e6..1a3b40ebc12b 100644
692 @@ -1450,6 +1450,7 @@ void d_walk(struct dentry *parent, void *data,
696 +EXPORT_SYMBOL_GPL(d_walk);
699 struct vfsmount *mnt;
700 @@ -3061,6 +3062,7 @@ void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
702 write_sequnlock(&rename_lock);
704 +EXPORT_SYMBOL_GPL(d_exchange);
707 * d_ancestor - search for an ancestor
708 diff --git a/fs/exec.c b/fs/exec.c
709 index 9a5ca7b82bfc..22cda82470f0 100644
712 @@ -112,6 +112,7 @@ bool path_noexec(const struct path *path)
713 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
714 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
716 +EXPORT_SYMBOL_GPL(path_noexec);
720 diff --git a/fs/fcntl.c b/fs/fcntl.c
721 index c980a0524118..40087d11a974 100644
724 @@ -86,6 +86,7 @@ int setfl(int fd, struct file *filp, unsigned long arg)
728 +EXPORT_SYMBOL_GPL(setfl);
730 static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
732 diff --git a/fs/file_table.c b/fs/file_table.c
733 index 99c6796c9f28..ad533dcedfec 100644
734 --- a/fs/file_table.c
735 +++ b/fs/file_table.c
736 @@ -198,6 +198,7 @@ struct file *alloc_empty_file(int flags, const struct cred *cred)
738 return ERR_PTR(-ENFILE);
740 +EXPORT_SYMBOL_GPL(alloc_empty_file);
743 * Variant of alloc_empty_file() that doesn't check and modify nr_files.
744 diff --git a/fs/namespace.c b/fs/namespace.c
745 index faae7f8644ac..0310c99a83dd 100644
748 @@ -456,6 +456,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
749 mnt_dec_writers(real_mount(mnt));
752 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
755 * mnt_drop_write - give up write access to a mount
756 @@ -857,6 +858,7 @@ int is_current_mnt_ns(struct vfsmount *mnt)
758 return check_mnt(real_mount(mnt));
760 +EXPORT_SYMBOL_GPL(is_current_mnt_ns);
763 * vfsmount lock must be held for write
764 @@ -2060,6 +2062,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
768 +EXPORT_SYMBOL_GPL(iterate_mounts);
770 static void lock_mnt_tree(struct mount *mnt)
772 diff --git a/fs/notify/group.c b/fs/notify/group.c
773 index 1de6631a3925..3008eb37a18d 100644
774 --- a/fs/notify/group.c
775 +++ b/fs/notify/group.c
776 @@ -100,6 +100,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
778 refcount_inc(&group->refcnt);
780 +EXPORT_SYMBOL_GPL(fsnotify_get_group);
783 * Drop a reference to a group. Free it if it's through.
784 diff --git a/fs/open.c b/fs/open.c
785 index 8a813fa5ca56..29f5960ec109 100644
788 @@ -66,6 +66,7 @@ int do_truncate(struct user_namespace *mnt_userns, struct dentry *dentry,
789 inode_unlock(dentry->d_inode);
792 +EXPORT_SYMBOL_GPL(do_truncate);
794 long vfs_truncate(const struct path *path, loff_t length)
796 diff --git a/fs/read_write.c b/fs/read_write.c
797 index 1a261dcf1778..937646c50ea4 100644
798 --- a/fs/read_write.c
799 +++ b/fs/read_write.c
800 @@ -477,6 +477,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
804 +EXPORT_SYMBOL_GPL(vfs_read);
806 static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
808 @@ -586,6 +587,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
809 file_end_write(file);
812 +EXPORT_SYMBOL_GPL(vfs_write);
814 /* file_ppos returns &file->f_pos or NULL if file is stream */
815 static inline loff_t *file_ppos(struct file *file)
816 diff --git a/fs/splice.c b/fs/splice.c
817 index d83f9e5cddbc..ac355ebeb9f0 100644
820 @@ -763,6 +763,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
821 return warn_unsupported(out, "write");
822 return out->f_op->splice_write(pipe, out, ppos, len, flags);
824 +EXPORT_SYMBOL_GPL(do_splice_from);
827 * Attempt to initiate a splice from a file to a pipe.
828 @@ -792,6 +793,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
829 return warn_unsupported(in, "read");
830 return in->f_op->splice_read(in, ppos, pipe, len, flags);
832 +EXPORT_SYMBOL_GPL(do_splice_to);
835 * splice_direct_to_actor - splices data directly between two non-pipes
836 diff --git a/fs/xattr.c b/fs/xattr.c
837 index a1f4998bc6be..284d4ca54893 100644
840 @@ -396,6 +396,7 @@ vfs_getxattr_alloc(struct user_namespace *mnt_userns, struct dentry *dentry,
841 *xattr_value = value;
844 +EXPORT_SYMBOL_GPL(vfs_getxattr_alloc);
847 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
848 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
849 index c9f30e12aa32..d04fedf1e4d9 100644
850 --- a/kernel/locking/lockdep.c
851 +++ b/kernel/locking/lockdep.c
852 @@ -238,6 +238,7 @@ inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
854 return lock_classes + class_idx;
856 +EXPORT_SYMBOL_GPL(lockdep_hlock_class);
857 #define hlock_class(hlock) lockdep_hlock_class(hlock)
859 #ifdef CONFIG_LOCK_STAT
860 diff --git a/kernel/task_work.c b/kernel/task_work.c
861 index dff75bcde151..30b814618535 100644
862 --- a/kernel/task_work.c
863 +++ b/kernel/task_work.c
864 @@ -180,3 +180,4 @@ void task_work_run(void)
868 +EXPORT_SYMBOL_GPL(task_work_run);
869 diff --git a/security/security.c b/security/security.c
870 index 4b95de24bc8d..139e1443c49b 100644
871 --- a/security/security.c
872 +++ b/security/security.c
873 @@ -1165,6 +1165,7 @@ int security_path_rmdir(const struct path *dir, struct dentry *dentry)
875 return call_int_hook(path_rmdir, 0, dir, dentry);
877 +EXPORT_SYMBOL_GPL(security_path_rmdir);
879 int security_path_unlink(const struct path *dir, struct dentry *dentry)
881 @@ -1181,6 +1182,7 @@ int security_path_symlink(const struct path *dir, struct dentry *dentry,
883 return call_int_hook(path_symlink, 0, dir, dentry, old_name);
885 +EXPORT_SYMBOL_GPL(security_path_symlink);
887 int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
888 struct dentry *new_dentry)
889 @@ -1189,6 +1191,7 @@ int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
891 return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
893 +EXPORT_SYMBOL_GPL(security_path_link);
895 int security_path_rename(const struct path *old_dir, struct dentry *old_dentry,
896 const struct path *new_dir, struct dentry *new_dentry,
897 @@ -1209,6 +1212,7 @@ int security_path_truncate(const struct path *path)
899 return call_int_hook(path_truncate, 0, path);
901 +EXPORT_SYMBOL_GPL(security_path_truncate);
903 int security_path_chmod(const struct path *path, umode_t mode)
905 @@ -1216,6 +1220,7 @@ int security_path_chmod(const struct path *path, umode_t mode)
907 return call_int_hook(path_chmod, 0, path, mode);
909 +EXPORT_SYMBOL_GPL(security_path_chmod);
911 int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
913 @@ -1223,6 +1228,7 @@ int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
915 return call_int_hook(path_chown, 0, path, uid, gid);
917 +EXPORT_SYMBOL_GPL(security_path_chown);
919 int security_path_chroot(const struct path *path)
921 @@ -1323,6 +1329,7 @@ int security_inode_permission(struct inode *inode, int mask)
923 return call_int_hook(inode_permission, 0, inode, mask);
925 +EXPORT_SYMBOL_GPL(security_inode_permission);
927 int security_inode_setattr(struct user_namespace *mnt_userns,
928 struct dentry *dentry, struct iattr *attr)
929 @@ -1521,6 +1528,7 @@ int security_file_permission(struct file *file, int mask)
931 return fsnotify_perm(file, mask);
933 +EXPORT_SYMBOL_GPL(security_file_permission);
935 int security_file_alloc(struct file *file)
937 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
938 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
939 +++ linux/Documentation/ABI/testing/debugfs-aufs 2022-03-21 14:49:05.713299677 +0100
941 +What: /debug/aufs/si_<id>/
943 +Contact: J. R. Okajima <hooanon05g@gmail.com>
945 + Under /debug/aufs, a directory named si_<id> is created
946 + per aufs mount, where <id> is a unique id generated
949 +What: /debug/aufs/si_<id>/plink
951 +Contact: J. R. Okajima <hooanon05g@gmail.com>
953 + It has three lines and shows the information about the
954 + pseudo-link. The first line is a single number
955 + representing a number of buckets. The second line is a
956 + number of pseudo-links per buckets (separated by a
957 + blank). The last line is a single number representing a
958 + total number of psedo-links.
959 + When the aufs mount option 'noplink' is specified, it
960 + will show "1\n0\n0\n".
962 +What: /debug/aufs/si_<id>/xib
964 +Contact: J. R. Okajima <hooanon05g@gmail.com>
966 + It shows the consumed blocks by xib (External Inode Number
967 + Bitmap), its block size and file size.
968 + When the aufs mount option 'noxino' is specified, it
969 + will be empty. About XINO files, see the aufs manual.
971 +What: /debug/aufs/si_<id>/xi<branch-index>
973 +Contact: J. R. Okajima <hooanon05g@gmail.com>
975 + It shows the consumed blocks by xino (External Inode Number
976 + Translation Table), its link count, block size and file
978 + Due to the file size limit, there may exist multiple
979 + xino files per branch. In this case, "-N" is added to
980 + the filename and it corresponds to the index of the
981 + internal xino array. "-0" is omitted.
982 + When the aufs mount option 'noxino' is specified, Those
983 + entries won't exist. About XINO files, see the aufs
986 +What: /debug/aufs/si_<id>/xigen
988 +Contact: J. R. Okajima <hooanon05g@gmail.com>
990 + It shows the consumed blocks by xigen (External Inode
991 + Generation Table), its block size and file size.
992 + If CONFIG_AUFS_EXPORT is disabled, this entry will not
994 + When the aufs mount option 'noxino' is specified, it
995 + will be empty. About XINO files, see the aufs manual.
996 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
997 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
998 +++ linux/Documentation/ABI/testing/sysfs-aufs 2022-03-21 14:49:05.713299677 +0100
1000 +What: /sys/fs/aufs/si_<id>/
1002 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1004 + Under /sys/fs/aufs, a directory named si_<id> is created
1005 + per aufs mount, where <id> is a unique id generated
1008 +What: /sys/fs/aufs/si_<id>/br<idx>
1010 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1012 + It shows the abolute path of a member directory (which
1013 + is called branch) in aufs, and its permission.
1015 +What: /sys/fs/aufs/si_<id>/brid<idx>
1017 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1019 + It shows the id of a member directory (which is called
1022 +What: /sys/fs/aufs/si_<id>/xi_path
1024 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1026 + It shows the abolute path of XINO (External Inode Number
1027 + Bitmap, Translation Table and Generation Table) file
1028 + even if it is the default path.
1029 + When the aufs mount option 'noxino' is specified, it
1030 + will be empty. About XINO files, see the aufs manual.
1031 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1032 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
1033 +++ linux/Documentation/filesystems/aufs/design/01intro.txt 2022-03-21 14:49:05.713299677 +0100
1036 +# Copyright (C) 2005-2021 Junjiro R. Okajima
1038 +# This program is free software; you can redistribute it and/or modify
1039 +# it under the terms of the GNU General Public License as published by
1040 +# the Free Software Foundation; either version 2 of the License, or
1041 +# (at your option) any later version.
1043 +# This program is distributed in the hope that it will be useful,
1044 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1045 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1046 +# GNU General Public License for more details.
1048 +# You should have received a copy of the GNU General Public License
1049 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1052 +----------------------------------------
1054 +aufs [ei ju: ef es] | /ey-yoo-ef-es/ | [a u f s]
1055 +1. abbrev. for "advanced multi-layered unification filesystem".
1056 +2. abbrev. for "another unionfs".
1057 +3. abbrev. for "auf das" in German which means "on the" in English.
1058 + Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1059 + But "Filesystem aufs Filesystem" is hard to understand.
1060 +4. abbrev. for "African Urban Fashion Show".
1062 +AUFS is a filesystem with features:
1063 +- multi layered stackable unification filesystem, the member directory
1064 + is called as a branch.
1065 +- branch permission and attribute, 'readonly', 'real-readonly',
1066 + 'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1068 +- internal "file copy-on-write".
1069 +- logical deletion, whiteout.
1070 +- dynamic branch manipulation, adding, deleting and changing permission.
1071 +- allow bypassing aufs, user's direct branch access.
1072 +- external inode number translation table and bitmap which maintains the
1073 + persistent aufs inode number.
1074 +- seekable directory, including NFS readdir.
1075 +- file mapping, mmap and sharing pages.
1076 +- pseudo-link, hardlink over branches.
1077 +- loopback mounted filesystem as a branch.
1078 +- several policies to select one among multiple writable branches.
1079 +- revert a single systemcall when an error occurs in aufs.
1083 +Multi Layered Stackable Unification Filesystem
1084 +----------------------------------------------------------------------
1085 +Most people already knows what it is.
1086 +It is a filesystem which unifies several directories and provides a
1087 +merged single directory. When users access a file, the access will be
1088 +passed/re-directed/converted (sorry, I am not sure which English word is
1089 +correct) to the real file on the member filesystem. The member
1090 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1091 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1092 +readonly branch is handled by creating 'whiteout' on the upper writable
1095 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1096 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1097 +different approaches to implement the merged-view.
1098 +The former tries putting it into VFS, and the latter implements as a
1099 +separate filesystem.
1100 +(If I misunderstand about these implementations, please let me know and
1101 +I shall correct it. Because it is a long time ago when I read their
1102 +source files last time).
1104 +UnionMount's approach will be able to small, but may be hard to share
1105 +branches between several UnionMount since the whiteout in it is
1106 +implemented in the inode on branch filesystem and always
1107 +shared. According to Bharata's post, readdir does not seems to be
1109 +There are several missing features known in this implementations such as
1110 +- for users, the inode number may change silently. eg. copy-up.
1111 +- link(2) may break by copy-up.
1112 +- read(2) may get an obsoleted filedata (fstat(2) too).
1113 +- fcntl(F_SETLK) may be broken by copy-up.
1114 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1117 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1118 +merged into mainline. This is another implementation of UnionMount as a
1119 +separated filesystem. All the limitations and known problems which
1120 +UnionMount are equally inherited to "overlay" filesystem.
1122 +Unionfs has a longer history. When I started implementing a stackable
1123 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1124 +inode, dentry and file objects and they have an array pointing lower
1125 +same kind objects. After contributing many patches for Unionfs, I
1126 +re-started my project AUFS (Jun 2006).
1128 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1129 +implemented my own ideas, approaches and enhancements and it became
1130 +totally different one.
1132 +Comparing DM snapshot and fs based implementation
1133 +- the number of bytes to be copied between devices is much smaller.
1134 +- the type of filesystem must be one and only.
1135 +- the fs must be writable, no readonly fs, even for the lower original
1136 + device. so the compression fs will not be usable. but if we use
1137 + loopback mount, we may address this issue.
1139 + mount /cdrom/squashfs.img /sq
1140 + losetup /sq/ext2.img
1141 + losetup /somewhere/cow
1142 + dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1143 +- it will be difficult (or needs more operations) to extract the
1144 + difference between the original device and COW.
1145 +- DM snapshot-merge may help a lot when users try merging. in the
1146 + fs-layer union, users will use rsync(1).
1148 +You may want to read my old paper "Filesystems in LiveCD"
1149 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1152 +Several characters/aspects/persona of aufs
1153 +----------------------------------------------------------------------
1155 +Aufs has several characters, aspects or persona.
1156 +1. a filesystem, callee of VFS helper
1157 +2. sub-VFS, caller of VFS helper for branches
1158 +3. a virtual filesystem which maintains persistent inode number
1159 +4. reader/writer of files on branches such like an application
1161 +1. Callee of VFS Helper
1162 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1163 +unlink(2) from an application reaches sys_unlink() kernel function and
1164 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1165 +calls filesystem specific unlink operation. Actually aufs implements the
1166 +unlink operation but it behaves like a redirector.
1168 +2. Caller of VFS Helper for Branches
1169 +aufs_unlink() passes the unlink request to the branch filesystem as if
1170 +it were called from VFS. So the called unlink operation of the branch
1171 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1172 +every necessary pre/post operation for the branch filesystem.
1173 +- acquire the lock for the parent dir on a branch
1174 +- lookup in a branch
1175 +- revalidate dentry on a branch
1176 +- mnt_want_write() for a branch
1177 +- vfs_unlink() for a branch
1178 +- mnt_drop_write() for a branch
1179 +- release the lock on a branch
1181 +3. Persistent Inode Number
1182 +One of the most important issue for a filesystem is to maintain inode
1183 +numbers. This is particularly important to support exporting a
1184 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1185 +backend block device for its own. But some storage is necessary to
1186 +keep and maintain the inode numbers. It may be a large space and may not
1187 +suit to keep in memory. Aufs rents some space from its first writable
1188 +branch filesystem (by default) and creates file(s) on it. These files
1189 +are created by aufs internally and removed soon (currently) keeping
1191 +Note: Because these files are removed, they are totally gone after
1192 + unmounting aufs. It means the inode numbers are not persistent
1193 + across unmount or reboot. I have a plan to make them really
1194 + persistent which will be important for aufs on NFS server.
1196 +4. Read/Write Files Internally (copy-on-write)
1197 +Because a branch can be readonly, when you write a file on it, aufs will
1198 +"copy-up" it to the upper writable branch internally. And then write the
1199 +originally requested thing to the file. Generally kernel doesn't
1200 +open/read/write file actively. In aufs, even a single write may cause a
1201 +internal "file copy". This behaviour is very similar to cp(1) command.
1203 +Some people may think it is better to pass such work to user space
1204 +helper, instead of doing in kernel space. Actually I am still thinking
1205 +about it. But currently I have implemented it in kernel space.
1206 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1207 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1208 +++ linux/Documentation/filesystems/aufs/design/02struct.txt 2022-03-21 14:49:05.713299677 +0100
1211 +# Copyright (C) 2005-2021 Junjiro R. Okajima
1213 +# This program is free software; you can redistribute it and/or modify
1214 +# it under the terms of the GNU General Public License as published by
1215 +# the Free Software Foundation; either version 2 of the License, or
1216 +# (at your option) any later version.
1218 +# This program is distributed in the hope that it will be useful,
1219 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1220 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1221 +# GNU General Public License for more details.
1223 +# You should have received a copy of the GNU General Public License
1224 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1226 +Basic Aufs Internal Structure
1228 +Superblock/Inode/Dentry/File Objects
1229 +----------------------------------------------------------------------
1230 +As like an ordinary filesystem, aufs has its own
1231 +superblock/inode/dentry/file objects. All these objects have a
1232 +dynamically allocated array and store the same kind of pointers to the
1233 +lower filesystem, branch.
1234 +For example, when you build a union with one readwrite branch and one
1235 +readonly, mounted /au, /rw and /ro respectively.
1237 +- /ro/fileA exists but /rw/fileA
1239 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1240 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1241 +- [0] = NULL (because /rw/fileA doesn't exist)
1244 +This style of an array is essentially same to the aufs
1245 +superblock/inode/dentry/file objects.
1247 +Because aufs supports manipulating branches, ie. add/delete/change
1248 +branches dynamically, these objects has its own generation. When
1249 +branches are changed, the generation in aufs superblock is
1250 +incremented. And a generation in other object are compared when it is
1251 +accessed. When a generation in other objects are obsoleted, aufs
1252 +refreshes the internal array.
1256 +----------------------------------------------------------------------
1257 +Additionally aufs superblock has some data for policies to select one
1258 +among multiple writable branches, XIB files, pseudo-links and kobject.
1259 +See below in detail.
1260 +About the policies which supports copy-down a directory, see
1261 +wbr_policy.txt too.
1264 +Branch and XINO(External Inode Number Translation Table)
1265 +----------------------------------------------------------------------
1266 +Every branch has its own xino (external inode number translation table)
1267 +file. The xino file is created and unlinked by aufs internally. When two
1268 +members of a union exist on the same filesystem, they share the single
1270 +The struct of a xino file is simple, just a sequence of aufs inode
1271 +numbers which is indexed by the lower inode number.
1272 +In the above sample, assume the inode number of /ro/fileA is i111 and
1273 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1274 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1276 +When the inode numbers are not contiguous, the xino file will be sparse
1277 +which has a hole in it and doesn't consume as much disk space as it
1278 +might appear. If your branch filesystem consumes disk space for such
1279 +holes, then you should specify 'xino=' option at mounting aufs.
1281 +Aufs has a mount option to free the disk blocks for such holes in XINO
1282 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1283 +meet a problem of disk shortage due to XINO files, then you should try
1284 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1285 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1286 +the holes in XINO files.
1288 +Also a writable branch has three kinds of "whiteout bases". All these
1289 +are existed when the branch is joined to aufs, and their names are
1290 +whiteout-ed doubly, so that users will never see their names in aufs
1292 +1. a regular file which will be hardlinked to all whiteouts.
1293 +2. a directory to store a pseudo-link.
1294 +3. a directory to store an "orphan"-ed file temporary.
1297 + When you remove a file on a readonly branch, aufs handles it as a
1298 + logical deletion and creates a whiteout on the upper writable branch
1299 + as a hardlink of this file in order not to consume inode on the
1302 + See below, Pseudo-link.
1304 + When "fileC" exists on the lower readonly branch only and it is
1305 + opened and removed with its parent dir, and then user writes
1306 + something into it, then aufs copies-up fileC to this
1307 + directory. Because there is no other dir to store fileC. After
1308 + creating a file under this dir, the file is unlinked.
1310 +Because aufs supports manipulating branches, ie. add/delete/change
1311 +dynamically, a branch has its own id. When the branch order changes,
1312 +aufs finds the new index by searching the branch id.
1316 +----------------------------------------------------------------------
1317 +Assume "fileA" exists on the lower readonly branch only and it is
1318 +hardlinked to "fileB" on the branch. When you write something to fileA,
1319 +aufs copies-up it to the upper writable branch. Additionally aufs
1320 +creates a hardlink under the Pseudo-link Directory of the writable
1321 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1322 +simple list. If fileB is read after unlinking fileA, aufs returns
1323 +filedata from the pseudo-link instead of the lower readonly
1324 +branch. Because the pseudo-link is based upon the inode, to keep the
1325 +inode number by xino (see above) is essentially necessary.
1327 +All the hardlinks under the Pseudo-link Directory of the writable branch
1328 +should be restored in a proper location later. Aufs provides a utility
1329 +to do this. The userspace helpers executed at remounting and unmounting
1331 +During this utility is running, it puts aufs into the pseudo-link
1332 +maintenance mode. In this mode, only the process which began the
1333 +maintenance mode (and its child processes) is allowed to operate in
1334 +aufs. Some other processes which are not related to the pseudo-link will
1335 +be allowed to run too, but the rest have to return an error or wait
1336 +until the maintenance mode ends. If a process already acquires an inode
1337 +mutex (in VFS), it has to return an error.
1340 +XIB(external inode number bitmap)
1341 +----------------------------------------------------------------------
1342 +Addition to the xino file per a branch, aufs has an external inode number
1343 +bitmap in a superblock object. It is also an internal file such like a
1345 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1347 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1349 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1350 +reduce the number of consumed disk blocks for these files.
1353 +Virtual or Vertical Dir, and Readdir in Userspace
1354 +----------------------------------------------------------------------
1355 +In order to support multiple layers (branches), aufs readdir operation
1356 +constructs a virtual dir block on memory. For readdir, aufs calls
1357 +vfs_readdir() internally for each dir on branches, merges their entries
1358 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1359 +object. So the file object has its entry list until it is closed. The
1360 +entry list will be updated when the file position is zero and becomes
1361 +obsoleted. This decision is made in aufs automatically.
1363 +The dynamically allocated memory block for the name of entries has a
1364 +unit of 512 bytes (by default) and stores the names contiguously (no
1365 +padding). Another block for each entry is handled by kmem_cache too.
1366 +During building dir blocks, aufs creates hash list and judging whether
1367 +the entry is whiteouted by its upper branch or already listed.
1368 +The merged result is cached in the corresponding inode object and
1369 +maintained by a customizable life-time option.
1371 +Some people may call it can be a security hole or invite DoS attack
1372 +since the opened and once readdir-ed dir (file object) holds its entry
1373 +list and becomes a pressure for system memory. But I'd say it is similar
1374 +to files under /proc or /sys. The virtual files in them also holds a
1375 +memory page (generally) while they are opened. When an idea to reduce
1376 +memory for them is introduced, it will be applied to aufs too.
1377 +For those who really hate this situation, I've developed readdir(3)
1378 +library which operates this merging in userspace. You just need to set
1379 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1380 +kernel space for readdir(3).
1384 +----------------------------------------------------------------------
1385 +Aufs sometimes requires privilege access to a branch. For instance,
1386 +in copy-up/down operation. When a user process is going to make changes
1387 +to a file which exists in the lower readonly branch only, and the mode
1388 +of one of ancestor directories may not be writable by a user
1389 +process. Here aufs copy-up the file with its ancestors and they may
1390 +require privilege to set its owner/group/mode/etc.
1391 +This is a typical case of a application character of aufs (see
1394 +Aufs uses workqueue synchronously for this case. It creates its own
1395 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1396 +passes the request to call mkdir or write (for example), and wait for
1397 +its completion. This approach solves a problem of a signal handler
1399 +If aufs didn't adopt the workqueue and changed the privilege of the
1400 +process, then the process may receive the unexpected SIGXFSZ or other
1403 +Also aufs uses the system global workqueue ("events" kernel thread) too
1404 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1405 +whiteout base and etc. This is unrelated to a privilege.
1406 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1407 +superblock at the beginning, at the same time waits for the completion
1408 +of all queued asynchronous tasks.
1412 +----------------------------------------------------------------------
1413 +The whiteout in aufs is very similar to Unionfs's. That is represented
1414 +by its filename. UnionMount takes an approach of a file mode, but I am
1415 +afraid several utilities (find(1) or something) will have to support it.
1417 +Basically the whiteout represents "logical deletion" which stops aufs to
1418 +lookup further, but also it represents "dir is opaque" which also stop
1421 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1422 +In order to make several functions in a single systemcall to be
1423 +revertible, aufs adopts an approach to rename a directory to a temporary
1424 +unique whiteouted name.
1425 +For example, in rename(2) dir where the target dir already existed, aufs
1426 +renames the target dir to a temporary unique whiteouted name before the
1427 +actual rename on a branch, and then handles other actions (make it opaque,
1428 +update the attributes, etc). If an error happens in these actions, aufs
1429 +simply renames the whiteouted name back and returns an error. If all are
1430 +succeeded, aufs registers a function to remove the whiteouted unique
1431 +temporary name completely and asynchronously to the system global
1436 +----------------------------------------------------------------------
1437 +It is a well-known feature or concept.
1438 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1439 +internally and makes change to the new file on the upper writable branch.
1440 +When the trigger systemcall does not update the timestamps of the parent
1441 +dir, aufs reverts it after copy-up.
1444 +Move-down (aufs3.9 and later)
1445 +----------------------------------------------------------------------
1446 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1447 +the lower readonly branch to the upper writable branch when a user
1448 +changes something about the file.
1449 +"Move-down" is an opposite action of copy-up. Basically this action is
1450 +ran manually instead of automatically and internally.
1451 +For desgin and implementation, aufs has to consider these issues.
1452 +- whiteout for the file may exist on the lower branch.
1453 +- ancestor directories may not exist on the lower branch.
1454 +- diropq for the ancestor directories may exist on the upper branch.
1455 +- free space on the lower branch will reduce.
1456 +- another access to the file may happen during moving-down, including
1457 + UDBA (see "Revalidate Dentry and UDBA").
1458 +- the file should not be hard-linked nor pseudo-linked. they should be
1459 + handled by auplink utility later.
1461 +Sometimes users want to move-down a file from the upper writable branch
1462 +to the lower readonly or writable branch. For instance,
1463 +- the free space of the upper writable branch is going to run out.
1464 +- create a new intermediate branch between the upper and lower branch.
1467 +For this purpose, use "aumvdown" command in aufs-util.git.
1468 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1469 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt 1970-01-01 01:00:00.000000000 +0100
1470 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt 2022-03-21 14:49:05.713299677 +0100
1473 +# Copyright (C) 2015-2021 Junjiro R. Okajima
1475 +# This program is free software; you can redistribute it and/or modify
1476 +# it under the terms of the GNU General Public License as published by
1477 +# the Free Software Foundation; either version 2 of the License, or
1478 +# (at your option) any later version.
1480 +# This program is distributed in the hope that it will be useful,
1481 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1482 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1483 +# GNU General Public License for more details.
1485 +# You should have received a copy of the GNU General Public License
1486 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1488 +Support for a branch who has its ->atomic_open()
1489 +----------------------------------------------------------------------
1490 +The filesystems who implement its ->atomic_open() are not majority. For
1491 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1492 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1493 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1494 +sure whether all filesystems who have ->atomic_open() behave like this,
1495 +but NFSv4 surely returns the error.
1497 +In order to support ->atomic_open() for aufs, there are a few
1500 +A. Introduce aufs_atomic_open()
1501 + - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1503 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1504 + an aufs user Pip Cet's approach
1505 + - calls aufs_create(), VFS finish_open() and notify_change().
1506 + - pass fake-mode to finish_open(), and then correct the mode by
1508 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1509 + - no aufs_atomic_open().
1510 + - aufs_lookup() registers the TID to an aufs internal object.
1511 + - aufs_create() does nothing when the matching TID is registered, but
1512 + registers the mode.
1513 + - aufs_open() calls branch fs's ->atomic_open() when the matching
1514 + TID is registered.
1515 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1517 + - no aufs_atomic_open().
1518 + - aufs_create() registers the TID to an internal object. this info
1519 + represents "this process created this file just now."
1520 + - when aufs gets EACCES from branch fs's ->open(), then confirm the
1521 + registered TID and re-try open() with superuser's credential.
1523 +Pros and cons for each approach.
1526 + - straightforward but highly depends upon VFS internal.
1527 + - the atomic behavaiour is kept.
1528 + - some of parameters such as nameidata are hard to reproduce for
1532 + - easy to implement.
1533 + - the atomic behavaiour is lost.
1535 + - the atomic behavaiour is kept.
1536 + - dirty and tricky.
1537 + - VFS checks whether the file is created correctly after calling
1538 + ->create(), which means this approach doesn't work.
1540 + - easy to implement.
1541 + - the atomic behavaiour is lost.
1542 + - to open a file with superuser's credential and give it to a user
1543 + process is a bad idea, since the file object keeps the credential
1544 + in it. It may affect LSM or something. This approach doesn't work
1547 +The approach A is ideal, but it hard to implement. So here is a
1548 +variation of A, which is to be implemented.
1550 +A-1. Introduce aufs_atomic_open()
1551 + - calls branch fs ->atomic_open() if exists. otherwise calls
1552 + vfs_create() and finish_open().
1553 + - the demerit is that the several checks after branch fs
1554 + ->atomic_open() are lost. in the ordinary case, the checks are
1555 + done by VFS:do_last(), lookup_open() and atomic_open(). some can
1556 + be implemented in aufs, but not all I am afraid.
1557 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1558 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1559 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2022-03-21 14:49:05.713299677 +0100
1562 +# Copyright (C) 2005-2021 Junjiro R. Okajima
1564 +# This program is free software; you can redistribute it and/or modify
1565 +# it under the terms of the GNU General Public License as published by
1566 +# the Free Software Foundation; either version 2 of the License, or
1567 +# (at your option) any later version.
1569 +# This program is distributed in the hope that it will be useful,
1570 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1571 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1572 +# GNU General Public License for more details.
1574 +# You should have received a copy of the GNU General Public License
1575 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1578 +----------------------------------------------------------------------
1579 +Since aufs has a character of sub-VFS (see Introduction), it operates
1580 +lookup for branches as VFS does. It may be a heavy work. But almost all
1581 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1582 +directly connected to its parent. Digging down the directory hierarchy
1583 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1586 +When a branch is a remote filesystem, aufs basically relies upon its
1587 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1589 +For d_revalidate, aufs implements three levels of revalidate tests. See
1590 +"Revalidate Dentry and UDBA" in detail.
1593 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1594 +----------------------------------------------------------------------
1595 +Let's try case study.
1596 +- aufs has two branches, upper readwrite and lower readonly.
1598 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1599 +- user invoked "chmod a+rx /au/dirA"
1600 +- the internal copy-up is activated and "/rw/dirA" is created and its
1601 + permission bits are set to world readable.
1602 +- then "/au/dirA" becomes world readable?
1604 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1605 +or it may be a natively readonly filesystem. If aufs respects the lower
1606 +branch, it should not respond readdir request from other users. But user
1607 +allowed it by chmod. Should really aufs rejects showing the entries
1610 +To be honest, I don't have a good solution for this case. So aufs
1611 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1613 +When dirperm1 is specified, aufs checks only the highest one for the
1614 +directory permission, and shows the entries. Otherwise, as usual, checks
1615 +every dir existing on all branches and rejects the request.
1617 +As a side effect, dirperm1 option improves the performance of aufs
1618 +because the number of permission check is reduced when the number of
1622 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1623 +----------------------------------------------------------------------
1624 +Generally VFS helpers re-validate a dentry as a part of lookup.
1625 +0. digging down the directory hierarchy.
1626 +1. lock the parent dir by its i_mutex.
1627 +2. lookup the final (child) entry.
1629 +4. call the actual operation (create, unlink, etc.)
1630 +5. unlock the parent dir
1632 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1633 +called. Actually aufs implements it and checks the dentry on a branch is
1635 +But it is not enough. Because aufs has to release the lock for the
1636 +parent dir on a branch at the end of ->lookup() (step 2) and
1637 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1639 +If the file on a branch is changed directly, eg. bypassing aufs, after
1640 +aufs released the lock, then the subsequent operation may cause
1641 +something unpleasant result.
1643 +This situation is a result of VFS architecture, ->lookup() and
1644 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1645 +design from VFS's point of view. It is just not suitable for sub-VFS
1648 +Aufs supports such case by three level of revalidation which is
1649 +selectable by user.
1650 +1. Simple Revalidate
1651 + Addition to the native flow in VFS's, confirm the child-parent
1652 + relationship on the branch just after locking the parent dir on the
1653 + branch in the "actual operation" (step 4). When this validation
1654 + fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1655 + checks the validation of the dentry on branches.
1656 +2. Monitor Changes Internally by Inotify/Fsnotify
1657 + Addition to above, in the "actual operation" (step 4) aufs re-lookup
1658 + the dentry on the branch, and returns EBUSY if it finds different
1660 + Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1661 + during it is in cache. When the event is notified, aufs registers a
1662 + function to kernel 'events' thread by schedule_work(). And the
1663 + function sets some special status to the cached aufs dentry and inode
1664 + private data. If they are not cached, then aufs has nothing to
1665 + do. When the same file is accessed through aufs (step 0-3) later,
1666 + aufs will detect the status and refresh all necessary data.
1667 + In this mode, aufs has to ignore the event which is fired by aufs
1669 +3. No Extra Validation
1670 + This is the simplest test and doesn't add any additional revalidation
1671 + test, and skip the revalidation in step 4. It is useful and improves
1672 + aufs performance when system surely hide the aufs branches from user,
1673 + by over-mounting something (or another method).
1674 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1675 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1676 +++ linux/Documentation/filesystems/aufs/design/04branch.txt 2022-03-21 14:49:05.713299677 +0100
1679 +# Copyright (C) 2005-2021 Junjiro R. Okajima
1681 +# This program is free software; you can redistribute it and/or modify
1682 +# it under the terms of the GNU General Public License as published by
1683 +# the Free Software Foundation; either version 2 of the License, or
1684 +# (at your option) any later version.
1686 +# This program is distributed in the hope that it will be useful,
1687 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1688 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1689 +# GNU General Public License for more details.
1691 +# You should have received a copy of the GNU General Public License
1692 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1694 +Branch Manipulation
1696 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1697 +and changing its permission/attribute, there are a lot of works to do.
1701 +----------------------------------------------------------------------
1702 +o Confirm the adding dir exists outside of aufs, including loopback
1703 + mount, and its various attributes.
1704 +o Initialize the xino file and whiteout bases if necessary.
1707 +o Check the owner/group/mode of the directory
1708 + When the owner/group/mode of the adding directory differs from the
1709 + existing branch, aufs issues a warning because it may impose a
1711 + For example, when a upper writable branch has a world writable empty
1712 + top directory, a malicious user can create any files on the writable
1713 + branch directly, like copy-up and modify manually. If something like
1714 + /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1715 + writable branch, and the writable branch is world-writable, then a
1716 + malicious guy may create /etc/passwd on the writable branch directly
1717 + and the infected file will be valid in aufs.
1718 + I am afraid it can be a security issue, but aufs can do nothing except
1719 + producing a warning.
1723 +----------------------------------------------------------------------
1724 +o Confirm the deleting branch is not busy
1725 + To be general, there is one merit to adopt "remount" interface to
1726 + manipulate branches. It is to discard caches. At deleting a branch,
1727 + aufs checks the still cached (and connected) dentries and inodes. If
1728 + there are any, then they are all in-use. An inode without its
1729 + corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1731 + For the cached one, aufs checks whether the same named entry exists on
1733 + If the cached one is a directory, because aufs provides a merged view
1734 + to users, as long as one dir is left on any branch aufs can show the
1735 + dir to users. In this case, the branch can be removed from aufs.
1736 + Otherwise aufs rejects deleting the branch.
1738 + If any file on the deleting branch is opened by aufs, then aufs
1742 +Modify the Permission of a Branch
1743 +----------------------------------------------------------------------
1744 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1747 +o rw --> ro: Confirm the modifying branch is not busy
1748 + Aufs rejects the request if any of these conditions are true.
1749 + - a file on the branch is mmap-ed.
1750 + - a regular file on the branch is opened for write and there is no
1751 + same named entry on the upper branch.
1752 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1753 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
1754 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2022-03-21 14:49:05.713299677 +0100
1757 +# Copyright (C) 2005-2021 Junjiro R. Okajima
1759 +# This program is free software; you can redistribute it and/or modify
1760 +# it under the terms of the GNU General Public License as published by
1761 +# the Free Software Foundation; either version 2 of the License, or
1762 +# (at your option) any later version.
1764 +# This program is distributed in the hope that it will be useful,
1765 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1766 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1767 +# GNU General Public License for more details.
1769 +# You should have received a copy of the GNU General Public License
1770 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1772 +Policies to Select One among Multiple Writable Branches
1773 +----------------------------------------------------------------------
1774 +When the number of writable branch is more than one, aufs has to decide
1775 +the target branch for file creation or copy-up. By default, the highest
1776 +writable branch which has the parent (or ancestor) dir of the target
1777 +file is chosen (top-down-parent policy).
1778 +By user's request, aufs implements some other policies to select the
1779 +writable branch, for file creation several policies, round-robin,
1780 +most-free-space, and other policies. For copy-up, top-down-parent,
1781 +bottom-up-parent, bottom-up and others.
1783 +As expected, the round-robin policy selects the branch in circular. When
1784 +you have two writable branches and creates 10 new files, 5 files will be
1785 +created for each branch. mkdir(2) systemcall is an exception. When you
1786 +create 10 new directories, all will be created on the same branch.
1787 +And the most-free-space policy selects the one which has most free
1788 +space among the writable branches. The amount of free space will be
1789 +checked by aufs internally, and users can specify its time interval.
1791 +The policies for copy-up is more simple,
1792 +top-down-parent is equivalent to the same named on in create policy,
1793 +bottom-up-parent selects the writable branch where the parent dir
1794 +exists and the nearest upper one from the copyup-source,
1795 +bottom-up selects the nearest upper writable branch from the
1796 +copyup-source, regardless the existence of the parent dir.
1798 +There are some rules or exceptions to apply these policies.
1799 +- If there is a readonly branch above the policy-selected branch and
1800 + the parent dir is marked as opaque (a variation of whiteout), or the
1801 + target (creating) file is whiteout-ed on the upper readonly branch,
1802 + then the result of the policy is ignored and the target file will be
1803 + created on the nearest upper writable branch than the readonly branch.
1804 +- If there is a writable branch above the policy-selected branch and
1805 + the parent dir is marked as opaque or the target file is whiteouted
1806 + on the branch, then the result of the policy is ignored and the target
1807 + file will be created on the highest one among the upper writable
1808 + branches who has diropq or whiteout. In case of whiteout, aufs removes
1810 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1811 + They try selecting the branch where the source exists as possible
1812 + since copyup a large file will take long time. If it can't be,
1813 + ie. the branch where the source exists is readonly, then they will
1814 + follow the copyup policy.
1815 +- There is an exception for rename(2) when the target exists.
1816 + If the rename target exists, aufs compares the index of the branches
1817 + where the source and the target exists and selects the higher
1818 + one. If the selected branch is readonly, then aufs follows the
1820 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot linux/Documentation/filesystems/aufs/design/06dirren.dot
1821 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot 1970-01-01 01:00:00.000000000 +0100
1822 +++ linux/Documentation/filesystems/aufs/design/06dirren.dot 2022-03-21 14:49:05.713299677 +0100
1825 +// to view this graph, run dot(1) command in GRAPHVIZ.
1829 +whinfo [label="detailed info file\n(lower_brid_root-hinum, h_inum, namelen, old name)"];
1833 +aufs_rename -> whinfo [label="store/remove"];
1836 +inode_list [label="h_inum list in branch\ncache"];
1839 +whinode [label="h_inum list file"];
1842 +brmgmt [label="br_add/del/mod/umount"];
1844 +brmgmt -> inode_list [label="create/remove"];
1845 +brmgmt -> whinode [label="load/store"];
1847 +inode_list -> whinode [style=dashed,dir=both];
1849 +aufs_rename -> inode_list [label="add/del"];
1851 +aufs_lookup -> inode_list [label="search"];
1853 +aufs_lookup -> whinfo [label="load/remove"];
1855 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt linux/Documentation/filesystems/aufs/design/06dirren.txt
1856 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt 1970-01-01 01:00:00.000000000 +0100
1857 +++ linux/Documentation/filesystems/aufs/design/06dirren.txt 2022-10-03 07:51:14.926209260 +0200
1860 +# Copyright (C) 2017-2021 Junjiro R. Okajima
1862 +# This program is free software; you can redistribute it and/or modify
1863 +# it under the terms of the GNU General Public License as published by
1864 +# the Free Software Foundation; either version 2 of the License, or
1865 +# (at your option) any later version.
1867 +# This program is distributed in the hope that it will be useful,
1868 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1869 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1870 +# GNU General Public License for more details.
1872 +# You should have received a copy of the GNU General Public License
1873 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1875 +Special handling for renaming a directory (DIRREN)
1876 +----------------------------------------------------------------------
1877 +First, let's assume we have a simple usecase.
1881 +- /ro/dirA and /ro/dirA/file exist too
1882 +- there is no dirB on both branches
1883 +- a user issues rename("dirA", "dirB")
1885 +Now, what should aufs behave against this rename(2)?
1886 +There are a few possible cases.
1889 + since dirA exists on a readonly branch which cannot be renamed.
1891 + it is possible to copy-up dirA (only the dir itself), but the child
1892 + entries ("file" in this case) should not be. it must be a bad
1893 + approach to copy-up recursively.
1894 +C. returns a success.
1895 + even the branch /ro is readonly, aufs tries renaming it. Obviously it
1896 + is a violation of aufs' policy.
1897 +D. construct an extra information which indicates that /ro/dirA should
1898 + be handled as the name of dirB.
1899 + overlayfs has a similar feature called REDIRECT.
1901 +Until now, aufs implements the case B only which returns EXDEV, and
1902 +expects the userspace application behaves like mv(1) which tries
1903 +issueing rename(2) recursively.
1905 +A new aufs feature called DIRREN is introduced which implements the case
1906 +D. There are several "extra information" added.
1908 +1. detailed info per renamed directory
1909 + path: /rw/dirB/$AUFS_WH_DR_INFO_PFX.<lower branch-id>
1910 +2. the inode-number list of directories on a branch
1911 + path: /rw/dirB/$AUFS_WH_DR_BRHINO
1913 +The filename of "detailed info per directory" represents the lower
1914 +branch, and its format is
1915 +- a type of the branch id
1917 + + uuid (not implemented yet)
1920 +- the inode-number of the branch root dir
1922 +And it contains these info in a single regular file.
1924 +- branch's inode-number of the logically renamed dir
1925 +- the name of the before-renamed dir
1927 +The "detailed info per directory" file is created in aufs rename(2), and
1928 +loaded in any lookup.
1929 +The info is considered in lookup for the matching case only. Here
1930 +"matching" means that the root of branch (in the info filename) is same
1931 +to the current looking-up branch. After looking-up the before-renamed
1932 +name, the inode-number is compared. And the matched dentry is used.
1934 +The "inode-number list of directories" is a regular file which contains
1935 +simply the inode-numbers on the branch. The file is created or updated
1936 +in removing the branch, and loaded in adding the branch. Its lifetime is
1937 +equal to the branch.
1938 +The list is referred in lookup, and when the current target inode is
1939 +found in the list, the aufs tries loading the "detailed info per
1940 +directory" and get the changed and valid name of the dir.
1942 +Theoretically these "extra informaiton" may be able to be put into XATTR
1943 +in the dir inode. But aufs doesn't choose this way because
1944 +1. XATTR may not be supported by the branch (or its configuration)
1945 +2. XATTR may have its size limit.
1946 +3. XATTR may be less easy to convert than a regular file, when the
1947 + format of the info is changed in the future.
1948 +At the same time, I agree that the regular file approach is much slower
1949 +than XATTR approach. So, in the future, aufs may take the XATTR or other
1952 +This DIRREN feature is enabled by aufs configuration, and is activated
1953 +by a new mount option.
1955 +For the more complicated case, there is a work with UDBA option, which
1956 +is to dected the direct access to the branches (by-passing aufs) and to
1957 +maintain the cashes in aufs. Since a single cached aufs dentry may
1958 +contains two names, before- and after-rename, the name comparision in
1959 +UDBA handler may not work correctly. In this case, the behaviour will be
1960 +equivalen to udba=reval case.
1961 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1962 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt 1970-01-01 01:00:00.000000000 +0100
1963 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt 2022-03-21 14:49:05.716633010 +0100
1966 +# Copyright (C) 2011-2021 Junjiro R. Okajima
1968 +# This program is free software; you can redistribute it and/or modify
1969 +# it under the terms of the GNU General Public License as published by
1970 +# the Free Software Foundation; either version 2 of the License, or
1971 +# (at your option) any later version.
1973 +# This program is distributed in the hope that it will be useful,
1974 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1975 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1976 +# GNU General Public License for more details.
1978 +# You should have received a copy of the GNU General Public License
1979 +# along with this program; if not, write to the Free Software
1980 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1983 +File-based Hierarchical Storage Management (FHSM)
1984 +----------------------------------------------------------------------
1985 +Hierarchical Storage Management (or HSM) is a well-known feature in the
1986 +storage world. Aufs provides this feature as file-based with multiple
1987 +writable branches, based upon the principle of "Colder, the Lower".
1988 +Here the word "colder" means that the less used files, and "lower" means
1989 +that the position in the order of the stacked branches vertically.
1990 +These multiple writable branches are prioritized, ie. the topmost one
1991 +should be the fastest drive and be used heavily.
1993 +o Characters in aufs FHSM story
1994 +- aufs itself and a new branch attribute.
1995 +- a new ioctl interface to move-down and to establish a connection with
1996 + the daemon ("move-down" is a converse of "copy-up").
1997 +- userspace tool and daemon.
1999 +The userspace daemon establishes a connection with aufs and waits for
2000 +the notification. The notified information is very similar to struct
2001 +statfs containing the number of consumed blocks and inodes.
2002 +When the consumed blocks/inodes of a branch exceeds the user-specified
2003 +upper watermark, the daemon activates its move-down process until the
2004 +consumed blocks/inodes reaches the user-specified lower watermark.
2006 +The actual move-down is done by aufs based upon the request from
2007 +user-space since we need to maintain the inode number and the internal
2008 +pointer arrays in aufs.
2010 +Currently aufs FHSM handles the regular files only. Additionally they
2011 +must not be hard-linked nor pseudo-linked.
2014 +o Cowork of aufs and the user-space daemon
2015 + During the userspace daemon established the connection, aufs sends a
2016 + small notification to it whenever aufs writes something into the
2017 + writable branch. But it may cost high since aufs issues statfs(2)
2018 + internally. So user can specify a new option to cache the
2019 + info. Actually the notification is controlled by these factors.
2020 + + the specified cache time.
2021 + + classified as "force" by aufs internally.
2022 + Until the specified time expires, aufs doesn't send the info
2023 + except the forced cases. When aufs decide forcing, the info is always
2024 + notified to userspace.
2025 + For example, the number of free inodes is generally large enough and
2026 + the shortage of it happens rarely. So aufs doesn't force the
2027 + notification when creating a new file, directory and others. This is
2028 + the typical case which aufs doesn't force.
2029 + When aufs writes the actual filedata and the files consumes any of new
2030 + blocks, the aufs forces notifying.
2033 +o Interfaces in aufs
2034 +- New branch attribute.
2036 + Specifies that the branch is managed by FHSM feature. In other word,
2037 + participant in the FHSM.
2038 + When nofhsm is set to the branch, it will not be the source/target
2039 + branch of the move-down operation. This attribute is set
2040 + independently from coo and moo attributes, and if you want full
2041 + FHSM, you should specify them as well.
2042 +- New mount option.
2044 + Specifies a second to suppress many less important info to be
2047 + + AUFS_CTL_FHSM_FD
2048 + create a new file descriptor which userspace can read the notification
2049 + (a subset of struct statfs) from aufs.
2050 +- Module parameter 'brs'
2051 + It has to be set to 1. Otherwise the new mount option 'fhsm' will not
2053 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
2054 + When there are two or more branches with fhsm attributes,
2055 + /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
2056 + terminates it. As a result of remounting and branch-manipulation, the
2057 + number of branches with fhsm attribute can be one. In this case,
2058 + /sbin/mount.aufs will terminate the user-space daemon.
2061 +Finally the operation is done as these steps in kernel-space.
2063 + + no one else is using the file.
2064 + + the file is not hard-linked.
2065 + + the file is not pseudo-linked.
2066 + + the file is a regular file.
2067 + + the parent dir is not opaqued.
2068 +- find the target writable branch.
2069 +- make sure the file is not whiteout-ed by the upper (than the target)
2071 +- make the parent dir on the target branch.
2072 +- mutex lock the inode on the branch.
2073 +- unlink the whiteout on the target branch (if exists).
2074 +- lookup and create the whiteout-ed temporary name on the target branch.
2075 +- copy the file as the whiteout-ed temporary name on the target branch.
2076 +- rename the whiteout-ed temporary name to the original name.
2077 +- unlink the file on the source branch.
2078 +- maintain the internal pointer array and the external inode number
2080 +- maintain the timestamps and other attributes of the parent dir and the
2083 +And of course, in every step, an error may happen. So the operation
2084 +should restore the original file state after an error happens.
2085 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
2086 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
2087 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2022-03-21 14:49:05.716633010 +0100
2090 +# Copyright (C) 2005-2021 Junjiro R. Okajima
2092 +# This program is free software; you can redistribute it and/or modify
2093 +# it under the terms of the GNU General Public License as published by
2094 +# the Free Software Foundation; either version 2 of the License, or
2095 +# (at your option) any later version.
2097 +# This program is distributed in the hope that it will be useful,
2098 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2099 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2100 +# GNU General Public License for more details.
2102 +# You should have received a copy of the GNU General Public License
2103 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2105 +mmap(2) -- File Memory Mapping
2106 +----------------------------------------------------------------------
2107 +In aufs, the file-mapped pages are handled by a branch fs directly, no
2108 +interaction with aufs. It means aufs_mmap() calls the branch fs's
2110 +This approach is simple and good, but there is one problem.
2111 +Under /proc, several entries show the mmapped files by its path (with
2112 +device and inode number), and the printed path will be the path on the
2113 +branch fs's instead of virtual aufs's.
2114 +This is not a problem in most cases, but some utilities lsof(1) (and its
2115 +user) may expect the path on aufs.
2117 +To address this issue, aufs adds a new member called vm_prfile in struct
2118 +vm_area_struct (and struct vm_region). The original vm_file points to
2119 +the file on the branch fs in order to handle everything correctly as
2120 +usual. The new vm_prfile points to a virtual file in aufs, and the
2121 +show-functions in procfs refers to vm_prfile if it is set.
2122 +Also we need to maintain several other places where touching vm_file
2124 +- fork()/clone() copies vma and the reference count of vm_file is
2126 +- merging vma maintains the ref count too.
2128 +This is not a good approach. It just fakes the printed path. But it
2129 +leaves all behaviour around f_mapping unchanged. This is surely an
2131 +Actually aufs had adopted another complicated approach which calls
2132 +generic_file_mmap() and handles struct vm_operations_struct. In this
2133 +approach, aufs met a hard problem and I could not solve it without
2134 +switching the approach.
2136 +There may be one more another approach which is
2137 +- bind-mount the branch-root onto the aufs-root internally
2138 +- grab the new vfsmount (ie. struct mount)
2139 +- lazy-umount the branch-root internally
2140 +- in open(2) the aufs-file, open the branch-file with the hidden
2141 + vfsmount (instead of the original branch's vfsmount)
2142 +- ideally this "bind-mount and lazy-umount" should be done atomically,
2143 + but it may be possible from userspace by the mount helper.
2145 +Adding the internal hidden vfsmount and using it in opening a file, the
2146 +file path under /proc will be printed correctly. This approach looks
2147 +smarter, but is not possible I am afraid.
2148 +- aufs-root may be bind-mount later. when it happens, another hidden
2149 + vfsmount will be required.
2150 +- it is hard to get the chance to bind-mount and lazy-umount
2151 + + in kernel-space, FS can have vfsmount in open(2) via
2152 + file->f_path, and aufs can know its vfsmount. But several locks are
2153 + already acquired, and if aufs tries to bind-mount and lazy-umount
2154 + here, then it may cause a deadlock.
2155 + + in user-space, bind-mount doesn't invoke the mount helper.
2156 +- since /proc shows dev and ino, aufs has to give vma these info. it
2157 + means a new member vm_prinode will be necessary. this is essentially
2158 + equivalent to vm_prfile described above.
2160 +I have to give up this "looks-smater" approach.
2161 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2162 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt 1970-01-01 01:00:00.000000000 +0100
2163 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt 2022-03-21 14:49:05.716633010 +0100
2166 +# Copyright (C) 2014-2021 Junjiro R. Okajima
2168 +# This program is free software; you can redistribute it and/or modify
2169 +# it under the terms of the GNU General Public License as published by
2170 +# the Free Software Foundation; either version 2 of the License, or
2171 +# (at your option) any later version.
2173 +# This program is distributed in the hope that it will be useful,
2174 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2175 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2176 +# GNU General Public License for more details.
2178 +# You should have received a copy of the GNU General Public License
2179 +# along with this program; if not, write to the Free Software
2180 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2183 +Listing XATTR/EA and getting the value
2184 +----------------------------------------------------------------------
2185 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2186 +shows the values from the topmost existing file. This behaviour is good
2187 +for the non-dir entries since the bahaviour exactly matches the shown
2188 +information. But for the directories, aufs considers all the same named
2189 +entries on the lower branches. Which means, if one of the lower entry
2190 +rejects readdir call, then aufs returns an error even if the topmost
2191 +entry allows it. This behaviour is necessary to respect the branch fs's
2192 +security, but can make users confused since the user-visible standard
2193 +attributes don't match the behaviour.
2194 +To address this issue, aufs has a mount option called dirperm1 which
2195 +checks the permission for the topmost entry only, and ignores the lower
2196 +entry's permission.
2198 +A similar issue can happen around XATTR.
2199 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2200 +always set. Otherwise these very unpleasant situation would happen.
2201 +- listxattr(2) may return the duplicated entries.
2202 +- users may not be able to remove or reset the XATTR forever,
2205 +XATTR/EA support in the internal (copy,move)-(up,down)
2206 +----------------------------------------------------------------------
2207 +Generally the extended attributes of inode are categorized as these.
2208 +- "security" for LSM and capability.
2209 +- "system" for posix ACL, 'acl' mount option is required for the branch
2211 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2212 +- "user" for userspace, 'user_xattr' mount option is required for the
2213 + branch fs generally.
2215 +Moreover there are some other categories. Aufs handles these rather
2216 +unpopular categories as the ordinary ones, ie. there is no special
2217 +condition nor exception.
2219 +In copy-up, the support for XATTR on the dst branch may differ from the
2220 +src branch. In this case, the copy-up operation will get an error and
2221 +the original user operation which triggered the copy-up will fail. It
2222 +can happen that even all copy-up will fail.
2223 +When both of src and dst branches support XATTR and if an error occurs
2224 +during copying XATTR, then the copy-up should fail obviously. That is a
2225 +good reason and aufs should return an error to userspace. But when only
2226 +the src branch support that XATTR, aufs should not return an error.
2227 +For example, the src branch supports ACL but the dst branch doesn't
2228 +because the dst branch may natively un-support it or temporary
2229 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2230 +may NOT return an error even if the XATTR is not supported. It is
2231 +totally up to the branch fs.
2233 +Anyway when the aufs internal copy-up gets an error from the dst branch
2234 +fs, then aufs tries removing the just copied entry and returns the error
2235 +to the userspace. The worst case of this situation will be all copy-up
2238 +For the copy-up operation, there two basic approaches.
2239 +- copy the specified XATTR only (by category above), and return the
2240 + error unconditionally if it happens.
2241 +- copy all XATTR, and ignore the error on the specified category only.
2243 +In order to support XATTR and to implement the correct behaviour, aufs
2244 +chooses the latter approach and introduces some new branch attributes,
2245 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2246 +They correspond to the XATTR namespaces (see above). Additionally, to be
2247 +convenient, "icex" is also provided which means all "icex*" attributes
2248 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2250 +The meaning of these attributes is to ignore the error from setting
2251 +XATTR on that branch.
2252 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2253 +error from the dst branch according to the specified attributes.
2255 +Some XATTR may have its default value. The default value may come from
2256 +the parent dir or the environment. If the default value is set at the
2257 +file creating-time, it will be overwritten by copy-up.
2258 +Some contradiction may happen I am afraid.
2259 +Do we need another attribute to stop copying XATTR? I am unsure. For
2260 +now, aufs implements the branch attributes to ignore the error.
2261 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2262 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2263 +++ linux/Documentation/filesystems/aufs/design/07export.txt 2022-03-21 14:49:05.716633010 +0100
2266 +# Copyright (C) 2005-2021 Junjiro R. Okajima
2268 +# This program is free software; you can redistribute it and/or modify
2269 +# it under the terms of the GNU General Public License as published by
2270 +# the Free Software Foundation; either version 2 of the License, or
2271 +# (at your option) any later version.
2273 +# This program is distributed in the hope that it will be useful,
2274 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2275 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2276 +# GNU General Public License for more details.
2278 +# You should have received a copy of the GNU General Public License
2279 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2281 +Export Aufs via NFS
2282 +----------------------------------------------------------------------
2283 +Here is an approach.
2284 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2286 +- iget_locked(): initialize aufs inode generation for a new inode, and
2287 + store it in xigen file.
2288 +- destroy_inode(): increment aufs inode generation and store it in xigen
2289 + file. it is necessary even if it is not unlinked, because any data of
2290 + inode may be changed by UDBA.
2291 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2292 + build file handle by
2293 + + branch id (4 bytes)
2294 + + superblock generation (4 bytes)
2295 + + inode number (4 or 8 bytes)
2296 + + parent dir inode number (4 or 8 bytes)
2297 + + inode generation (4 bytes))
2298 + + return value of exportfs_encode_fh() for the parent on a branch (4
2300 + + file handle for a branch (by exportfs_encode_fh())
2302 + + find the index of a branch from its id in handle, and check it is
2303 + still exist in aufs.
2304 + + 1st level: get the inode number from handle and search it in cache.
2305 + + 2nd level: if not found in cache, get the parent inode number from
2306 + the handle and search it in cache. and then open the found parent
2307 + dir, find the matching inode number by vfs_readdir() and get its
2308 + name, and call lookup_one_len() for the target dentry.
2309 + + 3rd level: if the parent dir is not cached, call
2310 + exportfs_decode_fh() for a branch and get the parent on a branch,
2311 + build a pathname of it, convert it a pathname in aufs, call
2312 + path_lookup(). now aufs gets a parent dir dentry, then handle it as
2314 + + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2315 + for every branch, but not itself. to get this, (currently) aufs
2316 + searches in current->nsproxy->mnt_ns list. it may not be a good
2317 + idea, but I didn't get other approach.
2318 + + test the generation of the gotten inode.
2319 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2320 + convert it into ESTALE for NFSD.
2321 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2322 + lookup_one_len(), vfs_getattr(), encode_fh() and others.
2323 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2324 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
2325 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2022-03-21 14:49:05.716633010 +0100
2328 +# Copyright (C) 2005-2021 Junjiro R. Okajima
2330 +# This program is free software; you can redistribute it and/or modify
2331 +# it under the terms of the GNU General Public License as published by
2332 +# the Free Software Foundation; either version 2 of the License, or
2333 +# (at your option) any later version.
2335 +# This program is distributed in the hope that it will be useful,
2336 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2337 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2338 +# GNU General Public License for more details.
2340 +# You should have received a copy of the GNU General Public License
2341 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2343 +Show Whiteout Mode (shwh)
2344 +----------------------------------------------------------------------
2345 +Generally aufs hides the name of whiteouts. But in some cases, to show
2346 +them is very useful for users. For instance, creating a new middle layer
2347 +(branch) by merging existing layers.
2349 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2350 +When you have three branches,
2351 +- Bottom: 'system', squashfs (underlying base system), read-only
2352 +- Middle: 'mods', squashfs, read-only
2353 +- Top: 'overlay', ram (tmpfs), read-write
2355 +The top layer is loaded at boot time and saved at shutdown, to preserve
2356 +the changes made to the system during the session.
2357 +When larger changes have been made, or smaller changes have accumulated,
2358 +the size of the saved top layer data grows. At this point, it would be
2359 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2360 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2361 +restoring save and load speed.
2363 +This merging is simplified by the use of another aufs mount, of just the
2364 +two overlay branches using the 'shwh' option.
2365 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2366 + aufs /livesys/merge_union
2368 +A merged view of these two branches is then available at
2369 +/livesys/merge_union, and the new feature is that the whiteouts are
2371 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2372 +writing to all branches. Also the default mode for all branches is 'ro'.
2373 +It is now possible to save the combined contents of the two overlay
2374 +branches to a new squashfs, e.g.:
2375 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2377 +This new squashfs archive can be stored on the boot device and the
2378 +initramfs will use it to replace the old one at the next boot.
2379 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2380 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
2381 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2022-03-21 14:49:05.716633010 +0100
2384 +# Copyright (C) 2010-2021 Junjiro R. Okajima
2386 +# This program is free software; you can redistribute it and/or modify
2387 +# it under the terms of the GNU General Public License as published by
2388 +# the Free Software Foundation; either version 2 of the License, or
2389 +# (at your option) any later version.
2391 +# This program is distributed in the hope that it will be useful,
2392 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2393 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2394 +# GNU General Public License for more details.
2396 +# You should have received a copy of the GNU General Public License
2397 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2399 +Dynamically customizable FS operations
2400 +----------------------------------------------------------------------
2401 +Generally FS operations (struct inode_operations, struct
2402 +address_space_operations, struct file_operations, etc.) are defined as
2403 +"static const", but it never means that FS have only one set of
2404 +operation. Some FS have multiple sets of them. For instance, ext2 has
2405 +three sets, one for XIP, for NOBH, and for normal.
2406 +Since aufs overrides and redirects these operations, sometimes aufs has
2407 +to change its behaviour according to the branch FS type. More importantly
2408 +VFS acts differently if a function (member in the struct) is set or
2409 +not. It means aufs should have several sets of operations and select one
2410 +among them according to the branch FS definition.
2412 +In order to solve this problem and not to affect the behaviour of VFS,
2413 +aufs defines these operations dynamically. For instance, aufs defines
2414 +dummy direct_IO function for struct address_space_operations, but it may
2415 +not be set to the address_space_operations actually. When the branch FS
2416 +doesn't have it, aufs doesn't set it to its address_space_operations
2417 +while the function definition itself is still alive. So the behaviour
2418 +itself will not change, and it will return an error when direct_IO is
2421 +The lifetime of these dynamically generated operation object is
2422 +maintained by aufs branch object. When the branch is removed from aufs,
2423 +the reference counter of the object is decremented. When it reaches
2424 +zero, the dynamically generated operation object will be freed.
2426 +This approach is designed to support AIO (io_submit), Direct I/O and
2428 +Currently this approach is applied to address_space_operations for
2429 +regular files only.
2430 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2431 --- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
2432 +++ linux/Documentation/filesystems/aufs/README 2022-10-03 07:51:14.922875926 +0200
2435 +Aufs5 -- advanced multi layered unification filesystem version 5.x
2441 +----------------------------------------
2442 +In the early days, aufs was entirely re-designed and re-implemented
2443 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2444 +improvements and implementations, it became totally different from
2445 +Unionfs while keeping the basic features.
2446 +Later, Unionfs Version 2.x series began taking some of the same
2447 +approaches to aufs1's.
2448 +Unionfs was being developed by Professor Erez Zadok at Stony Brook
2449 +University and his team.
2451 +Aufs5 supports linux-v5.0 and later, If you want older kernel version
2453 +- for linux-v4.x series, try aufs4-linux.git or aufs4-standalone.git
2454 +- for linux-v3.x series, try aufs3-linux.git or aufs3-standalone.git
2455 +- for linux-v2.6.16 and later, try aufs2-2.6.git, aufs2-standalone.git
2456 + or aufs1 from CVS on SourceForge.
2458 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2459 + According to Christoph Hellwig, linux rejects all union-type
2460 + filesystems but UnionMount.
2461 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2463 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2464 + UnionMount, and he pointed out an issue around a directory mutex
2465 + lock and aufs addressed it. But it is still unsure whether aufs will
2466 + be merged (or any other union solution).
2467 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2471 +----------------------------------------
2472 +- unite several directories into a single virtual filesystem. The member
2473 + directory is called as a branch.
2474 +- you can specify the permission flags to the branch, which are 'readonly',
2475 + 'readwrite' and 'whiteout-able.'
2476 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2477 + readonly branch are modifiable logically.
2478 +- dynamic branch manipulation, add, del.
2481 +Also there are many enhancements in aufs, such as:
2482 +- test only the highest one for the directory permission (dirperm1)
2483 +- copyup on open (coo=)
2484 +- 'move' policy for copy-up between two writable branches, after
2485 + checking free space.
2487 +- readdir(3) in userspace.
2488 +- keep inode number by external inode number table
2489 +- keep the timestamps of file/dir in internal copyup operation
2490 +- seekable directory, supporting NFS readdir.
2491 +- whiteout is hardlinked in order to reduce the consumption of inodes
2493 +- do not copyup, nor create a whiteout when it is unnecessary
2494 +- revert a single systemcall when an error occurs in aufs
2495 +- remount interface instead of ioctl
2496 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2497 +- loopback mounted filesystem as a branch
2498 +- kernel thread for removing the dir who has a plenty of whiteouts
2499 +- support copyup sparse file (a file which has a 'hole' in it)
2500 +- default permission flags for branches
2501 +- selectable permission flags for ro branch, whether whiteout can
2504 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2505 +- support multiple writable branches, some policies to select one
2506 + among multiple writable branches.
2507 +- a new semantics for link(2) and rename(2) to support multiple
2508 + writable branches.
2509 +- no glibc changes are required.
2510 +- pseudo hardlink (hardlink over branches)
2511 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2512 + including NFS or remote filesystem branch.
2513 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2516 +Currently these features are dropped temporary from aufs5.
2517 +See design/08plan.txt in detail.
2518 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2520 +- statistics of aufs thread (/sys/fs/aufs/stat)
2522 +Features or just an idea in the future (see also design/*.txt),
2523 +- reorder the branch index without del/re-add.
2524 +- permanent xino files for NFSD
2525 +- an option for refreshing the opened files after add/del branches
2526 +- light version, without branch manipulation. (unnecessary?)
2527 +- copyup in userspace
2528 +- inotify in userspace
2533 +----------------------------------------
2534 +There are three GIT trees for aufs5, aufs5-linux.git,
2535 +aufs5-standalone.git, and aufs-util.git. Note that there is no "5" in
2537 +While the aufs-util is always necessary, you need either of aufs5-linux
2538 +or aufs5-standalone.
2540 +The aufs5-linux tree includes the whole linux mainline GIT tree,
2541 +git://git.kernel.org/.../torvalds/linux.git.
2542 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2543 +build aufs5 as an external kernel module.
2544 +Several extra patches are not included in this tree. Only
2545 +aufs5-standalone tree contains them. They are described in the later
2546 +section "Configuration and Compilation."
2548 +On the other hand, the aufs5-standalone tree has only aufs source files
2549 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2550 +But you need to apply all aufs patches manually.
2552 +You will find GIT branches whose name is in form of "aufs5.x" where "x"
2553 +represents the linux kernel version, "linux-5.x". For instance,
2554 +"aufs5.0" is for linux-5.0. For latest "linux-5.x-rcN", use
2555 +"aufs5.x-rcN" branch.
2558 +$ git clone --reference /your/linux/git/tree \
2559 + git://github.com/sfjro/aufs5-linux.git aufs5-linux.git
2560 +- if you don't have linux GIT tree, then remove "--reference ..."
2561 +$ cd aufs5-linux.git
2562 +$ git checkout origin/aufs5.0
2564 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2565 +leave the patch-work to GIT.
2566 +$ cd /your/linux/git/tree
2567 +$ git remote add aufs5 git://github.com/sfjro/aufs5-linux.git
2569 +$ git checkout -b my5.0 v5.0
2570 +$ (add your local change...)
2571 +$ git pull aufs5 aufs5.0
2572 +- now you have v5.0 + your_changes + aufs5.0 in you my5.0 branch.
2573 +- you may need to solve some conflicts between your_changes and
2574 + aufs5.0. in this case, git-rerere is recommended so that you can
2575 + solve the similar conflicts automatically when you upgrade to 5.1 or
2576 + later in the future.
2578 +o aufs5-standalone tree
2579 +$ git clone git://github.com/sfjro/aufs5-standalone.git aufs5-standalone.git
2580 +$ cd aufs5-standalone.git
2581 +$ git checkout origin/aufs5.0
2584 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2585 +- note that the public aufs-util.git is on SourceForge instead of
2588 +$ git checkout origin/aufs5.0
2590 +Note: The 5.x-rcN branch is to be used with `rc' kernel versions ONLY.
2591 +The minor version number, 'x' in '5.x', of aufs may not always
2592 +follow the minor version number of the kernel.
2593 +Because changes in the kernel that cause the use of a new
2594 +minor version number do not always require changes to aufs-util.
2596 +Since aufs-util has its own minor version number, you may not be
2597 +able to find a GIT branch in aufs-util for your kernel's
2598 +exact minor version number.
2599 +In this case, you should git-checkout the branch for the
2600 +nearest lower number.
2602 +For (an unreleased) example:
2603 +If you are using "linux-5.10" and the "aufs5.10" branch
2604 +does not exist in aufs-util repository, then "aufs5.9", "aufs5.8"
2605 +or something numerically smaller is the branch for your kernel.
2607 +Also you can view all branches by
2611 +3. Configuration and Compilation
2612 +----------------------------------------
2613 +Make sure you have git-checkout'ed the correct branch.
2615 +For aufs5-linux tree,
2616 +- enable CONFIG_AUFS_FS.
2617 +- set other aufs configurations if necessary.
2618 +- for aufs5.13 and later
2619 + Because aufs is not only an ordinary filesystem (callee of VFS), but
2620 + also a caller of VFS functions for branch filesystems, subclassing of
2621 + the internal locks for LOCKDEP is necessary. LOCKDEP is a debugging
2622 + feature of linux kernel. If you enable CONFIG_LOCKDEP, then you will
2623 + need to customize some LOCKDEP numbers. Here are what I use on my
2625 + CONFIG_LOCKDEP_BITS=21
2626 + CONFIG_LOCKDEP_CHAINS_BITS=21
2627 + CONFIG_LOCKDEP_STACK_TRACE_BITS=24
2629 +For aufs5-standalone tree,
2630 +There are several ways to build.
2633 +- apply ./aufs5-kbuild.patch to your kernel source files.
2634 +- apply ./aufs5-base.patch too.
2635 +- apply ./aufs5-mmap.patch too.
2636 +- apply ./aufs5-standalone.patch too, if you have a plan to set
2637 + CONFIG_AUFS_FS=m. otherwise you don't need ./aufs5-standalone.patch.
2638 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2639 + kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2640 +- enable CONFIG_AUFS_FS, you can select either
2642 +- and build your kernel as usual.
2643 +- install the built kernel.
2644 +- install the header files too by "make headers_install" to the
2645 + directory where you specify. By default, it is $PWD/usr.
2646 + "make help" shows a brief note for headers_install.
2647 +- and reboot your system.
2650 +- module only (CONFIG_AUFS_FS=m).
2651 +- apply ./aufs5-base.patch to your kernel source files.
2652 +- apply ./aufs5-mmap.patch too.
2653 +- apply ./aufs5-standalone.patch too.
2654 +- build your kernel, don't forget "make headers_install", and reboot.
2655 +- edit ./config.mk and set other aufs configurations if necessary.
2656 + Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2657 + every aufs configurations.
2658 +- build the module by simple "make".
2659 +- you can specify ${KDIR} make variable which points to your kernel
2661 +- install the files
2662 + + run "make install" to install the aufs module, or copy the built
2663 + $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2664 + + run "make install_headers" (instead of headers_install) to install
2665 + the modified aufs header file (you can specify DESTDIR which is
2666 + available in aufs standalone version's Makefile only), or copy
2667 + $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2668 + you like manually. By default, the target directory is $PWD/usr.
2669 +- no need to apply aufs5-kbuild.patch, nor copying source files to your
2670 + kernel source tree.
2672 +Note: The header file aufs_type.h is necessary to build aufs-util
2673 + as well as "make headers_install" in the kernel source tree.
2674 + headers_install is subject to be forgotten, but it is essentially
2675 + necessary, not only for building aufs-util.
2676 + You may not meet problems without headers_install in some older
2680 +- read README in aufs-util, build and install it
2681 +- note that your distribution may contain an obsoleted version of
2682 + aufs_type.h in /usr/include/linux or something. When you build aufs
2683 + utilities, make sure that your compiler refers the correct aufs header
2684 + file which is built by "make headers_install."
2685 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2686 + then run "make install_ulib" too. And refer to the aufs manual in
2689 +There several other patches in aufs5-standalone.git. They are all
2690 +optional. When you meet some problems, they will help you.
2691 +- aufs5-loopback.patch
2692 + Supports a nested loopback mount in a branch-fs. This patch is
2693 + unnecessary until aufs produces a message like "you may want to try
2694 + another patch for loopback file".
2696 + Modifies a system global kernel internal function get_next_ino() in
2697 + order to stop assigning 0 for an inode-number. Not directly related to
2698 + aufs, but recommended generally.
2700 + Keeps the tmpfs inode number as the lowest value. Effective to reduce
2701 + the size of aufs XINO files for tmpfs branch. Also it prevents the
2702 + duplication of inode number, which is important for backup tools and
2703 + other utilities. When you find aufs XINO files for tmpfs branch
2704 + growing too much, try this patch.
2708 +----------------------------------------
2709 +At first, make sure aufs-util are installed, and please read the aufs
2710 +manual, aufs.5 in aufs-util.git tree.
2714 +$ mkdir /tmp/rw /tmp/aufs
2715 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2717 +Here is another example. The result is equivalent.
2718 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2720 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2721 +# mount -o remount,append:${HOME} /tmp/aufs
2723 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2724 +you modify a file under /tmp/aufs, the one on your home directory is
2725 +not affected, instead the same named file will be newly created under
2726 +/tmp/rw. And all of your modification to a file will be applied to
2727 +the one under /tmp/rw. This is called the file based Copy on Write
2729 +Aufs mount options are described in aufs.5.
2730 +If you run chroot or something and make your aufs as a root directory,
2731 +then you need to customize the shutdown script. See the aufs manual in
2734 +Additionally, there are some sample usages of aufs which are a
2735 +diskless system with network booting, and LiveCD over NFS.
2736 +See sample dir in CVS tree on SourceForge.
2740 +----------------------------------------
2741 +When you have any problems or strange behaviour in aufs, please let me
2743 +- /proc/mounts (instead of the output of mount(8))
2744 +- /sys/module/aufs/*
2745 +- /sys/fs/aufs/* (if you have them)
2746 +- /debug/aufs/* (if you have them)
2747 +- linux kernel version
2748 + if your kernel is not plain, for example modified by distributor,
2749 + the url where i can download its source is necessary too.
2750 +- aufs version which was printed at loading the module or booting the
2751 + system, instead of the date you downloaded.
2752 +- configuration (define/undefine CONFIG_AUFS_xxx)
2753 +- kernel configuration or /proc/config.gz (if you have it)
2754 +- LSM (linux security module, if you are using)
2755 +- behaviour which you think to be incorrect
2756 +- actual operation, reproducible one is better
2757 +- mailto: aufs-users at lists.sourceforge.net
2759 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2760 +and Feature Requests) on SourceForge. Please join and write to
2764 +6. Acknowledgements
2765 +----------------------------------------
2766 +Thanks to everyone who have tried and are using aufs, whoever
2767 +have reported a bug or any feedback.
2769 +Especially donators:
2770 +Tomas Matejicek(slax.org) made a donation (much more than once).
2771 + Since Apr 2010, Tomas M (the author of Slax and Linux Live
2772 + scripts) is making "doubling" donations.
2773 + Unfortunately I cannot list all of the donators, but I really
2775 + It ends Aug 2010, but the ordinary donation URL is still available.
2776 + <http://sourceforge.net/donate/index.php?group_id=167503>
2777 +Dai Itasaka made a donation (2007/8).
2778 +Chuck Smith made a donation (2008/4, 10 and 12).
2779 +Henk Schoneveld made a donation (2008/9).
2780 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2781 +Francois Dupoux made a donation (2008/11).
2782 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2783 + aufs2 GIT tree (2009/2).
2784 +William Grant made a donation (2009/3).
2785 +Patrick Lane made a donation (2009/4).
2786 +The Mail Archive (mail-archive.com) made donations (2009/5).
2787 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2788 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2789 +Pavel Pronskiy made a donation (2011/2).
2790 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2791 + Networks (Ed Wildgoose) made a donation for hardware (2011/3).
2792 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2794 +Sam Liddicott made a donation (2011/9).
2795 +Era Scarecrow made a donation (2013/4).
2796 +Bor Ratajc made a donation (2013/4).
2797 +Alessandro Gorreta made a donation (2013/4).
2798 +POIRETTE Marc made a donation (2013/4).
2799 +Alessandro Gorreta made a donation (2013/4).
2800 +lauri kasvandik made a donation (2013/5).
2801 +"pemasu from Finland" made a donation (2013/7).
2802 +The Parted Magic Project made a donation (2013/9 and 11).
2803 +Pavel Barta made a donation (2013/10).
2804 +Nikolay Pertsev made a donation (2014/5).
2805 +James B made a donation (2014/7, 2015/7, and 2021/12).
2806 +Stefano Di Biase made a donation (2014/8).
2807 +Daniel Epellei made a donation (2015/1).
2808 +OmegaPhil made a donation (2016/1, 2018/4).
2809 +Tomasz Szewczyk made a donation (2016/4).
2810 +James Burry made a donation (2016/12).
2811 +Carsten Rose made a donation (2018/9).
2812 +Porteus Kiosk made a donation (2018/10).
2813 +Enya Quetzalli Gomez Rodriguez made a donation (2022/5).
2815 +Thank you very much.
2816 +Donations are always, including future donations, very important and
2817 +helpful for me to keep on developing aufs.
2821 +----------------------------------------
2822 +If you are an experienced user, no explanation is needed. Aufs is
2823 +just a linux filesystem.
2828 +# Local variables: ;
2831 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2832 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
2833 +++ linux/fs/aufs/aufs.h 2022-03-21 14:49:05.716633010 +0100
2835 +/* SPDX-License-Identifier: GPL-2.0 */
2837 + * Copyright (C) 2005-2021 Junjiro R. Okajima
2839 + * This program, aufs is free software; you can redistribute it and/or modify
2840 + * it under the terms of the GNU General Public License as published by
2841 + * the Free Software Foundation; either version 2 of the License, or
2842 + * (at your option) any later version.
2844 + * This program is distributed in the hope that it will be useful,
2845 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2846 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2847 + * GNU General Public License for more details.
2849 + * You should have received a copy of the GNU General Public License
2850 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2854 + * all header files
2862 +#define AuStub(type, name, body, ...) \
2863 + static inline type name(__VA_ARGS__) { body; }
2865 +#define AuStubVoid(name, ...) \
2866 + AuStub(void, name, , __VA_ARGS__)
2867 +#define AuStubInt0(name, ...) \
2868 + AuStub(int, name, return 0, __VA_ARGS__)
2872 +#include "branch.h"
2875 +#include "dbgaufs.h"
2876 +#include "dentry.h"
2878 +#include "dirren.h"
2881 +#include "fstype.h"
2886 +#include "module.h"
2890 +#include "sysaufs.h"
2895 +#endif /* __KERNEL__ */
2896 +#endif /* __AUFS_H__ */
2897 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2898 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
2899 +++ linux/fs/aufs/branch.c 2022-10-03 07:51:14.926209260 +0200
2901 +// SPDX-License-Identifier: GPL-2.0
2903 + * Copyright (C) 2005-2021 Junjiro R. Okajima
2905 + * This program, aufs is free software; you can redistribute it and/or modify
2906 + * it under the terms of the GNU General Public License as published by
2907 + * the Free Software Foundation; either version 2 of the License, or
2908 + * (at your option) any later version.
2910 + * This program is distributed in the hope that it will be useful,
2911 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2912 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2913 + * GNU General Public License for more details.
2915 + * You should have received a copy of the GNU General Public License
2916 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2920 + * branch management
2923 +#include <linux/compat.h>
2924 +#include <linux/statfs.h>
2928 + * free a single branch
2930 +static void au_br_do_free(struct au_branch *br)
2933 + struct au_wbr *wbr;
2934 + struct au_dykey **key;
2936 + au_hnotify_fin_br(br);
2937 + /* always, regardless the mount option */
2938 + au_dr_hino_free(&br->br_dirren);
2941 + AuLCntZero(au_lcnt_read(&br->br_nfiles, /*do_rev*/0));
2942 + au_lcnt_fin(&br->br_nfiles, /*do_sync*/0);
2943 + AuLCntZero(au_lcnt_read(&br->br_count, /*do_rev*/0));
2944 + au_lcnt_fin(&br->br_count, /*do_sync*/0);
2948 + for (i = 0; i < AuBrWh_Last; i++)
2949 + dput(wbr->wbr_wh[i]);
2950 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2951 + AuRwDestroy(&wbr->wbr_wh_rwsem);
2954 + if (br->br_fhsm) {
2955 + au_br_fhsm_fin(br->br_fhsm);
2956 + au_kfree_try_rcu(br->br_fhsm);
2959 + key = br->br_dykey;
2960 + for (i = 0; i < AuBrDynOp; i++, key++)
2966 + /* recursive lock, s_umount of branch's */
2967 + /* synchronize_rcu(); */ /* why? */
2969 + path_put(&br->br_path);
2971 + au_kfree_rcu(wbr);
2972 + au_lcnt_wait_for_fin(&br->br_nfiles);
2973 + au_lcnt_wait_for_fin(&br->br_count);
2974 + /* I don't know why, but percpu_refcount requires this */
2975 + /* synchronize_rcu(); */
2980 + * frees all branches
2982 +void au_br_free(struct au_sbinfo *sbinfo)
2984 + aufs_bindex_t bmax;
2985 + struct au_branch **br;
2987 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2989 + bmax = sbinfo->si_bbot + 1;
2990 + br = sbinfo->si_branch;
2992 + au_br_do_free(*br++);
2996 + * find the index of a branch which is specified by @br_id.
2998 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
3000 + aufs_bindex_t bindex, bbot;
3002 + bbot = au_sbbot(sb);
3003 + for (bindex = 0; bindex <= bbot; bindex++)
3004 + if (au_sbr_id(sb, bindex) == br_id)
3009 +/* ---------------------------------------------------------------------- */
3015 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
3016 + struct dentry *h_root)
3018 + if (unlikely(h_adding == h_root
3019 + || au_test_loopback_overlap(sb, h_adding)))
3021 + if (h_adding->d_sb != h_root->d_sb)
3023 + return au_test_subdir(h_adding, h_root)
3024 + || au_test_subdir(h_root, h_adding);
3028 + * returns a newly allocated branch. @new_nbranch is a number of branches
3029 + * after adding a branch.
3031 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
3034 + struct au_branch *add_branch;
3035 + struct dentry *root;
3036 + struct inode *inode;
3040 + add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
3041 + if (unlikely(!add_branch))
3043 + add_branch->br_xino = au_xino_alloc(/*nfile*/1);
3044 + if (unlikely(!add_branch->br_xino))
3046 + err = au_hnotify_init_br(add_branch, perm);
3047 + if (unlikely(err))
3050 + if (au_br_writable(perm)) {
3051 + /* may be freed separately at changing the branch permission */
3052 + add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
3054 + if (unlikely(!add_branch->br_wbr))
3058 + if (au_br_fhsm(perm)) {
3059 + err = au_fhsm_br_alloc(add_branch);
3060 + if (unlikely(err))
3064 + root = sb->s_root;
3065 + err = au_sbr_realloc(au_sbi(sb), new_nbranch, /*may_shrink*/0);
3067 + err = au_di_realloc(au_di(root), new_nbranch, /*may_shrink*/0);
3069 + inode = d_inode(root);
3070 + err = au_hinode_realloc(au_ii(inode), new_nbranch,
3074 + return add_branch; /* success */
3077 + au_kfree_rcu(add_branch->br_wbr);
3079 + au_hnotify_fin_br(add_branch);
3081 + au_xino_put(add_branch);
3083 + au_kfree_rcu(add_branch);
3085 + return ERR_PTR(err);
3089 + * test if the branch permission is legal or not.
3091 +static int test_br(struct inode *inode, int brperm, char *path)
3095 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
3100 + pr_err("write permission for readonly mount or inode, %s\n", path);
3108 + * 0: success, the caller will add it
3109 + * plus: success, it is already unified, the caller should ignore it
3112 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
3115 + aufs_bindex_t bbot, bindex;
3116 + struct dentry *root, *h_dentry;
3117 + struct inode *inode, *h_inode;
3119 + root = sb->s_root;
3120 + bbot = au_sbbot(sb);
3121 + if (unlikely(bbot >= 0
3122 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
3126 + pr_err("%s duplicated\n", add->pathname);
3131 + err = -ENOSPC; /* -E2BIG; */
3132 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
3133 + || AUFS_BRANCH_MAX - 1 <= bbot)) {
3134 + pr_err("number of branches exceeded %s\n", add->pathname);
3139 + if (unlikely(add->bindex < 0 || bbot + 1 < add->bindex)) {
3140 + pr_err("bad index %d\n", add->bindex);
3144 + inode = d_inode(add->path.dentry);
3146 + if (unlikely(!inode->i_nlink)) {
3147 + pr_err("no existence %s\n", add->pathname);
3152 + if (unlikely(inode->i_sb == sb)) {
3153 + pr_err("%s must be outside\n", add->pathname);
3157 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
3158 + pr_err("unsupported filesystem, %s (%s)\n",
3159 + add->pathname, au_sbtype(inode->i_sb));
3163 + if (unlikely(inode->i_sb->s_stack_depth)) {
3164 + pr_err("already stacked, %s (%s)\n",
3165 + add->pathname, au_sbtype(inode->i_sb));
3169 + err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3170 + if (unlikely(err))
3174 + return 0; /* success */
3177 + for (bindex = 0; bindex <= bbot; bindex++)
3178 + if (unlikely(test_overlap(sb, add->path.dentry,
3179 + au_h_dptr(root, bindex)))) {
3180 + pr_err("%s is overlapped\n", add->pathname);
3185 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3186 + h_dentry = au_h_dptr(root, 0);
3187 + h_inode = d_inode(h_dentry);
3188 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3189 + || !uid_eq(h_inode->i_uid, inode->i_uid)
3190 + || !gid_eq(h_inode->i_gid, inode->i_gid))
3191 + pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3193 + i_uid_read(inode), i_gid_read(inode),
3194 + (inode->i_mode & S_IALLUGO),
3195 + i_uid_read(h_inode), i_gid_read(h_inode),
3196 + (h_inode->i_mode & S_IALLUGO));
3204 + * initialize or clean the whiteouts for an adding branch
3206 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3209 + int err, old_perm;
3210 + aufs_bindex_t bindex;
3211 + struct inode *h_inode;
3212 + struct au_wbr *wbr;
3213 + struct au_hinode *hdir;
3214 + struct dentry *h_dentry;
3216 + err = vfsub_mnt_want_write(au_br_mnt(br));
3217 + if (unlikely(err))
3221 + old_perm = br->br_perm;
3222 + br->br_perm = new_perm;
3225 + bindex = au_br_index(sb, br->br_id);
3226 + if (0 <= bindex) {
3227 + hdir = au_hi(d_inode(sb->s_root), bindex);
3228 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
3230 + h_dentry = au_br_dentry(br);
3231 + h_inode = d_inode(h_dentry);
3232 + inode_lock_nested(h_inode, AuLsc_I_PARENT);
3235 + err = au_wh_init(br, sb);
3237 + wbr_wh_write_lock(wbr);
3238 + err = au_wh_init(br, sb);
3239 + wbr_wh_write_unlock(wbr);
3242 + au_hn_inode_unlock(hdir);
3244 + inode_unlock(h_inode);
3245 + vfsub_mnt_drop_write(au_br_mnt(br));
3246 + br->br_perm = old_perm;
3248 + if (!err && wbr && !au_br_writable(new_perm)) {
3249 + au_kfree_rcu(wbr);
3250 + br->br_wbr = NULL;
3257 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3261 + struct kstatfs kst;
3262 + struct au_wbr *wbr;
3265 + au_rw_init(&wbr->wbr_wh_rwsem);
3266 + atomic_set(&wbr->wbr_wh_running, 0);
3269 + * a limit for rmdir/rename a dir
3270 + * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3272 + err = vfs_statfs(&br->br_path, &kst);
3273 + if (unlikely(err))
3276 + if (kst.f_namelen >= NAME_MAX)
3277 + err = au_br_init_wh(sb, br, perm);
3279 + pr_err("%pd(%s), unsupported namelen %ld\n",
3281 + au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3287 +/* initialize a new branch */
3288 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3289 + struct au_opt_add *add)
3292 + struct au_branch *brbase;
3294 + struct inode *h_inode;
3297 + br->br_perm = add->perm;
3298 + br->br_path = add->path; /* set first, path_get() later */
3299 + spin_lock_init(&br->br_dykey_lock);
3300 + au_lcnt_init(&br->br_nfiles, /*release*/NULL);
3301 + au_lcnt_init(&br->br_count, /*release*/NULL);
3302 + br->br_id = au_new_br_id(sb);
3303 + AuDebugOn(br->br_id < 0);
3305 + /* always, regardless the given option */
3306 + err = au_dr_br_init(sb, br, &add->path);
3307 + if (unlikely(err))
3310 + if (au_br_writable(add->perm)) {
3311 + err = au_wbr_init(br, sb, add->perm);
3312 + if (unlikely(err))
3316 + if (au_opt_test(au_mntflags(sb), XINO)) {
3317 + brbase = au_sbr(sb, 0);
3318 + xf = au_xino_file(brbase->br_xino, /*idx*/-1);
3320 + h_inode = d_inode(add->path.dentry);
3321 + err = au_xino_init_br(sb, br, h_inode->i_ino, &xf->f_path);
3322 + if (unlikely(err)) {
3323 + AuDebugOn(au_xino_file(br->br_xino, /*idx*/-1));
3328 + sysaufs_br_init(br);
3329 + path_get(&br->br_path);
3330 + goto out; /* success */
3333 + memset(&br->br_path, 0, sizeof(br->br_path));
3338 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3339 + struct au_branch *br, aufs_bindex_t bbot,
3340 + aufs_bindex_t amount)
3342 + struct au_branch **brp;
3344 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3346 + brp = sbinfo->si_branch + bindex;
3347 + memmove(brp + 1, brp, sizeof(*brp) * amount);
3349 + sbinfo->si_bbot++;
3350 + if (unlikely(bbot < 0))
3351 + sbinfo->si_bbot = 0;
3354 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3355 + aufs_bindex_t bbot, aufs_bindex_t amount)
3357 + struct au_hdentry *hdp;
3359 + AuRwMustWriteLock(&dinfo->di_rwsem);
3361 + hdp = au_hdentry(dinfo, bindex);
3362 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3363 + au_h_dentry_init(hdp);
3365 + if (unlikely(bbot < 0))
3366 + dinfo->di_btop = 0;
3369 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3370 + aufs_bindex_t bbot, aufs_bindex_t amount)
3372 + struct au_hinode *hip;
3374 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3376 + hip = au_hinode(iinfo, bindex);
3377 + memmove(hip + 1, hip, sizeof(*hip) * amount);
3378 + au_hinode_init(hip);
3380 + if (unlikely(bbot < 0))
3381 + iinfo->ii_btop = 0;
3384 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3385 + aufs_bindex_t bindex)
3387 + struct dentry *root, *h_dentry;
3388 + struct inode *root_inode, *h_inode;
3389 + aufs_bindex_t bbot, amount;
3391 + root = sb->s_root;
3392 + root_inode = d_inode(root);
3393 + bbot = au_sbbot(sb);
3394 + amount = bbot + 1 - bindex;
3395 + h_dentry = au_br_dentry(br);
3396 + au_sbilist_lock();
3397 + au_br_do_add_brp(au_sbi(sb), bindex, br, bbot, amount);
3398 + au_br_do_add_hdp(au_di(root), bindex, bbot, amount);
3399 + au_br_do_add_hip(au_ii(root_inode), bindex, bbot, amount);
3400 + au_set_h_dptr(root, bindex, dget(h_dentry));
3401 + h_inode = d_inode(h_dentry);
3402 + au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3403 + au_sbilist_unlock();
3406 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3409 + aufs_bindex_t bbot, add_bindex;
3410 + struct dentry *root, *h_dentry;
3411 + struct inode *root_inode;
3412 + struct au_branch *add_branch;
3414 + root = sb->s_root;
3415 + root_inode = d_inode(root);
3416 + IMustLock(root_inode);
3417 + IiMustWriteLock(root_inode);
3418 + err = test_add(sb, add, remount);
3419 + if (unlikely(err < 0))
3423 + goto out; /* success */
3426 + bbot = au_sbbot(sb);
3427 + add_branch = au_br_alloc(sb, bbot + 2, add->perm);
3428 + err = PTR_ERR(add_branch);
3429 + if (IS_ERR(add_branch))
3432 + err = au_br_init(add_branch, sb, add);
3433 + if (unlikely(err)) {
3434 + au_br_do_free(add_branch);
3438 + add_bindex = add->bindex;
3439 + sysaufs_brs_del(sb, add_bindex); /* remove successors */
3440 + au_br_do_add(sb, add_branch, add_bindex);
3441 + sysaufs_brs_add(sb, add_bindex); /* append successors */
3442 + dbgaufs_brs_add(sb, add_bindex, /*topdown*/0); /* rename successors */
3444 + h_dentry = add->path.dentry;
3445 + if (!add_bindex) {
3446 + au_cpup_attr_all(root_inode, /*force*/1);
3447 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3449 + au_add_nlink(root_inode, d_inode(h_dentry));
3455 +/* ---------------------------------------------------------------------- */
3457 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3458 + unsigned long long max __maybe_unused,
3461 + unsigned long long n;
3462 + struct file **p, *f;
3463 + struct hlist_bl_head *files;
3464 + struct hlist_bl_node *pos;
3465 + struct au_finfo *finfo;
3469 + files = &au_sbi(sb)->si_files;
3470 + hlist_bl_lock(files);
3471 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
3472 + f = finfo->fi_file;
3474 + && !special_file(file_inode(f)->i_mode)) {
3478 + AuDebugOn(n > max);
3481 + hlist_bl_unlock(files);
3486 +static struct file **au_farray_alloc(struct super_block *sb,
3487 + unsigned long long *max)
3489 + struct au_sbinfo *sbi;
3492 + *max = au_lcnt_read(&sbi->si_nfiles, /*do_rev*/1);
3493 + return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3496 +static void au_farray_free(struct file **a, unsigned long long max)
3498 + unsigned long long ull;
3500 + for (ull = 0; ull < max; ull++)
3506 +/* ---------------------------------------------------------------------- */
3512 +/* to show the line number, do not make it inlined function */
3513 +#define AuVerbose(do_info, fmt, ...) do { \
3515 + pr_info(fmt, ##__VA_ARGS__); \
3518 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t btop,
3519 + aufs_bindex_t bbot)
3521 + return (inode && !S_ISDIR(inode->i_mode)) || btop == bbot;
3524 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t btop,
3525 + aufs_bindex_t bbot)
3527 + return au_test_ibusy(d_inode(dentry), btop, bbot);
3531 + * test if the branch is deletable or not.
3533 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3534 + unsigned int sigen, const unsigned int verbose)
3536 + int err, i, j, ndentry;
3537 + aufs_bindex_t btop, bbot;
3538 + struct au_dcsub_pages dpages;
3539 + struct au_dpage *dpage;
3542 + err = au_dpages_init(&dpages, GFP_NOFS);
3543 + if (unlikely(err))
3545 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
3546 + if (unlikely(err))
3549 + for (i = 0; !err && i < dpages.ndpage; i++) {
3550 + dpage = dpages.dpages + i;
3551 + ndentry = dpage->ndentry;
3552 + for (j = 0; !err && j < ndentry; j++) {
3553 + d = dpage->dentries[j];
3554 + AuDebugOn(au_dcount(d) <= 0);
3555 + if (!au_digen_test(d, sigen)) {
3556 + di_read_lock_child(d, AuLock_IR);
3557 + if (unlikely(au_dbrange_test(d))) {
3558 + di_read_unlock(d, AuLock_IR);
3562 + di_write_lock_child(d);
3563 + if (unlikely(au_dbrange_test(d))) {
3564 + di_write_unlock(d);
3567 + err = au_reval_dpath(d, sigen);
3569 + di_downgrade_lock(d, AuLock_IR);
3571 + di_write_unlock(d);
3576 + /* AuDbgDentry(d); */
3577 + btop = au_dbtop(d);
3578 + bbot = au_dbbot(d);
3579 + if (btop <= bindex
3581 + && au_h_dptr(d, bindex)
3582 + && au_test_dbusy(d, btop, bbot)) {
3584 + AuVerbose(verbose, "busy %pd\n", d);
3587 + di_read_unlock(d, AuLock_IR);
3592 + au_dpages_free(&dpages);
3597 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3598 + unsigned int sigen, const unsigned int verbose)
3601 + unsigned long long max, ull;
3602 + struct inode *i, **array;
3603 + aufs_bindex_t btop, bbot;
3605 + array = au_iarray_alloc(sb, &max);
3606 + err = PTR_ERR(array);
3607 + if (IS_ERR(array))
3611 + AuDbg("b%d\n", bindex);
3612 + for (ull = 0; !err && ull < max; ull++) {
3616 + if (i->i_ino == AUFS_ROOT_INO)
3619 + /* AuDbgInode(i); */
3620 + if (au_iigen(i, NULL) == sigen)
3621 + ii_read_lock_child(i);
3623 + ii_write_lock_child(i);
3624 + err = au_refresh_hinode_self(i);
3627 + ii_downgrade_lock(i);
3629 + ii_write_unlock(i);
3634 + btop = au_ibtop(i);
3635 + bbot = au_ibbot(i);
3636 + if (btop <= bindex
3638 + && au_h_iptr(i, bindex)
3639 + && au_test_ibusy(i, btop, bbot)) {
3641 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3644 + ii_read_unlock(i);
3646 + au_iarray_free(array, max);
3652 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3653 + const unsigned int verbose)
3656 + unsigned int sigen;
3658 + sigen = au_sigen(root->d_sb);
3659 + DiMustNoWaiters(root);
3660 + IiMustNoWaiters(d_inode(root));
3661 + di_write_unlock(root);
3662 + err = test_dentry_busy(root, bindex, sigen, verbose);
3664 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3665 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3670 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3671 + struct file **to_free, int *idx)
3674 + unsigned char matched, root;
3675 + aufs_bindex_t bindex, bbot;
3676 + struct au_fidir *fidir;
3677 + struct au_hfile *hfile;
3680 + root = IS_ROOT(file->f_path.dentry);
3683 + to_free[*idx] = file;
3689 + fidir = au_fi(file)->fi_hdir;
3690 + AuDebugOn(!fidir);
3691 + bbot = au_fbbot_dir(file);
3692 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++) {
3693 + hfile = fidir->fd_hfile + bindex;
3694 + if (!hfile->hf_file)
3697 + if (hfile->hf_br->br_id == br_id) {
3709 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3710 + struct file **to_free, int opened)
3713 + unsigned long long ull, max;
3714 + aufs_bindex_t btop;
3715 + struct file *file, **array;
3716 + struct dentry *root;
3717 + struct au_hfile *hfile;
3719 + array = au_farray_alloc(sb, &max);
3720 + err = PTR_ERR(array);
3721 + if (IS_ERR(array))
3726 + root = sb->s_root;
3727 + di_write_unlock(root);
3728 + for (ull = 0; ull < max; ull++) {
3729 + file = array[ull];
3730 + if (unlikely(!file))
3733 + /* AuDbg("%pD\n", file); */
3734 + fi_read_lock(file);
3735 + btop = au_fbtop(file);
3736 + if (!d_is_dir(file->f_path.dentry)) {
3737 + hfile = &au_fi(file)->fi_htop;
3738 + if (hfile->hf_br->br_id == br_id)
3741 + err = test_dir_busy(file, br_id, to_free, &idx);
3742 + fi_read_unlock(file);
3743 + if (unlikely(err))
3746 + di_write_lock_child(root);
3747 + au_farray_free(array, max);
3748 + AuDebugOn(idx > opened);
3754 +static void br_del_file(struct file **to_free, unsigned long long opened,
3755 + aufs_bindex_t br_id)
3757 + unsigned long long ull;
3758 + aufs_bindex_t bindex, btop, bbot, bfound;
3759 + struct file *file;
3760 + struct au_fidir *fidir;
3761 + struct au_hfile *hfile;
3763 + for (ull = 0; ull < opened; ull++) {
3764 + file = to_free[ull];
3765 + if (unlikely(!file))
3768 + /* AuDbg("%pD\n", file); */
3769 + AuDebugOn(!d_is_dir(file->f_path.dentry));
3771 + fidir = au_fi(file)->fi_hdir;
3772 + AuDebugOn(!fidir);
3773 + fi_write_lock(file);
3774 + btop = au_fbtop(file);
3775 + bbot = au_fbbot_dir(file);
3776 + for (bindex = btop; bindex <= bbot; bindex++) {
3777 + hfile = fidir->fd_hfile + bindex;
3778 + if (!hfile->hf_file)
3781 + if (hfile->hf_br->br_id == br_id) {
3786 + AuDebugOn(bfound < 0);
3787 + au_set_h_fptr(file, bfound, NULL);
3788 + if (bfound == btop) {
3789 + for (btop++; btop <= bbot; btop++)
3790 + if (au_hf_dir(file, btop)) {
3791 + au_set_fbtop(file, btop);
3795 + fi_write_unlock(file);
3799 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3800 + const aufs_bindex_t bindex,
3801 + const aufs_bindex_t bbot)
3803 + struct au_branch **brp, **p;
3805 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3807 + brp = sbinfo->si_branch + bindex;
3808 + if (bindex < bbot)
3809 + memmove(brp, brp + 1, sizeof(*brp) * (bbot - bindex));
3810 + sbinfo->si_branch[0 + bbot] = NULL;
3811 + sbinfo->si_bbot--;
3813 + p = au_krealloc(sbinfo->si_branch, sizeof(*p) * bbot, AuGFP_SBILIST,
3816 + sbinfo->si_branch = p;
3817 + /* harmless error */
3820 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3821 + const aufs_bindex_t bbot)
3823 + struct au_hdentry *hdp, *p;
3825 + AuRwMustWriteLock(&dinfo->di_rwsem);
3827 + hdp = au_hdentry(dinfo, bindex);
3828 + if (bindex < bbot)
3829 + memmove(hdp, hdp + 1, sizeof(*hdp) * (bbot - bindex));
3830 + /* au_h_dentry_init(au_hdentry(dinfo, bbot); */
3833 + p = au_krealloc(dinfo->di_hdentry, sizeof(*p) * bbot, AuGFP_SBILIST,
3836 + dinfo->di_hdentry = p;
3837 + /* harmless error */
3840 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3841 + const aufs_bindex_t bbot)
3843 + struct au_hinode *hip, *p;
3845 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3847 + hip = au_hinode(iinfo, bindex);
3848 + if (bindex < bbot)
3849 + memmove(hip, hip + 1, sizeof(*hip) * (bbot - bindex));
3850 + /* au_hinode_init(au_hinode(iinfo, bbot)); */
3853 + p = au_krealloc(iinfo->ii_hinode, sizeof(*p) * bbot, AuGFP_SBILIST,
3856 + iinfo->ii_hinode = p;
3857 + /* harmless error */
3860 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3861 + struct au_branch *br)
3863 + aufs_bindex_t bbot;
3864 + struct au_sbinfo *sbinfo;
3865 + struct dentry *root, *h_root;
3866 + struct inode *inode, *h_inode;
3867 + struct au_hinode *hinode;
3869 + SiMustWriteLock(sb);
3871 + root = sb->s_root;
3872 + inode = d_inode(root);
3873 + sbinfo = au_sbi(sb);
3874 + bbot = sbinfo->si_bbot;
3876 + h_root = au_h_dptr(root, bindex);
3877 + hinode = au_hi(inode, bindex);
3878 + h_inode = au_igrab(hinode->hi_inode);
3881 + au_sbilist_lock();
3882 + au_br_do_del_brp(sbinfo, bindex, bbot);
3883 + au_br_do_del_hdp(au_di(root), bindex, bbot);
3884 + au_br_do_del_hip(au_ii(inode), bindex, bbot);
3885 + au_sbilist_unlock();
3887 + /* ignore an error */
3888 + au_dr_br_fin(sb, br); /* always, regardless the mount option */
3892 + au_br_do_free(br);
3895 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3896 + unsigned long long max, void *arg)
3901 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3904 + unsigned long long opened;
3905 + unsigned int mnt_flags;
3906 + aufs_bindex_t bindex, bbot, br_id;
3907 + unsigned char do_wh, verbose;
3908 + struct au_branch *br;
3909 + struct au_wbr *wbr;
3910 + struct dentry *root;
3911 + struct file **to_free;
3916 + root = sb->s_root;
3917 + bindex = au_find_dbindex(root, del->h_path.dentry);
3920 + goto out; /* success */
3922 + pr_err("%s no such branch\n", del->pathname);
3925 + AuDbg("bindex b%d\n", bindex);
3928 + mnt_flags = au_mntflags(sb);
3929 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3930 + bbot = au_sbbot(sb);
3931 + if (unlikely(!bbot)) {
3932 + AuVerbose(verbose, "no more branches left\n");
3936 + br = au_sbr(sb, bindex);
3937 + AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3938 + if (unlikely(au_lcnt_read(&br->br_count, /*do_rev*/1))) {
3939 + AuVerbose(verbose, "br %pd2 is busy now\n", del->h_path.dentry);
3943 + br_id = br->br_id;
3944 + opened = au_lcnt_read(&br->br_nfiles, /*do_rev*/1);
3945 + if (unlikely(opened)) {
3946 + to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3947 + err = PTR_ERR(to_free);
3948 + if (IS_ERR(to_free))
3951 + err = test_file_busy(sb, br_id, to_free, opened);
3952 + if (unlikely(err)) {
3953 + AuVerbose(verbose, "%llu file(s) opened\n", opened);
3959 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3961 + /* instead of WbrWhMustWriteLock(wbr) */
3962 + SiMustWriteLock(sb);
3963 + for (i = 0; i < AuBrWh_Last; i++) {
3964 + dput(wbr->wbr_wh[i]);
3965 + wbr->wbr_wh[i] = NULL;
3969 + err = test_children_busy(root, bindex, verbose);
3970 + if (unlikely(err)) {
3979 + * now we confirmed the branch is deletable.
3980 + * let's free the remaining opened dirs on the branch.
3982 + di_write_unlock(root);
3983 + br_del_file(to_free, opened, br_id);
3984 + di_write_lock_child(root);
3987 + sysaufs_brs_del(sb, bindex); /* remove successors */
3988 + dbgaufs_xino_del(br); /* remove one */
3989 + au_br_do_del(sb, bindex, br);
3990 + sysaufs_brs_add(sb, bindex); /* append successors */
3991 + dbgaufs_brs_add(sb, bindex, /*topdown*/1); /* rename successors */
3994 + au_cpup_attr_all(d_inode(root), /*force*/1);
3995 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3997 + au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
3998 + if (au_opt_test(mnt_flags, PLINK))
3999 + au_plink_half_refresh(sb, br_id);
4001 + goto out; /* success */
4005 + rerr = au_br_init_wh(sb, br, br->br_perm);
4007 + pr_warn("failed re-creating base whiteout, %s. (%d)\n",
4008 + del->pathname, rerr);
4011 + au_farray_free(to_free, opened);
4015 +/* ---------------------------------------------------------------------- */
4017 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
4020 + aufs_bindex_t btop, bbot;
4021 + struct aufs_ibusy ibusy;
4022 + struct inode *inode, *h_inode;
4025 + if (unlikely(!capable(CAP_SYS_ADMIN)))
4028 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
4030 + /* VERIFY_WRITE */
4031 + err = !access_ok(&arg->h_ino, sizeof(arg->h_ino));
4032 + if (unlikely(err)) {
4039 + si_read_lock(sb, AuLock_FLUSH);
4040 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbbot(sb)))
4044 + ibusy.h_ino = 0; /* invalid */
4045 + inode = ilookup(sb, ibusy.ino);
4047 + || inode->i_ino == AUFS_ROOT_INO
4048 + || au_is_bad_inode(inode))
4051 + ii_read_lock_child(inode);
4052 + btop = au_ibtop(inode);
4053 + bbot = au_ibbot(inode);
4054 + if (btop <= ibusy.bindex && ibusy.bindex <= bbot) {
4055 + h_inode = au_h_iptr(inode, ibusy.bindex);
4056 + if (h_inode && au_test_ibusy(inode, btop, bbot))
4057 + ibusy.h_ino = h_inode->i_ino;
4059 + ii_read_unlock(inode);
4063 + si_read_unlock(sb);
4065 + err = __put_user(ibusy.h_ino, &arg->h_ino);
4066 + if (unlikely(err)) {
4075 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
4077 + return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
4080 +#ifdef CONFIG_COMPAT
4081 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
4083 + return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
4087 +/* ---------------------------------------------------------------------- */
4090 + * change a branch permission
4093 +static void au_warn_ima(void)
4096 + /* since it doesn't support mark_files_ro() */
4097 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
4101 +static int do_need_sigen_inc(int a, int b)
4103 + return au_br_whable(a) && !au_br_whable(b);
4106 +static int need_sigen_inc(int old, int new)
4108 + return do_need_sigen_inc(old, new)
4109 + || do_need_sigen_inc(new, old);
4112 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
4115 + unsigned int mnt_flags;
4116 + unsigned long long ull, max;
4117 + aufs_bindex_t br_id;
4118 + unsigned char verbose, writer;
4119 + struct file *file, *hf, **array;
4120 + struct au_hfile *hfile;
4121 + struct inode *h_inode;
4123 + mnt_flags = au_mntflags(sb);
4124 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4126 + array = au_farray_alloc(sb, &max);
4127 + err = PTR_ERR(array);
4128 + if (IS_ERR(array))
4132 + br_id = au_sbr_id(sb, bindex);
4133 + for (ull = 0; ull < max; ull++) {
4134 + file = array[ull];
4135 + if (unlikely(!file))
4138 + /* AuDbg("%pD\n", file); */
4139 + fi_read_lock(file);
4140 + if (unlikely(au_test_mmapped(file))) {
4142 + AuVerbose(verbose, "mmapped %pD\n", file);
4144 + FiMustNoWaiters(file);
4145 + fi_read_unlock(file);
4149 + hfile = &au_fi(file)->fi_htop;
4150 + hf = hfile->hf_file;
4151 + if (!d_is_reg(file->f_path.dentry)
4152 + || !(file->f_mode & FMODE_WRITE)
4153 + || hfile->hf_br->br_id != br_id
4154 + || !(hf->f_mode & FMODE_WRITE))
4155 + array[ull] = NULL;
4161 + FiMustNoWaiters(file);
4162 + fi_read_unlock(file);
4170 + for (ull = 0; ull < max; ull++) {
4171 + file = array[ull];
4175 + /* todo: already flushed? */
4177 + * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4178 + * approach which resets f_mode and calls mnt_drop_write() and
4179 + * file_release_write() for each file, because the branch
4180 + * attribute in aufs world is totally different from the native
4183 + /* fi_read_lock(file); */
4184 + hfile = &au_fi(file)->fi_htop;
4185 + hf = hfile->hf_file;
4186 + /* fi_read_unlock(file); */
4187 + spin_lock(&hf->f_lock);
4188 + writer = !!(hf->f_mode & FMODE_WRITER);
4189 + hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4190 + spin_unlock(&hf->f_lock);
4192 + h_inode = file_inode(hf);
4193 + if (hf->f_mode & FMODE_READ)
4194 + i_readcount_inc(h_inode);
4195 + put_write_access(h_inode);
4196 + __mnt_drop_write(hf->f_path.mnt);
4201 + au_farray_free(array, max);
4207 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4211 + aufs_bindex_t bindex;
4212 + struct dentry *root;
4213 + struct au_branch *br;
4214 + struct au_br_fhsm *bf;
4216 + root = sb->s_root;
4217 + bindex = au_find_dbindex(root, mod->h_root);
4220 + return 0; /* success */
4222 + pr_err("%s no such branch\n", mod->path);
4225 + AuDbg("bindex b%d\n", bindex);
4227 + err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4228 + if (unlikely(err))
4231 + br = au_sbr(sb, bindex);
4232 + AuDebugOn(mod->h_root != au_br_dentry(br));
4233 + if (br->br_perm == mod->perm)
4234 + return 0; /* success */
4236 + /* pre-allocate for non-fhsm --> fhsm */
4238 + if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4239 + err = au_fhsm_br_alloc(br);
4240 + if (unlikely(err))
4243 + br->br_fhsm = NULL;
4246 + if (au_br_writable(br->br_perm)) {
4247 + /* remove whiteout base */
4248 + err = au_br_init_wh(sb, br, mod->perm);
4249 + if (unlikely(err))
4252 + if (!au_br_writable(mod->perm)) {
4253 + /* rw --> ro, file might be mmapped */
4254 + DiMustNoWaiters(root);
4255 + IiMustNoWaiters(d_inode(root));
4256 + di_write_unlock(root);
4257 + err = au_br_mod_files_ro(sb, bindex);
4258 + /* aufs_write_lock() calls ..._child() */
4259 + di_write_lock_child(root);
4261 + if (unlikely(err)) {
4263 + br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4266 + rerr = au_wbr_init(br, sb, br->br_perm);
4267 + if (unlikely(rerr)) {
4268 + AuIOErr("nested error %d (%d)\n",
4270 + br->br_perm = mod->perm;
4274 + } else if (au_br_writable(mod->perm)) {
4277 + br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4279 + err = au_wbr_init(br, sb, mod->perm);
4280 + if (unlikely(err)) {
4281 + au_kfree_rcu(br->br_wbr);
4282 + br->br_wbr = NULL;
4286 + if (unlikely(err))
4289 + if (au_br_fhsm(br->br_perm)) {
4290 + if (!au_br_fhsm(mod->perm)) {
4291 + /* fhsm --> non-fhsm */
4292 + au_br_fhsm_fin(br->br_fhsm);
4293 + au_kfree_rcu(br->br_fhsm);
4294 + br->br_fhsm = NULL;
4296 + } else if (au_br_fhsm(mod->perm))
4297 + /* non-fhsm --> fhsm */
4300 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4301 + br->br_perm = mod->perm;
4302 + goto out; /* success */
4305 + au_kfree_try_rcu(bf);
4311 +/* ---------------------------------------------------------------------- */
4313 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4316 + struct kstatfs kstfs;
4318 + err = vfs_statfs(&br->br_path, &kstfs);
4320 + stfs->f_blocks = kstfs.f_blocks;
4321 + stfs->f_bavail = kstfs.f_bavail;
4322 + stfs->f_files = kstfs.f_files;
4323 + stfs->f_ffree = kstfs.f_ffree;
4328 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4329 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
4330 +++ linux/fs/aufs/branch.h 2022-03-21 14:49:05.719966344 +0100
4332 +/* SPDX-License-Identifier: GPL-2.0 */
4334 + * Copyright (C) 2005-2021 Junjiro R. Okajima
4336 + * This program, aufs is free software; you can redistribute it and/or modify
4337 + * it under the terms of the GNU General Public License as published by
4338 + * the Free Software Foundation; either version 2 of the License, or
4339 + * (at your option) any later version.
4341 + * This program is distributed in the hope that it will be useful,
4342 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4343 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4344 + * GNU General Public License for more details.
4346 + * You should have received a copy of the GNU General Public License
4347 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4351 + * branch filesystems and xino for them
4354 +#ifndef __AUFS_BRANCH_H__
4355 +#define __AUFS_BRANCH_H__
4359 +#include <linux/mount.h>
4360 +#include "dirren.h"
4366 +/* ---------------------------------------------------------------------- */
4370 + struct file **xi_file;
4371 + unsigned int xi_nfile;
4377 + /* reserved for future use */
4378 + /* unsigned long *bitmap; */
4379 + wait_queue_head_t wqh;
4382 + struct mutex xi_mtx; /* protects xi_file array */
4383 + struct hlist_bl_head xi_writing;
4385 + atomic_t xi_truncating;
4387 + struct kref xi_kref;
4390 +/* File-based Hierarchical Storage Management */
4391 +struct au_br_fhsm {
4392 +#ifdef CONFIG_AUFS_FHSM
4393 + struct mutex bf_lock;
4394 + unsigned long bf_jiffy;
4395 + struct aufs_stfs bf_stfs;
4400 +/* members for writable branch only */
4401 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4403 + struct au_rwsem wbr_wh_rwsem;
4404 + struct dentry *wbr_wh[AuBrWh_Last];
4405 + atomic_t wbr_wh_running;
4406 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4407 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4408 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4411 + unsigned long long wbr_bytes;
4414 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4415 +#define AuBrDynOp (AuDyLast * 4)
4417 +#ifdef CONFIG_AUFS_HFSNOTIFY
4418 +/* support for asynchronous destruction */
4419 +struct au_br_hfsnotify {
4420 + struct fsnotify_group *hfsn_group;
4424 +/* sysfs entries */
4425 +struct au_brsysfs {
4427 + struct attribute attr;
4436 +/* protected by superblock rwsem */
4438 + struct au_xino *br_xino;
4440 + aufs_bindex_t br_id;
4443 + struct path br_path;
4444 + spinlock_t br_dykey_lock;
4445 + struct au_dykey *br_dykey[AuBrDynOp];
4446 + au_lcnt_t br_nfiles; /* opened files */
4447 + au_lcnt_t br_count; /* in-use for other */
4449 + struct au_wbr *br_wbr;
4450 + struct au_br_fhsm *br_fhsm;
4452 +#ifdef CONFIG_AUFS_HFSNOTIFY
4453 + struct au_br_hfsnotify *br_hfsn;
4456 +#ifdef CONFIG_SYSFS
4457 + /* entries under sysfs per mount-point */
4458 + struct au_brsysfs br_sysfs[AuBrSysfs_Last];
4461 +#ifdef CONFIG_DEBUG_FS
4462 + struct dentry *br_dbgaufs; /* xino */
4465 + struct au_dr_br br_dirren;
4468 +/* ---------------------------------------------------------------------- */
4470 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4472 + return br->br_path.mnt;
4475 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4477 + return br->br_path.dentry;
4480 +static inline struct user_namespace *au_br_userns(struct au_branch *br)
4482 + return mnt_user_ns(br->br_path.mnt);
4485 +static inline struct super_block *au_br_sb(struct au_branch *br)
4487 + return au_br_mnt(br)->mnt_sb;
4490 +static inline int au_br_rdonly(struct au_branch *br)
4492 + return (sb_rdonly(au_br_sb(br))
4493 + || !au_br_writable(br->br_perm))
4497 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4499 +#ifdef CONFIG_AUFS_HNOTIFY
4500 + return !(brperm & AuBrPerm_RR);
4506 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4508 + int err, exec_flag;
4511 + exec_flag = oflag & __FMODE_EXEC;
4512 + if (unlikely(exec_flag && path_noexec(&br->br_path)))
4518 +static inline void au_xino_get(struct au_branch *br)
4520 + struct au_xino *xi;
4524 + kref_get(&xi->xi_kref);
4527 +static inline int au_xino_count(struct au_branch *br)
4530 + struct au_xino *xi;
4535 + v = kref_read(&xi->xi_kref);
4540 +/* ---------------------------------------------------------------------- */
4544 +void au_br_free(struct au_sbinfo *sinfo);
4545 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4547 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4549 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4550 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4551 +#ifdef CONFIG_COMPAT
4552 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4555 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4558 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4561 +static const loff_t au_loff_max = LLONG_MAX;
4563 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry);
4564 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
4566 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
4567 + struct file *copy_src);
4569 + struct au_xino *xi; /* switch between xino and xigen */
4571 + struct path *base;
4572 + struct file *copy_src;
4574 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew);
4576 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4578 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4580 +ssize_t xino_fread(struct file *file, void *buf, size_t size, loff_t *pos);
4581 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos);
4583 +int au_xib_trunc(struct super_block *sb);
4584 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin);
4586 +struct au_xino *au_xino_alloc(unsigned int nfile);
4587 +int au_xino_put(struct au_branch *br);
4588 +struct file *au_xino_file1(struct au_xino *xi);
4590 +struct au_opt_xino;
4591 +void au_xino_clr(struct super_block *sb);
4592 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount);
4593 +struct file *au_xino_def(struct super_block *sb);
4594 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4595 + struct path *base);
4597 +ino_t au_xino_new_ino(struct super_block *sb);
4598 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4600 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
4601 + ino_t h_ino, int idx);
4602 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4605 +int au_xino_path(struct seq_file *seq, struct file *file);
4607 +/* ---------------------------------------------------------------------- */
4609 +/* @idx is signed to accept -1 meaning the first file */
4610 +static inline struct file *au_xino_file(struct au_xino *xi, int idx)
4612 + struct file *file;
4619 + if (idx < xi->xi_nfile)
4620 + file = xi->xi_file[idx];
4622 + file = au_xino_file1(xi);
4628 +/* ---------------------------------------------------------------------- */
4630 +/* Superblock to branch */
4632 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4634 + return au_sbr(sb, bindex)->br_id;
4638 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4640 + return au_br_mnt(au_sbr(sb, bindex));
4644 +struct user_namespace *au_sbr_userns(struct super_block *sb, aufs_bindex_t bindex)
4646 + return au_br_userns(au_sbr(sb, bindex));
4650 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4652 + return au_br_sb(au_sbr(sb, bindex));
4655 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4657 + return au_sbr(sb, bindex)->br_perm;
4660 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4662 + return au_br_whable(au_sbr_perm(sb, bindex));
4665 +/* ---------------------------------------------------------------------- */
4667 +#define wbr_wh_read_lock(wbr) au_rw_read_lock(&(wbr)->wbr_wh_rwsem)
4668 +#define wbr_wh_write_lock(wbr) au_rw_write_lock(&(wbr)->wbr_wh_rwsem)
4669 +#define wbr_wh_read_trylock(wbr) au_rw_read_trylock(&(wbr)->wbr_wh_rwsem)
4670 +#define wbr_wh_write_trylock(wbr) au_rw_write_trylock(&(wbr)->wbr_wh_rwsem)
4672 +#define wbr_wh_read_trylock_nested(wbr) \
4673 + au_rw_read_trylock_nested(&(wbr)->wbr_wh_rwsem)
4674 +#define wbr_wh_write_trylock_nested(wbr) \
4675 + au_rw_write_trylock_nested(&(wbr)->wbr_wh_rwsem)
4678 +#define wbr_wh_read_unlock(wbr) au_rw_read_unlock(&(wbr)->wbr_wh_rwsem)
4679 +#define wbr_wh_write_unlock(wbr) au_rw_write_unlock(&(wbr)->wbr_wh_rwsem)
4680 +#define wbr_wh_downgrade_lock(wbr) au_rw_dgrade_lock(&(wbr)->wbr_wh_rwsem)
4682 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&(wbr)->wbr_wh_rwsem)
4683 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&(wbr)->wbr_wh_rwsem)
4684 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&(wbr)->wbr_wh_rwsem)
4686 +/* ---------------------------------------------------------------------- */
4688 +#ifdef CONFIG_AUFS_FHSM
4689 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4691 + mutex_init(&brfhsm->bf_lock);
4692 + brfhsm->bf_jiffy = 0;
4693 + brfhsm->bf_readable = 0;
4696 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4698 + mutex_destroy(&brfhsm->bf_lock);
4701 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4702 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4705 +#endif /* __KERNEL__ */
4706 +#endif /* __AUFS_BRANCH_H__ */
4707 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4708 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
4709 +++ linux/fs/aufs/conf.mk 2022-03-21 14:49:05.719966344 +0100
4711 +# SPDX-License-Identifier: GPL-2.0
4713 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4717 +AuConfStr += ${1}=${${1}}
4721 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4723 + HNOTIFY HFSNOTIFY \
4735 +$(foreach i, ${AuConfAll}, \
4736 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
4738 +AuConfName = ${obj}/conf.str
4739 +${AuConfName}.tmp: FORCE
4740 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4741 +${AuConfName}: ${AuConfName}.tmp
4742 + @diff -q $< $@ > /dev/null 2>&1 || { \
4743 + echo ' GEN ' $@; \
4747 +clean-files += ${AuConfName} ${AuConfName}.tmp
4748 +${obj}/sysfs.o: ${AuConfName}
4750 +-include ${srctree}/${src}/conf_priv.mk
4751 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4752 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
4753 +++ linux/fs/aufs/cpup.c 2022-03-21 14:49:05.719966344 +0100
4755 +// SPDX-License-Identifier: GPL-2.0
4757 + * Copyright (C) 2005-2021 Junjiro R. Okajima
4759 + * This program, aufs is free software; you can redistribute it and/or modify
4760 + * it under the terms of the GNU General Public License as published by
4761 + * the Free Software Foundation; either version 2 of the License, or
4762 + * (at your option) any later version.
4764 + * This program is distributed in the hope that it will be useful,
4765 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4766 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4767 + * GNU General Public License for more details.
4769 + * You should have received a copy of the GNU General Public License
4770 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4774 + * copy-up functions, see wbr_policy.c for copy-down
4777 +#include <linux/fs_stack.h>
4778 +#include <linux/mm.h>
4779 +#include <linux/task_work.h>
4782 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4784 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4785 + | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4787 + BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4789 + dst->i_flags |= iflags & ~mask;
4790 + if (au_test_fs_notime(dst->i_sb))
4791 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
4794 +void au_cpup_attr_timesizes(struct inode *inode)
4796 + struct inode *h_inode;
4798 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4799 + fsstack_copy_attr_times(inode, h_inode);
4800 + fsstack_copy_inode_size(inode, h_inode);
4803 +void au_cpup_attr_nlink(struct inode *inode, int force)
4805 + struct inode *h_inode;
4806 + struct super_block *sb;
4807 + aufs_bindex_t bindex, bbot;
4810 + bindex = au_ibtop(inode);
4811 + h_inode = au_h_iptr(inode, bindex);
4813 + && !S_ISDIR(h_inode->i_mode)
4814 + && au_opt_test(au_mntflags(sb), PLINK)
4815 + && au_plink_test(inode))
4819 + * 0 can happen in revalidating.
4820 + * h_inode->i_mutex may not be held here, but it is harmless since once
4821 + * i_nlink reaches 0, it will never become positive except O_TMPFILE
4823 + * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4824 + * the incorrect link count.
4826 + set_nlink(inode, h_inode->i_nlink);
4829 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4830 + * it may includes whplink directory.
4832 + if (S_ISDIR(h_inode->i_mode)) {
4833 + bbot = au_ibbot(inode);
4834 + for (bindex++; bindex <= bbot; bindex++) {
4835 + h_inode = au_h_iptr(inode, bindex);
4837 + au_add_nlink(inode, h_inode);
4842 +void au_cpup_attr_changeable(struct inode *inode)
4844 + struct inode *h_inode;
4846 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4847 + inode->i_mode = h_inode->i_mode;
4848 + inode->i_uid = h_inode->i_uid;
4849 + inode->i_gid = h_inode->i_gid;
4850 + au_cpup_attr_timesizes(inode);
4851 + au_cpup_attr_flags(inode, h_inode->i_flags);
4854 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4856 + struct au_iinfo *iinfo = au_ii(inode);
4858 + IiMustWriteLock(inode);
4860 + iinfo->ii_higen = h_inode->i_generation;
4861 + iinfo->ii_hsb1 = h_inode->i_sb;
4864 +void au_cpup_attr_all(struct inode *inode, int force)
4866 + struct inode *h_inode;
4868 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4869 + au_cpup_attr_changeable(inode);
4870 + if (inode->i_nlink > 0)
4871 + au_cpup_attr_nlink(inode, force);
4872 + inode->i_rdev = h_inode->i_rdev;
4873 + inode->i_blkbits = h_inode->i_blkbits;
4874 + au_cpup_igen(inode, h_inode);
4877 +/* ---------------------------------------------------------------------- */
4879 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4881 +/* keep the timestamps of the parent dir when cpup */
4882 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4883 + struct path *h_path)
4885 + struct inode *h_inode;
4887 + dt->dt_dentry = dentry;
4888 + dt->dt_h_path = *h_path;
4889 + h_inode = d_inode(h_path->dentry);
4890 + dt->dt_atime = h_inode->i_atime;
4891 + dt->dt_mtime = h_inode->i_mtime;
4895 +void au_dtime_revert(struct au_dtime *dt)
4897 + struct iattr attr;
4900 + attr.ia_atime = dt->dt_atime;
4901 + attr.ia_mtime = dt->dt_mtime;
4902 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4903 + | ATTR_ATIME | ATTR_ATIME_SET;
4905 + /* no delegation since this is a directory */
4906 + err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4907 + if (unlikely(err))
4908 + pr_warn("restoring timestamps failed(%d). ignored\n", err);
4911 +/* ---------------------------------------------------------------------- */
4913 +/* internal use only */
4914 +struct au_cpup_reg_attr {
4917 + unsigned int iflags; /* inode->i_flags */
4920 +static noinline_for_stack
4921 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct path *h_src,
4922 + struct au_cpup_reg_attr *h_src_attr)
4924 + int err, sbits, icex;
4925 + unsigned int mnt_flags;
4926 + unsigned char verbose;
4928 + struct path h_path;
4929 + struct inode *h_isrc, *h_idst;
4930 + struct kstat *h_st;
4931 + struct au_branch *br;
4933 + br = au_sbr(dst->d_sb, bindex);
4934 + h_path.mnt = au_br_mnt(br);
4935 + h_path.dentry = au_h_dptr(dst, bindex);
4936 + h_idst = d_inode(h_path.dentry);
4937 + h_isrc = d_inode(h_src->dentry);
4938 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4939 + | ATTR_ATIME | ATTR_MTIME
4940 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
4941 + if (h_src_attr && h_src_attr->valid) {
4942 + h_st = &h_src_attr->st;
4943 + ia.ia_uid = h_st->uid;
4944 + ia.ia_gid = h_st->gid;
4945 + ia.ia_atime = h_st->atime;
4946 + ia.ia_mtime = h_st->mtime;
4947 + if (h_idst->i_mode != h_st->mode
4948 + && !S_ISLNK(h_idst->i_mode)) {
4949 + ia.ia_valid |= ATTR_MODE;
4950 + ia.ia_mode = h_st->mode;
4952 + sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4953 + au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4955 + ia.ia_uid = h_isrc->i_uid;
4956 + ia.ia_gid = h_isrc->i_gid;
4957 + ia.ia_atime = h_isrc->i_atime;
4958 + ia.ia_mtime = h_isrc->i_mtime;
4959 + if (h_idst->i_mode != h_isrc->i_mode
4960 + && !S_ISLNK(h_idst->i_mode)) {
4961 + ia.ia_valid |= ATTR_MODE;
4962 + ia.ia_mode = h_isrc->i_mode;
4964 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4965 + au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4967 + /* no delegation since it is just created */
4968 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4970 + /* is this nfs only? */
4971 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4972 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4973 + ia.ia_mode = h_isrc->i_mode;
4974 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4977 + icex = br->br_perm & AuBrAttr_ICEX;
4979 + mnt_flags = au_mntflags(dst->d_sb);
4980 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4981 + err = au_cpup_xattr(&h_path, h_src, icex, verbose);
4987 +/* ---------------------------------------------------------------------- */
4989 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4990 + char *buf, unsigned long blksize)
4993 + size_t sz, rbytes, wbytes;
4994 + unsigned char all_zero;
4996 + struct inode *h_inode;
4997 + /* reduce stack usage */
5000 + zp = page_address(ZERO_PAGE(0));
5001 + if (unlikely(!zp))
5002 + return -ENOMEM; /* possible? */
5007 + AuDbg("len %lld\n", len);
5009 + if (len < blksize)
5013 + /* todo: signal_pending? */
5014 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
5015 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
5018 + if (unlikely(err < 0))
5022 + if (len >= rbytes && rbytes == blksize)
5023 + all_zero = !memcmp(buf, zp, rbytes);
5030 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
5032 + /* todo: signal_pending? */
5033 + if (unlikely(err == -EAGAIN || err == -EINTR))
5035 + if (unlikely(err < 0))
5040 + if (unlikely(err < 0))
5046 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
5048 + if (unlikely(res < 0))
5055 + /* the last block may be a hole */
5056 + if (!err && all_zero) {
5057 + AuLabel(last hole);
5060 + if (au_test_nfs(dst->f_path.dentry->d_sb)) {
5061 + /* nfs requires this step to make last hole */
5062 + /* is this only nfs? */
5064 + /* todo: signal_pending? */
5065 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
5066 + } while (err == -EAGAIN || err == -EINTR);
5073 + ia->ia_size = dst->f_pos;
5074 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
5075 + ia->ia_file = dst;
5076 + h_inode = file_inode(dst);
5077 + inode_lock_nested(h_inode, AuLsc_I_CHILD2);
5078 + /* no delegation since it is just created */
5079 + err = vfsub_notify_change(&dst->f_path, ia,
5080 + /*delegated*/NULL);
5081 + inode_unlock(h_inode);
5088 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
5091 + unsigned long blksize;
5092 + unsigned char do_kfree;
5094 + struct super_block *h_sb;
5097 + h_sb = file_inode(dst)->i_sb;
5098 + blksize = h_sb->s_blocksize;
5099 + if (!blksize || PAGE_SIZE < blksize)
5100 + blksize = PAGE_SIZE;
5101 + AuDbg("blksize %lu\n", blksize);
5102 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
5104 + buf = kmalloc(blksize, GFP_NOFS);
5106 + buf = (void *)__get_free_page(GFP_NOFS);
5107 + if (unlikely(!buf))
5110 + if (len > (1 << 22))
5111 + AuDbg("copying a large file %lld\n", (long long)len);
5115 + err = au_do_copy_file(dst, src, len, buf, blksize);
5117 + AuDebugOn(!au_kfree_do_sz_test(blksize));
5118 + au_kfree_do_rcu(buf);
5120 + free_page((unsigned long)buf);
5126 +static int au_do_copy(struct file *dst, struct file *src, loff_t len)
5129 + struct super_block *h_src_sb;
5130 + struct inode *h_src_inode;
5132 + h_src_inode = file_inode(src);
5133 + h_src_sb = h_src_inode->i_sb;
5135 + /* XFS acquires inode_lock */
5136 + if (!au_test_xfs(h_src_sb))
5137 + err = au_copy_file(dst, src, len);
5139 + inode_unlock_shared(h_src_inode);
5140 + err = au_copy_file(dst, src, len);
5141 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5147 +static int au_clone_or_copy(struct file *dst, struct file *src, loff_t len)
5151 + struct super_block *h_src_sb;
5152 + struct inode *h_src_inode;
5154 + h_src_inode = file_inode(src);
5155 + h_src_sb = h_src_inode->i_sb;
5156 + if (h_src_sb != file_inode(dst)->i_sb
5157 + || !dst->f_op->remap_file_range) {
5158 + err = au_do_copy(dst, src, len);
5162 + if (!au_test_nfs(h_src_sb)) {
5163 + inode_unlock_shared(h_src_inode);
5164 + lo = vfsub_clone_file_range(src, dst, len);
5165 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5167 + lo = vfsub_clone_file_range(src, dst, len);
5170 + goto out; /* success */
5171 + } else if (lo >= 0)
5172 + /* todo: possible? */
5173 + /* paritially succeeded */
5174 + AuDbg("lo %lld, len %lld. Retrying.\n", lo, len);
5175 + else if (lo != -EOPNOTSUPP) {
5176 + /* older XFS has a condition in cloning */
5181 + /* the backend fs on NFS may not support cloning */
5182 + err = au_do_copy(dst, src, len);
5190 + * to support a sparse file which is opened with O_APPEND,
5191 + * we need to close the file.
5193 +static int au_cp_regular(struct au_cp_generic *cpg)
5196 + enum { SRC, DST };
5198 + aufs_bindex_t bindex;
5199 + unsigned int flags;
5200 + struct dentry *dentry;
5202 + struct file *file;
5205 + .bindex = cpg->bsrc,
5206 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
5209 + .bindex = cpg->bdst,
5210 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
5211 + .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
5214 + struct au_branch *br;
5215 + struct super_block *sb, *h_src_sb;
5216 + struct inode *h_src_inode;
5217 + struct task_struct *tsk = current;
5219 + /* bsrc branch can be ro/rw. */
5220 + sb = cpg->dentry->d_sb;
5222 + for (i = 0; i < 2; i++, f++) {
5223 + f->dentry = au_h_dptr(cpg->dentry, f->bindex);
5224 + f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
5225 + /*file*/NULL, f->force_wr);
5226 + if (IS_ERR(f->file)) {
5227 + err = PTR_ERR(f->file);
5235 + /* try stopping to update while we copyup */
5236 + h_src_inode = d_inode(file[SRC].dentry);
5237 + h_src_sb = h_src_inode->i_sb;
5238 + if (!au_test_nfs(h_src_sb))
5239 + IMustLock(h_src_inode);
5240 + err = au_clone_or_copy(file[DST].file, file[SRC].file, cpg->len);
5242 + /* i wonder if we had O_NO_DELAY_FPUT flag */
5243 + if (tsk->flags & PF_KTHREAD)
5244 + __fput_sync(file[DST].file);
5246 + /* it happened actually */
5247 + fput(file[DST].file);
5250 + * we have to call both since we don't know which place the file
5254 + flush_delayed_fput();
5256 + br = au_sbr(sb, file[DST].bindex);
5257 + au_lcnt_dec(&br->br_nfiles);
5260 + fput(file[SRC].file);
5261 + br = au_sbr(sb, file[SRC].bindex);
5262 + au_lcnt_dec(&br->br_nfiles);
5267 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
5268 + struct au_cpup_reg_attr *h_src_attr)
5272 + struct path h_path;
5273 + struct inode *h_src_inode, *h_dst_inode;
5276 + h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
5277 + l = i_size_read(h_src_inode);
5278 + if (cpg->len == -1 || l < cpg->len)
5281 + /* try stopping to update while we are referencing */
5282 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5283 + au_pin_hdir_unlock(cpg->pin);
5285 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5286 + h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
5287 + h_src_attr->iflags = h_src_inode->i_flags;
5288 + if (!au_test_nfs(h_src_inode->i_sb))
5289 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5291 + inode_unlock_shared(h_src_inode);
5292 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5293 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5295 + if (unlikely(err)) {
5296 + inode_unlock_shared(h_src_inode);
5299 + h_src_attr->valid = 1;
5300 + if (!au_test_nfs(h_src_inode->i_sb)) {
5301 + err = au_cp_regular(cpg);
5302 + inode_unlock_shared(h_src_inode);
5304 + inode_unlock_shared(h_src_inode);
5305 + err = au_cp_regular(cpg);
5307 + rerr = au_pin_hdir_relock(cpg->pin);
5311 + if (!err && (h_src_inode->i_state & I_LINKABLE)) {
5312 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5313 + h_dst_inode = d_inode(h_path.dentry);
5314 + spin_lock(&h_dst_inode->i_lock);
5315 + h_dst_inode->i_state |= I_LINKABLE;
5316 + spin_unlock(&h_dst_inode->i_lock);
5323 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
5324 + struct inode *h_dir)
5327 + DEFINE_DELAYED_CALL(done);
5330 + sym = vfs_get_link(h_src, &done);
5331 + err = PTR_ERR(sym);
5335 + err = vfsub_symlink(h_dir, h_path, sym);
5338 + do_delayed_call(&done);
5343 + * regardless 'acl' option, reset all ACL.
5344 + * All ACL will be copied up later from the original entry on the lower branch.
5346 +static int au_reset_acl(struct inode *h_dir, struct path *h_path, umode_t mode)
5349 + struct dentry *h_dentry;
5350 + struct inode *h_inode;
5351 + struct user_namespace *h_userns;
5353 + h_userns = mnt_user_ns(h_path->mnt);
5354 + h_dentry = h_path->dentry;
5355 + h_inode = d_inode(h_dentry);
5356 + /* forget_all_cached_acls(h_inode)); */
5357 + err = vfsub_removexattr(h_userns, h_dentry, XATTR_NAME_POSIX_ACL_ACCESS);
5359 + if (err == -EOPNOTSUPP)
5362 + err = vfsub_acl_chmod(h_userns, h_inode, mode);
5368 +static int au_do_cpup_dir(struct au_cp_generic *cpg, struct dentry *dst_parent,
5369 + struct inode *h_dir, struct path *h_path)
5372 + struct inode *dir, *inode;
5373 + struct user_namespace *h_userns;
5375 + h_userns = mnt_user_ns(h_path->mnt);
5376 + err = vfsub_removexattr(h_userns, h_path->dentry,
5377 + XATTR_NAME_POSIX_ACL_DEFAULT);
5379 + if (err == -EOPNOTSUPP)
5381 + if (unlikely(err))
5385 + * strange behaviour from the users view,
5386 + * particularly setattr case
5388 + dir = d_inode(dst_parent);
5389 + if (au_ibtop(dir) == cpg->bdst)
5390 + au_cpup_attr_nlink(dir, /*force*/1);
5391 + inode = d_inode(cpg->dentry);
5392 + au_cpup_attr_nlink(inode, /*force*/1);
5398 +static noinline_for_stack
5399 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5400 + struct au_cpup_reg_attr *h_src_attr)
5404 + unsigned int mnt_flags;
5405 + unsigned char isdir, isreg, force;
5406 + const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5407 + struct au_dtime dt;
5408 + struct path h_path;
5409 + struct dentry *h_src, *h_dst, *h_parent;
5410 + struct inode *h_inode, *h_dir;
5411 + struct super_block *sb;
5413 + /* bsrc branch can be ro/rw. */
5414 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5415 + h_inode = d_inode(h_src);
5416 + AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5418 + /* try stopping to be referenced while we are creating */
5419 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5420 + if (au_ftest_cpup(cpg->flags, RENAME))
5421 + AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5422 + AUFS_WH_PFX_LEN));
5423 + h_parent = h_dst->d_parent; /* dir inode is locked */
5424 + h_dir = d_inode(h_parent);
5426 + AuDebugOn(h_parent != h_dst->d_parent);
5428 + sb = cpg->dentry->d_sb;
5429 + h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5431 + h_path.dentry = h_parent;
5432 + au_dtime_store(&dt, dst_parent, &h_path);
5434 + h_path.dentry = h_dst;
5438 + mode = h_inode->i_mode;
5439 + switch (mode & S_IFMT) {
5442 + err = vfsub_create(h_dir, &h_path, 0600, /*want_excl*/true);
5444 + err = au_do_cpup_regular(cpg, h_src_attr);
5448 + err = vfsub_mkdir(h_dir, &h_path, mode);
5450 + err = au_do_cpup_dir(cpg, dst_parent, h_dir, &h_path);
5453 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5457 + AuDebugOn(!capable(CAP_MKNOD));
5461 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5464 + AuIOErr("Unknown inode type 0%o\n", mode);
5468 + err = au_reset_acl(h_dir, &h_path, mode);
5470 + mnt_flags = au_mntflags(sb);
5471 + if (!au_opt_test(mnt_flags, UDBA_NONE)
5473 + && au_opt_test(mnt_flags, XINO)
5474 + && (h_inode->i_nlink == 1
5475 + || (h_inode->i_state & I_LINKABLE))
5476 + /* todo: unnecessary? */
5477 + /* && d_inode(cpg->dentry)->i_nlink == 1 */
5478 + && cpg->bdst < cpg->bsrc
5479 + && !au_ftest_cpup(cpg->flags, KEEPLINO))
5480 + au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5481 + /* ignore this error */
5486 + force = !!cpg->len;
5487 + if (cpg->len == -1)
5488 + force = !!i_size_read(h_inode);
5490 + au_fhsm_wrote(sb, cpg->bdst, force);
5494 + au_dtime_revert(&dt);
5498 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5501 + struct dentry *dentry, *h_dentry, *h_parent, *parent;
5502 + struct path h_ppath;
5503 + struct inode *h_dir;
5504 + aufs_bindex_t bdst;
5506 + dentry = cpg->dentry;
5508 + h_ppath.mnt = au_sbr_mnt(dentry->d_sb, bdst);
5509 + h_dentry = au_h_dptr(dentry, bdst);
5510 + if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5512 + au_set_h_dptr(dentry, bdst, NULL);
5513 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
5515 + h_path->dentry = dget(au_h_dptr(dentry, bdst));
5516 + au_set_h_dptr(dentry, bdst, h_dentry);
5519 + parent = dget_parent(dentry);
5520 + h_ppath.dentry = au_h_dptr(parent, bdst);
5522 + h_path->dentry = vfsub_lkup_one(&dentry->d_name, &h_ppath);
5523 + if (IS_ERR(h_path->dentry))
5524 + err = PTR_ERR(h_path->dentry);
5526 + if (unlikely(err))
5529 + h_parent = h_dentry->d_parent; /* dir inode is locked */
5530 + h_dir = d_inode(h_parent);
5532 + AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5533 + /* no delegation since it is just created */
5534 + err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL,
5536 + dput(h_path->dentry);
5543 + * copyup the @dentry from @bsrc to @bdst.
5544 + * the caller must set the both of lower dentries.
5545 + * @len is for truncating when it is -1 copyup the entire file.
5546 + * in link/rename cases, @dst_parent may be different from the real one.
5547 + * basic->bsrc can be larger than basic->bdst.
5548 + * aufs doesn't touch the credential so
5549 + * security_inode_copy_up{,_xattr}() are unnecessary.
5551 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5554 + aufs_bindex_t old_ibtop;
5555 + unsigned char isdir, plink;
5556 + struct dentry *h_src, *h_dst, *h_parent;
5557 + struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5558 + struct super_block *sb;
5559 + struct au_branch *br;
5560 + struct path h_src_path;
5561 + /* to reduce stack size */
5563 + struct au_dtime dt;
5564 + struct path h_path;
5565 + struct au_cpup_reg_attr h_src_attr;
5569 + a = kmalloc(sizeof(*a), GFP_NOFS);
5572 + a->h_src_attr.valid = 0;
5574 + sb = cpg->dentry->d_sb;
5575 + br = au_sbr(sb, cpg->bdst);
5576 + a->h_path.mnt = au_br_mnt(br);
5577 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5578 + h_parent = h_dst->d_parent; /* dir inode is locked */
5579 + h_dir = d_inode(h_parent);
5582 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5583 + inode = d_inode(cpg->dentry);
5586 + dst_parent = dget_parent(cpg->dentry);
5590 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
5591 + dst_inode = au_h_iptr(inode, cpg->bdst);
5593 + if (unlikely(!plink)) {
5595 + AuIOErr("hi%lu(i%lu) exists on b%d "
5596 + "but plink is disabled\n",
5597 + dst_inode->i_ino, inode->i_ino, cpg->bdst);
5601 + if (dst_inode->i_nlink) {
5602 + const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5604 + h_src = au_plink_lkup(inode, cpg->bdst);
5605 + err = PTR_ERR(h_src);
5606 + if (IS_ERR(h_src))
5608 + if (unlikely(d_is_negative(h_src))) {
5610 + AuIOErr("i%lu exists on b%d "
5611 + "but not pseudo-linked\n",
5612 + inode->i_ino, cpg->bdst);
5618 + a->h_path.dentry = h_parent;
5619 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5622 + a->h_path.dentry = h_dst;
5624 + err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5625 + if (!err && au_ftest_cpup(cpg->flags, RENAME))
5626 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5628 + au_dtime_revert(&a->dt);
5629 + if (unlikely(err == -EWOULDBLOCK)) {
5630 + pr_warn("cannot retry for NFSv4 delegation"
5631 + " for an internal link\n");
5637 + /* todo: cpup_wh_file? */
5639 + au_update_ibrange(inode, /*do_put_zero*/1);
5642 + isdir = S_ISDIR(inode->i_mode);
5643 + old_ibtop = au_ibtop(inode);
5644 + err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5645 + if (unlikely(err))
5647 + dst_inode = d_inode(h_dst);
5648 + inode_lock_nested(dst_inode, AuLsc_I_CHILD2);
5649 + /* todo: necessary? */
5650 + /* au_pin_hdir_unlock(cpg->pin); */
5652 + h_src_path.dentry = h_src;
5653 + h_src_path.mnt = au_sbr_mnt(sb, cpg->bsrc);
5654 + err = cpup_iattr(cpg->dentry, cpg->bdst, &h_src_path, &a->h_src_attr);
5655 + if (unlikely(err)) {
5656 + /* todo: necessary? */
5657 + /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5658 + inode_unlock(dst_inode);
5662 + if (cpg->bdst < old_ibtop) {
5663 + if (S_ISREG(inode->i_mode)) {
5664 + err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5665 + if (unlikely(err)) {
5666 + /* ignore an error */
5667 + /* au_pin_hdir_relock(cpg->pin); */
5668 + inode_unlock(dst_inode);
5672 + au_set_ibtop(inode, cpg->bdst);
5674 + au_set_ibbot(inode, cpg->bdst);
5675 + au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5676 + au_hi_flags(inode, isdir));
5678 + /* todo: necessary? */
5679 + /* err = au_pin_hdir_relock(cpg->pin); */
5680 + inode_unlock(dst_inode);
5681 + if (unlikely(err))
5684 + src_inode = d_inode(h_src);
5686 + && (src_inode->i_nlink > 1
5687 + || src_inode->i_state & I_LINKABLE)
5689 + au_plink_append(inode, cpg->bdst, h_dst);
5691 + if (au_ftest_cpup(cpg->flags, RENAME)) {
5692 + a->h_path.dentry = h_dst;
5693 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5696 + goto out_parent; /* success */
5700 + a->h_path.dentry = h_parent;
5701 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5702 + a->h_path.dentry = h_dst;
5704 + if (d_is_positive(h_dst)) {
5706 + /* no delegation since it is just created */
5707 + rerr = vfsub_unlink(h_dir, &a->h_path,
5708 + /*delegated*/NULL, /*force*/0);
5710 + rerr = vfsub_rmdir(h_dir, &a->h_path);
5712 + au_dtime_revert(&a->dt);
5714 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5724 +#if 0 /* reserved */
5725 +struct au_cpup_single_args {
5727 + struct au_cp_generic *cpg;
5728 + struct dentry *dst_parent;
5731 +static void au_call_cpup_single(void *args)
5733 + struct au_cpup_single_args *a = args;
5735 + au_pin_hdir_acquire_nest(a->cpg->pin);
5736 + *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5737 + au_pin_hdir_release(a->cpg->pin);
5742 + * prevent SIGXFSZ in copy-up.
5743 + * testing CAP_MKNOD is for generic fs,
5744 + * but CAP_FSETID is for xfs only, currently.
5746 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5749 + struct super_block *sb;
5750 + struct inode *h_dir;
5753 + sb = au_pinned_parent(pin)->d_sb;
5754 + if (!au_wkq_test()
5755 + && (!au_sbi(sb)->si_plink_maint_pid
5756 + || au_plink_maint(sb, AuLock_NOPLM))) {
5757 + switch (mode & S_IFMT) {
5759 + /* no condition about RLIMIT_FSIZE and the file size */
5764 + do_sio = !capable(CAP_MKNOD);
5768 + do_sio = ((mode & (S_ISUID | S_ISGID))
5769 + && !capable(CAP_FSETID));
5770 + /* this workaround may be removed in the future */
5772 + h_dir = au_pinned_h_dir(pin);
5773 + do_sio = h_dir->i_mode & S_ISVTX;
5780 +#if 0 /* reserved */
5781 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5784 + struct dentry *h_dentry;
5786 + h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5787 + if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5788 + err = au_cpup_single(cpg, dst_parent);
5790 + struct au_cpup_single_args args = {
5793 + .dst_parent = dst_parent
5795 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5796 + if (unlikely(wkq_err))
5805 + * copyup the @dentry from the first active lower branch to @bdst,
5806 + * using au_cpup_single().
5808 +static int au_cpup_simple(struct au_cp_generic *cpg)
5811 + unsigned int flags_orig;
5812 + struct dentry *dentry;
5814 + AuDebugOn(cpg->bsrc < 0);
5816 + dentry = cpg->dentry;
5817 + DiMustWriteLock(dentry);
5819 + err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5821 + flags_orig = cpg->flags;
5822 + au_fset_cpup(cpg->flags, RENAME);
5823 + err = au_cpup_single(cpg, NULL);
5824 + cpg->flags = flags_orig;
5826 + return 0; /* success */
5829 + au_set_h_dptr(dentry, cpg->bdst, NULL);
5830 + au_set_dbtop(dentry, cpg->bsrc);
5836 +struct au_cpup_simple_args {
5838 + struct au_cp_generic *cpg;
5841 +static void au_call_cpup_simple(void *args)
5843 + struct au_cpup_simple_args *a = args;
5845 + au_pin_hdir_acquire_nest(a->cpg->pin);
5846 + *a->errp = au_cpup_simple(a->cpg);
5847 + au_pin_hdir_release(a->cpg->pin);
5850 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5853 + struct dentry *dentry, *parent;
5854 + struct file *h_file;
5855 + struct inode *h_dir;
5856 + struct user_namespace *h_userns;
5858 + dentry = cpg->dentry;
5860 + if (au_ftest_cpup(cpg->flags, HOPEN)) {
5861 + AuDebugOn(cpg->bsrc < 0);
5862 + h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5863 + err = PTR_ERR(h_file);
5864 + if (IS_ERR(h_file))
5868 + parent = dget_parent(dentry);
5869 + h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5870 + h_userns = au_sbr_userns(dentry->d_sb, cpg->bdst);
5871 + if (!au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC | MAY_WRITE)
5872 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5873 + err = au_cpup_simple(cpg);
5875 + struct au_cpup_simple_args args = {
5879 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5880 + if (unlikely(wkq_err))
5886 + au_h_open_post(dentry, cpg->bsrc, h_file);
5892 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5894 + aufs_bindex_t bsrc, bbot;
5895 + struct dentry *dentry, *h_dentry;
5897 + if (cpg->bsrc < 0) {
5898 + dentry = cpg->dentry;
5899 + bbot = au_dbbot(dentry);
5900 + for (bsrc = cpg->bdst + 1; bsrc <= bbot; bsrc++) {
5901 + h_dentry = au_h_dptr(dentry, bsrc);
5903 + AuDebugOn(d_is_negative(h_dentry));
5907 + AuDebugOn(bsrc > bbot);
5910 + AuDebugOn(cpg->bsrc <= cpg->bdst);
5911 + return au_do_sio_cpup_simple(cpg);
5914 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5916 + AuDebugOn(cpg->bdst <= cpg->bsrc);
5917 + return au_do_sio_cpup_simple(cpg);
5920 +/* ---------------------------------------------------------------------- */
5923 + * copyup the deleted file for writing.
5925 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5926 + struct file *file)
5929 + unsigned int flags_orig;
5930 + aufs_bindex_t bsrc_orig;
5931 + struct au_dinfo *dinfo;
5933 + struct au_hdentry *hd;
5934 + struct dentry *h_dentry;
5937 + dinfo = au_di(cpg->dentry);
5938 + AuRwMustWriteLock(&dinfo->di_rwsem);
5940 + bsrc_orig = cpg->bsrc;
5941 + cpg->bsrc = dinfo->di_btop;
5942 + hdst.hd = au_hdentry(dinfo, cpg->bdst);
5943 + hdst.h_dentry = hdst.hd->hd_dentry;
5944 + hdst.hd->hd_dentry = wh_dentry;
5945 + dinfo->di_btop = cpg->bdst;
5947 + hsrc.h_dentry = NULL;
5949 + hsrc.hd = au_hdentry(dinfo, cpg->bsrc);
5950 + hsrc.h_dentry = hsrc.hd->hd_dentry;
5951 + hsrc.hd->hd_dentry = au_hf_top(file)->f_path.dentry;
5953 + flags_orig = cpg->flags;
5954 + cpg->flags = !AuCpup_DTIME;
5955 + err = au_cpup_single(cpg, /*h_parent*/NULL);
5956 + cpg->flags = flags_orig;
5959 + err = au_reopen_nondir(file);
5960 + hsrc.hd->hd_dentry = hsrc.h_dentry;
5962 + hdst.hd->hd_dentry = hdst.h_dentry;
5963 + dinfo->di_btop = cpg->bsrc;
5964 + cpg->bsrc = bsrc_orig;
5969 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5972 + aufs_bindex_t bdst;
5973 + struct au_dtime dt;
5974 + struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5975 + struct au_branch *br;
5976 + struct path h_path;
5978 + dentry = cpg->dentry;
5980 + br = au_sbr(dentry->d_sb, bdst);
5981 + parent = dget_parent(dentry);
5982 + h_parent = au_h_dptr(parent, bdst);
5983 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5984 + err = PTR_ERR(wh_dentry);
5985 + if (IS_ERR(wh_dentry))
5988 + h_path.dentry = h_parent;
5989 + h_path.mnt = au_br_mnt(br);
5990 + au_dtime_store(&dt, parent, &h_path);
5991 + err = au_do_cpup_wh(cpg, wh_dentry, file);
5992 + if (unlikely(err))
5996 + h_path.dentry = wh_dentry;
5997 + if (!d_is_dir(wh_dentry)) {
5998 + /* no delegation since it is just created */
5999 + err = vfsub_unlink(d_inode(h_parent), &h_path,
6000 + /*delegated*/NULL, /*force*/0);
6002 + err = vfsub_rmdir(d_inode(h_parent), &h_path);
6003 + if (unlikely(err)) {
6004 + AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
6008 + au_dtime_revert(&dt);
6009 + au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
6018 +struct au_cpup_wh_args {
6020 + struct au_cp_generic *cpg;
6021 + struct file *file;
6024 +static void au_call_cpup_wh(void *args)
6026 + struct au_cpup_wh_args *a = args;
6028 + au_pin_hdir_acquire_nest(a->cpg->pin);
6029 + *a->errp = au_cpup_wh(a->cpg, a->file);
6030 + au_pin_hdir_release(a->cpg->pin);
6033 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6036 + aufs_bindex_t bdst;
6037 + struct dentry *dentry, *parent, *h_orph, *h_parent;
6038 + struct inode *dir, *h_dir, *h_tmpdir;
6039 + struct au_wbr *wbr;
6040 + struct au_pin wh_pin, *pin_orig;
6041 + struct user_namespace *h_userns;
6043 + dentry = cpg->dentry;
6045 + parent = dget_parent(dentry);
6046 + dir = d_inode(parent);
6049 + h_dir = au_igrab(au_h_iptr(dir, bdst));
6052 + if (!h_dir->i_nlink) {
6053 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
6054 + h_orph = wbr->wbr_orph;
6056 + h_parent = dget(au_h_dptr(parent, bdst));
6057 + au_set_h_dptr(parent, bdst, dget(h_orph));
6058 + h_tmpdir = d_inode(h_orph);
6059 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
6061 + inode_lock_nested(h_tmpdir, AuLsc_I_PARENT3);
6062 + /* todo: au_h_open_pre()? */
6064 + pin_orig = cpg->pin;
6065 + au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
6066 + AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
6067 + cpg->pin = &wh_pin;
6070 + h_userns = au_sbr_userns(dentry->d_sb, bdst);
6071 + if (!au_test_h_perm_sio(h_userns, h_tmpdir, MAY_EXEC | MAY_WRITE)
6072 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6073 + err = au_cpup_wh(cpg, file);
6075 + struct au_cpup_wh_args args = {
6080 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
6081 + if (unlikely(wkq_err))
6086 + inode_unlock(h_tmpdir);
6087 + /* todo: au_h_open_post()? */
6088 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
6089 + au_set_h_dptr(parent, bdst, h_parent);
6090 + AuDebugOn(!pin_orig);
6091 + cpg->pin = pin_orig;
6099 +/* ---------------------------------------------------------------------- */
6102 + * generic routine for both of copy-up and copy-down.
6104 +/* cf. revalidate function in file.c */
6105 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6106 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6107 + struct au_pin *pin,
6108 + struct dentry *h_parent, void *arg),
6112 + struct au_pin pin;
6113 + struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
6116 + parent = dget_parent(dentry);
6117 + if (IS_ROOT(parent))
6120 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
6121 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
6123 + /* do not use au_dpage */
6124 + real_parent = parent;
6127 + parent = dget_parent(dentry);
6128 + h_parent = au_h_dptr(parent, bdst);
6130 + goto out; /* success */
6132 + /* find top dir which is necessary to cpup */
6136 + parent = dget_parent(d);
6137 + di_read_lock_parent3(parent, !AuLock_IR);
6138 + h_parent = au_h_dptr(parent, bdst);
6139 + di_read_unlock(parent, !AuLock_IR);
6140 + } while (!h_parent);
6142 + if (d != real_parent)
6143 + di_write_lock_child3(d);
6145 + /* somebody else might create while we were sleeping */
6146 + h_dentry = au_h_dptr(d, bdst);
6147 + if (!h_dentry || d_is_negative(h_dentry)) {
6149 + au_update_dbtop(d);
6151 + au_pin_set_dentry(&pin, d);
6152 + err = au_do_pin(&pin);
6154 + err = cp(d, bdst, &pin, h_parent, arg);
6159 + if (d != real_parent)
6160 + di_write_unlock(d);
6161 + if (unlikely(err))
6170 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
6171 + struct au_pin *pin,
6172 + struct dentry *h_parent __maybe_unused,
6173 + void *arg __maybe_unused)
6175 + struct au_cp_generic cpg = {
6181 + .flags = AuCpup_DTIME
6183 + return au_sio_cpup_simple(&cpg);
6186 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6188 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
6191 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6194 + struct dentry *parent;
6195 + struct inode *dir;
6197 + parent = dget_parent(dentry);
6198 + dir = d_inode(parent);
6200 + if (au_h_iptr(dir, bdst))
6203 + di_read_unlock(parent, AuLock_IR);
6204 + di_write_lock_parent(parent);
6205 + /* someone else might change our inode while we were sleeping */
6206 + if (!au_h_iptr(dir, bdst))
6207 + err = au_cpup_dirs(dentry, bdst);
6208 + di_downgrade_lock(parent, AuLock_IR);
6214 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
6215 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
6216 +++ linux/fs/aufs/cpup.h 2022-03-21 14:49:05.719966344 +0100
6218 +/* SPDX-License-Identifier: GPL-2.0 */
6220 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6222 + * This program, aufs is free software; you can redistribute it and/or modify
6223 + * it under the terms of the GNU General Public License as published by
6224 + * the Free Software Foundation; either version 2 of the License, or
6225 + * (at your option) any later version.
6227 + * This program is distributed in the hope that it will be useful,
6228 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6229 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6230 + * GNU General Public License for more details.
6232 + * You should have received a copy of the GNU General Public License
6233 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6237 + * copy-up/down functions
6240 +#ifndef __AUFS_CPUP_H__
6241 +#define __AUFS_CPUP_H__
6245 +#include <linux/path.h>
6251 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
6252 +void au_cpup_attr_timesizes(struct inode *inode);
6253 +void au_cpup_attr_nlink(struct inode *inode, int force);
6254 +void au_cpup_attr_changeable(struct inode *inode);
6255 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
6256 +void au_cpup_attr_all(struct inode *inode, int force);
6258 +/* ---------------------------------------------------------------------- */
6260 +struct au_cp_generic {
6261 + struct dentry *dentry;
6262 + aufs_bindex_t bdst, bsrc;
6264 + struct au_pin *pin;
6265 + unsigned int flags;
6269 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
6270 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
6272 +#define AuCpup_RENAME (1 << 2) /* rename after cpup */
6273 +#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
6275 +#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
6277 +#define AuCpup_RWDST (1 << 5) /* force write target even if
6278 + the branch is marked as RO */
6280 +#ifndef CONFIG_AUFS_BR_HFSPLUS
6281 +#undef AuCpup_HOPEN
6282 +#define AuCpup_HOPEN 0
6285 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
6286 +#define au_fset_cpup(flags, name) \
6287 + do { (flags) |= AuCpup_##name; } while (0)
6288 +#define au_fclr_cpup(flags, name) \
6289 + do { (flags) &= ~AuCpup_##name; } while (0)
6291 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
6292 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
6293 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
6294 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
6296 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6297 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6298 + struct au_pin *pin,
6299 + struct dentry *h_parent, void *arg),
6301 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6302 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6304 +/* ---------------------------------------------------------------------- */
6306 +/* keep timestamps when copyup */
6308 + struct dentry *dt_dentry;
6309 + struct path dt_h_path;
6310 + struct timespec64 dt_atime, dt_mtime;
6312 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
6313 + struct path *h_path);
6314 +void au_dtime_revert(struct au_dtime *dt);
6316 +#endif /* __KERNEL__ */
6317 +#endif /* __AUFS_CPUP_H__ */
6318 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
6319 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
6320 +++ linux/fs/aufs/dbgaufs.c 2022-03-21 14:49:05.719966344 +0100
6322 +// SPDX-License-Identifier: GPL-2.0
6324 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6326 + * This program, aufs is free software; you can redistribute it and/or modify
6327 + * it under the terms of the GNU General Public License as published by
6328 + * the Free Software Foundation; either version 2 of the License, or
6329 + * (at your option) any later version.
6331 + * This program is distributed in the hope that it will be useful,
6332 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6333 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6334 + * GNU General Public License for more details.
6336 + * You should have received a copy of the GNU General Public License
6337 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6341 + * debugfs interface
6344 +#include <linux/debugfs.h>
6347 +#ifndef CONFIG_SYSFS
6348 +#error DEBUG_FS depends upon SYSFS
6351 +static struct dentry *dbgaufs;
6352 +static const mode_t dbgaufs_mode = 0444;
6354 +/* 20 is max digits length of ulong 64 */
6355 +struct dbgaufs_arg {
6361 + * common function for all XINO files
6363 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
6364 + struct file *file)
6368 + p = file->private_data;
6370 + /* this is struct dbgaufs_arg */
6371 + AuDebugOn(!au_kfree_sz_test(p));
6372 + au_kfree_do_rcu(p);
6377 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt,
6382 + struct dbgaufs_arg *p;
6385 + p = kmalloc(sizeof(*p), GFP_NOFS);
6391 + file->private_data = p;
6395 + err = vfsub_getattr(&xf->f_path, &st);
6399 + (p->a, sizeof(p->a), "%d, %llux%u %lld\n",
6400 + cnt, st.blocks, st.blksize,
6401 + (long long)st.size);
6403 + p->n = snprintf(p->a, sizeof(p->a), "%llux%u %lld\n",
6404 + st.blocks, st.blksize,
6405 + (long long)st.size);
6406 + AuDebugOn(p->n >= sizeof(p->a));
6408 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
6416 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
6417 + size_t count, loff_t *ppos)
6419 + struct dbgaufs_arg *p;
6421 + p = file->private_data;
6422 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6425 +/* ---------------------------------------------------------------------- */
6427 +struct dbgaufs_plink_arg {
6432 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6433 + struct file *file)
6435 + free_page((unsigned long)file->private_data);
6439 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6441 + int err, i, limit;
6442 + unsigned long n, sum;
6443 + struct dbgaufs_plink_arg *p;
6444 + struct au_sbinfo *sbinfo;
6445 + struct super_block *sb;
6446 + struct hlist_bl_head *hbl;
6449 + p = (void *)get_zeroed_page(GFP_NOFS);
6454 + sbinfo = inode->i_private;
6455 + sb = sbinfo->si_sb;
6456 + si_noflush_read_lock(sb);
6457 + if (au_opt_test(au_mntflags(sb), PLINK)) {
6458 + limit = PAGE_SIZE - sizeof(p->n);
6460 + /* the number of buckets */
6461 + n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6466 + for (i = 0, hbl = sbinfo->si_plink; i < AuPlink_NHASH;
6468 + n = au_hbl_count(hbl);
6471 + n = snprintf(p->a + p->n, limit, "%lu ", n);
6474 + if (unlikely(limit <= 0))
6477 + p->a[p->n - 1] = '\n';
6479 + /* the sum of plinks */
6480 + n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6483 + if (unlikely(limit <= 0))
6486 +#define str "1\n0\n0\n"
6487 + p->n = sizeof(str) - 1;
6488 + strcpy(p->a, str);
6491 + si_read_unlock(sb);
6494 + file->private_data = p;
6495 + goto out; /* success */
6498 + free_page((unsigned long)p);
6503 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6504 + size_t count, loff_t *ppos)
6506 + struct dbgaufs_plink_arg *p;
6508 + p = file->private_data;
6509 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6512 +static const struct file_operations dbgaufs_plink_fop = {
6513 + .owner = THIS_MODULE,
6514 + .open = dbgaufs_plink_open,
6515 + .release = dbgaufs_plink_release,
6516 + .read = dbgaufs_plink_read
6519 +/* ---------------------------------------------------------------------- */
6521 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6524 + struct au_sbinfo *sbinfo;
6525 + struct super_block *sb;
6527 + sbinfo = inode->i_private;
6528 + sb = sbinfo->si_sb;
6529 + si_noflush_read_lock(sb);
6530 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0, /*cnt*/0);
6531 + si_read_unlock(sb);
6535 +static const struct file_operations dbgaufs_xib_fop = {
6536 + .owner = THIS_MODULE,
6537 + .open = dbgaufs_xib_open,
6538 + .release = dbgaufs_xi_release,
6539 + .read = dbgaufs_xi_read
6542 +/* ---------------------------------------------------------------------- */
6544 +#define DbgaufsXi_PREFIX "xi"
6546 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6550 + aufs_bindex_t bindex;
6551 + char *p, a[sizeof(DbgaufsXi_PREFIX) + 8];
6552 + struct au_sbinfo *sbinfo;
6553 + struct super_block *sb;
6554 + struct au_xino *xi;
6556 + struct qstr *name;
6557 + struct au_branch *br;
6560 + name = &file->f_path.dentry->d_name;
6561 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6562 + || memcmp(name->name, DbgaufsXi_PREFIX,
6563 + sizeof(DbgaufsXi_PREFIX) - 1)))
6566 + AuDebugOn(name->len >= sizeof(a));
6567 + memcpy(a, name->name, name->len);
6568 + a[name->len] = '\0';
6569 + p = strchr(a, '-');
6572 + err = kstrtol(a + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6573 + if (unlikely(err))
6578 + err = kstrtol(p + 1, 10, &l);
6579 + if (unlikely(err))
6585 + sbinfo = inode->i_private;
6586 + sb = sbinfo->si_sb;
6587 + si_noflush_read_lock(sb);
6588 + if (unlikely(bindex < 0 || bindex > au_sbbot(sb)))
6590 + br = au_sbr(sb, bindex);
6592 + if (unlikely(idx >= xi->xi_nfile))
6594 + xf = au_xino_file(xi, idx);
6596 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1,
6597 + au_xino_count(br));
6600 + si_read_unlock(sb);
6606 +static const struct file_operations dbgaufs_xino_fop = {
6607 + .owner = THIS_MODULE,
6608 + .open = dbgaufs_xino_open,
6609 + .release = dbgaufs_xi_release,
6610 + .read = dbgaufs_xi_read
6613 +void dbgaufs_xino_del(struct au_branch *br)
6615 + struct dentry *dbgaufs;
6617 + dbgaufs = br->br_dbgaufs;
6621 + br->br_dbgaufs = NULL;
6622 + /* debugfs acquires the parent i_mutex */
6624 + debugfs_remove(dbgaufs);
6628 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6630 + aufs_bindex_t bbot;
6631 + struct au_branch *br;
6633 + if (!au_sbi(sb)->si_dbgaufs)
6636 + bbot = au_sbbot(sb);
6637 + for (; bindex <= bbot; bindex++) {
6638 + br = au_sbr(sb, bindex);
6639 + dbgaufs_xino_del(br);
6643 +static void dbgaufs_br_do_add(struct super_block *sb, aufs_bindex_t bindex,
6644 + unsigned int idx, struct dentry *parent,
6645 + struct au_sbinfo *sbinfo)
6647 + struct au_branch *br;
6649 + /* "xi" bindex(5) "-" idx(2) NULL */
6650 + char name[sizeof(DbgaufsXi_PREFIX) + 8];
6653 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6655 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d-%u",
6657 + br = au_sbr(sb, bindex);
6658 + if (br->br_dbgaufs) {
6659 + struct qstr qstr = QSTR_INIT(name, strlen(name));
6661 + if (!au_qstreq(&br->br_dbgaufs->d_name, &qstr)) {
6662 + /* debugfs acquires the parent i_mutex */
6664 + d = debugfs_rename(parent, br->br_dbgaufs, parent,
6668 + pr_warn("failed renaming %pd/%s, ignored.\n",
6673 + br->br_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6674 + sbinfo, &dbgaufs_xino_fop);
6676 + if (unlikely(!br->br_dbgaufs))
6677 + pr_warn("failed creating %pd/%s, ignored.\n",
6682 +static void dbgaufs_br_add(struct super_block *sb, aufs_bindex_t bindex,
6683 + struct dentry *parent, struct au_sbinfo *sbinfo)
6685 + struct au_branch *br;
6686 + struct au_xino *xi;
6689 + br = au_sbr(sb, bindex);
6691 + for (u = 0; u < xi->xi_nfile; u++)
6692 + dbgaufs_br_do_add(sb, bindex, u, parent, sbinfo);
6695 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown)
6697 + struct au_sbinfo *sbinfo;
6698 + struct dentry *parent;
6699 + aufs_bindex_t bbot;
6701 + if (!au_opt_test(au_mntflags(sb), XINO))
6704 + sbinfo = au_sbi(sb);
6705 + parent = sbinfo->si_dbgaufs;
6709 + bbot = au_sbbot(sb);
6711 + for (; bindex <= bbot; bindex++)
6712 + dbgaufs_br_add(sb, bindex, parent, sbinfo);
6714 + for (; bbot >= bindex; bbot--)
6715 + dbgaufs_br_add(sb, bbot, parent, sbinfo);
6718 +/* ---------------------------------------------------------------------- */
6720 +#ifdef CONFIG_AUFS_EXPORT
6721 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6724 + struct au_sbinfo *sbinfo;
6725 + struct super_block *sb;
6727 + sbinfo = inode->i_private;
6728 + sb = sbinfo->si_sb;
6729 + si_noflush_read_lock(sb);
6730 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0, /*cnt*/0);
6731 + si_read_unlock(sb);
6735 +static const struct file_operations dbgaufs_xigen_fop = {
6736 + .owner = THIS_MODULE,
6737 + .open = dbgaufs_xigen_open,
6738 + .release = dbgaufs_xi_release,
6739 + .read = dbgaufs_xi_read
6742 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6747 + * This function is a dynamic '__init' function actually,
6748 + * so the tiny check for si_rwsem is unnecessary.
6750 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6753 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
6754 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6755 + &dbgaufs_xigen_fop);
6756 + if (sbinfo->si_dbgaufs_xigen)
6762 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6766 +#endif /* CONFIG_AUFS_EXPORT */
6768 +/* ---------------------------------------------------------------------- */
6770 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6773 + * This function is a dynamic '__fin' function actually,
6774 + * so the tiny check for si_rwsem is unnecessary.
6776 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6778 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
6779 + sbinfo->si_dbgaufs = NULL;
6782 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6785 + char name[SysaufsSiNameLen];
6788 + * This function is a dynamic '__init' function actually,
6789 + * so the tiny check for si_rwsem is unnecessary.
6791 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6795 + AuErr1("/debug/aufs is uninitialized\n");
6800 + sysaufs_name(sbinfo, name);
6801 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6802 + if (unlikely(!sbinfo->si_dbgaufs))
6805 + /* regardless plink/noplink option */
6806 + sbinfo->si_dbgaufs_plink = debugfs_create_file
6807 + ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6808 + &dbgaufs_plink_fop);
6809 + if (unlikely(!sbinfo->si_dbgaufs_plink))
6812 + /* regardless xino/noxino option */
6813 + sbinfo->si_dbgaufs_xib = debugfs_create_file
6814 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6815 + &dbgaufs_xib_fop);
6816 + if (unlikely(!sbinfo->si_dbgaufs_xib))
6819 + err = dbgaufs_xigen_init(sbinfo);
6821 + goto out; /* success */
6824 + dbgaufs_si_fin(sbinfo);
6826 + if (unlikely(err))
6827 + pr_err("debugfs/aufs failed\n");
6831 +/* ---------------------------------------------------------------------- */
6833 +void dbgaufs_fin(void)
6835 + debugfs_remove(dbgaufs);
6838 +int __init dbgaufs_init(void)
6843 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6848 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6849 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
6850 +++ linux/fs/aufs/dbgaufs.h 2022-03-21 14:49:05.719966344 +0100
6852 +/* SPDX-License-Identifier: GPL-2.0 */
6854 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6856 + * This program, aufs is free software; you can redistribute it and/or modify
6857 + * it under the terms of the GNU General Public License as published by
6858 + * the Free Software Foundation; either version 2 of the License, or
6859 + * (at your option) any later version.
6861 + * This program is distributed in the hope that it will be useful,
6862 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6863 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6864 + * GNU General Public License for more details.
6866 + * You should have received a copy of the GNU General Public License
6867 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6871 + * debugfs interface
6874 +#ifndef __DBGAUFS_H__
6875 +#define __DBGAUFS_H__
6879 +struct super_block;
6883 +#ifdef CONFIG_DEBUG_FS
6885 +void dbgaufs_xino_del(struct au_branch *br);
6886 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6887 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown);
6888 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6889 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6890 +void dbgaufs_fin(void);
6891 +int __init dbgaufs_init(void);
6893 +AuStubVoid(dbgaufs_xino_del, struct au_branch *br)
6894 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6895 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex,
6897 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6898 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6899 +AuStubVoid(dbgaufs_fin, void)
6900 +AuStubInt0(__init dbgaufs_init, void)
6901 +#endif /* CONFIG_DEBUG_FS */
6903 +#endif /* __KERNEL__ */
6904 +#endif /* __DBGAUFS_H__ */
6905 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6906 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
6907 +++ linux/fs/aufs/dcsub.c 2022-03-21 14:49:05.719966344 +0100
6909 +// SPDX-License-Identifier: GPL-2.0
6911 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6913 + * This program, aufs is free software; you can redistribute it and/or modify
6914 + * it under the terms of the GNU General Public License as published by
6915 + * the Free Software Foundation; either version 2 of the License, or
6916 + * (at your option) any later version.
6918 + * This program is distributed in the hope that it will be useful,
6919 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6920 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6921 + * GNU General Public License for more details.
6923 + * You should have received a copy of the GNU General Public License
6924 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6928 + * sub-routines for dentry cache
6933 +static void au_dpage_free(struct au_dpage *dpage)
6936 + struct dentry **p;
6938 + p = dpage->dentries;
6939 + for (i = 0; i < dpage->ndentry; i++)
6941 + free_page((unsigned long)dpage->dentries);
6944 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6950 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6951 + if (unlikely(!dpages->dpages))
6954 + p = (void *)__get_free_page(gfp);
6958 + dpages->dpages[0].ndentry = 0;
6959 + dpages->dpages[0].dentries = p;
6960 + dpages->ndpage = 1;
6961 + return 0; /* success */
6964 + au_kfree_try_rcu(dpages->dpages);
6969 +void au_dpages_free(struct au_dcsub_pages *dpages)
6972 + struct au_dpage *p;
6974 + p = dpages->dpages;
6975 + for (i = 0; i < dpages->ndpage; i++)
6976 + au_dpage_free(p++);
6977 + au_kfree_try_rcu(dpages->dpages);
6980 +static int au_dpages_append(struct au_dcsub_pages *dpages,
6981 + struct dentry *dentry, gfp_t gfp)
6984 + struct au_dpage *dpage;
6987 + dpage = dpages->dpages + dpages->ndpage - 1;
6988 + sz = PAGE_SIZE / sizeof(dentry);
6989 + if (unlikely(dpage->ndentry >= sz)) {
6990 + AuLabel(new dpage);
6992 + sz = dpages->ndpage * sizeof(*dpages->dpages);
6993 + p = au_kzrealloc(dpages->dpages, sz,
6994 + sz + sizeof(*dpages->dpages), gfp,
6999 + dpages->dpages = p;
7000 + dpage = dpages->dpages + dpages->ndpage;
7001 + p = (void *)__get_free_page(gfp);
7005 + dpage->ndentry = 0;
7006 + dpage->dentries = p;
7010 + AuDebugOn(au_dcount(dentry) <= 0);
7011 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
7012 + return 0; /* success */
7018 +/* todo: BAD approach */
7019 +/* copied from linux/fs/dcache.c */
7027 +extern void d_walk(struct dentry *parent, void *data,
7028 + enum d_walk_ret (*enter)(void *, struct dentry *));
7030 +struct ac_dpages_arg {
7032 + struct au_dcsub_pages *dpages;
7033 + struct super_block *sb;
7034 + au_dpages_test test;
7038 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
7040 + enum d_walk_ret ret;
7041 + struct ac_dpages_arg *arg = _arg;
7043 + ret = D_WALK_CONTINUE;
7044 + if (dentry->d_sb == arg->sb
7045 + && !IS_ROOT(dentry)
7046 + && au_dcount(dentry) > 0
7048 + && (!arg->test || arg->test(dentry, arg->arg))) {
7049 + arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
7050 + if (unlikely(arg->err))
7051 + ret = D_WALK_QUIT;
7057 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7058 + au_dpages_test test, void *arg)
7060 + struct ac_dpages_arg args = {
7068 + d_walk(root, &args, au_call_dpages_append);
7073 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7074 + int do_include, au_dpages_test test, void *arg)
7079 + write_seqlock(&rename_lock);
7080 + spin_lock(&dentry->d_lock);
7082 + && au_dcount(dentry) > 0
7083 + && (!test || test(dentry, arg)))
7084 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7085 + spin_unlock(&dentry->d_lock);
7086 + if (unlikely(err))
7090 + * RCU for vfsmount is unnecessary since this is a traverse in a single
7093 + while (!IS_ROOT(dentry)) {
7094 + dentry = dentry->d_parent; /* rename_lock is locked */
7095 + spin_lock(&dentry->d_lock);
7096 + if (au_dcount(dentry) > 0
7097 + && (!test || test(dentry, arg)))
7098 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7099 + spin_unlock(&dentry->d_lock);
7100 + if (unlikely(err))
7105 + write_sequnlock(&rename_lock);
7109 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
7111 + return au_di(dentry) && dentry->d_sb == arg;
7114 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7115 + struct dentry *dentry, int do_include)
7117 + return au_dcsub_pages_rev(dpages, dentry, do_include,
7118 + au_dcsub_dpages_aufs, dentry->d_sb);
7121 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
7123 + struct path path[2] = {
7132 + return path_is_under(path + 0, path + 1);
7134 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
7135 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
7136 +++ linux/fs/aufs/dcsub.h 2022-03-21 14:49:05.719966344 +0100
7138 +/* SPDX-License-Identifier: GPL-2.0 */
7140 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7142 + * This program, aufs is free software; you can redistribute it and/or modify
7143 + * it under the terms of the GNU General Public License as published by
7144 + * the Free Software Foundation; either version 2 of the License, or
7145 + * (at your option) any later version.
7147 + * This program is distributed in the hope that it will be useful,
7148 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7149 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7150 + * GNU General Public License for more details.
7152 + * You should have received a copy of the GNU General Public License
7153 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7157 + * sub-routines for dentry cache
7160 +#ifndef __AUFS_DCSUB_H__
7161 +#define __AUFS_DCSUB_H__
7165 +#include <linux/dcache.h>
7166 +#include <linux/fs.h>
7170 + struct dentry **dentries;
7173 +struct au_dcsub_pages {
7175 + struct au_dpage *dpages;
7178 +/* ---------------------------------------------------------------------- */
7181 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
7182 +void au_dpages_free(struct au_dcsub_pages *dpages);
7183 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
7184 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7185 + au_dpages_test test, void *arg);
7186 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7187 + int do_include, au_dpages_test test, void *arg);
7188 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7189 + struct dentry *dentry, int do_include);
7190 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
7192 +/* ---------------------------------------------------------------------- */
7195 + * todo: in linux-3.13, several similar (but faster) helpers are added to
7196 + * include/linux/dcache.h. Try them (in the future).
7199 +static inline int au_d_hashed_positive(struct dentry *d)
7202 + struct inode *inode = d_inode(d);
7205 + if (unlikely(d_unhashed(d)
7206 + || d_is_negative(d)
7207 + || !inode->i_nlink))
7212 +static inline int au_d_linkable(struct dentry *d)
7215 + struct inode *inode = d_inode(d);
7217 + err = au_d_hashed_positive(d);
7219 + && d_is_positive(d)
7220 + && (inode->i_state & I_LINKABLE))
7225 +static inline int au_d_alive(struct dentry *d)
7228 + struct inode *inode;
7232 + err = au_d_hashed_positive(d);
7234 + inode = d_inode(d);
7235 + if (unlikely(d_unlinked(d)
7236 + || d_is_negative(d)
7237 + || !inode->i_nlink))
7243 +static inline int au_alive_dir(struct dentry *d)
7247 + err = au_d_alive(d);
7248 + if (unlikely(err || IS_DEADDIR(d_inode(d))))
7253 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
7255 + return a->len == b->len
7256 + && !memcmp(a->name, b->name, a->len);
7261 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
7263 + * the type of d_lockref.count became int, but the inlined function d_count()
7264 + * still returns unsigned int.
7265 + * I don't know why. Maybe it is for every d_count() users?
7266 + * Anyway au_dcount() lives on.
7268 +static inline int au_dcount(struct dentry *d)
7270 + return (int)d_count(d);
7273 +#endif /* __KERNEL__ */
7274 +#endif /* __AUFS_DCSUB_H__ */
7275 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
7276 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
7277 +++ linux/fs/aufs/debug.c 2022-10-03 07:51:14.926209260 +0200
7279 +// SPDX-License-Identifier: GPL-2.0
7281 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7283 + * This program, aufs is free software; you can redistribute it and/or modify
7284 + * it under the terms of the GNU General Public License as published by
7285 + * the Free Software Foundation; either version 2 of the License, or
7286 + * (at your option) any later version.
7288 + * This program is distributed in the hope that it will be useful,
7289 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7290 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7291 + * GNU General Public License for more details.
7293 + * You should have received a copy of the GNU General Public License
7294 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7298 + * debug print functions
7301 +#include <linux/iversion.h>
7304 +/* Returns 0, or -errno. arg is in kp->arg. */
7305 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
7309 + err = kstrtoint(val, 0, &n);
7319 +/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
7320 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
7325 + return sprintf(buffer, "%d", atomic_read(a));
7328 +static const struct kernel_param_ops param_ops_atomic_t = {
7329 + .set = param_atomic_t_set,
7330 + .get = param_atomic_t_get
7331 + /* void (*free)(void *arg) */
7334 +atomic_t aufs_debug = ATOMIC_INIT(0);
7335 +MODULE_PARM_DESC(debug, "debug print");
7336 +module_param_named(debug, aufs_debug, atomic_t, 0664);
7338 +DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
7339 +char *au_plevel = KERN_DEBUG;
7340 +#define dpri(fmt, ...) do { \
7342 + && strcmp(au_plevel, KERN_DEBUG)) \
7343 + || au_debug_test()) \
7344 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
7347 +/* ---------------------------------------------------------------------- */
7349 +void au_dpri_whlist(struct au_nhash *whlist)
7351 + unsigned long ul, n;
7352 + struct hlist_head *head;
7353 + struct au_vdir_wh *pos;
7355 + n = whlist->nh_num;
7356 + head = whlist->nh_head;
7357 + for (ul = 0; ul < n; ul++) {
7358 + hlist_for_each_entry(pos, head, wh_hash)
7359 + dpri("b%d, %.*s, %d\n",
7361 + pos->wh_str.len, pos->wh_str.name,
7367 +void au_dpri_vdir(struct au_vdir *vdir)
7370 + union au_vdir_deblk_p p;
7373 + if (!vdir || IS_ERR(vdir)) {
7374 + dpri("err %ld\n", PTR_ERR(vdir));
7378 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %llu\n",
7379 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
7380 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
7381 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
7382 + p.deblk = vdir->vd_deblk[ul];
7384 + dpri("[%lu]: %p\n", ul, o);
7388 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
7389 + struct dentry *wh)
7394 + if (!inode || IS_ERR(inode)) {
7395 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
7399 + /* the type of i_blocks depends upon CONFIG_LBDAF */
7400 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
7401 + && sizeof(inode->i_blocks) != sizeof(u64));
7403 + n = (void *)wh->d_name.name;
7404 + l = wh->d_name.len;
7407 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
7408 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
7410 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
7411 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
7412 + i_size_read(inode), (unsigned long long)inode->i_blocks,
7413 + hn, (long long)timespec64_to_ns(&inode->i_ctime) & 0x0ffff,
7414 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
7415 + inode->i_state, inode->i_flags, inode_peek_iversion(inode),
7416 + inode->i_generation,
7417 + l ? ", wh " : "", l, n);
7421 +void au_dpri_inode(struct inode *inode)
7423 + struct au_iinfo *iinfo;
7424 + struct au_hinode *hi;
7425 + aufs_bindex_t bindex;
7428 + err = do_pri_inode(-1, inode, -1, NULL);
7429 + if (err || !au_test_aufs(inode->i_sb) || au_is_bad_inode(inode))
7432 + iinfo = au_ii(inode);
7433 + dpri("i-1: btop %d, bbot %d, gen %d\n",
7434 + iinfo->ii_btop, iinfo->ii_bbot, au_iigen(inode, NULL));
7435 + if (iinfo->ii_btop < 0)
7438 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot; bindex++) {
7439 + hi = au_hinode(iinfo, bindex);
7441 + do_pri_inode(bindex, hi->hi_inode, hn, hi->hi_whdentry);
7445 +void au_dpri_dalias(struct inode *inode)
7449 + spin_lock(&inode->i_lock);
7450 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
7451 + au_dpri_dentry(d);
7452 + spin_unlock(&inode->i_lock);
7455 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
7457 + struct dentry *wh = NULL;
7459 + struct inode *inode;
7460 + struct au_iinfo *iinfo;
7461 + struct au_hinode *hi;
7463 + if (!dentry || IS_ERR(dentry)) {
7464 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
7467 + /* do not call dget_parent() here */
7468 + /* note: access d_xxx without d_lock */
7469 + dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
7470 + bindex, dentry, dentry,
7471 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
7472 + au_dcount(dentry), dentry->d_flags,
7473 + d_unhashed(dentry) ? "un" : "");
7476 + if (d_is_positive(dentry))
7477 + inode = d_inode(dentry);
7479 + && au_test_aufs(dentry->d_sb)
7481 + && !au_is_bad_inode(inode)) {
7482 + iinfo = au_ii(inode);
7483 + hi = au_hinode(iinfo, bindex);
7485 + wh = hi->hi_whdentry;
7487 + do_pri_inode(bindex, inode, hn, wh);
7491 +void au_dpri_dentry(struct dentry *dentry)
7493 + struct au_dinfo *dinfo;
7494 + aufs_bindex_t bindex;
7497 + err = do_pri_dentry(-1, dentry);
7498 + if (err || !au_test_aufs(dentry->d_sb))
7501 + dinfo = au_di(dentry);
7504 + dpri("d-1: btop %d, bbot %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
7505 + dinfo->di_btop, dinfo->di_bbot,
7506 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
7507 + dinfo->di_tmpfile);
7508 + if (dinfo->di_btop < 0)
7510 + for (bindex = dinfo->di_btop; bindex <= dinfo->di_bbot; bindex++)
7511 + do_pri_dentry(bindex, au_hdentry(dinfo, bindex)->hd_dentry);
7514 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7518 + if (!file || IS_ERR(file)) {
7519 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7524 + && !IS_ERR_OR_NULL(file->f_path.dentry)
7525 + && au_test_aufs(file->f_path.dentry->d_sb)
7527 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7528 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7529 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7530 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
7531 + file->f_version, file->f_pos, a);
7532 + if (!IS_ERR_OR_NULL(file->f_path.dentry))
7533 + do_pri_dentry(bindex, file->f_path.dentry);
7537 +void au_dpri_file(struct file *file)
7539 + struct au_finfo *finfo;
7540 + struct au_fidir *fidir;
7541 + struct au_hfile *hfile;
7542 + aufs_bindex_t bindex;
7545 + err = do_pri_file(-1, file);
7547 + || IS_ERR_OR_NULL(file->f_path.dentry)
7548 + || !au_test_aufs(file->f_path.dentry->d_sb))
7551 + finfo = au_fi(file);
7554 + if (finfo->fi_btop < 0)
7556 + fidir = finfo->fi_hdir;
7558 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7560 + for (bindex = finfo->fi_btop;
7561 + bindex >= 0 && bindex <= fidir->fd_bbot;
7563 + hfile = fidir->fd_hfile + bindex;
7564 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7568 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7570 + struct vfsmount *mnt;
7571 + struct super_block *sb;
7573 + if (!br || IS_ERR(br))
7575 + mnt = au_br_mnt(br);
7576 + if (!mnt || IS_ERR(mnt))
7579 + if (!sb || IS_ERR(sb))
7582 + dpri("s%d: {perm 0x%x, id %d, wbr %p}, "
7583 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7585 + bindex, br->br_perm, br->br_id, br->br_wbr,
7586 + au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7587 + sb->s_flags, sb->s_count,
7588 + atomic_read(&sb->s_active),
7589 + !!au_xino_file(br->br_xino, /*idx*/-1));
7593 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7597 +void au_dpri_sb(struct super_block *sb)
7599 + struct au_sbinfo *sbinfo;
7600 + aufs_bindex_t bindex;
7602 + /* to reduce stack size */
7604 + struct vfsmount mnt;
7605 + struct au_branch fake;
7608 + /* this function can be called from magic sysrq */
7609 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
7610 + if (unlikely(!a)) {
7611 + dpri("no memory\n");
7615 + a->mnt.mnt_sb = sb;
7616 + a->fake.br_path.mnt = &a->mnt;
7617 + err = do_pri_br(-1, &a->fake);
7619 + dpri("dev 0x%x\n", sb->s_dev);
7620 + if (err || !au_test_aufs(sb))
7623 + sbinfo = au_sbi(sb);
7626 + dpri("nw %d, gen %u, kobj %d\n",
7627 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7628 + kref_read(&sbinfo->si_kobj.kref));
7629 + for (bindex = 0; bindex <= sbinfo->si_bbot; bindex++)
7630 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7633 +/* ---------------------------------------------------------------------- */
7635 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7637 + struct inode *h_inode, *inode = d_inode(dentry);
7638 + struct dentry *h_dentry;
7639 + aufs_bindex_t bindex, bbot, bi;
7641 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7644 + bbot = au_dbbot(dentry);
7645 + bi = au_ibbot(inode);
7648 + bindex = au_dbtop(dentry);
7649 + bi = au_ibtop(inode);
7653 + for (; bindex <= bbot; bindex++) {
7654 + h_dentry = au_h_dptr(dentry, bindex);
7657 + h_inode = au_h_iptr(inode, bindex);
7658 + if (unlikely(h_inode != d_inode(h_dentry))) {
7660 + AuDbg("b%d, %s:%d\n", bindex, func, line);
7661 + AuDbgDentry(dentry);
7662 + AuDbgInode(inode);
7664 + if (au_test_fuse(h_inode->i_sb))
7672 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7675 + struct au_dcsub_pages dpages;
7676 + struct au_dpage *dpage;
7677 + struct dentry **dentries;
7679 + err = au_dpages_init(&dpages, GFP_NOFS);
7681 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7683 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7684 + dpage = dpages.dpages + i;
7685 + dentries = dpage->dentries;
7686 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7687 + AuDebugOn(au_digen_test(dentries[j], sigen));
7689 + au_dpages_free(&dpages);
7692 +void au_dbg_verify_kthread(void)
7694 + if (au_wkq_test()) {
7697 + * It may be recursive, but udba=notify between two aufs mounts,
7698 + * where a single ro branch is shared, is not a problem.
7704 +/* ---------------------------------------------------------------------- */
7706 +int __init au_debug_init(void)
7708 + aufs_bindex_t bindex;
7709 + struct au_vdir_destr destr;
7712 + AuDebugOn(bindex >= 0);
7715 + AuDebugOn(destr.len < NAME_MAX);
7717 +#ifdef CONFIG_4KSTACKS
7718 + pr_warn("CONFIG_4KSTACKS is defined.\n");
7723 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7724 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
7725 +++ linux/fs/aufs/debug.h 2022-03-21 14:49:05.719966344 +0100
7727 +/* SPDX-License-Identifier: GPL-2.0 */
7729 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7731 + * This program, aufs is free software; you can redistribute it and/or modify
7732 + * it under the terms of the GNU General Public License as published by
7733 + * the Free Software Foundation; either version 2 of the License, or
7734 + * (at your option) any later version.
7736 + * This program is distributed in the hope that it will be useful,
7737 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7738 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7739 + * GNU General Public License for more details.
7741 + * You should have received a copy of the GNU General Public License
7742 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7746 + * debug print functions
7749 +#ifndef __AUFS_DEBUG_H__
7750 +#define __AUFS_DEBUG_H__
7754 +#include <linux/atomic.h>
7755 +#include <linux/module.h>
7756 +#include <linux/kallsyms.h>
7757 +#include <linux/sysrq.h>
7759 +#ifdef CONFIG_AUFS_DEBUG
7760 +#define AuDebugOn(a) BUG_ON(a)
7762 +/* module parameter */
7763 +extern atomic_t aufs_debug;
7764 +static inline void au_debug_on(void)
7766 + atomic_inc(&aufs_debug);
7768 +static inline void au_debug_off(void)
7770 + atomic_dec_if_positive(&aufs_debug);
7773 +static inline int au_debug_test(void)
7775 + return atomic_read(&aufs_debug) > 0;
7778 +#define AuDebugOn(a) do {} while (0)
7779 +AuStubVoid(au_debug_on, void)
7780 +AuStubVoid(au_debug_off, void)
7781 +AuStubInt0(au_debug_test, void)
7782 +#endif /* CONFIG_AUFS_DEBUG */
7784 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7786 +/* ---------------------------------------------------------------------- */
7790 +#define AuDbg(fmt, ...) do { \
7791 + if (au_debug_test()) \
7792 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7794 +#define AuLabel(l) AuDbg(#l "\n")
7795 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7796 +#define AuWarn1(fmt, ...) do { \
7797 + static unsigned char _c; \
7799 + pr_warn(fmt, ##__VA_ARGS__); \
7802 +#define AuErr1(fmt, ...) do { \
7803 + static unsigned char _c; \
7805 + pr_err(fmt, ##__VA_ARGS__); \
7808 +#define AuIOErr1(fmt, ...) do { \
7809 + static unsigned char _c; \
7811 + AuIOErr(fmt, ##__VA_ARGS__); \
7814 +#define AuUnsupportMsg "This operation is not supported." \
7815 + " Please report this application to aufs-users ML."
7816 +#define AuUnsupport(fmt, ...) do { \
7817 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7821 +#define AuTraceErr(e) do { \
7822 + if (unlikely((e) < 0)) \
7823 + AuDbg("err %d\n", (int)(e)); \
7826 +#define AuTraceErrPtr(p) do { \
7828 + AuDbg("err %ld\n", PTR_ERR(p)); \
7831 +/* dirty macros for debug print, use with "%.*s" and caution */
7832 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
7834 +/* ---------------------------------------------------------------------- */
7837 +#ifdef CONFIG_AUFS_DEBUG
7838 +extern struct mutex au_dbg_mtx;
7839 +extern char *au_plevel;
7841 +void au_dpri_whlist(struct au_nhash *whlist);
7843 +void au_dpri_vdir(struct au_vdir *vdir);
7845 +void au_dpri_inode(struct inode *inode);
7846 +void au_dpri_dalias(struct inode *inode);
7847 +void au_dpri_dentry(struct dentry *dentry);
7849 +void au_dpri_file(struct file *filp);
7850 +struct super_block;
7851 +void au_dpri_sb(struct super_block *sb);
7853 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7854 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7855 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7856 +void au_dbg_verify_kthread(void);
7858 +int __init au_debug_init(void);
7860 +#define AuDbgWhlist(w) do { \
7861 + mutex_lock(&au_dbg_mtx); \
7863 + au_dpri_whlist(w); \
7864 + mutex_unlock(&au_dbg_mtx); \
7867 +#define AuDbgVdir(v) do { \
7868 + mutex_lock(&au_dbg_mtx); \
7870 + au_dpri_vdir(v); \
7871 + mutex_unlock(&au_dbg_mtx); \
7874 +#define AuDbgInode(i) do { \
7875 + mutex_lock(&au_dbg_mtx); \
7877 + au_dpri_inode(i); \
7878 + mutex_unlock(&au_dbg_mtx); \
7881 +#define AuDbgDAlias(i) do { \
7882 + mutex_lock(&au_dbg_mtx); \
7884 + au_dpri_dalias(i); \
7885 + mutex_unlock(&au_dbg_mtx); \
7888 +#define AuDbgDentry(d) do { \
7889 + mutex_lock(&au_dbg_mtx); \
7891 + au_dpri_dentry(d); \
7892 + mutex_unlock(&au_dbg_mtx); \
7895 +#define AuDbgFile(f) do { \
7896 + mutex_lock(&au_dbg_mtx); \
7898 + au_dpri_file(f); \
7899 + mutex_unlock(&au_dbg_mtx); \
7902 +#define AuDbgSb(sb) do { \
7903 + mutex_lock(&au_dbg_mtx); \
7904 + AuDbg(#sb "\n"); \
7906 + mutex_unlock(&au_dbg_mtx); \
7909 +#define AuDbgSym(addr) do { \
7910 + char sym[KSYM_SYMBOL_LEN]; \
7911 + sprint_symbol(sym, (unsigned long)addr); \
7912 + AuDbg("%s\n", sym); \
7915 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7916 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7917 +AuStubVoid(au_dbg_verify_kthread, void)
7918 +AuStubInt0(__init au_debug_init, void)
7920 +#define AuDbgWhlist(w) do {} while (0)
7921 +#define AuDbgVdir(v) do {} while (0)
7922 +#define AuDbgInode(i) do {} while (0)
7923 +#define AuDbgDAlias(i) do {} while (0)
7924 +#define AuDbgDentry(d) do {} while (0)
7925 +#define AuDbgFile(f) do {} while (0)
7926 +#define AuDbgSb(sb) do {} while (0)
7927 +#define AuDbgSym(addr) do {} while (0)
7928 +#endif /* CONFIG_AUFS_DEBUG */
7930 +/* ---------------------------------------------------------------------- */
7932 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7933 +int __init au_sysrq_init(void);
7934 +void au_sysrq_fin(void);
7936 +#ifdef CONFIG_HW_CONSOLE
7937 +#define au_dbg_blocked() do { \
7939 + handle_sysrq('w'); \
7942 +AuStubVoid(au_dbg_blocked, void)
7946 +AuStubInt0(__init au_sysrq_init, void)
7947 +AuStubVoid(au_sysrq_fin, void)
7948 +AuStubVoid(au_dbg_blocked, void)
7949 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7951 +#endif /* __KERNEL__ */
7952 +#endif /* __AUFS_DEBUG_H__ */
7953 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7954 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
7955 +++ linux/fs/aufs/dentry.c 2022-03-21 14:49:05.719966344 +0100
7957 +// SPDX-License-Identifier: GPL-2.0
7959 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7961 + * This program, aufs is free software; you can redistribute it and/or modify
7962 + * it under the terms of the GNU General Public License as published by
7963 + * the Free Software Foundation; either version 2 of the License, or
7964 + * (at your option) any later version.
7966 + * This program is distributed in the hope that it will be useful,
7967 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7968 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7969 + * GNU General Public License for more details.
7971 + * You should have received a copy of the GNU General Public License
7972 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7976 + * lookup and dentry operations
7979 +#include <linux/iversion.h>
7983 + * returns positive/negative dentry, NULL or an error.
7984 + * NULL means whiteout-ed or not-found.
7986 +static struct dentry*
7987 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7988 + aufs_bindex_t bindex, struct au_do_lookup_args *args)
7990 + struct dentry *h_dentry;
7991 + struct inode *h_inode;
7992 + struct au_branch *br;
7993 + struct user_namespace *h_userns;
7994 + struct path h_path;
7995 + int wh_found, opq;
7996 + unsigned char wh_able;
7997 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
7998 + const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
8002 + br = au_sbr(dentry->d_sb, bindex);
8003 + h_path.dentry = h_parent;
8004 + h_path.mnt = au_br_mnt(br);
8005 + h_userns = au_br_userns(br);
8006 + wh_able = !!au_br_whable(br->br_perm);
8008 + wh_found = au_wh_test(h_userns, &h_path, &args->whname,
8010 + h_dentry = ERR_PTR(wh_found);
8013 + if (unlikely(wh_found < 0))
8016 + /* We found a whiteout */
8017 + /* au_set_dbbot(dentry, bindex); */
8018 + au_set_dbwh(dentry, bindex);
8020 + return NULL; /* success */
8024 + h_dentry = vfsub_lkup_one(args->name, &h_path);
8026 + h_dentry = au_sio_lkup_one(h_userns, args->name, &h_path);
8027 + if (IS_ERR(h_dentry)) {
8028 + if (PTR_ERR(h_dentry) == -ENAMETOOLONG
8034 + h_inode = d_inode(h_dentry);
8035 + if (d_is_negative(h_dentry)) {
8038 + } else if (wh_found
8039 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
8041 + else if (au_ftest_lkup(args->flags, DIRREN)
8043 + && !au_dr_lkup_h_ino(args, bindex, h_inode->i_ino)) {
8044 + AuDbg("b%d %pd ignored hi%llu\n", bindex, h_dentry,
8045 + (unsigned long long)h_inode->i_ino);
8049 + if (au_dbbot(dentry) <= bindex)
8050 + au_set_dbbot(dentry, bindex);
8051 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
8052 + au_set_dbtop(dentry, bindex);
8053 + au_set_h_dptr(dentry, bindex, h_dentry);
8055 + if (!d_is_dir(h_dentry)
8057 + || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
8058 + goto out; /* success */
8060 + h_path.dentry = h_dentry;
8061 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
8062 + opq = au_diropq_test(h_userns, &h_path);
8063 + inode_unlock_shared(h_inode);
8065 + au_set_dbdiropq(dentry, bindex);
8066 + else if (unlikely(opq < 0)) {
8067 + au_set_h_dptr(dentry, bindex, NULL);
8068 + h_dentry = ERR_PTR(opq);
8079 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
8081 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
8082 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
8088 + * returns the number of lower positive dentries,
8089 + * otherwise an error.
8090 + * can be called at unlinking with @type is zero.
8092 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
8093 + unsigned int flags)
8095 + int npositive, err;
8096 + aufs_bindex_t bindex, btail, bdiropq;
8097 + unsigned char isdir, dirperm1, dirren;
8098 + struct au_do_lookup_args args = {
8100 + .name = &dentry->d_name
8102 + struct dentry *parent;
8103 + struct super_block *sb;
8105 + sb = dentry->d_sb;
8106 + err = au_test_shwh(sb, args.name);
8107 + if (unlikely(err))
8110 + err = au_wh_name_alloc(&args.whname, args.name);
8111 + if (unlikely(err))
8114 + isdir = !!d_is_dir(dentry);
8115 + dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
8116 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
8118 + au_fset_lkup(args.flags, DIRREN);
8121 + parent = dget_parent(dentry);
8122 + btail = au_dbtaildir(parent);
8123 + for (bindex = btop; bindex <= btail; bindex++) {
8124 + struct dentry *h_parent, *h_dentry;
8125 + struct inode *h_inode, *h_dir;
8126 + struct au_branch *br;
8128 + h_dentry = au_h_dptr(dentry, bindex);
8130 + if (d_is_positive(h_dentry))
8134 + h_parent = au_h_dptr(parent, bindex);
8135 + if (!h_parent || !d_is_dir(h_parent))
8139 + /* if the inum matches, then use the prepared name */
8140 + err = au_dr_lkup_name(&args, bindex);
8141 + if (unlikely(err))
8145 + h_dir = d_inode(h_parent);
8146 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
8147 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &args);
8148 + inode_unlock_shared(h_dir);
8149 + err = PTR_ERR(h_dentry);
8150 + if (IS_ERR(h_dentry))
8153 + au_fclr_lkup(args.flags, ALLOW_NEG);
8155 + au_fset_lkup(args.flags, IGNORE_PERM);
8157 + if (au_dbwh(dentry) == bindex)
8161 + if (d_is_negative(h_dentry))
8163 + h_inode = d_inode(h_dentry);
8166 + args.type = h_inode->i_mode & S_IFMT;
8167 + if (args.type != S_IFDIR)
8170 + /* the type of lower may be different */
8171 + bdiropq = au_dbdiropq(dentry);
8172 + if (bdiropq >= 0 && bdiropq <= bindex)
8175 + br = au_sbr(sb, bindex);
8177 + && au_dr_hino_test_add(&br->br_dirren, h_inode->i_ino,
8178 + /*add_ent*/NULL)) {
8179 + /* prepare next name to lookup */
8180 + err = au_dr_lkup(&args, dentry, bindex);
8181 + if (unlikely(err))
8187 + AuLabel(positive);
8188 + au_update_dbtop(dentry);
8191 + if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
8192 + && au_dbtop(dentry) < 0)) {
8194 + AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
8200 + au_kfree_try_rcu(args.whname.name);
8202 + au_dr_lkup_fin(&args);
8207 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
8208 + struct path *ppath)
8210 + struct dentry *dentry;
8213 + if (!au_test_h_perm_sio(userns, d_inode(ppath->dentry), MAY_EXEC))
8214 + dentry = vfsub_lkup_one(name, ppath);
8216 + struct vfsub_lkup_one_args args = {
8222 + wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
8223 + if (unlikely(wkq_err))
8224 + dentry = ERR_PTR(wkq_err);
8231 + * lookup @dentry on @bindex which should be negative.
8233 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
8236 + struct dentry *parent, *h_dentry;
8237 + struct au_branch *br;
8238 + struct user_namespace *h_userns;
8239 + struct path h_ppath;
8241 + parent = dget_parent(dentry);
8242 + br = au_sbr(dentry->d_sb, bindex);
8243 + h_ppath.dentry = au_h_dptr(parent, bindex);
8244 + h_ppath.mnt = au_br_mnt(br);
8245 + h_userns = au_br_userns(br);
8247 + h_dentry = au_whtmp_lkup(h_ppath.dentry, br, &dentry->d_name);
8249 + h_dentry = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
8250 + err = PTR_ERR(h_dentry);
8251 + if (IS_ERR(h_dentry))
8253 + if (unlikely(d_is_positive(h_dentry))) {
8255 + AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
8261 + if (bindex < au_dbtop(dentry))
8262 + au_set_dbtop(dentry, bindex);
8263 + if (au_dbbot(dentry) < bindex)
8264 + au_set_dbbot(dentry, bindex);
8265 + au_set_h_dptr(dentry, bindex, h_dentry);
8272 +/* ---------------------------------------------------------------------- */
8274 +/* subset of struct inode */
8276 + unsigned long i_ino;
8277 + /* unsigned int i_nlink; */
8283 + blkcnt_t i_blocks;
8288 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
8290 + ia->i_ino = h_inode->i_ino;
8291 + /* ia->i_nlink = h_inode->i_nlink; */
8292 + ia->i_uid = h_inode->i_uid;
8293 + ia->i_gid = h_inode->i_gid;
8294 + ia->i_version = inode_query_iversion(h_inode);
8296 + ia->i_size = h_inode->i_size;
8297 + ia->i_blocks = h_inode->i_blocks;
8299 + ia->i_mode = (h_inode->i_mode & S_IFMT);
8302 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
8304 + return ia->i_ino != h_inode->i_ino
8305 + /* || ia->i_nlink != h_inode->i_nlink */
8306 + || !uid_eq(ia->i_uid, h_inode->i_uid)
8307 + || !gid_eq(ia->i_gid, h_inode->i_gid)
8308 + || !inode_eq_iversion(h_inode, ia->i_version)
8310 + || ia->i_size != h_inode->i_size
8311 + || ia->i_blocks != h_inode->i_blocks
8313 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
8316 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
8317 + struct au_branch *br)
8320 + struct au_iattr ia;
8321 + struct inode *h_inode;
8322 + struct dentry *h_d;
8323 + struct super_block *h_sb;
8324 + struct path h_ppath;
8327 + memset(&ia, -1, sizeof(ia));
8328 + h_sb = h_dentry->d_sb;
8330 + if (d_is_positive(h_dentry)) {
8331 + h_inode = d_inode(h_dentry);
8332 + au_iattr_save(&ia, h_inode);
8333 + } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
8334 + /* nfs d_revalidate may return 0 for negative dentry */
8335 + /* fuse d_revalidate always return 0 for negative dentry */
8338 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
8339 + h_ppath.dentry = h_parent;
8340 + h_ppath.mnt = au_br_mnt(br);
8341 + h_d = vfsub_lkup_one(&h_dentry->d_name, &h_ppath);
8342 + err = PTR_ERR(h_d);
8347 + if (unlikely(h_d != h_dentry
8348 + || d_inode(h_d) != h_inode
8349 + || (h_inode && au_iattr_test(&ia, h_inode))))
8350 + err = au_busy_or_stale();
8358 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8359 + struct dentry *h_parent, struct au_branch *br)
8364 + if (udba == AuOpt_UDBA_REVAL
8365 + && !au_test_fs_remote(h_dentry->d_sb)) {
8367 + err = (d_inode(h_dentry->d_parent) != h_dir);
8368 + } else if (udba != AuOpt_UDBA_NONE)
8369 + err = au_h_verify_dentry(h_dentry, h_parent, br);
8374 +/* ---------------------------------------------------------------------- */
8376 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
8379 + aufs_bindex_t new_bindex, bindex, bbot, bwh, bdiropq;
8380 + struct au_hdentry tmp, *p, *q;
8381 + struct au_dinfo *dinfo;
8382 + struct super_block *sb;
8384 + DiMustWriteLock(dentry);
8386 + sb = dentry->d_sb;
8387 + dinfo = au_di(dentry);
8388 + bbot = dinfo->di_bbot;
8389 + bwh = dinfo->di_bwh;
8390 + bdiropq = dinfo->di_bdiropq;
8391 + bindex = dinfo->di_btop;
8392 + p = au_hdentry(dinfo, bindex);
8393 + for (; bindex <= bbot; bindex++, p++) {
8394 + if (!p->hd_dentry)
8397 + new_bindex = au_br_index(sb, p->hd_id);
8398 + if (new_bindex == bindex)
8401 + if (dinfo->di_bwh == bindex)
8403 + if (dinfo->di_bdiropq == bindex)
8404 + bdiropq = new_bindex;
8405 + if (new_bindex < 0) {
8407 + p->hd_dentry = NULL;
8411 + /* swap two lower dentries, and loop again */
8412 + q = au_hdentry(dinfo, new_bindex);
8416 + if (tmp.hd_dentry) {
8422 + dinfo->di_bwh = -1;
8423 + if (bwh >= 0 && bwh <= au_sbbot(sb) && au_sbr_whable(sb, bwh))
8424 + dinfo->di_bwh = bwh;
8426 + dinfo->di_bdiropq = -1;
8428 + && bdiropq <= au_sbbot(sb)
8429 + && au_sbr_whable(sb, bdiropq))
8430 + dinfo->di_bdiropq = bdiropq;
8433 + dinfo->di_btop = -1;
8434 + dinfo->di_bbot = -1;
8435 + bbot = au_dbbot(parent);
8437 + p = au_hdentry(dinfo, bindex);
8438 + for (; bindex <= bbot; bindex++, p++)
8439 + if (p->hd_dentry) {
8440 + dinfo->di_btop = bindex;
8444 + if (dinfo->di_btop >= 0) {
8446 + p = au_hdentry(dinfo, bindex);
8447 + for (; bindex >= 0; bindex--, p--)
8448 + if (p->hd_dentry) {
8449 + dinfo->di_bbot = bindex;
8458 +static void au_do_hide(struct dentry *dentry)
8460 + struct inode *inode;
8462 + if (d_really_is_positive(dentry)) {
8463 + inode = d_inode(dentry);
8464 + if (!d_is_dir(dentry)) {
8465 + if (inode->i_nlink && !d_unhashed(dentry))
8466 + drop_nlink(inode);
8468 + clear_nlink(inode);
8469 + /* stop next lookup */
8470 + inode->i_flags |= S_DEAD;
8472 + smp_mb(); /* necessary? */
8477 +static int au_hide_children(struct dentry *parent)
8479 + int err, i, j, ndentry;
8480 + struct au_dcsub_pages dpages;
8481 + struct au_dpage *dpage;
8482 + struct dentry *dentry;
8484 + err = au_dpages_init(&dpages, GFP_NOFS);
8485 + if (unlikely(err))
8487 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
8488 + if (unlikely(err))
8491 + /* in reverse order */
8492 + for (i = dpages.ndpage - 1; i >= 0; i--) {
8493 + dpage = dpages.dpages + i;
8494 + ndentry = dpage->ndentry;
8495 + for (j = ndentry - 1; j >= 0; j--) {
8496 + dentry = dpage->dentries[j];
8497 + if (dentry != parent)
8498 + au_do_hide(dentry);
8503 + au_dpages_free(&dpages);
8508 +static void au_hide(struct dentry *dentry)
8512 + AuDbgDentry(dentry);
8513 + if (d_is_dir(dentry)) {
8514 + /* shrink_dcache_parent(dentry); */
8515 + err = au_hide_children(dentry);
8516 + if (unlikely(err))
8517 + AuIOErr("%pd, failed hiding children, ignored %d\n",
8520 + au_do_hide(dentry);
8524 + * By adding a dirty branch, a cached dentry may be affected in various ways.
8526 + * a dirty branch is added
8527 + * - on the top of layers
8528 + * - in the middle of layers
8529 + * - to the bottom of layers
8531 + * on the added branch there exists
8534 + * - a same named entry
8536 + * * negative --> positive
8537 + * * positive --> positive
8538 + * - type is unchanged
8539 + * - type is changed
8541 + * * negative --> negative
8542 + * * positive --> negative (rejected by au_br_del() for non-dir case)
8545 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8546 + struct au_dinfo *tmp)
8549 + aufs_bindex_t bindex, bbot;
8551 + struct dentry *dentry;
8552 + struct inode *inode;
8554 + } orig_h, tmp_h = {
8557 + struct au_hdentry *hd;
8558 + struct inode *inode, *h_inode;
8559 + struct dentry *h_dentry;
8562 + AuDebugOn(dinfo->di_btop < 0);
8564 + orig_h.dentry = au_hdentry(dinfo, dinfo->di_btop)->hd_dentry;
8565 + orig_h.inode = NULL;
8566 + if (d_is_positive(orig_h.dentry)) {
8567 + orig_h.inode = d_inode(orig_h.dentry);
8568 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8570 + if (tmp->di_btop >= 0) {
8571 + tmp_h.dentry = au_hdentry(tmp, tmp->di_btop)->hd_dentry;
8572 + if (d_is_positive(tmp_h.dentry)) {
8573 + tmp_h.inode = d_inode(tmp_h.dentry);
8574 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8579 + if (d_really_is_positive(dentry))
8580 + inode = d_inode(dentry);
8581 + if (!orig_h.inode) {
8582 + AuDbg("negative originally\n");
8588 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8589 + AuDebugOn(dinfo->di_bdiropq != -1);
8591 + if (!tmp_h.inode) {
8592 + AuDbg("negative --> negative\n");
8593 + /* should have only one negative lower */
8594 + if (tmp->di_btop >= 0
8595 + && tmp->di_btop < dinfo->di_btop) {
8596 + AuDebugOn(tmp->di_btop != tmp->di_bbot);
8597 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8598 + au_set_h_dptr(dentry, dinfo->di_btop, NULL);
8599 + au_di_cp(dinfo, tmp);
8600 + hd = au_hdentry(tmp, tmp->di_btop);
8601 + au_set_h_dptr(dentry, tmp->di_btop,
8602 + dget(hd->hd_dentry));
8604 + au_dbg_verify_dinode(dentry);
8606 + AuDbg("negative --> positive\n");
8608 + * similar to the behaviour of creating with bypassing
8610 + * unhash it in order to force an error in the
8611 + * succeeding create operation.
8612 + * we should not set S_DEAD here.
8615 + /* au_di_swap(tmp, dinfo); */
8616 + au_dbg_verify_dinode(dentry);
8619 + AuDbg("positive originally\n");
8620 + /* inode may be NULL */
8621 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8622 + if (!tmp_h.inode) {
8623 + AuDbg("positive --> negative\n");
8624 + /* or bypassing aufs */
8626 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_btop)
8627 + dinfo->di_bwh = tmp->di_bwh;
8629 + err = au_refresh_hinode_self(inode);
8630 + au_dbg_verify_dinode(dentry);
8631 + } else if (orig_h.mode == tmp_h.mode) {
8632 + AuDbg("positive --> positive, same type\n");
8633 + if (!S_ISDIR(orig_h.mode)
8634 + && dinfo->di_btop > tmp->di_btop) {
8636 + * similar to the behaviour of removing and
8641 + err = au_refresh_hinode_self(inode);
8642 + au_dbg_verify_dinode(dentry);
8644 + /* fill empty slots */
8645 + if (dinfo->di_btop > tmp->di_btop)
8646 + dinfo->di_btop = tmp->di_btop;
8647 + if (dinfo->di_bbot < tmp->di_bbot)
8648 + dinfo->di_bbot = tmp->di_bbot;
8649 + dinfo->di_bwh = tmp->di_bwh;
8650 + dinfo->di_bdiropq = tmp->di_bdiropq;
8651 + bbot = dinfo->di_bbot;
8652 + bindex = tmp->di_btop;
8653 + hd = au_hdentry(tmp, bindex);
8654 + for (; bindex <= bbot; bindex++, hd++) {
8655 + if (au_h_dptr(dentry, bindex))
8657 + h_dentry = hd->hd_dentry;
8660 + AuDebugOn(d_is_negative(h_dentry));
8661 + h_inode = d_inode(h_dentry);
8662 + AuDebugOn(orig_h.mode
8663 + != (h_inode->i_mode
8665 + au_set_h_dptr(dentry, bindex,
8669 + err = au_refresh_hinode(inode, dentry);
8670 + au_dbg_verify_dinode(dentry);
8673 + AuDbg("positive --> positive, different type\n");
8674 + /* similar to the behaviour of removing and creating */
8677 + err = au_refresh_hinode_self(inode);
8678 + au_dbg_verify_dinode(dentry);
8686 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8688 + const struct dentry_operations *dop
8689 + = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8690 + static const unsigned int mask
8691 + = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8693 + BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8695 + if (dentry->d_op == dop)
8698 + AuDbg("%pd\n", dentry);
8699 + spin_lock(&dentry->d_lock);
8700 + if (dop == &aufs_dop)
8701 + dentry->d_flags |= mask;
8703 + dentry->d_flags &= ~mask;
8704 + dentry->d_op = dop;
8705 + spin_unlock(&dentry->d_lock);
8708 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8710 + int err, ebrange, nbr;
8711 + unsigned int sigen;
8712 + struct au_dinfo *dinfo, *tmp;
8713 + struct super_block *sb;
8714 + struct inode *inode;
8716 + DiMustWriteLock(dentry);
8717 + AuDebugOn(IS_ROOT(dentry));
8718 + AuDebugOn(d_really_is_negative(parent));
8720 + sb = dentry->d_sb;
8721 + sigen = au_sigen(sb);
8722 + err = au_digen_test(parent, sigen);
8723 + if (unlikely(err))
8726 + nbr = au_sbbot(sb) + 1;
8727 + dinfo = au_di(dentry);
8728 + err = au_di_realloc(dinfo, nbr, /*may_shrink*/0);
8729 + if (unlikely(err))
8731 + ebrange = au_dbrange_test(dentry);
8733 + ebrange = au_do_refresh_hdentry(dentry, parent);
8735 + if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8736 + AuDebugOn(au_dbtop(dentry) < 0 && au_dbbot(dentry) >= 0);
8737 + if (d_really_is_positive(dentry)) {
8738 + inode = d_inode(dentry);
8739 + err = au_refresh_hinode_self(inode);
8741 + au_dbg_verify_dinode(dentry);
8743 + goto out_dgen; /* success */
8747 + /* temporary dinfo */
8748 + AuDbgDentry(dentry);
8750 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8751 + if (unlikely(!tmp))
8753 + au_di_swap(tmp, dinfo);
8754 + /* returns the number of positive dentries */
8756 + * if current working dir is removed, it returns an error.
8757 + * but the dentry is legal.
8759 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
8760 + AuDbgDentry(dentry);
8761 + au_di_swap(tmp, dinfo);
8762 + if (err == -ENOENT)
8765 + /* compare/refresh by dinfo */
8766 + AuDbgDentry(dentry);
8767 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8768 + au_dbg_verify_dinode(dentry);
8771 + au_di_realloc(dinfo, nbr, /*may_shrink*/1); /* harmless if err */
8772 + au_rw_write_unlock(&tmp->di_rwsem);
8774 + if (unlikely(err))
8778 + au_update_digen(dentry);
8780 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8781 + AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8782 + AuDbgDentry(dentry);
8788 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8789 + struct dentry *dentry, aufs_bindex_t bindex)
8794 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8797 + AuDbg("b%d\n", bindex);
8799 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8800 + * due to whiteout and branch permission.
8802 + flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8803 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
8804 + /* it may return tri-state */
8805 + valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8807 + if (unlikely(valid < 0))
8817 +/* todo: remove this */
8818 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8819 + unsigned int flags, int do_udba, int dirren)
8822 + umode_t mode, h_mode;
8823 + aufs_bindex_t bindex, btail, btop, ibs, ibe;
8824 + unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8825 + struct inode *h_inode, *h_cached_inode;
8826 + struct dentry *h_dentry;
8827 + struct qstr *name, *h_name;
8834 + unhashed = !!d_unhashed(dentry);
8835 + is_root = !!IS_ROOT(dentry);
8836 + name = &dentry->d_name;
8837 + tmpfile = au_di(dentry)->di_tmpfile;
8840 + * Theoretically, REVAL test should be unnecessary in case of
8842 + * But {fs,i}notify doesn't fire some necessary events,
8843 + * IN_ATTRIB for atime/nlink/pageio
8844 + * Let's do REVAL test too.
8846 + if (do_udba && inode) {
8847 + mode = (inode->i_mode & S_IFMT);
8848 + plus = (inode->i_nlink > 0);
8849 + ibs = au_ibtop(inode);
8850 + ibe = au_ibbot(inode);
8853 + btop = au_dbtop(dentry);
8855 + if (inode && S_ISDIR(inode->i_mode))
8856 + btail = au_dbtaildir(dentry);
8857 + for (bindex = btop; bindex <= btail; bindex++) {
8858 + h_dentry = au_h_dptr(dentry, bindex);
8862 + AuDbg("b%d, %pd\n", bindex, h_dentry);
8863 + h_nfs = !!au_test_nfs(h_dentry->d_sb);
8864 + spin_lock(&h_dentry->d_lock);
8865 + h_name = &h_dentry->d_name;
8866 + if (unlikely(do_udba
8869 + && (unhashed != !!d_unhashed(h_dentry)
8870 + || (!tmpfile && !dirren
8871 + && !au_qstreq(name, h_name))
8874 + && !(flags & LOOKUP_OPEN)
8875 + && (h_dentry->d_flags
8876 + & DCACHE_NFSFS_RENAMED)))
8880 + h_unhashed = d_unhashed(h_dentry);
8881 + spin_unlock(&h_dentry->d_lock);
8882 + AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8883 + unhashed, h_unhashed, dentry, h_dentry);
8886 + spin_unlock(&h_dentry->d_lock);
8888 + err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8889 + if (unlikely(err))
8890 + /* do not goto err, to keep the errno */
8893 + /* todo: plink too? */
8898 + if (unlikely(!!inode != d_is_positive(h_dentry)))
8902 + if (d_is_positive(h_dentry))
8903 + h_inode = d_inode(h_dentry);
8906 + h_cached_inode = h_inode;
8908 + h_mode = (h_inode->i_mode & S_IFMT);
8909 + h_plus = (h_inode->i_nlink > 0);
8911 + if (inode && ibs <= bindex && bindex <= ibe)
8912 + h_cached_inode = au_h_iptr(inode, bindex);
8915 + if (unlikely(plus != h_plus && !tmpfile))
8918 + if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8920 + && !IS_ROOT(h_dentry)
8921 + && unhashed != d_unhashed(h_dentry)))
8924 + if (unlikely(mode != h_mode
8925 + || h_cached_inode != h_inode))
8938 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8939 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8942 + struct dentry *parent;
8944 + if (!au_digen_test(dentry, sigen))
8947 + parent = dget_parent(dentry);
8948 + di_read_lock_parent(parent, AuLock_IR);
8949 + AuDebugOn(au_digen_test(parent, sigen));
8950 + au_dbg_verify_gen(parent, sigen);
8951 + err = au_refresh_dentry(dentry, parent);
8952 + di_read_unlock(parent, AuLock_IR);
8958 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8961 + struct dentry *d, *parent;
8963 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8964 + return simple_reval_dpath(dentry, sigen);
8966 + /* slow loop, keep it simple and stupid */
8967 + /* cf: au_cpup_dirs() */
8970 + while (au_digen_test(dentry, sigen)) {
8974 + parent = dget_parent(d);
8975 + if (!au_digen_test(parent, sigen))
8981 + di_write_lock_child2(d);
8983 + /* someone might update our dentry while we were sleeping */
8984 + if (au_digen_test(d, sigen)) {
8986 + * todo: consolidate with simple_reval_dpath(),
8987 + * do_refresh() and au_reval_for_attr().
8989 + di_read_lock_parent(parent, AuLock_IR);
8990 + err = au_refresh_dentry(d, parent);
8991 + di_read_unlock(parent, AuLock_IR);
8995 + di_write_unlock(d);
8997 + if (unlikely(err))
9005 + * if valid returns 1, otherwise 0.
9007 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
9010 + unsigned int sigen;
9011 + unsigned char do_udba, dirren;
9012 + struct super_block *sb;
9013 + struct inode *inode;
9015 + /* todo: support rcu-walk? */
9016 + if (flags & LOOKUP_RCU)
9020 + if (unlikely(!au_di(dentry)))
9024 + sb = dentry->d_sb;
9027 + * i_mutex of parent dir may be held,
9028 + * but we should not return 'invalid' due to busy.
9030 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
9031 + if (unlikely(err)) {
9037 + if (d_really_is_positive(dentry))
9038 + inode = d_inode(dentry);
9039 + if (unlikely(inode && au_is_bad_inode(inode))) {
9044 + if (unlikely(au_dbrange_test(dentry))) {
9050 + sigen = au_sigen(sb);
9051 + if (au_digen_test(dentry, sigen)) {
9052 + AuDebugOn(IS_ROOT(dentry));
9053 + err = au_reval_dpath(dentry, sigen);
9054 + if (unlikely(err)) {
9059 + di_downgrade_lock(dentry, AuLock_IR);
9062 + if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
9064 + && !(inode->i_state && I_LINKABLE)
9065 + && (IS_DEADDIR(inode) || !inode->i_nlink)) {
9070 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
9071 + if (do_udba && inode) {
9072 + aufs_bindex_t btop = au_ibtop(inode);
9073 + struct inode *h_inode;
9076 + h_inode = au_h_iptr(inode, btop);
9077 + if (h_inode && au_test_higen(inode, h_inode)) {
9084 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
9085 + err = h_d_revalidate(dentry, inode, flags, do_udba, dirren);
9086 + if (unlikely(!err && do_udba && au_dbtop(dentry) < 0)) {
9088 + AuDbg("both of real entry and whiteout found, %p, err %d\n",
9094 + di_downgrade_lock(dentry, AuLock_IR);
9096 + aufs_read_unlock(dentry, AuLock_IR);
9101 + AuDbg("%pd invalid, %d\n", dentry, valid);
9107 +static void aufs_d_release(struct dentry *dentry)
9109 + if (au_di(dentry)) {
9110 + au_di_fin(dentry);
9111 + au_hn_di_reinit(dentry);
9115 +const struct dentry_operations aufs_dop = {
9116 + .d_revalidate = aufs_d_revalidate,
9117 + .d_weak_revalidate = aufs_d_revalidate,
9118 + .d_release = aufs_d_release
9121 +/* aufs_dop without d_revalidate */
9122 +const struct dentry_operations aufs_dop_noreval = {
9123 + .d_release = aufs_d_release
9125 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
9126 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
9127 +++ linux/fs/aufs/dentry.h 2022-03-21 14:49:05.719966344 +0100
9129 +/* SPDX-License-Identifier: GPL-2.0 */
9131 + * Copyright (C) 2005-2021 Junjiro R. Okajima
9133 + * This program, aufs is free software; you can redistribute it and/or modify
9134 + * it under the terms of the GNU General Public License as published by
9135 + * the Free Software Foundation; either version 2 of the License, or
9136 + * (at your option) any later version.
9138 + * This program is distributed in the hope that it will be useful,
9139 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9140 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9141 + * GNU General Public License for more details.
9143 + * You should have received a copy of the GNU General Public License
9144 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9148 + * lookup and dentry operations
9151 +#ifndef __AUFS_DENTRY_H__
9152 +#define __AUFS_DENTRY_H__
9156 +#include <linux/dcache.h>
9157 +#include "dirren.h"
9160 +struct au_hdentry {
9161 + struct dentry *hd_dentry;
9162 + aufs_bindex_t hd_id;
9166 + atomic_t di_generation;
9168 + struct au_rwsem di_rwsem;
9169 + aufs_bindex_t di_btop, di_bbot, di_bwh, di_bdiropq;
9170 + unsigned char di_tmpfile; /* to allow the different name */
9171 + struct au_hdentry *di_hdentry;
9172 + struct rcu_head rcu;
9173 +} ____cacheline_aligned_in_smp;
9175 +/* ---------------------------------------------------------------------- */
9177 +/* flags for au_lkup_dentry() */
9178 +#define AuLkup_ALLOW_NEG 1
9179 +#define AuLkup_IGNORE_PERM (1 << 1)
9180 +#define AuLkup_DIRREN (1 << 2)
9181 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
9182 +#define au_fset_lkup(flags, name) \
9183 + do { (flags) |= AuLkup_##name; } while (0)
9184 +#define au_fclr_lkup(flags, name) \
9185 + do { (flags) &= ~AuLkup_##name; } while (0)
9187 +#ifndef CONFIG_AUFS_DIRREN
9188 +#undef AuLkup_DIRREN
9189 +#define AuLkup_DIRREN 0
9192 +struct au_do_lookup_args {
9193 + unsigned int flags;
9195 + struct qstr whname, *name;
9196 + struct au_dr_lookup dirren;
9199 +/* ---------------------------------------------------------------------- */
9202 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
9204 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
9205 + struct path *ppath);
9206 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
9207 + struct dentry *h_parent, struct au_branch *br);
9209 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
9210 + unsigned int flags);
9211 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
9212 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
9213 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
9214 +void au_refresh_dop(struct dentry *dentry, int force_reval);
9217 +void au_di_init_once(void *_di);
9218 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
9219 +void au_di_free(struct au_dinfo *dinfo);
9220 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
9221 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
9222 +int au_di_init(struct dentry *dentry);
9223 +void au_di_fin(struct dentry *dentry);
9224 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink);
9226 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
9227 +void di_read_unlock(struct dentry *d, int flags);
9228 +void di_downgrade_lock(struct dentry *d, int flags);
9229 +void di_write_lock(struct dentry *d, unsigned int lsc);
9230 +void di_write_unlock(struct dentry *d);
9231 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
9232 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
9233 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
9235 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
9236 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
9237 +aufs_bindex_t au_dbtail(struct dentry *dentry);
9238 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
9240 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9241 + struct dentry *h_dentry);
9242 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
9243 +int au_dbrange_test(struct dentry *dentry);
9244 +void au_update_digen(struct dentry *dentry);
9245 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
9246 +void au_update_dbtop(struct dentry *dentry);
9247 +void au_update_dbbot(struct dentry *dentry);
9248 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
9250 +/* ---------------------------------------------------------------------- */
9252 +static inline struct au_dinfo *au_di(struct dentry *dentry)
9254 + return dentry->d_fsdata;
9257 +/* ---------------------------------------------------------------------- */
9259 +/* lock subclass for dinfo */
9261 + AuLsc_DI_CHILD, /* child first */
9262 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
9263 + AuLsc_DI_CHILD3, /* copyup dirs */
9267 + AuLsc_DI_TMP /* temp for replacing dinfo */
9271 + * di_read_lock_child, di_write_lock_child,
9272 + * di_read_lock_child2, di_write_lock_child2,
9273 + * di_read_lock_child3, di_write_lock_child3,
9274 + * di_read_lock_parent, di_write_lock_parent,
9275 + * di_read_lock_parent2, di_write_lock_parent2,
9276 + * di_read_lock_parent3, di_write_lock_parent3,
9278 +#define AuReadLockFunc(name, lsc) \
9279 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
9280 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
9282 +#define AuWriteLockFunc(name, lsc) \
9283 +static inline void di_write_lock_##name(struct dentry *d) \
9284 +{ di_write_lock(d, AuLsc_DI_##lsc); }
9286 +#define AuRWLockFuncs(name, lsc) \
9287 + AuReadLockFunc(name, lsc) \
9288 + AuWriteLockFunc(name, lsc)
9290 +AuRWLockFuncs(child, CHILD);
9291 +AuRWLockFuncs(child2, CHILD2);
9292 +AuRWLockFuncs(child3, CHILD3);
9293 +AuRWLockFuncs(parent, PARENT);
9294 +AuRWLockFuncs(parent2, PARENT2);
9295 +AuRWLockFuncs(parent3, PARENT3);
9297 +#undef AuReadLockFunc
9298 +#undef AuWriteLockFunc
9299 +#undef AuRWLockFuncs
9301 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
9302 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
9303 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
9305 +/* ---------------------------------------------------------------------- */
9307 +/* todo: memory barrier? */
9308 +static inline unsigned int au_digen(struct dentry *d)
9310 + return atomic_read(&au_di(d)->di_generation);
9313 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
9315 + hdentry->hd_dentry = NULL;
9318 +static inline struct au_hdentry *au_hdentry(struct au_dinfo *di,
9319 + aufs_bindex_t bindex)
9321 + return di->di_hdentry + bindex;
9324 +static inline void au_hdput(struct au_hdentry *hd)
9327 + dput(hd->hd_dentry);
9330 +static inline aufs_bindex_t au_dbtop(struct dentry *dentry)
9332 + DiMustAnyLock(dentry);
9333 + return au_di(dentry)->di_btop;
9336 +static inline aufs_bindex_t au_dbbot(struct dentry *dentry)
9338 + DiMustAnyLock(dentry);
9339 + return au_di(dentry)->di_bbot;
9342 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
9344 + DiMustAnyLock(dentry);
9345 + return au_di(dentry)->di_bwh;
9348 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
9350 + DiMustAnyLock(dentry);
9351 + return au_di(dentry)->di_bdiropq;
9354 +/* todo: hard/soft set? */
9355 +static inline void au_set_dbtop(struct dentry *dentry, aufs_bindex_t bindex)
9357 + DiMustWriteLock(dentry);
9358 + au_di(dentry)->di_btop = bindex;
9361 +static inline void au_set_dbbot(struct dentry *dentry, aufs_bindex_t bindex)
9363 + DiMustWriteLock(dentry);
9364 + au_di(dentry)->di_bbot = bindex;
9367 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
9369 + DiMustWriteLock(dentry);
9370 + /* dbwh can be outside of btop - bbot range */
9371 + au_di(dentry)->di_bwh = bindex;
9374 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
9376 + DiMustWriteLock(dentry);
9377 + au_di(dentry)->di_bdiropq = bindex;
9380 +/* ---------------------------------------------------------------------- */
9382 +#ifdef CONFIG_AUFS_HNOTIFY
9383 +static inline void au_digen_dec(struct dentry *d)
9385 + atomic_dec(&au_di(d)->di_generation);
9388 +static inline void au_hn_di_reinit(struct dentry *dentry)
9390 + dentry->d_fsdata = NULL;
9393 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
9394 +#endif /* CONFIG_AUFS_HNOTIFY */
9396 +#endif /* __KERNEL__ */
9397 +#endif /* __AUFS_DENTRY_H__ */
9398 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
9399 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
9400 +++ linux/fs/aufs/dinfo.c 2022-03-21 14:49:05.719966344 +0100
9402 +// SPDX-License-Identifier: GPL-2.0
9404 + * Copyright (C) 2005-2021 Junjiro R. Okajima
9406 + * This program, aufs is free software; you can redistribute it and/or modify
9407 + * it under the terms of the GNU General Public License as published by
9408 + * the Free Software Foundation; either version 2 of the License, or
9409 + * (at your option) any later version.
9411 + * This program is distributed in the hope that it will be useful,
9412 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9413 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9414 + * GNU General Public License for more details.
9416 + * You should have received a copy of the GNU General Public License
9417 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9421 + * dentry private data
9426 +void au_di_init_once(void *_dinfo)
9428 + struct au_dinfo *dinfo = _dinfo;
9430 + au_rw_init(&dinfo->di_rwsem);
9433 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
9435 + struct au_dinfo *dinfo;
9438 + dinfo = au_cache_alloc_dinfo();
9439 + if (unlikely(!dinfo))
9442 + nbr = au_sbbot(sb) + 1;
9445 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
9446 + if (dinfo->di_hdentry) {
9447 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
9448 + dinfo->di_btop = -1;
9449 + dinfo->di_bbot = -1;
9450 + dinfo->di_bwh = -1;
9451 + dinfo->di_bdiropq = -1;
9452 + dinfo->di_tmpfile = 0;
9453 + for (i = 0; i < nbr; i++)
9454 + dinfo->di_hdentry[i].hd_id = -1;
9458 + au_cache_free_dinfo(dinfo);
9465 +void au_di_free(struct au_dinfo *dinfo)
9467 + struct au_hdentry *p;
9468 + aufs_bindex_t bbot, bindex;
9470 + /* dentry may not be revalidated */
9471 + bindex = dinfo->di_btop;
9472 + if (bindex >= 0) {
9473 + bbot = dinfo->di_bbot;
9474 + p = au_hdentry(dinfo, bindex);
9475 + while (bindex++ <= bbot)
9478 + au_kfree_try_rcu(dinfo->di_hdentry);
9479 + au_cache_free_dinfo(dinfo);
9482 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
9484 + struct au_hdentry *p;
9487 + AuRwMustWriteLock(&a->di_rwsem);
9488 + AuRwMustWriteLock(&b->di_rwsem);
9490 +#define DiSwap(v, name) \
9492 + v = a->di_##name; \
9493 + a->di_##name = b->di_##name; \
9494 + b->di_##name = v; \
9497 + DiSwap(p, hdentry);
9501 + DiSwap(bi, bdiropq);
9507 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
9509 + AuRwMustWriteLock(&dst->di_rwsem);
9510 + AuRwMustWriteLock(&src->di_rwsem);
9512 + dst->di_btop = src->di_btop;
9513 + dst->di_bbot = src->di_bbot;
9514 + dst->di_bwh = src->di_bwh;
9515 + dst->di_bdiropq = src->di_bdiropq;
9519 +int au_di_init(struct dentry *dentry)
9522 + struct super_block *sb;
9523 + struct au_dinfo *dinfo;
9526 + sb = dentry->d_sb;
9527 + dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
9529 + atomic_set(&dinfo->di_generation, au_sigen(sb));
9530 + /* smp_mb(); */ /* atomic_set */
9531 + dentry->d_fsdata = dinfo;
9538 +void au_di_fin(struct dentry *dentry)
9540 + struct au_dinfo *dinfo;
9542 + dinfo = au_di(dentry);
9543 + AuRwDestroy(&dinfo->di_rwsem);
9544 + au_di_free(dinfo);
9547 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink)
9550 + struct au_hdentry *hdp;
9552 + AuRwMustWriteLock(&dinfo->di_rwsem);
9555 + sz = sizeof(*hdp) * (dinfo->di_bbot + 1);
9557 + sz = sizeof(*hdp);
9558 + hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS,
9561 + dinfo->di_hdentry = hdp;
9568 +/* ---------------------------------------------------------------------- */
9570 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
9573 + case AuLsc_DI_CHILD:
9574 + ii_write_lock_child(inode);
9576 + case AuLsc_DI_CHILD2:
9577 + ii_write_lock_child2(inode);
9579 + case AuLsc_DI_CHILD3:
9580 + ii_write_lock_child3(inode);
9582 + case AuLsc_DI_PARENT:
9583 + ii_write_lock_parent(inode);
9585 + case AuLsc_DI_PARENT2:
9586 + ii_write_lock_parent2(inode);
9588 + case AuLsc_DI_PARENT3:
9589 + ii_write_lock_parent3(inode);
9596 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9599 + case AuLsc_DI_CHILD:
9600 + ii_read_lock_child(inode);
9602 + case AuLsc_DI_CHILD2:
9603 + ii_read_lock_child2(inode);
9605 + case AuLsc_DI_CHILD3:
9606 + ii_read_lock_child3(inode);
9608 + case AuLsc_DI_PARENT:
9609 + ii_read_lock_parent(inode);
9611 + case AuLsc_DI_PARENT2:
9612 + ii_read_lock_parent2(inode);
9614 + case AuLsc_DI_PARENT3:
9615 + ii_read_lock_parent3(inode);
9622 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9624 + struct inode *inode;
9626 + au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9627 + if (d_really_is_positive(d)) {
9628 + inode = d_inode(d);
9629 + if (au_ftest_lock(flags, IW))
9630 + do_ii_write_lock(inode, lsc);
9631 + else if (au_ftest_lock(flags, IR))
9632 + do_ii_read_lock(inode, lsc);
9636 +void di_read_unlock(struct dentry *d, int flags)
9638 + struct inode *inode;
9640 + if (d_really_is_positive(d)) {
9641 + inode = d_inode(d);
9642 + if (au_ftest_lock(flags, IW)) {
9643 + au_dbg_verify_dinode(d);
9644 + ii_write_unlock(inode);
9645 + } else if (au_ftest_lock(flags, IR)) {
9646 + au_dbg_verify_dinode(d);
9647 + ii_read_unlock(inode);
9650 + au_rw_read_unlock(&au_di(d)->di_rwsem);
9653 +void di_downgrade_lock(struct dentry *d, int flags)
9655 + if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9656 + ii_downgrade_lock(d_inode(d));
9657 + au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9660 +void di_write_lock(struct dentry *d, unsigned int lsc)
9662 + au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9663 + if (d_really_is_positive(d))
9664 + do_ii_write_lock(d_inode(d), lsc);
9667 +void di_write_unlock(struct dentry *d)
9669 + au_dbg_verify_dinode(d);
9670 + if (d_really_is_positive(d))
9671 + ii_write_unlock(d_inode(d));
9672 + au_rw_write_unlock(&au_di(d)->di_rwsem);
9675 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9677 + AuDebugOn(d1 == d2
9678 + || d_inode(d1) == d_inode(d2)
9679 + || d1->d_sb != d2->d_sb);
9681 + if ((isdir && au_test_subdir(d1, d2))
9683 + di_write_lock_child(d1);
9684 + di_write_lock_child2(d2);
9686 + di_write_lock_child(d2);
9687 + di_write_lock_child2(d1);
9691 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9693 + AuDebugOn(d1 == d2
9694 + || d_inode(d1) == d_inode(d2)
9695 + || d1->d_sb != d2->d_sb);
9697 + if ((isdir && au_test_subdir(d1, d2))
9699 + di_write_lock_parent(d1);
9700 + di_write_lock_parent2(d2);
9702 + di_write_lock_parent(d2);
9703 + di_write_lock_parent2(d1);
9707 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9709 + di_write_unlock(d1);
9710 + if (d_inode(d1) == d_inode(d2))
9711 + au_rw_write_unlock(&au_di(d2)->di_rwsem);
9713 + di_write_unlock(d2);
9716 +/* ---------------------------------------------------------------------- */
9718 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9722 + DiMustAnyLock(dentry);
9724 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
9726 + AuDebugOn(bindex < 0);
9727 + d = au_hdentry(au_di(dentry), bindex)->hd_dentry;
9728 + AuDebugOn(d && au_dcount(d) <= 0);
9733 + * extended version of au_h_dptr().
9734 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9737 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9739 + struct dentry *h_dentry;
9740 + struct inode *inode, *h_inode;
9742 + AuDebugOn(d_really_is_negative(dentry));
9745 + if (au_dbtop(dentry) <= bindex
9746 + && bindex <= au_dbbot(dentry))
9747 + h_dentry = au_h_dptr(dentry, bindex);
9748 + if (h_dentry && !au_d_linkable(h_dentry)) {
9750 + goto out; /* success */
9753 + inode = d_inode(dentry);
9754 + AuDebugOn(bindex < au_ibtop(inode));
9755 + AuDebugOn(au_ibbot(inode) < bindex);
9756 + h_inode = au_h_iptr(inode, bindex);
9757 + h_dentry = d_find_alias(h_inode);
9759 + if (!IS_ERR(h_dentry)) {
9760 + if (!au_d_linkable(h_dentry))
9761 + goto out; /* success */
9767 + if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9768 + h_dentry = au_plink_lkup(inode, bindex);
9769 + AuDebugOn(!h_dentry);
9770 + if (!IS_ERR(h_dentry)) {
9771 + if (!au_d_hashed_positive(h_dentry))
9772 + goto out; /* success */
9779 + AuDbgDentry(h_dentry);
9783 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9785 + aufs_bindex_t bbot, bwh;
9787 + bbot = au_dbbot(dentry);
9789 + bwh = au_dbwh(dentry);
9792 + if (0 < bwh && bwh < bbot)
9798 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9800 + aufs_bindex_t bbot, bopq;
9802 + bbot = au_dbtail(dentry);
9804 + bopq = au_dbdiropq(dentry);
9805 + if (0 <= bopq && bopq < bbot)
9811 +/* ---------------------------------------------------------------------- */
9813 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9814 + struct dentry *h_dentry)
9816 + struct au_dinfo *dinfo;
9817 + struct au_hdentry *hd;
9818 + struct au_branch *br;
9820 + DiMustWriteLock(dentry);
9822 + dinfo = au_di(dentry);
9823 + hd = au_hdentry(dinfo, bindex);
9825 + hd->hd_dentry = h_dentry;
9827 + br = au_sbr(dentry->d_sb, bindex);
9828 + hd->hd_id = br->br_id;
9832 +int au_dbrange_test(struct dentry *dentry)
9835 + aufs_bindex_t btop, bbot;
9838 + btop = au_dbtop(dentry);
9839 + bbot = au_dbbot(dentry);
9841 + AuDebugOn(bbot < 0 && btop > bbot);
9844 + AuDebugOn(bbot >= 0);
9850 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9855 + if (unlikely(au_digen(dentry) != sigen
9856 + || au_iigen_test(d_inode(dentry), sigen)))
9862 +void au_update_digen(struct dentry *dentry)
9864 + atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9865 + /* smp_mb(); */ /* atomic_set */
9868 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9870 + struct au_dinfo *dinfo;
9871 + struct dentry *h_d;
9872 + struct au_hdentry *hdp;
9873 + aufs_bindex_t bindex, bbot;
9875 + DiMustWriteLock(dentry);
9877 + dinfo = au_di(dentry);
9878 + if (!dinfo || dinfo->di_btop < 0)
9881 + if (do_put_zero) {
9882 + bbot = dinfo->di_bbot;
9883 + bindex = dinfo->di_btop;
9884 + hdp = au_hdentry(dinfo, bindex);
9885 + for (; bindex <= bbot; bindex++, hdp++) {
9886 + h_d = hdp->hd_dentry;
9887 + if (h_d && d_is_negative(h_d))
9888 + au_set_h_dptr(dentry, bindex, NULL);
9892 + dinfo->di_btop = 0;
9893 + hdp = au_hdentry(dinfo, dinfo->di_btop);
9894 + for (; dinfo->di_btop <= dinfo->di_bbot; dinfo->di_btop++, hdp++)
9895 + if (hdp->hd_dentry)
9897 + if (dinfo->di_btop > dinfo->di_bbot) {
9898 + dinfo->di_btop = -1;
9899 + dinfo->di_bbot = -1;
9903 + hdp = au_hdentry(dinfo, dinfo->di_bbot);
9904 + for (; dinfo->di_bbot >= 0; dinfo->di_bbot--, hdp--)
9905 + if (hdp->hd_dentry)
9907 + AuDebugOn(dinfo->di_btop > dinfo->di_bbot || dinfo->di_bbot < 0);
9910 +void au_update_dbtop(struct dentry *dentry)
9912 + aufs_bindex_t bindex, bbot;
9913 + struct dentry *h_dentry;
9915 + bbot = au_dbbot(dentry);
9916 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
9917 + h_dentry = au_h_dptr(dentry, bindex);
9920 + if (d_is_positive(h_dentry)) {
9921 + au_set_dbtop(dentry, bindex);
9924 + au_set_h_dptr(dentry, bindex, NULL);
9928 +void au_update_dbbot(struct dentry *dentry)
9930 + aufs_bindex_t bindex, btop;
9931 + struct dentry *h_dentry;
9933 + btop = au_dbtop(dentry);
9934 + for (bindex = au_dbbot(dentry); bindex >= btop; bindex--) {
9935 + h_dentry = au_h_dptr(dentry, bindex);
9938 + if (d_is_positive(h_dentry)) {
9939 + au_set_dbbot(dentry, bindex);
9942 + au_set_h_dptr(dentry, bindex, NULL);
9946 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9948 + aufs_bindex_t bindex, bbot;
9950 + bbot = au_dbbot(dentry);
9951 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++)
9952 + if (au_h_dptr(dentry, bindex) == h_dentry)
9956 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9957 --- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
9958 +++ linux/fs/aufs/dir.c 2022-03-21 14:49:05.719966344 +0100
9960 +// SPDX-License-Identifier: GPL-2.0
9962 + * Copyright (C) 2005-2021 Junjiro R. Okajima
9964 + * This program, aufs is free software; you can redistribute it and/or modify
9965 + * it under the terms of the GNU General Public License as published by
9966 + * the Free Software Foundation; either version 2 of the License, or
9967 + * (at your option) any later version.
9969 + * This program is distributed in the hope that it will be useful,
9970 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9971 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9972 + * GNU General Public License for more details.
9974 + * You should have received a copy of the GNU General Public License
9975 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9979 + * directory operations
9982 +#include <linux/fs_stack.h>
9983 +#include <linux/iversion.h>
9986 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
9988 + unsigned int nlink;
9990 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9992 + nlink = dir->i_nlink;
9993 + nlink += h_dir->i_nlink - 2;
9994 + if (h_dir->i_nlink < 2)
9996 + smp_mb(); /* for i_nlink */
9997 + /* 0 can happen in revaliding */
9998 + set_nlink(dir, nlink);
10001 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
10003 + unsigned int nlink;
10005 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10007 + nlink = dir->i_nlink;
10008 + nlink -= h_dir->i_nlink - 2;
10009 + if (h_dir->i_nlink < 2)
10011 + smp_mb(); /* for i_nlink */
10012 + /* nlink == 0 means the branch-fs is broken */
10013 + set_nlink(dir, nlink);
10016 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
10019 + aufs_bindex_t bindex, bbot;
10020 + struct file *h_file;
10021 + struct dentry *h_dentry;
10025 + AuDebugOn(!d_is_dir(file->f_path.dentry));
10027 + bbot = au_fbbot_dir(file);
10028 + for (bindex = au_fbtop(file);
10029 + bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10031 + h_file = au_hf_dir(file, bindex);
10032 + if (h_file && file_inode(h_file))
10033 + sz += vfsub_f_size_read(h_file);
10036 + AuDebugOn(!dentry);
10037 + AuDebugOn(!d_is_dir(dentry));
10039 + bbot = au_dbtaildir(dentry);
10040 + for (bindex = au_dbtop(dentry);
10041 + bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10043 + h_dentry = au_h_dptr(dentry, bindex);
10044 + if (h_dentry && d_is_positive(h_dentry))
10045 + sz += i_size_read(d_inode(h_dentry));
10048 + if (sz < KMALLOC_MAX_SIZE)
10049 + sz = roundup_pow_of_two(sz);
10050 + if (sz > KMALLOC_MAX_SIZE)
10051 + sz = KMALLOC_MAX_SIZE;
10052 + else if (sz < NAME_MAX) {
10053 + BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
10054 + sz = AUFS_RDBLK_DEF;
10059 +struct au_dir_ts_arg {
10060 + struct dentry *dentry;
10061 + aufs_bindex_t brid;
10064 +static void au_do_dir_ts(void *arg)
10066 + struct au_dir_ts_arg *a = arg;
10067 + struct au_dtime dt;
10068 + struct path h_path;
10069 + struct inode *dir, *h_dir;
10070 + struct super_block *sb;
10071 + struct au_branch *br;
10072 + struct au_hinode *hdir;
10074 + aufs_bindex_t btop, bindex;
10076 + sb = a->dentry->d_sb;
10077 + if (d_really_is_negative(a->dentry))
10079 + /* no dir->i_mutex lock */
10080 + aufs_read_lock(a->dentry, AuLock_DW); /* noflush */
10082 + dir = d_inode(a->dentry);
10083 + btop = au_ibtop(dir);
10084 + bindex = au_br_index(sb, a->brid);
10085 + if (bindex < btop)
10088 + br = au_sbr(sb, bindex);
10089 + h_path.dentry = au_h_dptr(a->dentry, bindex);
10090 + if (!h_path.dentry)
10092 + h_path.mnt = au_br_mnt(br);
10093 + au_dtime_store(&dt, a->dentry, &h_path);
10095 + br = au_sbr(sb, btop);
10096 + if (!au_br_writable(br->br_perm))
10098 + h_path.dentry = au_h_dptr(a->dentry, btop);
10099 + h_path.mnt = au_br_mnt(br);
10100 + err = vfsub_mnt_want_write(h_path.mnt);
10103 + hdir = au_hi(dir, btop);
10104 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
10105 + h_dir = au_h_iptr(dir, btop);
10106 + if (h_dir->i_nlink
10107 + && timespec64_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
10108 + dt.dt_h_path = h_path;
10109 + au_dtime_revert(&dt);
10111 + au_hn_inode_unlock(hdir);
10112 + vfsub_mnt_drop_write(h_path.mnt);
10113 + au_cpup_attr_timesizes(dir);
10116 + aufs_read_unlock(a->dentry, AuLock_DW);
10119 + au_nwt_done(&au_sbi(sb)->si_nowait);
10120 + au_kfree_try_rcu(arg);
10123 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
10125 + int perm, wkq_err;
10126 + aufs_bindex_t btop;
10127 + struct au_dir_ts_arg *arg;
10128 + struct dentry *dentry;
10129 + struct super_block *sb;
10133 + dentry = d_find_any_alias(dir);
10134 + AuDebugOn(!dentry);
10135 + sb = dentry->d_sb;
10136 + btop = au_ibtop(dir);
10137 + if (btop == bindex) {
10138 + au_cpup_attr_timesizes(dir);
10142 + perm = au_sbr_perm(sb, btop);
10143 + if (!au_br_writable(perm))
10146 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
10150 + arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
10151 + arg->brid = au_sbr_id(sb, bindex);
10152 + wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
10153 + if (unlikely(wkq_err)) {
10154 + pr_err("wkq %d\n", wkq_err);
10156 + au_kfree_try_rcu(arg);
10163 +/* ---------------------------------------------------------------------- */
10165 +static int reopen_dir(struct file *file)
10168 + unsigned int flags;
10169 + aufs_bindex_t bindex, btail, btop;
10170 + struct dentry *dentry, *h_dentry;
10171 + struct file *h_file;
10173 + /* open all lower dirs */
10174 + dentry = file->f_path.dentry;
10175 + btop = au_dbtop(dentry);
10176 + for (bindex = au_fbtop(file); bindex < btop; bindex++)
10177 + au_set_h_fptr(file, bindex, NULL);
10178 + au_set_fbtop(file, btop);
10180 + btail = au_dbtaildir(dentry);
10181 + for (bindex = au_fbbot_dir(file); btail < bindex; bindex--)
10182 + au_set_h_fptr(file, bindex, NULL);
10183 + au_set_fbbot_dir(file, btail);
10185 + flags = vfsub_file_flags(file);
10186 + for (bindex = btop; bindex <= btail; bindex++) {
10187 + h_dentry = au_h_dptr(dentry, bindex);
10190 + h_file = au_hf_dir(file, bindex);
10194 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10195 + err = PTR_ERR(h_file);
10196 + if (IS_ERR(h_file))
10197 + goto out; /* close all? */
10198 + au_set_h_fptr(file, bindex, h_file);
10200 + au_update_figen(file);
10201 + /* todo: necessary? */
10202 + /* file->f_ra = h_file->f_ra; */
10209 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
10212 + aufs_bindex_t bindex, btail;
10213 + struct dentry *dentry, *h_dentry;
10214 + struct vfsmount *mnt;
10216 + FiMustWriteLock(file);
10217 + AuDebugOn(h_file);
10220 + mnt = file->f_path.mnt;
10221 + dentry = file->f_path.dentry;
10222 + file->f_version = inode_query_iversion(d_inode(dentry));
10223 + bindex = au_dbtop(dentry);
10224 + au_set_fbtop(file, bindex);
10225 + btail = au_dbtaildir(dentry);
10226 + au_set_fbbot_dir(file, btail);
10227 + for (; !err && bindex <= btail; bindex++) {
10228 + h_dentry = au_h_dptr(dentry, bindex);
10232 + err = vfsub_test_mntns(mnt, h_dentry->d_sb);
10233 + if (unlikely(err))
10235 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10236 + if (IS_ERR(h_file)) {
10237 + err = PTR_ERR(h_file);
10240 + au_set_h_fptr(file, bindex, h_file);
10242 + au_update_figen(file);
10243 + /* todo: necessary? */
10244 + /* file->f_ra = h_file->f_ra; */
10246 + return 0; /* success */
10249 + for (bindex = au_fbtop(file); bindex <= btail; bindex++)
10250 + au_set_h_fptr(file, bindex, NULL);
10251 + au_set_fbtop(file, -1);
10252 + au_set_fbbot_dir(file, -1);
10257 +static int aufs_open_dir(struct inode *inode __maybe_unused,
10258 + struct file *file)
10261 + struct super_block *sb;
10262 + struct au_fidir *fidir;
10265 + sb = file->f_path.dentry->d_sb;
10266 + si_read_lock(sb, AuLock_FLUSH);
10267 + fidir = au_fidir_alloc(sb);
10269 + struct au_do_open_args args = {
10270 + .open = do_open_dir,
10273 + err = au_do_open(file, &args);
10274 + if (unlikely(err))
10275 + au_kfree_rcu(fidir);
10277 + si_read_unlock(sb);
10281 +static int aufs_release_dir(struct inode *inode __maybe_unused,
10282 + struct file *file)
10284 + struct au_vdir *vdir_cache;
10285 + struct au_finfo *finfo;
10286 + struct au_fidir *fidir;
10287 + struct au_hfile *hf;
10288 + aufs_bindex_t bindex, bbot;
10290 + finfo = au_fi(file);
10291 + fidir = finfo->fi_hdir;
10293 + au_hbl_del(&finfo->fi_hlist,
10294 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
10295 + vdir_cache = fidir->fd_vdir_cache; /* lock-free */
10297 + au_vdir_free(vdir_cache);
10299 + bindex = finfo->fi_btop;
10300 + if (bindex >= 0) {
10301 + hf = fidir->fd_hfile + bindex;
10303 + * calls fput() instead of filp_close(),
10304 + * since no dnotify or lock for the lower file.
10306 + bbot = fidir->fd_bbot;
10307 + for (; bindex <= bbot; bindex++, hf++)
10309 + au_hfput(hf, /*execed*/0);
10311 + au_kfree_rcu(fidir);
10312 + finfo->fi_hdir = NULL;
10314 + au_finfo_fin(file);
10318 +/* ---------------------------------------------------------------------- */
10320 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
10323 + aufs_bindex_t bindex, bbot;
10324 + struct file *h_file;
10327 + bbot = au_fbbot_dir(file);
10328 + for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10329 + h_file = au_hf_dir(file, bindex);
10331 + err = vfsub_flush(h_file, id);
10336 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
10338 + return au_do_flush(file, id, au_do_flush_dir);
10341 +/* ---------------------------------------------------------------------- */
10343 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
10346 + aufs_bindex_t bbot, bindex;
10347 + struct inode *inode;
10348 + struct super_block *sb;
10351 + sb = dentry->d_sb;
10352 + inode = d_inode(dentry);
10353 + IMustLock(inode);
10354 + bbot = au_dbbot(dentry);
10355 + for (bindex = au_dbtop(dentry); !err && bindex <= bbot; bindex++) {
10356 + struct path h_path;
10358 + if (au_test_ro(sb, bindex, inode))
10360 + h_path.dentry = au_h_dptr(dentry, bindex);
10361 + if (!h_path.dentry)
10364 + h_path.mnt = au_sbr_mnt(sb, bindex);
10365 + err = vfsub_fsync(NULL, &h_path, datasync);
10371 +static int au_do_fsync_dir(struct file *file, int datasync)
10374 + aufs_bindex_t bbot, bindex;
10375 + struct file *h_file;
10376 + struct super_block *sb;
10377 + struct inode *inode;
10379 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10380 + if (unlikely(err))
10383 + inode = file_inode(file);
10384 + sb = inode->i_sb;
10385 + bbot = au_fbbot_dir(file);
10386 + for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10387 + h_file = au_hf_dir(file, bindex);
10388 + if (!h_file || au_test_ro(sb, bindex, inode))
10391 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
10399 + * @file may be NULL
10401 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
10405 + struct dentry *dentry;
10406 + struct inode *inode;
10407 + struct super_block *sb;
10410 + dentry = file->f_path.dentry;
10411 + inode = d_inode(dentry);
10412 + inode_lock(inode);
10413 + sb = dentry->d_sb;
10414 + si_noflush_read_lock(sb);
10416 + err = au_do_fsync_dir(file, datasync);
10418 + di_write_lock_child(dentry);
10419 + err = au_do_fsync_dir_no_file(dentry, datasync);
10421 + au_cpup_attr_timesizes(inode);
10422 + di_write_unlock(dentry);
10424 + fi_write_unlock(file);
10426 + si_read_unlock(sb);
10427 + inode_unlock(inode);
10431 +/* ---------------------------------------------------------------------- */
10433 +static int aufs_iterate_shared(struct file *file, struct dir_context *ctx)
10436 + struct dentry *dentry;
10437 + struct inode *inode, *h_inode;
10438 + struct super_block *sb;
10440 + AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
10442 + dentry = file->f_path.dentry;
10443 + inode = d_inode(dentry);
10444 + IMustLock(inode);
10446 + sb = dentry->d_sb;
10447 + si_read_lock(sb, AuLock_FLUSH);
10448 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10449 + if (unlikely(err))
10451 + err = au_alive_dir(dentry);
10453 + err = au_vdir_init(file);
10454 + di_downgrade_lock(dentry, AuLock_IR);
10455 + if (unlikely(err))
10458 + h_inode = au_h_iptr(inode, au_ibtop(inode));
10459 + if (!au_test_nfsd()) {
10460 + err = au_vdir_fill_de(file, ctx);
10461 + fsstack_copy_attr_atime(inode, h_inode);
10464 + * nfsd filldir may call lookup_one_len(), vfs_getattr(),
10465 + * encode_fh() and others.
10467 + atomic_inc(&h_inode->i_count);
10468 + di_read_unlock(dentry, AuLock_IR);
10469 + si_read_unlock(sb);
10470 + err = au_vdir_fill_de(file, ctx);
10471 + fsstack_copy_attr_atime(inode, h_inode);
10472 + fi_write_unlock(file);
10480 + di_read_unlock(dentry, AuLock_IR);
10481 + fi_write_unlock(file);
10483 + si_read_unlock(sb);
10487 +/* ---------------------------------------------------------------------- */
10489 +#define AuTestEmpty_WHONLY 1
10490 +#define AuTestEmpty_CALLED (1 << 1)
10491 +#define AuTestEmpty_SHWH (1 << 2)
10492 +#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
10493 +#define au_fset_testempty(flags, name) \
10494 + do { (flags) |= AuTestEmpty_##name; } while (0)
10495 +#define au_fclr_testempty(flags, name) \
10496 + do { (flags) &= ~AuTestEmpty_##name; } while (0)
10498 +#ifndef CONFIG_AUFS_SHWH
10499 +#undef AuTestEmpty_SHWH
10500 +#define AuTestEmpty_SHWH 0
10503 +struct test_empty_arg {
10504 + struct dir_context ctx;
10505 + struct au_nhash *whlist;
10506 + unsigned int flags;
10508 + aufs_bindex_t bindex;
10511 +static int test_empty_cb(struct dir_context *ctx, const char *__name,
10512 + int namelen, loff_t offset __maybe_unused, u64 ino,
10513 + unsigned int d_type)
10515 + struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
10517 + char *name = (void *)__name;
10520 + au_fset_testempty(arg->flags, CALLED);
10522 + if (name[0] == '.'
10523 + && (namelen == 1 || (name[1] == '.' && namelen == 2)))
10524 + goto out; /* success */
10526 + if (namelen <= AUFS_WH_PFX_LEN
10527 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
10528 + if (au_ftest_testempty(arg->flags, WHONLY)
10529 + && !au_nhash_test_known_wh(arg->whlist, name, namelen))
10530 + arg->err = -ENOTEMPTY;
10534 + name += AUFS_WH_PFX_LEN;
10535 + namelen -= AUFS_WH_PFX_LEN;
10536 + if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
10537 + arg->err = au_nhash_append_wh
10538 + (arg->whlist, name, namelen, ino, d_type, arg->bindex,
10539 + au_ftest_testempty(arg->flags, SHWH));
10543 + AuTraceErr(arg->err);
10547 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10550 + struct file *h_file;
10551 + struct au_branch *br;
10553 + h_file = au_h_open(dentry, arg->bindex,
10554 + O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
10555 + /*file*/NULL, /*force_wr*/0);
10556 + err = PTR_ERR(h_file);
10557 + if (IS_ERR(h_file))
10561 + if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
10562 + && !file_inode(h_file)->i_nlink)
10567 + au_fclr_testempty(arg->flags, CALLED);
10569 + err = vfsub_iterate_dir(h_file, &arg->ctx);
10572 + } while (!err && au_ftest_testempty(arg->flags, CALLED));
10576 + br = au_sbr(dentry->d_sb, arg->bindex);
10577 + au_lcnt_dec(&br->br_nfiles);
10582 +struct do_test_empty_args {
10584 + struct dentry *dentry;
10585 + struct test_empty_arg *arg;
10588 +static void call_do_test_empty(void *args)
10590 + struct do_test_empty_args *a = args;
10591 + *a->errp = do_test_empty(a->dentry, a->arg);
10594 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10596 + int err, wkq_err;
10597 + struct dentry *h_dentry;
10598 + struct inode *h_inode;
10599 + struct user_namespace *h_userns;
10601 + h_userns = au_sbr_userns(dentry->d_sb, arg->bindex);
10602 + h_dentry = au_h_dptr(dentry, arg->bindex);
10603 + h_inode = d_inode(h_dentry);
10604 + /* todo: i_mode changes anytime? */
10605 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
10606 + err = au_test_h_perm_sio(h_userns, h_inode, MAY_EXEC | MAY_READ);
10607 + inode_unlock_shared(h_inode);
10609 + err = do_test_empty(dentry, arg);
10611 + struct do_test_empty_args args = {
10613 + .dentry = dentry,
10616 + unsigned int flags = arg->flags;
10618 + wkq_err = au_wkq_wait(call_do_test_empty, &args);
10619 + if (unlikely(wkq_err))
10621 + arg->flags = flags;
10627 +int au_test_empty_lower(struct dentry *dentry)
10630 + unsigned int rdhash;
10631 + aufs_bindex_t bindex, btop, btail;
10632 + struct au_nhash whlist;
10633 + struct test_empty_arg arg = {
10635 + .actor = test_empty_cb
10638 + int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10640 + SiMustAnyLock(dentry->d_sb);
10642 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10644 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10645 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10646 + if (unlikely(err))
10650 + arg.whlist = &whlist;
10651 + btop = au_dbtop(dentry);
10652 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10653 + au_fset_testempty(arg.flags, SHWH);
10654 + test_empty = do_test_empty;
10655 + if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10656 + test_empty = sio_test_empty;
10657 + arg.bindex = btop;
10658 + err = test_empty(dentry, &arg);
10659 + if (unlikely(err))
10662 + au_fset_testempty(arg.flags, WHONLY);
10663 + btail = au_dbtaildir(dentry);
10664 + for (bindex = btop + 1; !err && bindex <= btail; bindex++) {
10665 + struct dentry *h_dentry;
10667 + h_dentry = au_h_dptr(dentry, bindex);
10668 + if (h_dentry && d_is_positive(h_dentry)) {
10669 + arg.bindex = bindex;
10670 + err = test_empty(dentry, &arg);
10675 + au_nhash_wh_free(&whlist);
10680 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10683 + struct test_empty_arg arg = {
10685 + .actor = test_empty_cb
10688 + aufs_bindex_t bindex, btail;
10691 + arg.whlist = whlist;
10692 + arg.flags = AuTestEmpty_WHONLY;
10693 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10694 + au_fset_testempty(arg.flags, SHWH);
10695 + btail = au_dbtaildir(dentry);
10696 + for (bindex = au_dbtop(dentry); !err && bindex <= btail; bindex++) {
10697 + struct dentry *h_dentry;
10699 + h_dentry = au_h_dptr(dentry, bindex);
10700 + if (h_dentry && d_is_positive(h_dentry)) {
10701 + arg.bindex = bindex;
10702 + err = sio_test_empty(dentry, &arg);
10709 +/* ---------------------------------------------------------------------- */
10711 +const struct file_operations aufs_dir_fop = {
10712 + .owner = THIS_MODULE,
10713 + .llseek = default_llseek,
10714 + .read = generic_read_dir,
10715 + .iterate_shared = aufs_iterate_shared,
10716 + .unlocked_ioctl = aufs_ioctl_dir,
10717 +#ifdef CONFIG_COMPAT
10718 + .compat_ioctl = aufs_compat_ioctl_dir,
10720 + .open = aufs_open_dir,
10721 + .release = aufs_release_dir,
10722 + .flush = aufs_flush_dir,
10723 + .fsync = aufs_fsync_dir
10725 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10726 --- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
10727 +++ linux/fs/aufs/dir.h 2022-03-21 14:49:05.719966344 +0100
10729 +/* SPDX-License-Identifier: GPL-2.0 */
10731 + * Copyright (C) 2005-2021 Junjiro R. Okajima
10733 + * This program, aufs is free software; you can redistribute it and/or modify
10734 + * it under the terms of the GNU General Public License as published by
10735 + * the Free Software Foundation; either version 2 of the License, or
10736 + * (at your option) any later version.
10738 + * This program is distributed in the hope that it will be useful,
10739 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10740 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10741 + * GNU General Public License for more details.
10743 + * You should have received a copy of the GNU General Public License
10744 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10748 + * directory operations
10751 +#ifndef __AUFS_DIR_H__
10752 +#define __AUFS_DIR_H__
10756 +#include <linux/fs.h>
10758 +/* ---------------------------------------------------------------------- */
10760 +/* need to be faster and smaller */
10763 + unsigned int nh_num;
10764 + struct hlist_head *nh_head;
10767 +struct au_vdir_destr {
10768 + unsigned char len;
10769 + unsigned char name[];
10772 +struct au_vdir_dehstr {
10773 + struct hlist_node hash;
10774 + struct au_vdir_destr *str;
10775 + struct rcu_head rcu;
10776 +} ____cacheline_aligned_in_smp;
10778 +struct au_vdir_de {
10780 + unsigned char de_type;
10781 + /* caution: packed */
10782 + struct au_vdir_destr de_str;
10785 +struct au_vdir_wh {
10786 + struct hlist_node wh_hash;
10787 +#ifdef CONFIG_AUFS_SHWH
10789 + aufs_bindex_t wh_bindex;
10790 + unsigned char wh_type;
10792 + aufs_bindex_t wh_bindex;
10794 + /* caution: packed */
10795 + struct au_vdir_destr wh_str;
10798 +union au_vdir_deblk_p {
10799 + unsigned char *deblk;
10800 + struct au_vdir_de *de;
10804 + unsigned char **vd_deblk;
10805 + unsigned long vd_nblk;
10807 + unsigned long ul;
10808 + union au_vdir_deblk_p p;
10812 + unsigned int vd_deblk_sz;
10813 + unsigned long vd_jiffy;
10814 + struct rcu_head rcu;
10815 +} ____cacheline_aligned_in_smp;
10817 +/* ---------------------------------------------------------------------- */
10820 +extern const struct file_operations aufs_dir_fop;
10821 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10822 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10823 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10824 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10825 +int au_test_empty_lower(struct dentry *dentry);
10826 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10829 +unsigned int au_rdhash_est(loff_t sz);
10830 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10831 +void au_nhash_wh_free(struct au_nhash *whlist);
10832 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10834 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10835 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10836 + unsigned int d_type, aufs_bindex_t bindex,
10837 + unsigned char shwh);
10838 +void au_vdir_free(struct au_vdir *vdir);
10839 +int au_vdir_init(struct file *file);
10840 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10843 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10845 +#ifdef CONFIG_AUFS_RDU
10847 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10848 +#ifdef CONFIG_COMPAT
10849 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10850 + unsigned long arg);
10853 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10854 + unsigned int cmd, unsigned long arg)
10855 +#ifdef CONFIG_COMPAT
10856 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10857 + unsigned int cmd, unsigned long arg)
10861 +#endif /* __KERNEL__ */
10862 +#endif /* __AUFS_DIR_H__ */
10863 diff -urN /usr/share/empty/fs/aufs/dirren.c linux/fs/aufs/dirren.c
10864 --- /usr/share/empty/fs/aufs/dirren.c 1970-01-01 01:00:00.000000000 +0100
10865 +++ linux/fs/aufs/dirren.c 2022-03-21 14:49:05.723299677 +0100
10867 +// SPDX-License-Identifier: GPL-2.0
10869 + * Copyright (C) 2017-2021 Junjiro R. Okajima
10871 + * This program, aufs is free software; you can redistribute it and/or modify
10872 + * it under the terms of the GNU General Public License as published by
10873 + * the Free Software Foundation; either version 2 of the License, or
10874 + * (at your option) any later version.
10876 + * This program is distributed in the hope that it will be useful,
10877 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10878 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10879 + * GNU General Public License for more details.
10881 + * You should have received a copy of the GNU General Public License
10882 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10886 + * special handling in renaming a directory
10887 + * in order to support looking-up the before-renamed name on the lower readonly
10891 +#include <linux/byteorder/generic.h>
10894 +static void au_dr_hino_del(struct au_dr_br *dr, struct au_dr_hino *ent)
10898 + idx = au_dr_ihash(ent->dr_h_ino);
10899 + au_hbl_del(&ent->dr_hnode, dr->dr_h_ino + idx);
10902 +static int au_dr_hino_test_empty(struct au_dr_br *dr)
10905 + struct hlist_bl_head *hbl;
10908 + for (i = 0; ret && i < AuDirren_NHASH; i++) {
10909 + hbl = dr->dr_h_ino + i;
10910 + hlist_bl_lock(hbl);
10911 + ret &= hlist_bl_empty(hbl);
10912 + hlist_bl_unlock(hbl);
10918 +static struct au_dr_hino *au_dr_hino_find(struct au_dr_br *dr, ino_t ino)
10920 + struct au_dr_hino *found, *ent;
10921 + struct hlist_bl_head *hbl;
10922 + struct hlist_bl_node *pos;
10926 + idx = au_dr_ihash(ino);
10927 + hbl = dr->dr_h_ino + idx;
10928 + hlist_bl_lock(hbl);
10929 + hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10930 + if (ent->dr_h_ino == ino) {
10934 + hlist_bl_unlock(hbl);
10939 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t ino,
10940 + struct au_dr_hino *add_ent)
10943 + struct hlist_bl_head *hbl;
10944 + struct hlist_bl_node *pos;
10945 + struct au_dr_hino *ent;
10948 + idx = au_dr_ihash(ino);
10949 + hbl = dr->dr_h_ino + idx;
10950 +#if 0 /* debug print */
10952 + struct hlist_bl_node *tmp;
10954 + hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10955 + AuDbg("hi%llu\n", (unsigned long long)ent->dr_h_ino);
10958 + hlist_bl_lock(hbl);
10959 + hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10960 + if (ent->dr_h_ino == ino) {
10964 + if (!found && add_ent)
10965 + hlist_bl_add_head(&add_ent->dr_hnode, hbl);
10966 + hlist_bl_unlock(hbl);
10968 + if (!found && add_ent)
10969 + AuDbg("i%llu added\n", (unsigned long long)add_ent->dr_h_ino);
10974 +void au_dr_hino_free(struct au_dr_br *dr)
10977 + struct hlist_bl_head *hbl;
10978 + struct hlist_bl_node *pos, *tmp;
10979 + struct au_dr_hino *ent;
10981 + /* SiMustWriteLock(sb); */
10983 + for (i = 0; i < AuDirren_NHASH; i++) {
10984 + hbl = dr->dr_h_ino + i;
10985 + /* no spinlock since sbinfo must be write-locked */
10986 + hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10987 + au_kfree_rcu(ent);
10988 + INIT_HLIST_BL_HEAD(hbl);
10992 +/* returns the number of inodes or an error */
10993 +static int au_dr_hino_store(struct super_block *sb, struct au_branch *br,
10994 + struct file *hinofile)
10998 + loff_t pos, oldsize;
11000 + struct inode *hinoinode;
11001 + struct hlist_bl_head *hbl;
11002 + struct hlist_bl_node *n1, *n2;
11003 + struct au_dr_hino *ent;
11005 + SiMustWriteLock(sb);
11006 + AuDebugOn(!au_br_writable(br->br_perm));
11008 + hinoinode = file_inode(hinofile);
11009 + oldsize = i_size_read(hinoinode);
11013 + hbl = br->br_dirren.dr_h_ino;
11014 + for (i = 0; !err && i < AuDirren_NHASH; i++, hbl++) {
11015 + /* no bit-lock since sbinfo must be write-locked */
11016 + hlist_bl_for_each_entry_safe(ent, n1, n2, hbl, dr_hnode) {
11017 + AuDbg("hi%llu, %pD2\n",
11018 + (unsigned long long)ent->dr_h_ino, hinofile);
11019 + u64 = cpu_to_be64(ent->dr_h_ino);
11020 + ssz = vfsub_write_k(hinofile, &u64, sizeof(u64), &pos);
11021 + if (ssz == sizeof(u64))
11024 + /* write error */
11025 + pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11032 + /* regardless the error */
11033 + if (pos < oldsize) {
11034 + err = vfsub_trunc(&hinofile->f_path, pos, /*attr*/0, hinofile);
11042 +static int au_dr_hino_load(struct au_dr_br *dr, struct file *hinofile)
11049 + struct au_dr_hino *ent;
11050 + struct inode *hinoinode;
11051 + struct hlist_bl_head *hbl;
11055 + hbl = dr->dr_h_ino;
11056 + hinoinode = file_inode(hinofile);
11057 + sz = i_size_read(hinoinode);
11058 + AuDebugOn(sz % sizeof(u64));
11059 + n = sz / sizeof(u64);
11061 + ssz = vfsub_read_k(hinofile, &u64, sizeof(u64), &pos);
11062 + if (unlikely(ssz != sizeof(u64))) {
11063 + pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11070 + ent = kmalloc(sizeof(*ent), GFP_NOFS);
11076 + ent->dr_h_ino = be64_to_cpu((__force __be64)u64);
11077 + AuDbg("hi%llu, %pD2\n",
11078 + (unsigned long long)ent->dr_h_ino, hinofile);
11079 + hidx = au_dr_ihash(ent->dr_h_ino);
11080 + au_hbl_add(&ent->dr_hnode, hbl + hidx);
11082 + goto out; /* success */
11085 + au_dr_hino_free(dr);
11092 + * @bindex/@br is a switch to distinguish whether suspending hnotify or not.
11093 + * @path is a switch to distinguish load and store.
11095 +static int au_dr_hino(struct super_block *sb, aufs_bindex_t bindex,
11096 + struct au_branch *br, const struct path *path)
11099 + unsigned char load, suspend;
11100 + struct file *hinofile;
11101 + struct au_hinode *hdir;
11102 + struct inode *dir, *delegated;
11103 + struct path hinopath;
11104 + struct qstr hinoname = QSTR_INIT(AUFS_WH_DR_BRHINO,
11105 + sizeof(AUFS_WH_DR_BRHINO) - 1);
11107 + AuDebugOn(bindex < 0 && !br);
11108 + AuDebugOn(bindex >= 0 && br);
11113 + br = au_sbr(sb, bindex);
11116 + path = &br->br_path;
11117 + AuDebugOn(!au_br_writable(br->br_perm));
11118 + if (unlikely(!au_br_writable(br->br_perm)))
11124 + dir = d_inode(sb->s_root);
11125 + hdir = au_hinode(au_ii(dir), bindex);
11126 + dir = hdir->hi_inode;
11127 + au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11129 + dir = d_inode(path->dentry);
11130 + inode_lock_nested(dir, AuLsc_I_CHILD);
11132 + hinopath.mnt = path->mnt;
11133 + hinopath.dentry = vfsub_lkup_one(&hinoname, (struct path *)path);
11134 + err = PTR_ERR(hinopath.dentry);
11135 + if (IS_ERR(hinopath.dentry))
11139 + flags = O_RDONLY;
11141 + if (d_is_negative(hinopath.dentry))
11142 + goto out_dput; /* success */
11144 + if (au_dr_hino_test_empty(&br->br_dirren)) {
11145 + if (d_is_positive(hinopath.dentry)) {
11146 + delegated = NULL;
11147 + err = vfsub_unlink(dir, &hinopath, &delegated,
11150 + if (unlikely(err))
11151 + pr_err("ignored err %d, %pd2\n",
11152 + err, hinopath.dentry);
11153 + if (unlikely(err == -EWOULDBLOCK))
11158 + } else if (!d_is_positive(hinopath.dentry)) {
11159 + err = vfsub_create(dir, &hinopath, 0600,
11160 + /*want_excl*/false);
11162 + if (unlikely(err))
11165 + flags = O_WRONLY;
11167 + hinofile = vfsub_dentry_open(&hinopath, flags);
11169 + au_hn_inode_unlock(hdir);
11171 + inode_unlock(dir);
11172 + dput(hinopath.dentry);
11173 + AuTraceErrPtr(hinofile);
11174 + if (IS_ERR(hinofile)) {
11175 + err = PTR_ERR(hinofile);
11180 + err = au_dr_hino_load(&br->br_dirren, hinofile);
11182 + err = au_dr_hino_store(sb, br, hinofile);
11187 + dput(hinopath.dentry);
11190 + au_hn_inode_unlock(hdir);
11192 + inode_unlock(dir);
11198 +/* ---------------------------------------------------------------------- */
11200 +static int au_dr_brid_init(struct au_dr_brid *brid, const struct path *path)
11203 + struct kstatfs kstfs;
11205 + struct dentry *dentry;
11206 + struct super_block *sb;
11208 + err = vfs_statfs((void *)path, &kstfs);
11210 + if (unlikely(err))
11213 + /* todo: support for UUID */
11215 + if (kstfs.f_fsid.val[0] || kstfs.f_fsid.val[1]) {
11216 + brid->type = AuBrid_FSID;
11217 + brid->fsid = kstfs.f_fsid;
11219 + dentry = path->dentry;
11220 + sb = dentry->d_sb;
11223 + brid->type = AuBrid_DEV;
11232 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
11233 + const struct path *path)
11236 + struct au_dr_br *dr;
11237 + struct hlist_bl_head *hbl;
11239 + dr = &br->br_dirren;
11240 + hbl = dr->dr_h_ino;
11241 + for (i = 0; i < AuDirren_NHASH; i++, hbl++)
11242 + INIT_HLIST_BL_HEAD(hbl);
11244 + err = au_dr_brid_init(&dr->dr_brid, path);
11245 + if (unlikely(err))
11248 + if (au_opt_test(au_mntflags(sb), DIRREN))
11249 + err = au_dr_hino(sb, /*bindex*/-1, br, path);
11256 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br)
11261 + if (au_br_writable(br->br_perm))
11262 + err = au_dr_hino(sb, /*bindex*/-1, br, /*path*/NULL);
11264 + au_dr_hino_free(&br->br_dirren);
11269 +/* ---------------------------------------------------------------------- */
11271 +static int au_brid_str(struct au_dr_brid *brid, struct inode *h_inode,
11272 + char *buf, size_t sz)
11275 + unsigned int major, minor;
11279 + err = snprintf(p, sz, "%d_", brid->type);
11280 + AuDebugOn(err > sz);
11283 + switch (brid->type) {
11284 + case AuBrid_Unset:
11286 + case AuBrid_UUID:
11287 + err = snprintf(p, sz, "%pU", brid->uuid.b);
11289 + case AuBrid_FSID:
11290 + err = snprintf(p, sz, "%08x-%08x",
11291 + brid->fsid.val[0], brid->fsid.val[1]);
11294 + major = MAJOR(brid->dev);
11295 + minor = MINOR(brid->dev);
11296 + if (major <= 0xff && minor <= 0xff)
11297 + err = snprintf(p, sz, "%02x%02x", major, minor);
11299 + err = snprintf(p, sz, "%03x:%05x", major, minor);
11302 + AuDebugOn(err > sz);
11305 + err = snprintf(p, sz, "_%llu", (unsigned long long)h_inode->i_ino);
11306 + AuDebugOn(err > sz);
11313 +static int au_drinfo_name(struct au_branch *br, char *name, int len)
11316 + struct dentry *br_dentry;
11317 + struct inode *br_inode;
11319 + br_dentry = au_br_dentry(br);
11320 + br_inode = d_inode(br_dentry);
11321 + rlen = au_brid_str(&br->br_dirren.dr_brid, br_inode, name, len);
11322 + AuDebugOn(rlen >= AUFS_DIRREN_ENV_VAL_SZ);
11323 + AuDebugOn(rlen > len);
11328 +/* ---------------------------------------------------------------------- */
11331 + * from the given @h_dentry, construct drinfo at @*fdata.
11332 + * when the size of @*fdata is not enough, reallocate and return new @fdata and
11335 +static int au_drinfo_construct(struct au_drinfo_fdata **fdata,
11336 + struct dentry *h_dentry,
11337 + unsigned char *allocated)
11340 + struct au_drinfo_fdata *f, *p;
11341 + struct au_drinfo *drinfo;
11342 + struct inode *h_inode;
11343 + struct qstr *qname;
11347 + h_inode = d_inode(h_dentry);
11348 + qname = &h_dentry->d_name;
11349 + drinfo = &f->drinfo;
11350 + drinfo->ino = (__force uint64_t)cpu_to_be64(h_inode->i_ino);
11351 + drinfo->oldnamelen = qname->len;
11352 + if (*allocated < sizeof(*f) + qname->len) {
11353 + v = roundup_pow_of_two(*allocated + qname->len);
11354 + p = au_krealloc(f, v, GFP_NOFS, /*may_shrink*/0);
11355 + if (unlikely(!p)) {
11363 + drinfo = &f->drinfo;
11365 + memcpy(drinfo->oldname, qname->name, qname->len);
11366 + AuDbg("i%llu, %.*s\n",
11367 + be64_to_cpu((__force __be64)drinfo->ino), drinfo->oldnamelen,
11368 + drinfo->oldname);
11375 +/* callers have to free the return value */
11376 +static struct au_drinfo *au_drinfo_read_k(struct file *file, ino_t h_ino)
11378 + struct au_drinfo *ret, *drinfo;
11379 + struct au_drinfo_fdata fdata;
11384 + ret = ERR_PTR(-EIO);
11386 + ssz = vfsub_read_k(file, &fdata, sizeof(fdata), &pos);
11387 + if (unlikely(ssz != sizeof(fdata))) {
11388 + AuIOErr("ssz %zd, %u, %pD2\n",
11389 + ssz, (unsigned int)sizeof(fdata), file);
11393 + fdata.magic = ntohl((__force __be32)fdata.magic);
11394 + switch (fdata.magic) {
11395 + case AUFS_DRINFO_MAGIC_V1:
11398 + AuIOErr("magic-num 0x%x, 0x%x, %pD2\n",
11399 + fdata.magic, AUFS_DRINFO_MAGIC_V1, file);
11403 + drinfo = &fdata.drinfo;
11404 + len = drinfo->oldnamelen;
11406 + AuIOErr("broken drinfo %pD2\n", file);
11411 + drinfo->ino = be64_to_cpu((__force __be64)drinfo->ino);
11412 + if (unlikely(h_ino && drinfo->ino != h_ino)) {
11413 + AuDbg("ignored i%llu, i%llu, %pD2\n",
11414 + (unsigned long long)drinfo->ino,
11415 + (unsigned long long)h_ino, file);
11416 + goto out; /* success */
11419 + ret = kmalloc(sizeof(*ret) + len, GFP_NOFS);
11420 + if (unlikely(!ret)) {
11421 + ret = ERR_PTR(-ENOMEM);
11422 + AuTraceErrPtr(ret);
11427 + ssz = vfsub_read_k(file, (void *)ret->oldname, len, &pos);
11428 + if (unlikely(ssz != len)) {
11429 + au_kfree_rcu(ret);
11430 + ret = ERR_PTR(-EIO);
11431 + AuIOErr("ssz %zd, %u, %pD2\n", ssz, len, file);
11435 + AuDbg("oldname %.*s\n", ret->oldnamelen, ret->oldname);
11441 +/* ---------------------------------------------------------------------- */
11443 +/* in order to be revertible */
11444 +struct au_drinfo_rev_elm {
11446 + struct dentry *info_dentry;
11447 + struct au_drinfo *info_last;
11450 +struct au_drinfo_rev {
11451 + unsigned char already;
11452 + aufs_bindex_t nelm;
11453 + struct au_drinfo_rev_elm elm[];
11456 +/* todo: isn't it too large? */
11457 +struct au_drinfo_store {
11458 + struct path h_ppath;
11459 + struct dentry *h_dentry;
11460 + struct au_drinfo_fdata *fdata;
11461 + char *infoname; /* inside of whname, just after PFX */
11462 + char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ];
11463 + aufs_bindex_t btgt, btail;
11464 + unsigned char no_sio,
11465 + allocated, /* current size of *fdata */
11466 + infonamelen, /* room size for p */
11467 + whnamelen, /* length of the generated name */
11468 + renameback; /* renamed back */
11471 +/* on rename(2) error, the caller should revert it using @elm */
11472 +static int au_drinfo_do_store(struct au_drinfo_store *w,
11473 + struct au_drinfo_rev_elm *elm)
11478 + struct path infopath = {
11479 + .mnt = w->h_ppath.mnt
11481 + struct inode *h_dir, *h_inode, *delegated;
11482 + struct file *infofile;
11483 + struct qstr *qname;
11486 + && memcmp(elm, page_address(ZERO_PAGE(0)), sizeof(*elm)));
11488 + infopath.dentry = vfsub_lookup_one_len(w->whname, &w->h_ppath,
11490 + AuTraceErrPtr(infopath.dentry);
11491 + if (IS_ERR(infopath.dentry)) {
11492 + err = PTR_ERR(infopath.dentry);
11497 + h_dir = d_inode(w->h_ppath.dentry);
11498 + if (elm && d_is_negative(infopath.dentry)) {
11499 + err = vfsub_create(h_dir, &infopath, 0600, /*want_excl*/true);
11501 + if (unlikely(err))
11503 + elm->created = 1;
11504 + elm->info_dentry = dget(infopath.dentry);
11507 + infofile = vfsub_dentry_open(&infopath, O_RDWR);
11508 + AuTraceErrPtr(infofile);
11509 + if (IS_ERR(infofile)) {
11510 + err = PTR_ERR(infofile);
11514 + h_inode = d_inode(infopath.dentry);
11515 + if (elm && i_size_read(h_inode)) {
11516 + h_inode = d_inode(w->h_dentry);
11517 + elm->info_last = au_drinfo_read_k(infofile, h_inode->i_ino);
11518 + AuTraceErrPtr(elm->info_last);
11519 + if (IS_ERR(elm->info_last)) {
11520 + err = PTR_ERR(elm->info_last);
11521 + elm->info_last = NULL;
11522 + AuDebugOn(elm->info_dentry);
11527 + if (elm && w->renameback) {
11528 + delegated = NULL;
11529 + err = vfsub_unlink(h_dir, &infopath, &delegated, /*force*/0);
11531 + if (unlikely(err == -EWOULDBLOCK))
11537 + qname = &w->h_dentry->d_name;
11538 + len = sizeof(*w->fdata) + qname->len;
11540 + len = sizeof(*w->fdata) + w->fdata->drinfo.oldnamelen;
11541 + ssz = vfsub_write_k(infofile, w->fdata, len, &pos);
11542 + if (ssz == len) {
11543 + AuDbg("hi%llu, %.*s\n", w->fdata->drinfo.ino,
11544 + w->fdata->drinfo.oldnamelen, w->fdata->drinfo.oldname);
11545 + goto out_fput; /* success */
11550 + /* the caller should revert it using @elm */
11556 + dput(infopath.dentry);
11562 +struct au_call_drinfo_do_store_args {
11564 + struct au_drinfo_store *w;
11565 + struct au_drinfo_rev_elm *elm;
11568 +static void au_call_drinfo_do_store(void *args)
11570 + struct au_call_drinfo_do_store_args *a = args;
11572 + *a->errp = au_drinfo_do_store(a->w, a->elm);
11575 +static int au_drinfo_store_sio(struct au_drinfo_store *w,
11576 + struct au_drinfo_rev_elm *elm)
11578 + int err, wkq_err;
11581 + err = au_drinfo_do_store(w, elm);
11583 + struct au_call_drinfo_do_store_args a = {
11588 + wkq_err = au_wkq_wait(au_call_drinfo_do_store, &a);
11589 + if (unlikely(wkq_err))
11597 +static int au_drinfo_store_work_init(struct au_drinfo_store *w,
11598 + aufs_bindex_t btgt)
11602 + memset(w, 0, sizeof(*w));
11603 + w->allocated = roundup_pow_of_two(sizeof(*w->fdata) + 40);
11604 + strcpy(w->whname, AUFS_WH_DR_INFO_PFX);
11605 + w->infoname = w->whname + sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11606 + w->infonamelen = sizeof(w->whname) - sizeof(AUFS_WH_DR_INFO_PFX);
11608 + w->no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
11611 + w->fdata = kcalloc(1, w->allocated, GFP_NOFS);
11612 + if (unlikely(!w->fdata)) {
11616 + w->fdata->magic = (__force uint32_t)htonl(AUFS_DRINFO_MAGIC_V1);
11623 +static void au_drinfo_store_work_fin(struct au_drinfo_store *w)
11625 + au_kfree_rcu(w->fdata);
11628 +static void au_drinfo_store_rev(struct au_drinfo_rev *rev,
11629 + struct au_drinfo_store *w)
11631 + struct au_drinfo_rev_elm *elm;
11632 + struct inode *h_dir, *delegated;
11634 + struct path infopath = {
11635 + .mnt = w->h_ppath.mnt
11638 + h_dir = d_inode(w->h_ppath.dentry);
11639 + IMustLock(h_dir);
11643 + for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11644 + AuDebugOn(elm->created && elm->info_last);
11645 + if (elm->created) {
11647 + delegated = NULL;
11648 + infopath.dentry = elm->info_dentry;
11649 + err = vfsub_unlink(h_dir, &infopath, &delegated,
11652 + if (unlikely(err == -EWOULDBLOCK))
11654 + dput(elm->info_dentry);
11655 + } else if (elm->info_last) {
11657 + w->fdata->drinfo = *elm->info_last;
11658 + memcpy(w->fdata->drinfo.oldname,
11659 + elm->info_last->oldname,
11660 + elm->info_last->oldnamelen);
11661 + err = au_drinfo_store_sio(w, /*elm*/NULL);
11662 + au_kfree_rcu(elm->info_last);
11664 + if (unlikely(err))
11665 + AuIOErr("%d, %s\n", err, w->whname);
11666 + /* go on even if err */
11670 +/* caller has to call au_dr_rename_fin() later */
11671 +static int au_drinfo_store(struct dentry *dentry, aufs_bindex_t btgt,
11672 + struct qstr *dst_name, void *_rev)
11674 + int err, sz, nelm;
11675 + aufs_bindex_t bindex, btail;
11676 + struct au_drinfo_store work;
11677 + struct au_drinfo_rev *rev, **p;
11678 + struct au_drinfo_rev_elm *elm;
11679 + struct super_block *sb;
11680 + struct au_branch *br;
11681 + struct au_hinode *hdir;
11683 + err = au_drinfo_store_work_init(&work, btgt);
11685 + if (unlikely(err))
11689 + btail = au_dbtaildir(dentry);
11690 + nelm = btail - btgt;
11691 + sz = sizeof(*rev) + sizeof(*elm) * nelm;
11692 + rev = kcalloc(1, sz, GFP_NOFS);
11693 + if (unlikely(!rev)) {
11697 + rev->nelm = nelm;
11703 + sb = dentry->d_sb;
11704 + work.h_ppath.dentry = au_h_dptr(dentry, btgt);
11705 + work.h_ppath.mnt = au_sbr_mnt(sb, btgt);
11706 + hdir = au_hi(d_inode(dentry), btgt);
11707 + au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11708 + for (bindex = btgt + 1; bindex <= btail; bindex++, elm++) {
11709 + work.h_dentry = au_h_dptr(dentry, bindex);
11710 + if (!work.h_dentry)
11713 + err = au_drinfo_construct(&work.fdata, work.h_dentry,
11714 + &work.allocated);
11716 + if (unlikely(err))
11719 + work.renameback = au_qstreq(&work.h_dentry->d_name, dst_name);
11720 + br = au_sbr(sb, bindex);
11721 + work.whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11722 + work.whnamelen += au_drinfo_name(br, work.infoname,
11723 + work.infonamelen);
11724 + AuDbg("whname %.*s, i%llu, %.*s\n",
11725 + work.whnamelen, work.whname,
11726 + be64_to_cpu((__force __be64)work.fdata->drinfo.ino),
11727 + work.fdata->drinfo.oldnamelen,
11728 + work.fdata->drinfo.oldname);
11730 + err = au_drinfo_store_sio(&work, elm);
11732 + if (unlikely(err))
11735 + if (unlikely(err)) {
11736 + /* revert all drinfo */
11737 + au_drinfo_store_rev(rev, &work);
11738 + au_kfree_try_rcu(rev);
11741 + au_hn_inode_unlock(hdir);
11744 + au_drinfo_store_work_fin(&work);
11749 +/* ---------------------------------------------------------------------- */
11751 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
11752 + struct qstr *dst_name, void *_rev)
11754 + int err, already;
11756 + struct super_block *sb;
11757 + struct au_branch *br;
11758 + struct au_dr_br *dr;
11759 + struct dentry *h_dentry;
11760 + struct inode *h_inode;
11761 + struct au_dr_hino *ent;
11762 + struct au_drinfo_rev *rev, **p;
11764 + AuDbg("bindex %d\n", bindex);
11767 + ent = kmalloc(sizeof(*ent), GFP_NOFS);
11768 + if (unlikely(!ent))
11772 + br = au_sbr(sb, bindex);
11773 + dr = &br->br_dirren;
11774 + h_dentry = au_h_dptr(src, bindex);
11775 + h_inode = d_inode(h_dentry);
11776 + ino = h_inode->i_ino;
11777 + ent->dr_h_ino = ino;
11778 + already = au_dr_hino_test_add(dr, ino, ent);
11779 + AuDbg("b%d, hi%llu, already %d\n",
11780 + bindex, (unsigned long long)ino, already);
11782 + err = au_drinfo_store(src, bindex, dst_name, _rev);
11787 + rev->already = already;
11788 + goto out; /* success */
11793 + au_dr_hino_del(dr, ent);
11794 + au_kfree_rcu(ent);
11801 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11803 + struct au_drinfo_rev *rev;
11804 + struct au_drinfo_rev_elm *elm;
11809 + for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11810 + dput(elm->info_dentry);
11811 + au_kfree_rcu(elm->info_last);
11813 + au_kfree_try_rcu(rev);
11816 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11819 + struct au_drinfo_store work;
11820 + struct au_drinfo_rev *rev = _rev;
11821 + struct super_block *sb;
11822 + struct au_branch *br;
11823 + struct inode *h_inode;
11824 + struct au_dr_br *dr;
11825 + struct au_dr_hino *ent;
11827 + err = au_drinfo_store_work_init(&work, btgt);
11828 + if (unlikely(err))
11832 + br = au_sbr(sb, btgt);
11833 + work.h_ppath.dentry = au_h_dptr(src, btgt);
11834 + work.h_ppath.mnt = au_br_mnt(br);
11835 + au_drinfo_store_rev(rev, &work);
11836 + au_drinfo_store_work_fin(&work);
11837 + if (rev->already)
11840 + dr = &br->br_dirren;
11841 + h_inode = d_inode(work.h_ppath.dentry);
11842 + ent = au_dr_hino_find(dr, h_inode->i_ino);
11844 + au_dr_hino_del(dr, ent);
11845 + au_kfree_rcu(ent);
11848 + au_kfree_try_rcu(rev);
11849 + if (unlikely(err))
11850 + pr_err("failed to remove dirren info\n");
11853 +/* ---------------------------------------------------------------------- */
11855 +static struct au_drinfo *au_drinfo_do_load(struct path *h_ppath,
11856 + char *whname, int whnamelen,
11857 + struct dentry **info_dentry)
11859 + struct au_drinfo *drinfo;
11861 + struct inode *h_dir;
11862 + struct path infopath;
11865 + AuDbg("%pd/%.*s\n", h_ppath->dentry, whnamelen, whname);
11867 + *info_dentry = NULL;
11870 + h_dir = d_inode(h_ppath->dentry);
11871 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
11872 + infopath.dentry = vfsub_lookup_one_len(whname, h_ppath, whnamelen);
11873 + if (IS_ERR(infopath.dentry)) {
11874 + drinfo = (void *)infopath.dentry;
11878 + if (d_is_negative(infopath.dentry))
11879 + goto out_dput; /* success */
11881 + infopath.mnt = h_ppath->mnt;
11882 + f = vfsub_dentry_open(&infopath, O_RDONLY);
11883 + inode_unlock_shared(h_dir);
11886 + drinfo = (void *)f;
11890 + drinfo = au_drinfo_read_k(f, /*h_ino*/0);
11891 + if (IS_ERR_OR_NULL(drinfo))
11894 + AuDbg("oldname %.*s\n", drinfo->oldnamelen, drinfo->oldname);
11895 + *info_dentry = dget(infopath.dentry); /* keep it alive */
11900 + dput(infopath.dentry);
11903 + inode_unlock_shared(h_dir);
11904 + AuTraceErrPtr(drinfo);
11908 +struct au_drinfo_do_load_args {
11909 + struct au_drinfo **drinfop;
11910 + struct path *h_ppath;
11913 + struct dentry **info_dentry;
11916 +static void au_call_drinfo_do_load(void *args)
11918 + struct au_drinfo_do_load_args *a = args;
11920 + *a->drinfop = au_drinfo_do_load(a->h_ppath, a->whname, a->whnamelen,
11924 +struct au_drinfo_load {
11925 + struct path h_ppath;
11926 + struct qstr *qname;
11927 + unsigned char no_sio;
11929 + aufs_bindex_t ninfo;
11930 + struct au_drinfo **drinfo;
11933 +static int au_drinfo_load(struct au_drinfo_load *w, aufs_bindex_t bindex,
11934 + struct au_branch *br)
11936 + int err, wkq_err, whnamelen, e;
11937 + char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ]
11938 + = AUFS_WH_DR_INFO_PFX;
11939 + struct au_drinfo *drinfo;
11940 + struct qstr oldname;
11941 + struct inode *h_dir, *delegated;
11942 + struct dentry *info_dentry;
11943 + struct path infopath;
11945 + whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11946 + whnamelen += au_drinfo_name(br, whname + whnamelen,
11947 + sizeof(whname) - whnamelen);
11949 + drinfo = au_drinfo_do_load(&w->h_ppath, whname, whnamelen,
11952 + struct au_drinfo_do_load_args args = {
11953 + .drinfop = &drinfo,
11954 + .h_ppath = &w->h_ppath,
11955 + .whname = whname,
11956 + .whnamelen = whnamelen,
11957 + .info_dentry = &info_dentry
11959 + wkq_err = au_wkq_wait(au_call_drinfo_do_load, &args);
11960 + if (unlikely(wkq_err))
11961 + drinfo = ERR_PTR(wkq_err);
11963 + err = PTR_ERR(drinfo);
11964 + if (IS_ERR_OR_NULL(drinfo))
11968 + oldname.len = drinfo->oldnamelen;
11969 + oldname.name = drinfo->oldname;
11970 + if (au_qstreq(w->qname, &oldname)) {
11971 + /* the name is renamed back */
11972 + au_kfree_rcu(drinfo);
11975 + infopath.dentry = info_dentry;
11976 + infopath.mnt = w->h_ppath.mnt;
11977 + h_dir = d_inode(w->h_ppath.dentry);
11978 + delegated = NULL;
11979 + inode_lock_nested(h_dir, AuLsc_I_PARENT);
11980 + e = vfsub_unlink(h_dir, &infopath, &delegated, !w->no_sio);
11981 + inode_unlock(h_dir);
11983 + AuIOErr("ignored %d, %pd2\n", e, &infopath.dentry);
11984 + if (unlikely(e == -EWOULDBLOCK))
11987 + au_kfree_rcu(w->drinfo[bindex]);
11988 + w->drinfo[bindex] = drinfo;
11989 + dput(info_dentry);
11996 +/* ---------------------------------------------------------------------- */
11998 +static void au_dr_lkup_free(struct au_drinfo **drinfo, int n)
12000 + struct au_drinfo **p = drinfo;
12003 + au_kfree_rcu(*drinfo++);
12004 + au_kfree_try_rcu(p);
12007 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12008 + aufs_bindex_t btgt)
12011 + struct au_drinfo_load w;
12012 + aufs_bindex_t bindex, bbot;
12013 + struct au_branch *br;
12014 + struct inode *h_dir;
12015 + struct au_dr_hino *ent;
12016 + struct super_block *sb;
12018 + AuDbg("%.*s, name %.*s, whname %.*s, b%d\n",
12019 + AuLNPair(&dentry->d_name), AuLNPair(&lkup->dirren.dr_name),
12020 + AuLNPair(&lkup->whname), btgt);
12022 + sb = dentry->d_sb;
12023 + bbot = au_sbbot(sb);
12024 + w.ninfo = bbot + 1;
12025 + if (!lkup->dirren.drinfo) {
12026 + lkup->dirren.drinfo = kcalloc(w.ninfo,
12027 + sizeof(*lkup->dirren.drinfo),
12029 + if (unlikely(!lkup->dirren.drinfo)) {
12033 + lkup->dirren.ninfo = w.ninfo;
12035 + w.drinfo = lkup->dirren.drinfo;
12036 + w.no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
12037 + w.h_ppath.dentry = au_h_dptr(dentry, btgt);
12038 + AuDebugOn(!w.h_ppath.dentry);
12039 + w.h_ppath.mnt = au_sbr_mnt(sb, btgt);
12040 + w.qname = &dentry->d_name;
12043 + for (bindex = btgt + 1; bindex <= bbot; bindex++) {
12044 + br = au_sbr(sb, bindex);
12045 + err = au_drinfo_load(&w, bindex, br);
12046 + if (unlikely(err))
12048 + if (w.drinfo[bindex])
12052 + br = au_sbr(sb, btgt);
12053 + h_dir = d_inode(w.h_ppath.dentry);
12054 + ent = au_dr_hino_find(&br->br_dirren, h_dir->i_ino);
12056 + au_dr_hino_del(&br->br_dirren, ent);
12057 + au_kfree_rcu(ent);
12059 + goto out; /* success */
12062 + au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12063 + lkup->dirren.ninfo = 0;
12064 + lkup->dirren.drinfo = NULL;
12070 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup)
12072 + au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12075 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt)
12078 + struct au_drinfo *drinfo;
12081 + if (!lkup->dirren.drinfo)
12083 + AuDebugOn(lkup->dirren.ninfo <= btgt);
12084 + drinfo = lkup->dirren.drinfo[btgt];
12088 + au_kfree_try_rcu(lkup->whname.name);
12089 + lkup->whname.name = NULL;
12090 + lkup->dirren.dr_name.len = drinfo->oldnamelen;
12091 + lkup->dirren.dr_name.name = drinfo->oldname;
12092 + lkup->name = &lkup->dirren.dr_name;
12093 + err = au_wh_name_alloc(&lkup->whname, lkup->name);
12095 + AuDbg("name %.*s, whname %.*s, b%d\n",
12096 + AuLNPair(lkup->name), AuLNPair(&lkup->whname),
12104 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12108 + struct au_drinfo *drinfo;
12111 + if (!lkup->dirren.drinfo)
12113 + AuDebugOn(lkup->dirren.ninfo <= bindex);
12114 + drinfo = lkup->dirren.drinfo[bindex];
12118 + match = (drinfo->ino == h_ino);
12119 + AuDbg("match %d\n", match);
12125 +/* ---------------------------------------------------------------------- */
12127 +int au_dr_opt_set(struct super_block *sb)
12130 + aufs_bindex_t bindex, bbot;
12131 + struct au_branch *br;
12134 + bbot = au_sbbot(sb);
12135 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
12136 + br = au_sbr(sb, bindex);
12137 + err = au_dr_hino(sb, bindex, /*br*/NULL, &br->br_path);
12143 +int au_dr_opt_flush(struct super_block *sb)
12146 + aufs_bindex_t bindex, bbot;
12147 + struct au_branch *br;
12150 + bbot = au_sbbot(sb);
12151 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
12152 + br = au_sbr(sb, bindex);
12153 + if (au_br_writable(br->br_perm))
12154 + err = au_dr_hino(sb, bindex, /*br*/NULL, /*path*/NULL);
12160 +int au_dr_opt_clr(struct super_block *sb, int no_flush)
12163 + aufs_bindex_t bindex, bbot;
12164 + struct au_branch *br;
12168 + err = au_dr_opt_flush(sb);
12169 + if (unlikely(err))
12173 + bbot = au_sbbot(sb);
12174 + for (bindex = 0; bindex <= bbot; bindex++) {
12175 + br = au_sbr(sb, bindex);
12176 + au_dr_hino_free(&br->br_dirren);
12182 diff -urN /usr/share/empty/fs/aufs/dirren.h linux/fs/aufs/dirren.h
12183 --- /usr/share/empty/fs/aufs/dirren.h 1970-01-01 01:00:00.000000000 +0100
12184 +++ linux/fs/aufs/dirren.h 2022-03-21 14:49:05.723299677 +0100
12186 +/* SPDX-License-Identifier: GPL-2.0 */
12188 + * Copyright (C) 2017-2021 Junjiro R. Okajima
12190 + * This program, aufs is free software; you can redistribute it and/or modify
12191 + * it under the terms of the GNU General Public License as published by
12192 + * the Free Software Foundation; either version 2 of the License, or
12193 + * (at your option) any later version.
12195 + * This program is distributed in the hope that it will be useful,
12196 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12197 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12198 + * GNU General Public License for more details.
12200 + * You should have received a copy of the GNU General Public License
12201 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12205 + * renamed dir info
12208 +#ifndef __AUFS_DIRREN_H__
12209 +#define __AUFS_DIRREN_H__
12213 +#include <linux/dcache.h>
12214 +#include <linux/statfs.h>
12215 +#include <linux/uuid.h>
12218 +#define AuDirren_NHASH 100
12220 +#ifdef CONFIG_AUFS_DIRREN
12221 +enum au_brid_type {
12228 +struct au_dr_brid {
12229 + enum au_brid_type type;
12231 + uuid_t uuid; /* unimplemented yet */
12237 +/* 20 is the max digits length of ulong 64 */
12238 +/* brid-type "_" uuid "_" inum */
12239 +#define AUFS_DIRREN_FNAME_SZ (1 + 1 + UUID_STRING_LEN + 20)
12240 +#define AUFS_DIRREN_ENV_VAL_SZ (AUFS_DIRREN_FNAME_SZ + 1 + 20)
12242 +struct au_dr_hino {
12243 + struct hlist_bl_node dr_hnode;
12248 + struct hlist_bl_head dr_h_ino[AuDirren_NHASH];
12249 + struct au_dr_brid dr_brid;
12252 +struct au_dr_lookup {
12253 + /* dr_name is pointed by struct au_do_lookup_args.name */
12254 + struct qstr dr_name; /* subset of dr_info */
12255 + aufs_bindex_t ninfo;
12256 + struct au_drinfo **drinfo;
12259 +struct au_dr_hino;
12261 +struct au_dr_br { };
12262 +struct au_dr_lookup { };
12265 +/* ---------------------------------------------------------------------- */
12268 +struct au_do_lookup_args;
12270 +#ifdef CONFIG_AUFS_DIRREN
12271 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t h_ino,
12272 + struct au_dr_hino *add_ent);
12273 +void au_dr_hino_free(struct au_dr_br *dr);
12274 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
12275 + const struct path *path);
12276 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br);
12277 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
12278 + struct qstr *dst_name, void *_rev);
12279 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *rev);
12280 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t bindex, void *rev);
12281 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12282 + aufs_bindex_t bindex);
12283 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12284 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12286 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup);
12287 +int au_dr_opt_set(struct super_block *sb);
12288 +int au_dr_opt_flush(struct super_block *sb);
12289 +int au_dr_opt_clr(struct super_block *sb, int no_flush);
12291 +AuStubInt0(au_dr_hino_test_add, struct au_dr_br *dr, ino_t h_ino,
12292 + struct au_dr_hino *add_ent);
12293 +AuStubVoid(au_dr_hino_free, struct au_dr_br *dr);
12294 +AuStubInt0(au_dr_br_init, struct super_block *sb, struct au_branch *br,
12295 + const struct path *path);
12296 +AuStubInt0(au_dr_br_fin, struct super_block *sb, struct au_branch *br);
12297 +AuStubInt0(au_dr_rename, struct dentry *src, aufs_bindex_t bindex,
12298 + struct qstr *dst_name, void *_rev);
12299 +AuStubVoid(au_dr_rename_fin, struct dentry *src, aufs_bindex_t btgt, void *rev);
12300 +AuStubVoid(au_dr_rename_rev, struct dentry *src, aufs_bindex_t bindex,
12302 +AuStubInt0(au_dr_lkup, struct au_do_lookup_args *lkup, struct dentry *dentry,
12303 + aufs_bindex_t bindex);
12304 +AuStubInt0(au_dr_lkup_name, struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12305 +AuStubInt0(au_dr_lkup_h_ino, struct au_do_lookup_args *lkup,
12306 + aufs_bindex_t bindex, ino_t h_ino);
12307 +AuStubVoid(au_dr_lkup_fin, struct au_do_lookup_args *lkup);
12308 +AuStubInt0(au_dr_opt_set, struct super_block *sb);
12309 +AuStubInt0(au_dr_opt_flush, struct super_block *sb);
12310 +AuStubInt0(au_dr_opt_clr, struct super_block *sb, int no_flush);
12313 +/* ---------------------------------------------------------------------- */
12315 +#ifdef CONFIG_AUFS_DIRREN
12316 +static inline int au_dr_ihash(ino_t h_ino)
12318 + return h_ino % AuDirren_NHASH;
12321 +AuStubInt0(au_dr_ihash, ino_t h_ino);
12324 +#endif /* __KERNEL__ */
12325 +#endif /* __AUFS_DIRREN_H__ */
12326 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
12327 --- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
12328 +++ linux/fs/aufs/dynop.c 2022-10-03 07:51:14.926209260 +0200
12330 +// SPDX-License-Identifier: GPL-2.0
12332 + * Copyright (C) 2010-2021 Junjiro R. Okajima
12334 + * This program, aufs is free software; you can redistribute it and/or modify
12335 + * it under the terms of the GNU General Public License as published by
12336 + * the Free Software Foundation; either version 2 of the License, or
12337 + * (at your option) any later version.
12339 + * This program is distributed in the hope that it will be useful,
12340 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12341 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12342 + * GNU General Public License for more details.
12344 + * You should have received a copy of the GNU General Public License
12345 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12349 + * dynamically customizable operations for regular files
12354 +#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
12357 + * How large will these lists be?
12358 + * Usually just a few elements, 20-30 at most for each, I guess.
12360 +static struct hlist_bl_head dynop[AuDyLast];
12362 +static struct au_dykey *dy_gfind_get(struct hlist_bl_head *hbl,
12363 + const void *h_op)
12365 + struct au_dykey *key, *tmp;
12366 + struct hlist_bl_node *pos;
12369 + hlist_bl_lock(hbl);
12370 + hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12371 + if (tmp->dk_op.dy_hop == h_op) {
12372 + if (kref_get_unless_zero(&tmp->dk_kref))
12376 + hlist_bl_unlock(hbl);
12381 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
12383 + struct au_dykey **k, *found;
12384 + const void *h_op = key->dk_op.dy_hop;
12388 + k = br->br_dykey;
12389 + for (i = 0; i < AuBrDynOp; i++)
12391 + if (k[i]->dk_op.dy_hop == h_op) {
12398 + spin_lock(&br->br_dykey_lock);
12399 + for (; i < AuBrDynOp; i++)
12401 + if (k[i]->dk_op.dy_hop == h_op) {
12409 + spin_unlock(&br->br_dykey_lock);
12410 + BUG_ON(i == AuBrDynOp); /* expand the array */
12416 +/* kref_get() if @key is already added */
12417 +static struct au_dykey *dy_gadd(struct hlist_bl_head *hbl, struct au_dykey *key)
12419 + struct au_dykey *tmp, *found;
12420 + struct hlist_bl_node *pos;
12421 + const void *h_op = key->dk_op.dy_hop;
12424 + hlist_bl_lock(hbl);
12425 + hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12426 + if (tmp->dk_op.dy_hop == h_op) {
12427 + if (kref_get_unless_zero(&tmp->dk_kref))
12432 + hlist_bl_add_head(&key->dk_hnode, hbl);
12433 + hlist_bl_unlock(hbl);
12440 +static void dy_free_rcu(struct rcu_head *rcu)
12442 + struct au_dykey *key;
12444 + key = container_of(rcu, struct au_dykey, dk_rcu);
12449 +static void dy_free(struct kref *kref)
12451 + struct au_dykey *key;
12452 + struct hlist_bl_head *hbl;
12454 + key = container_of(kref, struct au_dykey, dk_kref);
12455 + hbl = dynop + key->dk_op.dy_type;
12456 + au_hbl_del(&key->dk_hnode, hbl);
12457 + call_rcu(&key->dk_rcu, dy_free_rcu);
12460 +void au_dy_put(struct au_dykey *key)
12462 + kref_put(&key->dk_kref, dy_free);
12465 +/* ---------------------------------------------------------------------- */
12467 +#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
12469 +#ifdef CONFIG_AUFS_DEBUG
12470 +#define DyDbgDeclare(cnt) unsigned int cnt = 0
12471 +#define DyDbgInc(cnt) do { cnt++; } while (0)
12473 +#define DyDbgDeclare(cnt) do {} while (0)
12474 +#define DyDbgInc(cnt) do {} while (0)
12477 +#define DySet(func, dst, src, h_op, h_sb) do { \
12479 + if (h_op->func) { \
12481 + dst.func = src.func; \
12483 + AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
12487 +#define DySetForce(func, dst, src) do { \
12488 + AuDebugOn(!src.func); \
12490 + dst.func = src.func; \
12493 +#define DySetAop(func) \
12494 + DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
12495 +#define DySetAopForce(func) \
12496 + DySetForce(func, dyaop->da_op, aufs_aop)
12498 +static void dy_aop(struct au_dykey *key, const void *h_op,
12499 + struct super_block *h_sb __maybe_unused)
12501 + struct au_dyaop *dyaop = (void *)key;
12502 + const struct address_space_operations *h_aop = h_op;
12503 + DyDbgDeclare(cnt);
12505 + AuDbg("%s\n", au_sbtype(h_sb));
12507 + DySetAop(writepage);
12508 + DySetAopForce(read_folio); /* force */
12509 + DySetAop(writepages);
12510 + DySetAop(dirty_folio);
12511 + DySetAop(invalidate_folio);
12512 + DySetAop(readahead);
12513 + DySetAop(write_begin);
12514 + DySetAop(write_end);
12516 + DySetAop(release_folio);
12517 + DySetAop(free_folio);
12518 + /* this one will be changed according to an aufs mount option */
12519 + DySetAop(direct_IO);
12520 + DySetAop(migrate_folio);
12521 + DySetAop(launder_folio);
12522 + DySetAop(is_partially_uptodate);
12523 + DySetAop(is_dirty_writeback);
12524 + DySetAop(error_remove_page);
12525 + DySetAop(swap_activate);
12526 + DySetAop(swap_deactivate);
12527 + DySetAop(swap_rw);
12529 + DyDbgSize(cnt, *h_aop);
12532 +/* ---------------------------------------------------------------------- */
12534 +static void dy_bug(struct kref *kref)
12539 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
12541 + struct au_dykey *key, *old;
12542 + struct hlist_bl_head *hbl;
12545 + void (*set)(struct au_dykey *key, const void *h_op,
12546 + struct super_block *h_sb __maybe_unused);
12548 + static const struct op a[] = {
12550 + .sz = sizeof(struct au_dyaop),
12554 + const struct op *p;
12556 + hbl = dynop + op->dy_type;
12557 + key = dy_gfind_get(hbl, op->dy_hop);
12559 + goto out_add; /* success */
12561 + p = a + op->dy_type;
12562 + key = kzalloc(p->sz, GFP_NOFS);
12563 + if (unlikely(!key)) {
12564 + key = ERR_PTR(-ENOMEM);
12568 + key->dk_op.dy_hop = op->dy_hop;
12569 + kref_init(&key->dk_kref);
12570 + p->set(key, op->dy_hop, au_br_sb(br));
12571 + old = dy_gadd(hbl, key);
12573 + au_kfree_rcu(key);
12578 + old = dy_bradd(br, key);
12580 + /* its ref-count should never be zero here */
12581 + kref_put(&key->dk_kref, dy_bug);
12586 +/* ---------------------------------------------------------------------- */
12588 + * Aufs prohibits O_DIRECT by default even if the branch supports it.
12589 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
12590 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
12591 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
12592 + * See the aufs manual in detail.
12594 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
12597 + dyaop->da_op.direct_IO = NULL;
12599 + dyaop->da_op.direct_IO = aufs_aop.direct_IO;
12602 +static struct au_dyaop *dy_aget(struct au_branch *br,
12603 + const struct address_space_operations *h_aop,
12606 + struct au_dyaop *dyaop;
12607 + struct au_dynop op;
12609 + op.dy_type = AuDy_AOP;
12610 + op.dy_haop = h_aop;
12611 + dyaop = (void *)dy_get(&op, br);
12612 + if (IS_ERR(dyaop))
12614 + dy_adx(dyaop, do_dx);
12620 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12621 + struct inode *h_inode)
12624 + struct super_block *sb;
12625 + struct au_branch *br;
12626 + struct au_dyaop *dyaop;
12628 + AuDebugOn(!S_ISREG(h_inode->i_mode));
12629 + IiMustWriteLock(inode);
12631 + sb = inode->i_sb;
12632 + br = au_sbr(sb, bindex);
12633 + do_dx = !!au_opt_test(au_mntflags(sb), DIO);
12634 + dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
12635 + err = PTR_ERR(dyaop);
12636 + if (IS_ERR(dyaop))
12637 + /* unnecessary to call dy_fput() */
12641 + inode->i_mapping->a_ops = &dyaop->da_op;
12648 + * Is it safe to replace a_ops during the inode/file is in operation?
12649 + * Yes, I hope so.
12651 +int au_dy_irefresh(struct inode *inode)
12654 + aufs_bindex_t btop;
12655 + struct inode *h_inode;
12658 + if (S_ISREG(inode->i_mode)) {
12659 + btop = au_ibtop(inode);
12660 + h_inode = au_h_iptr(inode, btop);
12661 + err = au_dy_iaop(inode, btop, h_inode);
12666 +void au_dy_arefresh(int do_dx)
12668 + struct hlist_bl_head *hbl;
12669 + struct hlist_bl_node *pos;
12670 + struct au_dykey *key;
12672 + hbl = dynop + AuDy_AOP;
12673 + hlist_bl_lock(hbl);
12674 + hlist_bl_for_each_entry(key, pos, hbl, dk_hnode)
12675 + dy_adx((void *)key, do_dx);
12676 + hlist_bl_unlock(hbl);
12679 +/* ---------------------------------------------------------------------- */
12681 +void __init au_dy_init(void)
12685 + for (i = 0; i < AuDyLast; i++)
12686 + INIT_HLIST_BL_HEAD(dynop + i);
12689 +void au_dy_fin(void)
12693 + for (i = 0; i < AuDyLast; i++)
12694 + WARN_ON(!hlist_bl_empty(dynop + i));
12696 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
12697 --- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
12698 +++ linux/fs/aufs/dynop.h 2022-03-21 14:49:05.723299677 +0100
12700 +/* SPDX-License-Identifier: GPL-2.0 */
12702 + * Copyright (C) 2010-2021 Junjiro R. Okajima
12704 + * This program, aufs is free software; you can redistribute it and/or modify
12705 + * it under the terms of the GNU General Public License as published by
12706 + * the Free Software Foundation; either version 2 of the License, or
12707 + * (at your option) any later version.
12709 + * This program is distributed in the hope that it will be useful,
12710 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12711 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12712 + * GNU General Public License for more details.
12714 + * You should have received a copy of the GNU General Public License
12715 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12719 + * dynamically customizable operations (for regular files only)
12722 +#ifndef __AUFS_DYNOP_H__
12723 +#define __AUFS_DYNOP_H__
12727 +#include <linux/fs.h>
12728 +#include <linux/kref.h>
12730 +enum {AuDy_AOP, AuDyLast};
12735 + const void *dy_hop;
12736 + const struct address_space_operations *dy_haop;
12742 + struct hlist_bl_node dk_hnode;
12743 + struct rcu_head dk_rcu;
12745 + struct au_dynop dk_op;
12748 + * during I am in the branch local array, kref is gotten. when the
12749 + * branch is removed, kref is put.
12751 + struct kref dk_kref;
12754 +/* stop unioning since their sizes are very different from each other */
12756 + struct au_dykey da_key;
12757 + struct address_space_operations da_op; /* not const */
12759 +/* make sure that 'struct au_dykey *' can be any type */
12760 +static_assert(!offsetof(struct au_dyaop, da_key));
12762 +/* ---------------------------------------------------------------------- */
12766 +void au_dy_put(struct au_dykey *key);
12767 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12768 + struct inode *h_inode);
12769 +int au_dy_irefresh(struct inode *inode);
12770 +void au_dy_arefresh(int do_dio);
12772 +void __init au_dy_init(void);
12773 +void au_dy_fin(void);
12775 +#endif /* __KERNEL__ */
12776 +#endif /* __AUFS_DYNOP_H__ */
12777 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
12778 --- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
12779 +++ linux/fs/aufs/export.c 2022-03-21 14:49:05.723299677 +0100
12781 +// SPDX-License-Identifier: GPL-2.0
12783 + * Copyright (C) 2005-2021 Junjiro R. Okajima
12785 + * This program, aufs is free software; you can redistribute it and/or modify
12786 + * it under the terms of the GNU General Public License as published by
12787 + * the Free Software Foundation; either version 2 of the License, or
12788 + * (at your option) any later version.
12790 + * This program is distributed in the hope that it will be useful,
12791 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12792 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12793 + * GNU General Public License for more details.
12795 + * You should have received a copy of the GNU General Public License
12796 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12803 +#include <linux/exportfs.h>
12804 +#include <linux/fs_struct.h>
12805 +#include <linux/nsproxy.h>
12806 +#include <linux/random.h>
12807 +#include <linux/writeback.h>
12811 +#ifdef CONFIG_AUFS_INO_T_64
12819 +static ino_t decode_ino(__u32 *a)
12823 + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
12825 +#ifdef CONFIG_AUFS_INO_T_64
12831 +static void encode_ino(__u32 *a, ino_t ino)
12837 +#ifdef CONFIG_AUFS_INO_T_64
12842 +/* NFS file handle */
12846 +#ifdef CONFIG_AUFS_INO_T_64
12847 + /* support 64bit inode number */
12860 + Fh_ino = Fh_ino1,
12861 + Fh_dir_ino = Fh_dir_ino1
12864 +static int au_test_anon(struct dentry *dentry)
12866 + /* note: read d_flags without d_lock */
12867 + return !!(dentry->d_flags & DCACHE_DISCONNECTED);
12870 +int au_test_nfsd(void)
12873 + struct task_struct *tsk = current;
12874 + char comm[sizeof(tsk->comm)];
12877 + if (tsk->flags & PF_KTHREAD) {
12878 + get_task_comm(comm, tsk);
12879 + ret = !strcmp(comm, "nfsd");
12885 +/* ---------------------------------------------------------------------- */
12886 +/* inode generation external table */
12888 +void au_xigen_inc(struct inode *inode)
12893 + struct super_block *sb;
12894 + struct au_sbinfo *sbinfo;
12896 + sb = inode->i_sb;
12897 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
12899 + sbinfo = au_sbi(sb);
12900 + pos = inode->i_ino;
12901 + pos *= sizeof(igen);
12902 + igen = inode->i_generation + 1;
12903 + sz = xino_fwrite(sbinfo->si_xigen, &igen, sizeof(igen), &pos);
12904 + if (sz == sizeof(igen))
12905 + return; /* success */
12907 + if (unlikely(sz >= 0))
12908 + AuIOErr("xigen error (%zd)\n", sz);
12911 +int au_xigen_new(struct inode *inode)
12916 + struct super_block *sb;
12917 + struct au_sbinfo *sbinfo;
12918 + struct file *file;
12921 + /* todo: dirty, at mount time */
12922 + if (inode->i_ino == AUFS_ROOT_INO)
12924 + sb = inode->i_sb;
12925 + SiMustAnyLock(sb);
12926 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
12930 + pos = inode->i_ino;
12931 + if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
12932 + AuIOErr1("too large i%lld\n", pos);
12935 + pos *= sizeof(inode->i_generation);
12938 + sbinfo = au_sbi(sb);
12939 + file = sbinfo->si_xigen;
12942 + if (vfsub_f_size_read(file)
12943 + < pos + sizeof(inode->i_generation)) {
12944 + inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
12945 + sz = xino_fwrite(file, &inode->i_generation,
12946 + sizeof(inode->i_generation), &pos);
12948 + sz = xino_fread(file, &inode->i_generation,
12949 + sizeof(inode->i_generation), &pos);
12950 + if (sz == sizeof(inode->i_generation))
12951 + goto out; /* success */
12954 + if (unlikely(sz >= 0)) {
12956 + AuIOErr("xigen error (%zd)\n", sz);
12963 +int au_xigen_set(struct super_block *sb, struct path *path)
12966 + struct au_sbinfo *sbinfo;
12967 + struct file *file;
12969 + SiMustWriteLock(sb);
12971 + sbinfo = au_sbi(sb);
12972 + file = au_xino_create2(sb, path, sbinfo->si_xigen);
12973 + err = PTR_ERR(file);
12974 + if (IS_ERR(file))
12977 + if (sbinfo->si_xigen)
12978 + fput(sbinfo->si_xigen);
12979 + sbinfo->si_xigen = file;
12986 +void au_xigen_clr(struct super_block *sb)
12988 + struct au_sbinfo *sbinfo;
12990 + SiMustWriteLock(sb);
12992 + sbinfo = au_sbi(sb);
12993 + if (sbinfo->si_xigen) {
12994 + fput(sbinfo->si_xigen);
12995 + sbinfo->si_xigen = NULL;
12999 +/* ---------------------------------------------------------------------- */
13001 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
13004 + struct dentry *dentry, *d;
13005 + struct inode *inode;
13006 + unsigned int sigen;
13009 + inode = ilookup(sb, ino);
13013 + dentry = ERR_PTR(-ESTALE);
13014 + sigen = au_sigen(sb);
13015 + if (unlikely(au_is_bad_inode(inode)
13016 + || IS_DEADDIR(inode)
13017 + || sigen != au_iigen(inode, NULL)))
13021 + if (!dir_ino || S_ISDIR(inode->i_mode))
13022 + dentry = d_find_alias(inode);
13024 + spin_lock(&inode->i_lock);
13025 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
13026 + spin_lock(&d->d_lock);
13027 + if (!au_test_anon(d)
13028 + && d_inode(d->d_parent)->i_ino == dir_ino) {
13029 + dentry = dget_dlock(d);
13030 + spin_unlock(&d->d_lock);
13033 + spin_unlock(&d->d_lock);
13035 + spin_unlock(&inode->i_lock);
13037 + if (unlikely(dentry && au_digen_test(dentry, sigen))) {
13038 + /* need to refresh */
13046 + AuTraceErrPtr(dentry);
13050 +/* ---------------------------------------------------------------------- */
13052 +/* todo: dirty? */
13053 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
13055 +struct au_compare_mnt_args {
13057 + struct super_block *sb;
13060 + struct vfsmount *mnt;
13063 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
13065 + struct au_compare_mnt_args *a = arg;
13067 + if (mnt->mnt_sb != a->sb)
13069 + a->mnt = mntget(mnt);
13073 +static struct vfsmount *au_mnt_get(struct super_block *sb)
13076 + struct path root;
13077 + struct au_compare_mnt_args args = {
13081 + get_fs_root(current->fs, &root);
13083 + err = iterate_mounts(au_compare_mnt, &args, root.mnt);
13084 + rcu_read_unlock();
13087 + AuDebugOn(!args.mnt);
13091 +struct au_nfsd_si_lock {
13092 + unsigned int sigen;
13093 + aufs_bindex_t bindex, br_id;
13094 + unsigned char force_lock;
13097 +static int si_nfsd_read_lock(struct super_block *sb,
13098 + struct au_nfsd_si_lock *nsi_lock)
13101 + aufs_bindex_t bindex;
13103 + si_read_lock(sb, AuLock_FLUSH);
13105 + /* branch id may be wrapped around */
13107 + bindex = au_br_index(sb, nsi_lock->br_id);
13108 + if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
13109 + goto out; /* success */
13113 + if (!nsi_lock->force_lock)
13114 + si_read_unlock(sb);
13117 + nsi_lock->bindex = bindex;
13121 +struct find_name_by_ino {
13122 + struct dir_context ctx;
13123 + int called, found;
13130 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
13131 + loff_t offset, u64 ino, unsigned int d_type)
13133 + struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
13137 + if (a->ino != ino)
13140 + memcpy(a->name, name, namelen);
13141 + a->namelen = namelen;
13146 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
13147 + struct au_nfsd_si_lock *nsi_lock)
13149 + struct dentry *dentry, *parent;
13150 + struct file *file;
13151 + struct inode *dir;
13152 + struct find_name_by_ino arg = {
13154 + .actor = find_name_by_ino
13159 + parent = path->dentry;
13161 + si_read_unlock(parent->d_sb);
13162 + file = vfsub_dentry_open(path, au_dir_roflags);
13163 + dentry = (void *)file;
13164 + if (IS_ERR(file))
13167 + dentry = ERR_PTR(-ENOMEM);
13168 + arg.name = (void *)__get_free_page(GFP_NOFS);
13169 + if (unlikely(!arg.name))
13176 + err = vfsub_iterate_dir(file, &arg.ctx);
13177 + } while (!err && !arg.found && arg.called);
13178 + dentry = ERR_PTR(err);
13179 + if (unlikely(err))
13181 + /* instead of ENOENT */
13182 + dentry = ERR_PTR(-ESTALE);
13186 + /* do not call vfsub_lkup_one() */
13187 + dir = d_inode(parent);
13188 + dentry = vfsub_lookup_one_len_unlocked(arg.name, path, arg.namelen);
13189 + AuTraceErrPtr(dentry);
13190 + if (IS_ERR(dentry))
13192 + AuDebugOn(au_test_anon(dentry));
13193 + if (unlikely(d_really_is_negative(dentry))) {
13195 + dentry = ERR_PTR(-ENOENT);
13199 + free_page((unsigned long)arg.name);
13203 + if (unlikely(nsi_lock
13204 + && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
13205 + if (!IS_ERR(dentry)) {
13207 + dentry = ERR_PTR(-ESTALE);
13209 + AuTraceErrPtr(dentry);
13213 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
13215 + struct au_nfsd_si_lock *nsi_lock)
13217 + struct dentry *dentry;
13218 + struct path path;
13220 + if (dir_ino != AUFS_ROOT_INO) {
13221 + path.dentry = decode_by_ino(sb, dir_ino, 0);
13222 + dentry = path.dentry;
13223 + if (!path.dentry || IS_ERR(path.dentry))
13225 + AuDebugOn(au_test_anon(path.dentry));
13227 + path.dentry = dget(sb->s_root);
13229 + path.mnt = au_mnt_get(sb);
13230 + dentry = au_lkup_by_ino(&path, ino, nsi_lock);
13234 + AuTraceErrPtr(dentry);
13238 +/* ---------------------------------------------------------------------- */
13240 +static int h_acceptable(void *expv, struct dentry *dentry)
13245 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
13246 + char *buf, int len, struct super_block *sb)
13250 + struct path path;
13252 + p = d_path(h_rootpath, buf, len);
13257 + path.mnt = h_rootpath->mnt;
13258 + path.dentry = h_parent;
13259 + p = d_path(&path, buf, len);
13265 + path.mnt = au_mnt_get(sb);
13266 + path.dentry = sb->s_root;
13267 + p = d_path(&path, buf, len - strlen(p));
13268 + mntput(path.mnt);
13272 + p[strlen(p)] = '/';
13275 + AuTraceErrPtr(p);
13280 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
13281 + int fh_len, struct au_nfsd_si_lock *nsi_lock)
13283 + struct dentry *dentry, *h_parent, *root;
13284 + struct super_block *h_sb;
13285 + char *pathname, *p;
13286 + struct vfsmount *h_mnt;
13287 + struct au_branch *br;
13289 + struct path path;
13291 + br = au_sbr(sb, nsi_lock->bindex);
13292 + h_mnt = au_br_mnt(br);
13293 + h_sb = h_mnt->mnt_sb;
13294 + /* todo: call lower fh_to_dentry()? fh_to_parent()? */
13296 + h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
13297 + fh_len - Fh_tail, fh[Fh_h_type],
13298 + h_acceptable, /*context*/NULL);
13300 + dentry = h_parent;
13301 + if (unlikely(!h_parent || IS_ERR(h_parent))) {
13302 + AuWarn1("%s decode_fh failed, %ld\n",
13303 + au_sbtype(h_sb), PTR_ERR(h_parent));
13307 + if (unlikely(au_test_anon(h_parent))) {
13308 + AuWarn1("%s decode_fh returned a disconnected dentry\n",
13309 + au_sbtype(h_sb));
13310 + goto out_h_parent;
13313 + dentry = ERR_PTR(-ENOMEM);
13314 + pathname = (void *)__get_free_page(GFP_NOFS);
13315 + if (unlikely(!pathname))
13316 + goto out_h_parent;
13318 + root = sb->s_root;
13319 + path.mnt = h_mnt;
13320 + di_read_lock_parent(root, !AuLock_IR);
13321 + path.dentry = au_h_dptr(root, nsi_lock->bindex);
13322 + di_read_unlock(root, !AuLock_IR);
13323 + p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
13324 + dentry = (void *)p;
13326 + goto out_pathname;
13328 + si_read_unlock(sb);
13329 + err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
13330 + dentry = ERR_PTR(err);
13331 + if (unlikely(err))
13334 + dentry = ERR_PTR(-ENOENT);
13335 + AuDebugOn(au_test_anon(path.dentry));
13336 + if (unlikely(d_really_is_negative(path.dentry)))
13339 + if (ino != d_inode(path.dentry)->i_ino)
13340 + dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
13342 + dentry = dget(path.dentry);
13347 + if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
13348 + if (!IS_ERR(dentry)) {
13350 + dentry = ERR_PTR(-ESTALE);
13353 + free_page((unsigned long)pathname);
13357 + AuTraceErrPtr(dentry);
13361 +/* ---------------------------------------------------------------------- */
13363 +static struct dentry *
13364 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
13367 + struct dentry *dentry;
13368 + __u32 *fh = fid->raw;
13369 + struct au_branch *br;
13370 + ino_t ino, dir_ino;
13371 + struct au_nfsd_si_lock nsi_lock = {
13375 + dentry = ERR_PTR(-ESTALE);
13376 + /* it should never happen, but the file handle is unreliable */
13377 + if (unlikely(fh_len < Fh_tail))
13379 + nsi_lock.sigen = fh[Fh_sigen];
13380 + nsi_lock.br_id = fh[Fh_br_id];
13382 + /* branch id may be wrapped around */
13384 + if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
13386 + nsi_lock.force_lock = 1;
13388 + /* is this inode still cached? */
13389 + ino = decode_ino(fh + Fh_ino);
13390 + /* it should never happen */
13391 + if (unlikely(ino == AUFS_ROOT_INO))
13394 + dir_ino = decode_ino(fh + Fh_dir_ino);
13395 + dentry = decode_by_ino(sb, ino, dir_ino);
13396 + if (IS_ERR(dentry))
13401 + /* is the parent dir cached? */
13402 + br = au_sbr(sb, nsi_lock.bindex);
13403 + au_lcnt_inc(&br->br_nfiles);
13404 + dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
13405 + if (IS_ERR(dentry))
13410 + /* lookup path */
13411 + dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
13412 + if (IS_ERR(dentry))
13414 + if (unlikely(!dentry))
13415 + /* todo?: make it ESTALE */
13419 + if (!au_digen_test(dentry, au_sigen(sb))
13420 + && d_inode(dentry)->i_generation == fh[Fh_igen])
13421 + goto out_unlock; /* success */
13424 + dentry = ERR_PTR(-ESTALE);
13427 + au_lcnt_dec(&br->br_nfiles);
13428 + si_read_unlock(sb);
13430 + AuTraceErrPtr(dentry);
13434 +#if 0 /* reserved for future use */
13435 +/* support subtreecheck option */
13436 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
13437 + int fh_len, int fh_type)
13439 + struct dentry *parent;
13440 + __u32 *fh = fid->raw;
13443 + dir_ino = decode_ino(fh + Fh_dir_ino);
13444 + parent = decode_by_ino(sb, dir_ino, 0);
13445 + if (IS_ERR(parent))
13448 + parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
13449 + dir_ino, fh, fh_len);
13452 + AuTraceErrPtr(parent);
13457 +/* ---------------------------------------------------------------------- */
13459 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
13460 + struct inode *dir)
13463 + aufs_bindex_t bindex;
13464 + struct super_block *sb, *h_sb;
13465 + struct dentry *dentry, *parent, *h_parent;
13466 + struct inode *h_dir;
13467 + struct au_branch *br;
13470 + if (unlikely(*max_len <= Fh_tail)) {
13471 + AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
13475 + err = FILEID_ROOT;
13476 + if (inode->i_ino == AUFS_ROOT_INO) {
13477 + AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
13482 + sb = inode->i_sb;
13483 + err = si_read_lock(sb, AuLock_FLUSH);
13484 + if (unlikely(err))
13487 +#ifdef CONFIG_AUFS_DEBUG
13488 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
13489 + AuWarn1("NFS-exporting requires xino\n");
13493 + ii_read_lock_child(inode);
13494 + bindex = au_ibtop(inode);
13496 + dentry = d_find_any_alias(inode);
13497 + if (unlikely(!dentry))
13499 + AuDebugOn(au_test_anon(dentry));
13500 + parent = dget_parent(dentry);
13502 + if (unlikely(!parent))
13504 + if (d_really_is_positive(parent))
13505 + dir = d_inode(parent);
13508 + ii_read_lock_parent(dir);
13509 + h_dir = au_h_iptr(dir, bindex);
13510 + ii_read_unlock(dir);
13511 + if (unlikely(!h_dir))
13513 + h_parent = d_find_any_alias(h_dir);
13514 + if (unlikely(!h_parent))
13515 + goto out_hparent;
13518 + br = au_sbr(sb, bindex);
13519 + h_sb = au_br_sb(br);
13520 + if (unlikely(!h_sb->s_export_op)) {
13521 + AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
13522 + goto out_hparent;
13525 + fh[Fh_br_id] = br->br_id;
13526 + fh[Fh_sigen] = au_sigen(sb);
13527 + encode_ino(fh + Fh_ino, inode->i_ino);
13528 + encode_ino(fh + Fh_dir_ino, dir->i_ino);
13529 + fh[Fh_igen] = inode->i_generation;
13531 + *max_len -= Fh_tail;
13532 + fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
13534 + /*connectable or subtreecheck*/0);
13535 + err = fh[Fh_h_type];
13536 + *max_len += Fh_tail;
13537 + /* todo: macros? */
13538 + if (err != FILEID_INVALID)
13541 + AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
13548 + ii_read_unlock(inode);
13549 + si_read_unlock(sb);
13551 + if (unlikely(err < 0))
13552 + err = FILEID_INVALID;
13556 +/* ---------------------------------------------------------------------- */
13558 +static int aufs_commit_metadata(struct inode *inode)
13561 + aufs_bindex_t bindex;
13562 + struct super_block *sb;
13563 + struct inode *h_inode;
13564 + int (*f)(struct inode *inode);
13566 + sb = inode->i_sb;
13567 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13568 + ii_write_lock_child(inode);
13569 + bindex = au_ibtop(inode);
13570 + AuDebugOn(bindex < 0);
13571 + h_inode = au_h_iptr(inode, bindex);
13573 + f = h_inode->i_sb->s_export_op->commit_metadata;
13575 + err = f(h_inode);
13577 + err = sync_inode_metadata(h_inode, /*wait*/1);
13579 + au_cpup_attr_timesizes(inode);
13580 + ii_write_unlock(inode);
13581 + si_read_unlock(sb);
13585 +/* ---------------------------------------------------------------------- */
13587 +static struct export_operations aufs_export_op = {
13588 + .fh_to_dentry = aufs_fh_to_dentry,
13589 + /* .fh_to_parent = aufs_fh_to_parent, */
13590 + .encode_fh = aufs_encode_fh,
13591 + .commit_metadata = aufs_commit_metadata
13594 +void au_export_init(struct super_block *sb)
13596 + struct au_sbinfo *sbinfo;
13599 + BUILD_BUG_ON_MSG(IS_BUILTIN(CONFIG_AUFS_FS)
13600 + && IS_MODULE(CONFIG_EXPORTFS),
13601 + AUFS_NAME ": unsupported configuration "
13602 + "CONFIG_EXPORTFS=m and CONFIG_AUFS_FS=y");
13604 + sb->s_export_op = &aufs_export_op;
13605 + sbinfo = au_sbi(sb);
13606 + sbinfo->si_xigen = NULL;
13607 + get_random_bytes(&u, sizeof(u));
13608 + BUILD_BUG_ON(sizeof(u) != sizeof(int));
13609 + atomic_set(&sbinfo->si_xigen_next, u);
13611 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
13612 --- /usr/share/empty/fs/aufs/fhsm.c 1970-01-01 01:00:00.000000000 +0100
13613 +++ linux/fs/aufs/fhsm.c 2022-03-21 14:49:05.723299677 +0100
13615 +// SPDX-License-Identifier: GPL-2.0
13617 + * Copyright (C) 2011-2021 Junjiro R. Okajima
13619 + * This program, aufs is free software; you can redistribute it and/or modify
13620 + * it under the terms of the GNU General Public License as published by
13621 + * the Free Software Foundation; either version 2 of the License, or
13622 + * (at your option) any later version.
13624 + * This program is distributed in the hope that it will be useful,
13625 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13626 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13627 + * GNU General Public License for more details.
13629 + * You should have received a copy of the GNU General Public License
13630 + * along with this program; if not, write to the Free Software
13631 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
13635 + * File-based Hierarchy Storage Management
13638 +#include <linux/anon_inodes.h>
13639 +#include <linux/poll.h>
13640 +#include <linux/seq_file.h>
13641 +#include <linux/statfs.h>
13644 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
13646 + struct au_sbinfo *sbinfo;
13647 + struct au_fhsm *fhsm;
13649 + SiMustAnyLock(sb);
13651 + sbinfo = au_sbi(sb);
13652 + fhsm = &sbinfo->si_fhsm;
13653 + AuDebugOn(!fhsm);
13654 + return fhsm->fhsm_bottom;
13657 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
13659 + struct au_sbinfo *sbinfo;
13660 + struct au_fhsm *fhsm;
13662 + SiMustWriteLock(sb);
13664 + sbinfo = au_sbi(sb);
13665 + fhsm = &sbinfo->si_fhsm;
13666 + AuDebugOn(!fhsm);
13667 + fhsm->fhsm_bottom = bindex;
13670 +/* ---------------------------------------------------------------------- */
13672 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
13674 + struct au_br_fhsm *bf;
13676 + bf = br->br_fhsm;
13677 + MtxMustLock(&bf->bf_lock);
13679 + return !bf->bf_readable
13680 + || time_after(jiffies,
13681 + bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
13684 +/* ---------------------------------------------------------------------- */
13686 +static void au_fhsm_notify(struct super_block *sb, int val)
13688 + struct au_sbinfo *sbinfo;
13689 + struct au_fhsm *fhsm;
13691 + SiMustAnyLock(sb);
13693 + sbinfo = au_sbi(sb);
13694 + fhsm = &sbinfo->si_fhsm;
13695 + if (au_fhsm_pid(fhsm)
13696 + && atomic_read(&fhsm->fhsm_readable) != -1) {
13697 + atomic_set(&fhsm->fhsm_readable, val);
13699 + wake_up(&fhsm->fhsm_wqh);
13703 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
13704 + struct aufs_stfs *rstfs, int do_lock, int do_notify)
13707 + struct au_branch *br;
13708 + struct au_br_fhsm *bf;
13710 + br = au_sbr(sb, bindex);
13711 + AuDebugOn(au_br_rdonly(br));
13712 + bf = br->br_fhsm;
13716 + mutex_lock(&bf->bf_lock);
13718 + MtxMustLock(&bf->bf_lock);
13720 + /* sb->s_root for NFS is unreliable */
13721 + err = au_br_stfs(br, &bf->bf_stfs);
13722 + if (unlikely(err)) {
13723 + AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
13727 + bf->bf_jiffy = jiffies;
13728 + bf->bf_readable = 1;
13730 + au_fhsm_notify(sb, /*val*/1);
13732 + *rstfs = bf->bf_stfs;
13736 + mutex_unlock(&bf->bf_lock);
13737 + au_fhsm_notify(sb, /*val*/1);
13742 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
13745 + struct au_sbinfo *sbinfo;
13746 + struct au_fhsm *fhsm;
13747 + struct au_branch *br;
13748 + struct au_br_fhsm *bf;
13750 + AuDbg("b%d, force %d\n", bindex, force);
13751 + SiMustAnyLock(sb);
13753 + sbinfo = au_sbi(sb);
13754 + fhsm = &sbinfo->si_fhsm;
13755 + if (!au_ftest_si(sbinfo, FHSM)
13756 + || fhsm->fhsm_bottom == bindex)
13759 + br = au_sbr(sb, bindex);
13760 + bf = br->br_fhsm;
13762 + mutex_lock(&bf->bf_lock);
13764 + || au_fhsm_pid(fhsm)
13765 + || au_fhsm_test_jiffy(sbinfo, br))
13766 + err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
13768 + mutex_unlock(&bf->bf_lock);
13771 +void au_fhsm_wrote_all(struct super_block *sb, int force)
13773 + aufs_bindex_t bindex, bbot;
13774 + struct au_branch *br;
13776 + /* exclude the bottom */
13777 + bbot = au_fhsm_bottom(sb);
13778 + for (bindex = 0; bindex < bbot; bindex++) {
13779 + br = au_sbr(sb, bindex);
13780 + if (au_br_fhsm(br->br_perm))
13781 + au_fhsm_wrote(sb, bindex, force);
13785 +/* ---------------------------------------------------------------------- */
13787 +static __poll_t au_fhsm_poll(struct file *file, struct poll_table_struct *wait)
13790 + struct au_sbinfo *sbinfo;
13791 + struct au_fhsm *fhsm;
13794 + sbinfo = file->private_data;
13795 + fhsm = &sbinfo->si_fhsm;
13796 + poll_wait(file, &fhsm->fhsm_wqh, wait);
13797 + if (atomic_read(&fhsm->fhsm_readable))
13798 + mask = EPOLLIN /* | EPOLLRDNORM */;
13801 + AuDbg("mask 0x%x\n", mask);
13805 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
13806 + struct aufs_stfs *stfs, __s16 brid)
13810 + err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
13812 + err = __put_user(brid, &stbr->brid);
13813 + if (unlikely(err))
13819 +static ssize_t au_fhsm_do_read(struct super_block *sb,
13820 + struct aufs_stbr __user *stbr, size_t count)
13824 + aufs_bindex_t bindex, bbot;
13825 + struct au_branch *br;
13826 + struct au_br_fhsm *bf;
13828 + /* except the bottom branch */
13831 + bbot = au_fhsm_bottom(sb);
13832 + for (bindex = 0; !err && bindex < bbot; bindex++) {
13833 + br = au_sbr(sb, bindex);
13834 + if (!au_br_fhsm(br->br_perm))
13837 + bf = br->br_fhsm;
13838 + mutex_lock(&bf->bf_lock);
13839 + if (bf->bf_readable) {
13841 + if (count >= sizeof(*stbr))
13842 + err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
13845 + bf->bf_readable = 0;
13846 + count -= sizeof(*stbr);
13850 + mutex_unlock(&bf->bf_lock);
13853 + err = sizeof(*stbr) * nstbr;
13858 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
13863 + aufs_bindex_t nfhsm, bindex, bbot;
13864 + struct au_sbinfo *sbinfo;
13865 + struct au_fhsm *fhsm;
13866 + struct au_branch *br;
13867 + struct super_block *sb;
13870 + sbinfo = file->private_data;
13871 + fhsm = &sbinfo->si_fhsm;
13873 + spin_lock_irq(&fhsm->fhsm_wqh.lock);
13874 + if (!atomic_read(&fhsm->fhsm_readable)) {
13875 + if (vfsub_file_flags(file) & O_NONBLOCK)
13878 + err = wait_event_interruptible_locked_irq
13880 + atomic_read(&fhsm->fhsm_readable));
13882 + spin_unlock_irq(&fhsm->fhsm_wqh.lock);
13883 + if (unlikely(err))
13886 + /* sb may already be dead */
13887 + au_rw_read_lock(&sbinfo->si_rwsem);
13888 + readable = atomic_read(&fhsm->fhsm_readable);
13889 + if (readable > 0) {
13890 + sb = sbinfo->si_sb;
13892 + /* exclude the bottom branch */
13894 + bbot = au_fhsm_bottom(sb);
13895 + for (bindex = 0; bindex < bbot; bindex++) {
13896 + br = au_sbr(sb, bindex);
13897 + if (au_br_fhsm(br->br_perm))
13901 + if (nfhsm * sizeof(struct aufs_stbr) <= count) {
13902 + atomic_set(&fhsm->fhsm_readable, 0);
13903 + err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
13907 + au_rw_read_unlock(&sbinfo->si_rwsem);
13915 +static int au_fhsm_release(struct inode *inode, struct file *file)
13917 + struct au_sbinfo *sbinfo;
13918 + struct au_fhsm *fhsm;
13920 + /* sb may already be dead */
13921 + sbinfo = file->private_data;
13922 + fhsm = &sbinfo->si_fhsm;
13923 + spin_lock(&fhsm->fhsm_spin);
13924 + fhsm->fhsm_pid = 0;
13925 + spin_unlock(&fhsm->fhsm_spin);
13926 + kobject_put(&sbinfo->si_kobj);
13931 +static const struct file_operations au_fhsm_fops = {
13932 + .owner = THIS_MODULE,
13933 + .llseek = noop_llseek,
13934 + .read = au_fhsm_read,
13935 + .poll = au_fhsm_poll,
13936 + .release = au_fhsm_release
13939 +int au_fhsm_fd(struct super_block *sb, int oflags)
13942 + struct au_sbinfo *sbinfo;
13943 + struct au_fhsm *fhsm;
13946 + if (unlikely(!capable(CAP_SYS_ADMIN)))
13950 + if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
13954 + sbinfo = au_sbi(sb);
13955 + fhsm = &sbinfo->si_fhsm;
13956 + spin_lock(&fhsm->fhsm_spin);
13957 + if (!fhsm->fhsm_pid)
13958 + fhsm->fhsm_pid = current->pid;
13961 + spin_unlock(&fhsm->fhsm_spin);
13962 + if (unlikely(err))
13965 + oflags |= O_RDONLY;
13966 + /* oflags |= FMODE_NONOTIFY; */
13967 + fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
13969 + if (unlikely(fd < 0))
13972 + /* succeed regardless 'fhsm' status */
13973 + kobject_get(&sbinfo->si_kobj);
13974 + si_noflush_read_lock(sb);
13975 + if (au_ftest_si(sbinfo, FHSM))
13976 + au_fhsm_wrote_all(sb, /*force*/0);
13977 + si_read_unlock(sb);
13978 + goto out; /* success */
13981 + spin_lock(&fhsm->fhsm_spin);
13982 + fhsm->fhsm_pid = 0;
13983 + spin_unlock(&fhsm->fhsm_spin);
13989 +/* ---------------------------------------------------------------------- */
13991 +int au_fhsm_br_alloc(struct au_branch *br)
13996 + br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
13998 + au_br_fhsm_init(br->br_fhsm);
14005 +/* ---------------------------------------------------------------------- */
14007 +void au_fhsm_fin(struct super_block *sb)
14009 + au_fhsm_notify(sb, /*val*/-1);
14012 +void au_fhsm_init(struct au_sbinfo *sbinfo)
14014 + struct au_fhsm *fhsm;
14016 + fhsm = &sbinfo->si_fhsm;
14017 + spin_lock_init(&fhsm->fhsm_spin);
14018 + init_waitqueue_head(&fhsm->fhsm_wqh);
14019 + atomic_set(&fhsm->fhsm_readable, 0);
14020 + fhsm->fhsm_expire
14021 + = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
14022 + fhsm->fhsm_bottom = -1;
14025 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
14027 + sbinfo->si_fhsm.fhsm_expire
14028 + = msecs_to_jiffies(sec * MSEC_PER_SEC);
14031 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
14035 + if (!au_ftest_si(sbinfo, FHSM))
14038 + u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
14039 + if (u != AUFS_FHSM_CACHE_DEF_SEC)
14040 + seq_printf(seq, ",fhsm_sec=%u", u);
14042 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
14043 --- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
14044 +++ linux/fs/aufs/file.c 2022-10-03 07:51:14.926209260 +0200
14046 +// SPDX-License-Identifier: GPL-2.0
14048 + * Copyright (C) 2005-2021 Junjiro R. Okajima
14050 + * This program, aufs is free software; you can redistribute it and/or modify
14051 + * it under the terms of the GNU General Public License as published by
14052 + * the Free Software Foundation; either version 2 of the License, or
14053 + * (at your option) any later version.
14055 + * This program is distributed in the hope that it will be useful,
14056 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14057 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14058 + * GNU General Public License for more details.
14060 + * You should have received a copy of the GNU General Public License
14061 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14065 + * handling file/dir, and address_space operation
14068 +#ifdef CONFIG_AUFS_DEBUG
14069 +#include <linux/migrate.h>
14071 +#include <linux/pagemap.h>
14074 +/* drop flags for writing */
14075 +unsigned int au_file_roflags(unsigned int flags)
14077 + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
14078 + flags |= O_RDONLY | O_NOATIME;
14082 +/* common functions to regular file and dir */
14083 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14084 + struct file *file, int force_wr)
14086 + struct file *h_file;
14087 + struct dentry *h_dentry;
14088 + struct inode *h_inode;
14089 + struct super_block *sb;
14090 + struct au_branch *br;
14091 + struct path h_path;
14094 + /* a race condition can happen between open and unlink/rmdir */
14095 + h_file = ERR_PTR(-ENOENT);
14096 + h_dentry = au_h_dptr(dentry, bindex);
14097 + if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
14099 + h_inode = d_inode(h_dentry);
14100 + spin_lock(&h_dentry->d_lock);
14101 + err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
14102 + /* || !d_inode(dentry)->i_nlink */
14104 + spin_unlock(&h_dentry->d_lock);
14105 + if (unlikely(err))
14108 + sb = dentry->d_sb;
14109 + br = au_sbr(sb, bindex);
14110 + err = au_br_test_oflag(flags, br);
14111 + h_file = ERR_PTR(err);
14112 + if (unlikely(err))
14115 + /* drop flags for writing */
14116 + if (au_test_ro(sb, bindex, d_inode(dentry))) {
14117 + if (force_wr && !(flags & O_WRONLY))
14119 + flags = au_file_roflags(flags);
14121 + h_file = ERR_PTR(-EROFS);
14122 + flags = au_file_roflags(flags);
14123 + if (unlikely(vfsub_native_ro(h_inode)
14124 + || IS_APPEND(h_inode)))
14126 + flags &= ~O_ACCMODE;
14127 + flags |= O_WRONLY;
14130 + flags &= ~O_CREAT;
14131 + au_lcnt_inc(&br->br_nfiles);
14132 + h_path.dentry = h_dentry;
14133 + h_path.mnt = au_br_mnt(br);
14134 + h_file = vfsub_dentry_open(&h_path, flags);
14135 + if (IS_ERR(h_file))
14138 + if (flags & __FMODE_EXEC) {
14139 + err = deny_write_access(h_file);
14140 + if (unlikely(err)) {
14142 + h_file = ERR_PTR(err);
14146 + fsnotify_open(h_file);
14147 + goto out; /* success */
14150 + au_lcnt_dec(&br->br_nfiles);
14155 +static int au_cmoo(struct dentry *dentry)
14157 + int err, cmoo, matched;
14158 + unsigned int udba;
14159 + struct path h_path;
14160 + struct au_pin pin;
14161 + struct au_cp_generic cpg = {
14162 + .dentry = dentry,
14167 + .flags = AuCpup_DTIME | AuCpup_HOPEN
14169 + struct inode *delegated;
14170 + struct super_block *sb;
14171 + struct au_sbinfo *sbinfo;
14172 + struct au_fhsm *fhsm;
14174 + struct au_branch *br;
14175 + struct dentry *parent;
14176 + struct au_hinode *hdir;
14178 + DiMustWriteLock(dentry);
14179 + IiMustWriteLock(d_inode(dentry));
14182 + if (IS_ROOT(dentry))
14184 + cpg.bsrc = au_dbtop(dentry);
14188 + sb = dentry->d_sb;
14189 + sbinfo = au_sbi(sb);
14190 + fhsm = &sbinfo->si_fhsm;
14191 + pid = au_fhsm_pid(fhsm);
14194 + && (current->pid == pid
14195 + || rcu_dereference(current->real_parent)->pid == pid));
14196 + rcu_read_unlock();
14200 + br = au_sbr(sb, cpg.bsrc);
14201 + cmoo = au_br_cmoo(br->br_perm);
14204 + if (!d_is_reg(dentry))
14205 + cmoo &= AuBrAttr_COO_ALL;
14209 + parent = dget_parent(dentry);
14210 + di_write_lock_parent(parent);
14211 + err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
14213 + if (unlikely(err < 0)) {
14214 + err = 0; /* there is no upper writable branch */
14217 + AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
14219 + /* do not respect the coo attrib for the target branch */
14220 + err = au_cpup_dirs(dentry, cpg.bdst);
14221 + if (unlikely(err))
14224 + di_downgrade_lock(parent, AuLock_IR);
14225 + udba = au_opt_udba(sb);
14226 + err = au_pin(&pin, dentry, cpg.bdst, udba,
14227 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14228 + if (unlikely(err))
14231 + err = au_sio_cpup_simple(&cpg);
14233 + if (unlikely(err))
14235 + if (!(cmoo & AuBrWAttr_MOO))
14236 + goto out_parent; /* success */
14238 + err = au_pin(&pin, dentry, cpg.bsrc, udba,
14239 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14240 + if (unlikely(err))
14243 + h_path.mnt = au_br_mnt(br);
14244 + h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
14245 + hdir = au_hi(d_inode(parent), cpg.bsrc);
14246 + delegated = NULL;
14247 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
14249 + /* todo: keep h_dentry or not? */
14250 + if (unlikely(err == -EWOULDBLOCK)) {
14251 + pr_warn("cannot retry for NFSv4 delegation"
14252 + " for an internal unlink\n");
14255 + if (unlikely(err)) {
14256 + pr_err("unlink %pd after coo failed (%d), ignored\n",
14260 + goto out_parent; /* success */
14263 + di_downgrade_lock(parent, AuLock_IR);
14265 + di_read_unlock(parent, AuLock_IR);
14272 +int au_do_open(struct file *file, struct au_do_open_args *args)
14274 + int err, aopen = args->aopen;
14275 + struct dentry *dentry;
14276 + struct au_finfo *finfo;
14279 + err = au_finfo_init(file, args->fidir);
14282 + err = au_finfo_init(file, args->fidir);
14285 + if (unlikely(err))
14288 + dentry = file->f_path.dentry;
14289 + AuDebugOn(IS_ERR_OR_NULL(dentry));
14290 + di_write_lock_child(dentry);
14291 + err = au_cmoo(dentry);
14292 + di_downgrade_lock(dentry, AuLock_IR);
14295 + err = args->open(file, vfsub_file_flags(file), NULL);
14298 + err = args->open(file, vfsub_file_flags(file),
14303 + di_read_unlock(dentry, AuLock_IR);
14305 + finfo = au_fi(file);
14307 + finfo->fi_file = file;
14308 + au_hbl_add(&finfo->fi_hlist,
14309 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
14312 + fi_write_unlock(file);
14315 + fi_write_unlock(file);
14318 + if (unlikely(err)) {
14319 + finfo->fi_hdir = NULL;
14320 + au_finfo_fin(file);
14328 +int au_reopen_nondir(struct file *file)
14331 + aufs_bindex_t btop;
14332 + struct dentry *dentry;
14333 + struct au_branch *br;
14334 + struct file *h_file, *h_file_tmp;
14336 + dentry = file->f_path.dentry;
14337 + btop = au_dbtop(dentry);
14338 + br = au_sbr(dentry->d_sb, btop);
14339 + h_file_tmp = NULL;
14340 + if (au_fbtop(file) == btop) {
14341 + h_file = au_hf_top(file);
14342 + if (file->f_mode == h_file->f_mode)
14343 + return 0; /* success */
14344 + h_file_tmp = h_file;
14345 + get_file(h_file_tmp);
14346 + au_lcnt_inc(&br->br_nfiles);
14347 + au_set_h_fptr(file, btop, NULL);
14349 + AuDebugOn(au_fi(file)->fi_hdir);
14352 + * file exists on both of rw and ro
14353 + * open --> dbtop and fbtop are both 0
14354 + * prepend a branch as rw, "rw" become ro
14356 + * delete the top branch, "rw" becomes rw again
14357 + * --> dbtop is 1, fbtop is still 0
14358 + * write --> fbtop is 0 but dbtop is 1
14360 + /* AuDebugOn(au_fbtop(file) < btop); */
14362 + h_file = au_h_open(dentry, btop, vfsub_file_flags(file) & ~O_TRUNC,
14363 + file, /*force_wr*/0);
14364 + err = PTR_ERR(h_file);
14365 + if (IS_ERR(h_file)) {
14366 + if (h_file_tmp) {
14368 + au_set_h_fptr(file, btop, h_file_tmp);
14369 + h_file_tmp = NULL;
14371 + goto out; /* todo: close all? */
14375 + au_set_fbtop(file, btop);
14376 + au_set_h_fptr(file, btop, h_file);
14377 + au_update_figen(file);
14378 + /* todo: necessary? */
14379 + /* file->f_ra = h_file->f_ra; */
14382 + if (h_file_tmp) {
14383 + fput(h_file_tmp);
14384 + au_lcnt_dec(&br->br_nfiles);
14389 +/* ---------------------------------------------------------------------- */
14391 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
14392 + struct dentry *hi_wh)
14395 + aufs_bindex_t btop;
14396 + struct au_dinfo *dinfo;
14397 + struct dentry *h_dentry;
14398 + struct au_hdentry *hdp;
14400 + dinfo = au_di(file->f_path.dentry);
14401 + AuRwMustWriteLock(&dinfo->di_rwsem);
14403 + btop = dinfo->di_btop;
14404 + dinfo->di_btop = btgt;
14405 + hdp = au_hdentry(dinfo, btgt);
14406 + h_dentry = hdp->hd_dentry;
14407 + hdp->hd_dentry = hi_wh;
14408 + err = au_reopen_nondir(file);
14409 + hdp->hd_dentry = h_dentry;
14410 + dinfo->di_btop = btop;
14415 +static int au_ready_to_write_wh(struct file *file, loff_t len,
14416 + aufs_bindex_t bcpup, struct au_pin *pin)
14419 + struct inode *inode, *h_inode;
14420 + struct dentry *h_dentry, *hi_wh;
14421 + struct au_cp_generic cpg = {
14422 + .dentry = file->f_path.dentry,
14429 + au_update_dbtop(cpg.dentry);
14430 + inode = d_inode(cpg.dentry);
14432 + if (au_dbtop(cpg.dentry) <= bcpup
14433 + && au_dbbot(cpg.dentry) >= bcpup) {
14434 + h_dentry = au_h_dptr(cpg.dentry, bcpup);
14435 + if (h_dentry && d_is_positive(h_dentry))
14436 + h_inode = d_inode(h_dentry);
14438 + hi_wh = au_hi_wh(inode, bcpup);
14439 + if (!hi_wh && !h_inode)
14440 + err = au_sio_cpup_wh(&cpg, file);
14442 + /* already copied-up after unlink */
14443 + err = au_reopen_wh(file, bcpup, hi_wh);
14446 + && (inode->i_nlink > 1
14447 + || (inode->i_state & I_LINKABLE))
14448 + && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
14449 + au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
14455 + * prepare the @file for writing.
14457 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
14460 + aufs_bindex_t dbtop;
14461 + struct dentry *parent;
14462 + struct inode *inode;
14463 + struct super_block *sb;
14464 + struct file *h_file;
14465 + struct au_cp_generic cpg = {
14466 + .dentry = file->f_path.dentry,
14471 + .flags = AuCpup_DTIME
14474 + sb = cpg.dentry->d_sb;
14475 + inode = d_inode(cpg.dentry);
14476 + cpg.bsrc = au_fbtop(file);
14477 + err = au_test_ro(sb, cpg.bsrc, inode);
14478 + if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
14479 + err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
14484 + /* need to cpup or reopen */
14485 + parent = dget_parent(cpg.dentry);
14486 + di_write_lock_parent(parent);
14487 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14489 + if (unlikely(err < 0))
14493 + if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
14494 + err = au_cpup_dirs(cpg.dentry, cpg.bdst);
14495 + if (unlikely(err))
14499 + err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14500 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14501 + if (unlikely(err))
14504 + dbtop = au_dbtop(cpg.dentry);
14505 + if (dbtop <= cpg.bdst)
14506 + cpg.bsrc = cpg.bdst;
14508 + if (dbtop <= cpg.bdst /* just reopen */
14509 + || !d_unhashed(cpg.dentry) /* copyup and reopen */
14511 + h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
14512 + if (IS_ERR(h_file))
14513 + err = PTR_ERR(h_file);
14515 + di_downgrade_lock(parent, AuLock_IR);
14516 + if (dbtop > cpg.bdst)
14517 + err = au_sio_cpup_simple(&cpg);
14519 + err = au_reopen_nondir(file);
14520 + au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
14522 + } else { /* copyup as wh and reopen */
14524 + * since writable hfsplus branch is not supported,
14525 + * h_open_pre/post() are unnecessary.
14527 + err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
14528 + di_downgrade_lock(parent, AuLock_IR);
14532 + au_pin_set_parent_lflag(pin, /*lflag*/0);
14533 + goto out_dput; /* success */
14539 + di_downgrade_lock(parent, AuLock_IR);
14541 + di_read_unlock(parent, AuLock_IR);
14548 +/* ---------------------------------------------------------------------- */
14550 +int au_do_flush(struct file *file, fl_owner_t id,
14551 + int (*flush)(struct file *file, fl_owner_t id))
14554 + struct super_block *sb;
14555 + struct inode *inode;
14557 + inode = file_inode(file);
14558 + sb = inode->i_sb;
14559 + si_noflush_read_lock(sb);
14560 + fi_read_lock(file);
14561 + ii_read_lock_child(inode);
14563 + err = flush(file, id);
14564 + au_cpup_attr_timesizes(inode);
14566 + ii_read_unlock(inode);
14567 + fi_read_unlock(file);
14568 + si_read_unlock(sb);
14572 +/* ---------------------------------------------------------------------- */
14574 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
14577 + struct au_pin pin;
14578 + struct au_finfo *finfo;
14579 + struct dentry *parent, *hi_wh;
14580 + struct inode *inode;
14581 + struct super_block *sb;
14582 + struct au_cp_generic cpg = {
14583 + .dentry = file->f_path.dentry,
14588 + .flags = AuCpup_DTIME
14591 + FiMustWriteLock(file);
14594 + finfo = au_fi(file);
14595 + sb = cpg.dentry->d_sb;
14596 + inode = d_inode(cpg.dentry);
14597 + cpg.bdst = au_ibtop(inode);
14598 + if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
14601 + parent = dget_parent(cpg.dentry);
14602 + if (au_test_ro(sb, cpg.bdst, inode)) {
14603 + di_read_lock_parent(parent, !AuLock_IR);
14604 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14606 + di_read_unlock(parent, !AuLock_IR);
14607 + if (unlikely(err < 0))
14612 + di_read_lock_parent(parent, AuLock_IR);
14613 + hi_wh = au_hi_wh(inode, cpg.bdst);
14614 + if (!S_ISDIR(inode->i_mode)
14615 + && au_opt_test(au_mntflags(sb), PLINK)
14616 + && au_plink_test(inode)
14617 + && !d_unhashed(cpg.dentry)
14618 + && cpg.bdst < au_dbtop(cpg.dentry)) {
14619 + err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
14620 + if (unlikely(err))
14623 + /* always superio. */
14624 + err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14625 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14627 + err = au_sio_cpup_simple(&cpg);
14630 + } else if (hi_wh) {
14631 + /* already copied-up after unlink */
14632 + err = au_reopen_wh(file, cpg.bdst, hi_wh);
14633 + *need_reopen = 0;
14637 + di_read_unlock(parent, AuLock_IR);
14644 +static void au_do_refresh_dir(struct file *file)
14646 + aufs_bindex_t bindex, bbot, new_bindex, brid;
14647 + struct au_hfile *p, tmp, *q;
14648 + struct au_finfo *finfo;
14649 + struct super_block *sb;
14650 + struct au_fidir *fidir;
14652 + FiMustWriteLock(file);
14654 + sb = file->f_path.dentry->d_sb;
14655 + finfo = au_fi(file);
14656 + fidir = finfo->fi_hdir;
14657 + AuDebugOn(!fidir);
14658 + p = fidir->fd_hfile + finfo->fi_btop;
14659 + brid = p->hf_br->br_id;
14660 + bbot = fidir->fd_bbot;
14661 + for (bindex = finfo->fi_btop; bindex <= bbot; bindex++, p++) {
14665 + new_bindex = au_br_index(sb, p->hf_br->br_id);
14666 + if (new_bindex == bindex)
14668 + if (new_bindex < 0) {
14669 + au_set_h_fptr(file, bindex, NULL);
14673 + /* swap two lower inode, and loop again */
14674 + q = fidir->fd_hfile + new_bindex;
14678 + if (tmp.hf_file) {
14684 + p = fidir->fd_hfile;
14685 + if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
14686 + bbot = au_sbbot(sb);
14687 + for (finfo->fi_btop = 0; finfo->fi_btop <= bbot;
14688 + finfo->fi_btop++, p++)
14689 + if (p->hf_file) {
14690 + if (file_inode(p->hf_file))
14692 + au_hfput(p, /*execed*/0);
14695 + bbot = au_br_index(sb, brid);
14696 + for (finfo->fi_btop = 0; finfo->fi_btop < bbot;
14697 + finfo->fi_btop++, p++)
14699 + au_hfput(p, /*execed*/0);
14700 + bbot = au_sbbot(sb);
14703 + p = fidir->fd_hfile + bbot;
14704 + for (fidir->fd_bbot = bbot; fidir->fd_bbot >= finfo->fi_btop;
14705 + fidir->fd_bbot--, p--)
14706 + if (p->hf_file) {
14707 + if (file_inode(p->hf_file))
14709 + au_hfput(p, /*execed*/0);
14711 + AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
14715 + * after branch manipulating, refresh the file.
14717 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
14719 + int err, need_reopen, nbr;
14720 + aufs_bindex_t bbot, bindex;
14721 + struct dentry *dentry;
14722 + struct super_block *sb;
14723 + struct au_finfo *finfo;
14724 + struct au_hfile *hfile;
14726 + dentry = file->f_path.dentry;
14727 + sb = dentry->d_sb;
14728 + nbr = au_sbbot(sb) + 1;
14729 + finfo = au_fi(file);
14730 + if (!finfo->fi_hdir) {
14731 + hfile = &finfo->fi_htop;
14732 + AuDebugOn(!hfile->hf_file);
14733 + bindex = au_br_index(sb, hfile->hf_br->br_id);
14734 + AuDebugOn(bindex < 0);
14735 + if (bindex != finfo->fi_btop)
14736 + au_set_fbtop(file, bindex);
14738 + err = au_fidir_realloc(finfo, nbr, /*may_shrink*/0);
14739 + if (unlikely(err))
14741 + au_do_refresh_dir(file);
14746 + if (!au_test_mmapped(file))
14747 + err = au_file_refresh_by_inode(file, &need_reopen);
14748 + if (finfo->fi_hdir)
14749 + /* harmless if err */
14750 + au_fidir_realloc(finfo, nbr, /*may_shrink*/1);
14751 + if (!err && need_reopen && !d_unlinked(dentry))
14752 + err = reopen(file);
14754 + au_update_figen(file);
14755 + goto out; /* success */
14758 + /* error, close all lower files */
14759 + if (finfo->fi_hdir) {
14760 + bbot = au_fbbot_dir(file);
14761 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++)
14762 + au_set_h_fptr(file, bindex, NULL);
14769 +/* common function to regular file and dir */
14770 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14771 + int wlock, unsigned int fi_lsc)
14774 + unsigned int sigen, figen;
14775 + aufs_bindex_t btop;
14776 + unsigned char pseudo_link;
14777 + struct dentry *dentry;
14778 + struct inode *inode;
14781 + dentry = file->f_path.dentry;
14782 + inode = d_inode(dentry);
14783 + sigen = au_sigen(dentry->d_sb);
14784 + fi_write_lock_nested(file, fi_lsc);
14785 + figen = au_figen(file);
14787 + di_write_lock_child(dentry);
14789 + di_write_lock_child2(dentry);
14790 + btop = au_dbtop(dentry);
14791 + pseudo_link = (btop != au_ibtop(inode));
14792 + if (sigen == figen && !pseudo_link && au_fbtop(file) == btop) {
14794 + di_downgrade_lock(dentry, AuLock_IR);
14795 + fi_downgrade_lock(file);
14797 + goto out; /* success */
14800 + AuDbg("sigen %d, figen %d\n", sigen, figen);
14801 + if (au_digen_test(dentry, sigen)) {
14802 + err = au_reval_dpath(dentry, sigen);
14803 + AuDebugOn(!err && au_digen_test(dentry, sigen));
14807 + err = refresh_file(file, reopen);
14810 + di_downgrade_lock(dentry, AuLock_IR);
14811 + fi_downgrade_lock(file);
14814 + di_write_unlock(dentry);
14815 + fi_write_unlock(file);
14822 +/* ---------------------------------------------------------------------- */
14824 +/* cf. aufs_nopage() */
14825 +/* for madvise(2) */
14826 +static int aufs_read_folio(struct file *file __maybe_unused, struct folio *folio)
14828 + folio_unlock(folio);
14832 +/* it will never be called, but necessary to support O_DIRECT */
14833 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
14834 +{ BUG(); return 0; }
14836 +/* they will never be called. */
14837 +#ifdef CONFIG_AUFS_DEBUG
14838 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
14839 + loff_t pos, unsigned len,
14840 + struct page **pagep, void **fsdata)
14841 +{ AuUnsupport(); return 0; }
14842 +static int aufs_write_end(struct file *file, struct address_space *mapping,
14843 + loff_t pos, unsigned len, unsigned copied,
14844 + struct page *page, void *fsdata)
14845 +{ AuUnsupport(); return 0; }
14846 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
14847 +{ AuUnsupport(); return 0; }
14849 +static bool aufs_dirty_folio(struct address_space *mapping, struct folio *folio)
14850 +{ AuUnsupport(); return true; }
14851 +static void aufs_invalidate_folio(struct folio *folio, size_t offset, size_t len)
14852 +{ AuUnsupport(); }
14853 +static bool aufs_release_folio(struct folio *folio, gfp_t gfp)
14854 +{ AuUnsupport(); return true; }
14855 +#if 0 /* called by memory compaction regardless file */
14856 +static int aufs_migrate_folio(struct address_space *mapping, struct folio *dst,
14857 + struct folio *src, enum migrate_mode mode)
14858 +{ AuUnsupport(); return 0; }
14860 +static int aufs_launder_folio(struct folio *folio)
14861 +{ AuUnsupport(); return 0; }
14862 +static bool aufs_is_partially_uptodate(struct folio *folio, size_t from,
14864 +{ AuUnsupport(); return true; }
14865 +static void aufs_is_dirty_writeback(struct folio *folio, bool *dirty,
14867 +{ AuUnsupport(); }
14868 +static int aufs_error_remove_page(struct address_space *mapping,
14869 + struct page *page)
14870 +{ AuUnsupport(); return 0; }
14871 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
14873 +{ AuUnsupport(); return 0; }
14874 +static void aufs_swap_deactivate(struct file *file)
14875 +{ AuUnsupport(); }
14876 +static int aufs_swap_rw(struct kiocb *iocb, struct iov_iter *iter)
14877 +{ AuUnsupport(); return 0; }
14878 +#endif /* CONFIG_AUFS_DEBUG */
14880 +const struct address_space_operations aufs_aop = {
14881 + .read_folio = aufs_read_folio,
14882 + .direct_IO = aufs_direct_IO,
14883 +#ifdef CONFIG_AUFS_DEBUG
14884 + .writepage = aufs_writepage,
14885 + /* no writepages, because of writepage */
14886 + .dirty_folio = aufs_dirty_folio,
14887 + /* no readpages, because of readpage */
14888 + .write_begin = aufs_write_begin,
14889 + .write_end = aufs_write_end,
14890 + /* no bmap, no block device */
14891 + .invalidate_folio = aufs_invalidate_folio,
14892 + .release_folio = aufs_release_folio,
14893 + /* is fallback_migrate_page ok? */
14894 + /* .migrate_folio = aufs_migrate_folio, */
14895 + .launder_folio = aufs_launder_folio,
14896 + .is_partially_uptodate = aufs_is_partially_uptodate,
14897 + .is_dirty_writeback = aufs_is_dirty_writeback,
14898 + .error_remove_page = aufs_error_remove_page,
14899 + .swap_activate = aufs_swap_activate,
14900 + .swap_deactivate = aufs_swap_deactivate,
14901 + .swap_rw = aufs_swap_rw
14902 +#endif /* CONFIG_AUFS_DEBUG */
14904 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
14905 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
14906 +++ linux/fs/aufs/file.h 2022-03-21 14:49:05.723299677 +0100
14908 +/* SPDX-License-Identifier: GPL-2.0 */
14910 + * Copyright (C) 2005-2021 Junjiro R. Okajima
14912 + * This program, aufs is free software; you can redistribute it and/or modify
14913 + * it under the terms of the GNU General Public License as published by
14914 + * the Free Software Foundation; either version 2 of the License, or
14915 + * (at your option) any later version.
14917 + * This program is distributed in the hope that it will be useful,
14918 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14919 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14920 + * GNU General Public License for more details.
14922 + * You should have received a copy of the GNU General Public License
14923 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14927 + * file operations
14930 +#ifndef __AUFS_FILE_H__
14931 +#define __AUFS_FILE_H__
14935 +#include <linux/file.h>
14936 +#include <linux/fs.h>
14937 +#include <linux/mm_types.h>
14938 +#include <linux/poll.h>
14939 +#include "rwsem.h"
14943 + struct file *hf_file;
14944 + struct au_branch *hf_br;
14949 + aufs_bindex_t fd_bbot;
14950 + aufs_bindex_t fd_nent;
14951 + struct au_vdir *fd_vdir_cache;
14952 + struct au_hfile fd_hfile[];
14955 +static inline int au_fidir_sz(int nent)
14957 + AuDebugOn(nent < 0);
14958 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
14962 + atomic_t fi_generation;
14964 + struct au_rwsem fi_rwsem;
14965 + aufs_bindex_t fi_btop;
14967 + /* do not union them */
14968 + struct { /* for non-dir */
14969 + struct au_hfile fi_htop;
14970 + atomic_t fi_mmapped;
14972 + struct au_fidir *fi_hdir; /* for dir only */
14974 + struct hlist_bl_node fi_hlist;
14975 + struct file *fi_file; /* very ugly */
14976 + struct rcu_head rcu;
14977 +} ____cacheline_aligned_in_smp;
14979 +/* ---------------------------------------------------------------------- */
14982 +extern const struct address_space_operations aufs_aop;
14983 +unsigned int au_file_roflags(unsigned int flags);
14984 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14985 + struct file *file, int force_wr);
14986 +struct au_do_open_args {
14988 + int (*open)(struct file *file, int flags,
14989 + struct file *h_file);
14990 + struct au_fidir *fidir;
14991 + struct file *h_file;
14993 +int au_do_open(struct file *file, struct au_do_open_args *args);
14994 +int au_reopen_nondir(struct file *file);
14996 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
14997 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14998 + int wlock, unsigned int fi_lsc);
14999 +int au_do_flush(struct file *file, fl_owner_t id,
15000 + int (*flush)(struct file *file, fl_owner_t id));
15003 +#ifdef CONFIG_AUFS_POLL
15004 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt);
15007 +#ifdef CONFIG_AUFS_BR_HFSPLUS
15009 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
15011 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
15012 + struct file *h_file);
15014 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
15015 + aufs_bindex_t bindex, int force_wr)
15016 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
15017 + struct file *h_file);
15021 +extern const struct file_operations aufs_file_fop;
15022 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
15023 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
15024 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc);
15027 +void au_hfput(struct au_hfile *hf, int execed);
15028 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
15029 + struct file *h_file);
15031 +void au_update_figen(struct file *file);
15032 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
15033 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink);
15035 +void au_fi_init_once(void *_fi);
15036 +void au_finfo_fin(struct file *file);
15037 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
15040 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
15041 +#ifdef CONFIG_COMPAT
15042 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15043 + unsigned long arg);
15044 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15045 + unsigned long arg);
15048 +/* ---------------------------------------------------------------------- */
15050 +static inline struct au_finfo *au_fi(struct file *file)
15052 + return file->private_data;
15055 +/* ---------------------------------------------------------------------- */
15057 +#define fi_read_lock(f) au_rw_read_lock(&au_fi(f)->fi_rwsem)
15058 +#define fi_write_lock(f) au_rw_write_lock(&au_fi(f)->fi_rwsem)
15059 +#define fi_read_trylock(f) au_rw_read_trylock(&au_fi(f)->fi_rwsem)
15060 +#define fi_write_trylock(f) au_rw_write_trylock(&au_fi(f)->fi_rwsem)
15062 +#define fi_read_trylock_nested(f) \
15063 + au_rw_read_trylock_nested(&au_fi(f)->fi_rwsem)
15064 +#define fi_write_trylock_nested(f) \
15065 + au_rw_write_trylock_nested(&au_fi(f)->fi_rwsem)
15068 +#define fi_read_unlock(f) au_rw_read_unlock(&au_fi(f)->fi_rwsem)
15069 +#define fi_write_unlock(f) au_rw_write_unlock(&au_fi(f)->fi_rwsem)
15070 +#define fi_downgrade_lock(f) au_rw_dgrade_lock(&au_fi(f)->fi_rwsem)
15072 +/* lock subclass for finfo */
15078 +static inline void fi_read_lock_nested(struct file *f, unsigned int lsc)
15080 + au_rw_read_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15083 +static inline void fi_write_lock_nested(struct file *f, unsigned int lsc)
15085 + au_rw_write_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15089 + * fi_read_lock_1, fi_write_lock_1,
15090 + * fi_read_lock_2, fi_write_lock_2
15092 +#define AuReadLockFunc(name) \
15093 +static inline void fi_read_lock_##name(struct file *f) \
15094 +{ fi_read_lock_nested(f, AuLsc_FI_##name); }
15096 +#define AuWriteLockFunc(name) \
15097 +static inline void fi_write_lock_##name(struct file *f) \
15098 +{ fi_write_lock_nested(f, AuLsc_FI_##name); }
15100 +#define AuRWLockFuncs(name) \
15101 + AuReadLockFunc(name) \
15102 + AuWriteLockFunc(name)
15107 +#undef AuReadLockFunc
15108 +#undef AuWriteLockFunc
15109 +#undef AuRWLockFuncs
15111 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
15112 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
15113 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
15115 +/* ---------------------------------------------------------------------- */
15117 +/* todo: hard/soft set? */
15118 +static inline aufs_bindex_t au_fbtop(struct file *file)
15120 + FiMustAnyLock(file);
15121 + return au_fi(file)->fi_btop;
15124 +static inline aufs_bindex_t au_fbbot_dir(struct file *file)
15126 + FiMustAnyLock(file);
15127 + AuDebugOn(!au_fi(file)->fi_hdir);
15128 + return au_fi(file)->fi_hdir->fd_bbot;
15131 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
15133 + FiMustAnyLock(file);
15134 + AuDebugOn(!au_fi(file)->fi_hdir);
15135 + return au_fi(file)->fi_hdir->fd_vdir_cache;
15138 +static inline void au_set_fbtop(struct file *file, aufs_bindex_t bindex)
15140 + FiMustWriteLock(file);
15141 + au_fi(file)->fi_btop = bindex;
15144 +static inline void au_set_fbbot_dir(struct file *file, aufs_bindex_t bindex)
15146 + FiMustWriteLock(file);
15147 + AuDebugOn(!au_fi(file)->fi_hdir);
15148 + au_fi(file)->fi_hdir->fd_bbot = bindex;
15151 +static inline void au_set_fvdir_cache(struct file *file,
15152 + struct au_vdir *vdir_cache)
15154 + FiMustWriteLock(file);
15155 + AuDebugOn(!au_fi(file)->fi_hdir);
15156 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
15159 +static inline struct file *au_hf_top(struct file *file)
15161 + FiMustAnyLock(file);
15162 + AuDebugOn(au_fi(file)->fi_hdir);
15163 + return au_fi(file)->fi_htop.hf_file;
15166 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
15168 + FiMustAnyLock(file);
15169 + AuDebugOn(!au_fi(file)->fi_hdir);
15170 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
15173 +/* todo: memory barrier? */
15174 +static inline unsigned int au_figen(struct file *f)
15176 + return atomic_read(&au_fi(f)->fi_generation);
15179 +static inline void au_set_mmapped(struct file *f)
15181 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
15183 + pr_warn("fi_mmapped wrapped around\n");
15184 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
15188 +static inline void au_unset_mmapped(struct file *f)
15190 + atomic_dec(&au_fi(f)->fi_mmapped);
15193 +static inline int au_test_mmapped(struct file *f)
15195 + return atomic_read(&au_fi(f)->fi_mmapped);
15198 +/* customize vma->vm_file */
15200 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
15201 + struct file *file)
15205 + f = vma->vm_file;
15207 + vma->vm_file = file;
15212 +#define AuDbgVmRegion(file, vma) do {} while (0)
15214 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15215 + struct file *file)
15217 + au_do_vm_file_reset(vma, file);
15220 +#define AuDbgVmRegion(file, vma) \
15221 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
15223 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15224 + struct file *file)
15228 + au_do_vm_file_reset(vma, file);
15229 + f = vma->vm_region->vm_file;
15231 + vma->vm_region->vm_file = file;
15234 +#endif /* CONFIG_MMU */
15236 +/* handle vma->vm_prfile */
15237 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
15238 + struct file *file)
15241 + vma->vm_prfile = file;
15242 +#ifndef CONFIG_MMU
15244 + vma->vm_region->vm_prfile = file;
15248 +#endif /* __KERNEL__ */
15249 +#endif /* __AUFS_FILE_H__ */
15250 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
15251 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
15252 +++ linux/fs/aufs/finfo.c 2022-03-21 14:49:05.723299677 +0100
15254 +// SPDX-License-Identifier: GPL-2.0
15256 + * Copyright (C) 2005-2021 Junjiro R. Okajima
15258 + * This program, aufs is free software; you can redistribute it and/or modify
15259 + * it under the terms of the GNU General Public License as published by
15260 + * the Free Software Foundation; either version 2 of the License, or
15261 + * (at your option) any later version.
15263 + * This program is distributed in the hope that it will be useful,
15264 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15265 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15266 + * GNU General Public License for more details.
15268 + * You should have received a copy of the GNU General Public License
15269 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15273 + * file private data
15278 +void au_hfput(struct au_hfile *hf, int execed)
15281 + allow_write_access(hf->hf_file);
15282 + fput(hf->hf_file);
15283 + hf->hf_file = NULL;
15284 + au_lcnt_dec(&hf->hf_br->br_nfiles);
15285 + hf->hf_br = NULL;
15288 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
15290 + struct au_finfo *finfo = au_fi(file);
15291 + struct au_hfile *hf;
15292 + struct au_fidir *fidir;
15294 + fidir = finfo->fi_hdir;
15296 + AuDebugOn(finfo->fi_btop != bindex);
15297 + hf = &finfo->fi_htop;
15299 + hf = fidir->fd_hfile + bindex;
15301 + if (hf && hf->hf_file)
15302 + au_hfput(hf, vfsub_file_execed(file));
15304 + FiMustWriteLock(file);
15305 + AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
15306 + hf->hf_file = val;
15307 + hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
15311 +void au_update_figen(struct file *file)
15313 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
15314 + /* smp_mb(); */ /* atomic_set */
15317 +/* ---------------------------------------------------------------------- */
15319 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
15321 + struct au_fidir *fidir;
15324 + nbr = au_sbbot(sb) + 1;
15326 + nbr = 2; /* initial allocate for 2 branches */
15327 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
15329 + fidir->fd_bbot = -1;
15330 + fidir->fd_nent = nbr;
15336 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink)
15339 + struct au_fidir *fidir, *p;
15341 + AuRwMustWriteLock(&finfo->fi_rwsem);
15342 + fidir = finfo->fi_hdir;
15343 + AuDebugOn(!fidir);
15346 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
15347 + GFP_NOFS, may_shrink);
15349 + p->fd_nent = nbr;
15350 + finfo->fi_hdir = p;
15357 +/* ---------------------------------------------------------------------- */
15359 +void au_finfo_fin(struct file *file)
15361 + struct au_finfo *finfo;
15363 + au_lcnt_dec(&au_sbi(file->f_path.dentry->d_sb)->si_nfiles);
15365 + finfo = au_fi(file);
15366 + AuDebugOn(finfo->fi_hdir);
15367 + AuRwDestroy(&finfo->fi_rwsem);
15368 + au_cache_free_finfo(finfo);
15371 +void au_fi_init_once(void *_finfo)
15373 + struct au_finfo *finfo = _finfo;
15375 + au_rw_init(&finfo->fi_rwsem);
15378 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
15381 + struct au_finfo *finfo;
15382 + struct dentry *dentry;
15385 + dentry = file->f_path.dentry;
15386 + finfo = au_cache_alloc_finfo();
15387 + if (unlikely(!finfo))
15391 + au_lcnt_inc(&au_sbi(dentry->d_sb)->si_nfiles);
15392 + au_rw_write_lock(&finfo->fi_rwsem);
15393 + finfo->fi_btop = -1;
15394 + finfo->fi_hdir = fidir;
15395 + atomic_set(&finfo->fi_generation, au_digen(dentry));
15396 + /* smp_mb(); */ /* atomic_set */
15398 + file->private_data = finfo;
15403 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
15404 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
15405 +++ linux/fs/aufs/f_op.c 2022-03-21 14:49:05.723299677 +0100
15407 +// SPDX-License-Identifier: GPL-2.0
15409 + * Copyright (C) 2005-2021 Junjiro R. Okajima
15411 + * This program, aufs is free software; you can redistribute it and/or modify
15412 + * it under the terms of the GNU General Public License as published by
15413 + * the Free Software Foundation; either version 2 of the License, or
15414 + * (at your option) any later version.
15416 + * This program is distributed in the hope that it will be useful,
15417 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15418 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15419 + * GNU General Public License for more details.
15421 + * You should have received a copy of the GNU General Public License
15422 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15426 + * file and vm operations
15429 +#include <linux/aio.h>
15430 +#include <linux/fs_stack.h>
15431 +#include <linux/mman.h>
15432 +#include <linux/security.h>
15435 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
15438 + aufs_bindex_t bindex;
15439 + struct dentry *dentry, *h_dentry;
15440 + struct au_finfo *finfo;
15441 + struct inode *h_inode;
15443 + FiMustWriteLock(file);
15446 + dentry = file->f_path.dentry;
15447 + AuDebugOn(IS_ERR_OR_NULL(dentry));
15448 + finfo = au_fi(file);
15449 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
15450 + atomic_set(&finfo->fi_mmapped, 0);
15451 + bindex = au_dbtop(dentry);
15453 + h_dentry = au_h_dptr(dentry, bindex);
15454 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15455 + if (unlikely(err))
15457 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
15458 + if (IS_ERR(h_file)) {
15459 + err = PTR_ERR(h_file);
15463 + h_dentry = h_file->f_path.dentry;
15464 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15465 + if (unlikely(err))
15467 + /* br ref is already inc-ed */
15470 + if ((flags & __O_TMPFILE)
15471 + && !(flags & O_EXCL)) {
15472 + h_inode = file_inode(h_file);
15473 + spin_lock(&h_inode->i_lock);
15474 + h_inode->i_state |= I_LINKABLE;
15475 + spin_unlock(&h_inode->i_lock);
15477 + au_set_fbtop(file, bindex);
15478 + au_set_h_fptr(file, bindex, h_file);
15479 + au_update_figen(file);
15480 + /* todo: necessary? */
15481 + /* file->f_ra = h_file->f_ra; */
15487 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
15488 + struct file *file)
15491 + struct super_block *sb;
15492 + struct au_do_open_args args = {
15493 + .open = au_do_open_nondir
15496 + AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
15497 + file, vfsub_file_flags(file), file->f_mode);
15499 + sb = file->f_path.dentry->d_sb;
15500 + si_read_lock(sb, AuLock_FLUSH);
15501 + err = au_do_open(file, &args);
15502 + si_read_unlock(sb);
15506 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
15508 + struct au_finfo *finfo;
15509 + aufs_bindex_t bindex;
15511 + finfo = au_fi(file);
15512 + au_hbl_del(&finfo->fi_hlist,
15513 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
15514 + bindex = finfo->fi_btop;
15516 + au_set_h_fptr(file, bindex, NULL);
15518 + au_finfo_fin(file);
15522 +/* ---------------------------------------------------------------------- */
15524 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
15527 + struct file *h_file;
15530 + h_file = au_hf_top(file);
15532 + err = vfsub_flush(h_file, id);
15536 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
15538 + return au_do_flush(file, id, au_do_flush_nondir);
15541 +/* ---------------------------------------------------------------------- */
15543 + * read and write functions acquire [fdi]_rwsem once, but release before
15544 + * mmap_sem. This is because to stop a race condition between mmap(2).
15545 + * Releasing these aufs-rwsem should be safe, no branch-management (by keeping
15546 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
15547 + * read functions after [fdi]_rwsem are released, but it should be harmless.
15550 +/* Callers should call au_read_post() or fput() in the end */
15551 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc)
15553 + struct file *h_file;
15556 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0, lsc);
15558 + di_read_unlock(file->f_path.dentry, AuLock_IR);
15559 + h_file = au_hf_top(file);
15560 + get_file(h_file);
15562 + fi_read_unlock(file);
15564 + h_file = ERR_PTR(err);
15569 +static void au_read_post(struct inode *inode, struct file *h_file)
15571 + /* update without lock, I don't think it a problem */
15572 + fsstack_copy_attr_atime(inode, file_inode(h_file));
15576 +struct au_write_pre {
15578 + unsigned int lsc;
15582 + aufs_bindex_t btop;
15586 + * return with iinfo is write-locked
15587 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
15590 +static struct file *au_write_pre(struct file *file, int do_ready,
15591 + struct au_write_pre *wpre)
15593 + struct file *h_file;
15594 + struct dentry *dentry;
15596 + unsigned int lsc;
15597 + struct au_pin pin;
15602 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1, lsc);
15603 + h_file = ERR_PTR(err);
15604 + if (unlikely(err))
15607 + dentry = file->f_path.dentry;
15609 + err = au_ready_to_write(file, -1, &pin);
15610 + if (unlikely(err)) {
15611 + h_file = ERR_PTR(err);
15612 + di_write_unlock(dentry);
15617 + di_downgrade_lock(dentry, /*flags*/0);
15619 + wpre->btop = au_fbtop(file);
15620 + h_file = au_hf_top(file);
15621 + get_file(h_file);
15623 + wpre->blks = file_inode(h_file)->i_blocks;
15626 + di_read_unlock(dentry, /*flags*/0);
15629 + fi_write_unlock(file);
15634 +static void au_write_post(struct inode *inode, struct file *h_file,
15635 + struct au_write_pre *wpre, ssize_t written)
15637 + struct inode *h_inode;
15639 + au_cpup_attr_timesizes(inode);
15640 + AuDebugOn(au_ibtop(inode) != wpre->btop);
15641 + h_inode = file_inode(h_file);
15642 + inode->i_mode = h_inode->i_mode;
15643 + ii_write_unlock(inode);
15644 + /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
15646 + au_fhsm_wrote(inode->i_sb, wpre->btop,
15647 + /*force*/h_inode->i_blocks > wpre->blks);
15652 + * todo: very ugly
15653 + * it locks both of i_mutex and si_rwsem for read in safe.
15654 + * if the plink maintenance mode continues forever (that is the problem),
15655 + * may loop forever.
15657 +static void au_mtx_and_read_lock(struct inode *inode)
15660 + struct super_block *sb = inode->i_sb;
15663 + inode_lock(inode);
15664 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
15667 + inode_unlock(inode);
15668 + si_read_lock(sb, AuLock_NOPLMW);
15669 + si_read_unlock(sb);
15673 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
15674 + struct iov_iter *iov_iter)
15677 + struct file *file;
15678 + ssize_t (*iter)(struct kiocb *, struct iov_iter *);
15680 + err = security_file_permission(h_file, rw);
15681 + if (unlikely(err))
15684 + err = -ENOSYS; /* the branch doesn't have its ->(read|write)_iter() */
15686 + if (rw == MAY_READ)
15687 + iter = h_file->f_op->read_iter;
15688 + else if (rw == MAY_WRITE)
15689 + iter = h_file->f_op->write_iter;
15691 + file = kio->ki_filp;
15692 + kio->ki_filp = h_file;
15695 + err = iter(kio, iov_iter);
15698 + /* currently there is no such fs */
15700 + kio->ki_filp = file;
15706 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15709 + struct file *file, *h_file;
15710 + struct inode *inode;
15711 + struct super_block *sb;
15713 + file = kio->ki_filp;
15714 + inode = file_inode(file);
15715 + sb = inode->i_sb;
15716 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15718 + h_file = au_read_pre(file, /*keep_fi*/1, /*lsc*/0);
15719 + err = PTR_ERR(h_file);
15720 + if (IS_ERR(h_file))
15723 + if (au_test_loopback_kthread()) {
15724 + au_warn_loopback(h_file->f_path.dentry->d_sb);
15725 + if (file->f_mapping != h_file->f_mapping) {
15726 + file->f_mapping = h_file->f_mapping;
15727 + smp_mb(); /* unnecessary? */
15730 + fi_read_unlock(file);
15732 + err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
15733 + /* todo: necessary? */
15734 + /* file->f_ra = h_file->f_ra; */
15735 + au_read_post(inode, h_file);
15738 + si_read_unlock(sb);
15742 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15745 + struct au_write_pre wpre;
15746 + struct inode *inode;
15747 + struct file *file, *h_file;
15749 + file = kio->ki_filp;
15750 + inode = file_inode(file);
15751 + au_mtx_and_read_lock(inode);
15754 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15755 + err = PTR_ERR(h_file);
15756 + if (IS_ERR(h_file))
15759 + err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
15760 + au_write_post(inode, h_file, &wpre, err);
15763 + si_read_unlock(inode->i_sb);
15764 + inode_unlock(inode);
15769 + * We may be able to remove aufs_splice_{read,write}() since almost all FSes
15770 + * don't have their own .splice_{read,write} implimentations, and they use
15771 + * generic_file_splice_read() and iter_file_splice_write() who can act like the
15772 + * simple converters to f_op->iter_read() and ->iter_write().
15773 + * But we keep our own implementations because some non-mainlined FSes may have
15774 + * their own .splice_{read,write} implimentations and aufs doesn't want to take
15775 + * away an opportunity to co-work with aufs from them.
15777 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
15778 + struct pipe_inode_info *pipe, size_t len,
15779 + unsigned int flags)
15782 + struct file *h_file;
15783 + struct inode *inode;
15784 + struct super_block *sb;
15786 + inode = file_inode(file);
15787 + sb = inode->i_sb;
15788 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15790 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15791 + err = PTR_ERR(h_file);
15792 + if (IS_ERR(h_file))
15795 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
15796 + /* todo: necessary? */
15797 + /* file->f_ra = h_file->f_ra; */
15798 + au_read_post(inode, h_file);
15801 + si_read_unlock(sb);
15806 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
15807 + size_t len, unsigned int flags)
15810 + struct au_write_pre wpre;
15811 + struct inode *inode;
15812 + struct file *h_file;
15814 + inode = file_inode(file);
15815 + au_mtx_and_read_lock(inode);
15818 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15819 + err = PTR_ERR(h_file);
15820 + if (IS_ERR(h_file))
15823 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
15824 + au_write_post(inode, h_file, &wpre, err);
15827 + si_read_unlock(inode->i_sb);
15828 + inode_unlock(inode);
15832 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
15836 + struct au_write_pre wpre;
15837 + struct inode *inode;
15838 + struct file *h_file;
15840 + inode = file_inode(file);
15841 + au_mtx_and_read_lock(inode);
15844 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15845 + err = PTR_ERR(h_file);
15846 + if (IS_ERR(h_file))
15850 + err = vfs_fallocate(h_file, mode, offset, len);
15852 + au_write_post(inode, h_file, &wpre, /*written*/1);
15855 + si_read_unlock(inode->i_sb);
15856 + inode_unlock(inode);
15860 +static ssize_t aufs_copy_file_range(struct file *src, loff_t src_pos,
15861 + struct file *dst, loff_t dst_pos,
15862 + size_t len, unsigned int flags)
15865 + struct au_write_pre wpre;
15866 + enum { SRC, DST };
15868 + struct inode *inode;
15869 + struct file *h_file;
15870 + struct super_block *h_sb;
15872 +#define a_src a[SRC]
15873 +#define a_dst a[DST]
15876 + a_src.inode = file_inode(src);
15877 + if (unlikely(!S_ISREG(a_src.inode->i_mode)))
15879 + a_dst.inode = file_inode(dst);
15880 + if (unlikely(!S_ISREG(a_dst.inode->i_mode)))
15883 + au_mtx_and_read_lock(a_dst.inode);
15885 + * in order to match the order in di_write_lock2_{child,parent}(),
15886 + * use f_path.dentry for this comparison.
15888 + if (src->f_path.dentry < dst->f_path.dentry) {
15889 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_1);
15890 + err = PTR_ERR(a_src.h_file);
15891 + if (IS_ERR(a_src.h_file))
15894 + wpre.lsc = AuLsc_FI_2;
15895 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15896 + err = PTR_ERR(a_dst.h_file);
15897 + if (IS_ERR(a_dst.h_file)) {
15898 + au_read_post(a_src.inode, a_src.h_file);
15902 + wpre.lsc = AuLsc_FI_1;
15903 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15904 + err = PTR_ERR(a_dst.h_file);
15905 + if (IS_ERR(a_dst.h_file))
15908 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_2);
15909 + err = PTR_ERR(a_src.h_file);
15910 + if (IS_ERR(a_src.h_file)) {
15911 + au_write_post(a_dst.inode, a_dst.h_file, &wpre,
15918 + a_src.h_sb = file_inode(a_src.h_file)->i_sb;
15919 + a_dst.h_sb = file_inode(a_dst.h_file)->i_sb;
15920 + if (unlikely(a_src.h_sb != a_dst.h_sb)) {
15926 + err = vfsub_copy_file_range(a_src.h_file, src_pos, a_dst.h_file,
15927 + dst_pos, len, flags);
15930 + au_write_post(a_dst.inode, a_dst.h_file, &wpre, err);
15931 + fi_read_unlock(src);
15932 + au_read_post(a_src.inode, a_src.h_file);
15934 + si_read_unlock(a_dst.inode->i_sb);
15935 + inode_unlock(a_dst.inode);
15942 +/* ---------------------------------------------------------------------- */
15945 + * The locking order around current->mmap_sem.
15946 + * - in most and regular cases
15947 + * file I/O syscall -- aufs_read() or something
15948 + * -- si_rwsem for read -- mmap_sem
15949 + * (Note that [fdi]i_rwsem are released before mmap_sem).
15951 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
15952 + * This AB-BA order is definitely bad, but is not a problem since "si_rwsem for
15953 + * read" allows multiple processes to acquire it and [fdi]i_rwsem are not held
15954 + * in file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
15955 + * It means that when aufs acquires si_rwsem for write, the process should never
15956 + * acquire mmap_sem.
15958 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
15959 + * problem either since any directory is not able to be mmap-ed.
15960 + * The similar scenario is applied to aufs_readlink() too.
15963 +#if 0 /* stop calling security_file_mmap() */
15964 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
15965 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
15967 +static unsigned long au_arch_prot_conv(unsigned long flags)
15969 + /* currently ppc64 only */
15970 +#ifdef CONFIG_PPC64
15971 + /* cf. linux/arch/powerpc/include/asm/mman.h */
15972 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
15973 + return AuConv_VM_PROT(flags, SAO);
15975 + AuDebugOn(arch_calc_vm_prot_bits(-1));
15980 +static unsigned long au_prot_conv(unsigned long flags)
15982 + return AuConv_VM_PROT(flags, READ)
15983 + | AuConv_VM_PROT(flags, WRITE)
15984 + | AuConv_VM_PROT(flags, EXEC)
15985 + | au_arch_prot_conv(flags);
15988 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
15989 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
15991 +static unsigned long au_flag_conv(unsigned long flags)
15993 + return AuConv_VM_MAP(flags, GROWSDOWN)
15994 + | AuConv_VM_MAP(flags, DENYWRITE)
15995 + | AuConv_VM_MAP(flags, LOCKED);
15999 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
16002 + const unsigned char wlock
16003 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
16004 + struct super_block *sb;
16005 + struct file *h_file;
16006 + struct inode *inode;
16008 + AuDbgVmRegion(file, vma);
16010 + inode = file_inode(file);
16011 + sb = inode->i_sb;
16013 + si_read_lock(sb, AuLock_NOPLMW);
16015 + h_file = au_write_pre(file, wlock, /*wpre*/NULL);
16017 + err = PTR_ERR(h_file);
16018 + if (IS_ERR(h_file))
16022 + au_set_mmapped(file);
16023 + au_vm_file_reset(vma, h_file);
16025 + * we cannot call security_mmap_file() here since it may acquire
16026 + * mmap_sem or i_mutex.
16028 + * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
16029 + * au_flag_conv(vma->vm_flags));
16032 + err = call_mmap(h_file, vma);
16034 + au_vm_prfile_set(vma, file);
16035 + fsstack_copy_attr_atime(inode, file_inode(h_file));
16036 + goto out_fput; /* success */
16038 + au_unset_mmapped(file);
16039 + au_vm_file_reset(vma, file);
16043 + ii_write_unlock(inode);
16048 + si_read_unlock(sb);
16054 +/* ---------------------------------------------------------------------- */
16056 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
16060 + struct au_write_pre wpre;
16061 + struct inode *inode;
16062 + struct file *h_file;
16064 + err = 0; /* -EBADF; */ /* posix? */
16065 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
16068 + inode = file_inode(file);
16069 + au_mtx_and_read_lock(inode);
16072 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
16073 + err = PTR_ERR(h_file);
16074 + if (IS_ERR(h_file))
16077 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
16078 + au_write_post(inode, h_file, &wpre, /*written*/0);
16081 + si_read_unlock(inode->i_sb);
16082 + inode_unlock(inode);
16087 +static int aufs_fasync(int fd, struct file *file, int flag)
16090 + struct file *h_file;
16091 + struct super_block *sb;
16093 + sb = file->f_path.dentry->d_sb;
16094 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16096 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16097 + err = PTR_ERR(h_file);
16098 + if (IS_ERR(h_file))
16101 + if (h_file->f_op->fasync)
16102 + err = h_file->f_op->fasync(fd, h_file, flag);
16103 + fput(h_file); /* instead of au_read_post() */
16106 + si_read_unlock(sb);
16110 +static int aufs_setfl(struct file *file, unsigned long arg)
16113 + struct file *h_file;
16114 + struct super_block *sb;
16116 + sb = file->f_path.dentry->d_sb;
16117 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16119 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16120 + err = PTR_ERR(h_file);
16121 + if (IS_ERR(h_file))
16124 + /* stop calling h_file->fasync */
16125 + arg |= vfsub_file_flags(file) & FASYNC;
16126 + err = setfl(/*unused fd*/-1, h_file, arg);
16127 + fput(h_file); /* instead of au_read_post() */
16130 + si_read_unlock(sb);
16134 +/* ---------------------------------------------------------------------- */
16136 +/* no one supports this operation, currently */
16137 +#if 0 /* reserved for future use */
16138 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
16139 + size_t len, loff_t *pos, int more)
16144 +/* ---------------------------------------------------------------------- */
16146 +const struct file_operations aufs_file_fop = {
16147 + .owner = THIS_MODULE,
16149 + .llseek = default_llseek,
16151 + .read_iter = aufs_read_iter,
16152 + .write_iter = aufs_write_iter,
16154 +#ifdef CONFIG_AUFS_POLL
16155 + .poll = aufs_poll,
16157 + .unlocked_ioctl = aufs_ioctl_nondir,
16158 +#ifdef CONFIG_COMPAT
16159 + .compat_ioctl = aufs_compat_ioctl_nondir,
16161 + .mmap = aufs_mmap,
16162 + .open = aufs_open_nondir,
16163 + .flush = aufs_flush_nondir,
16164 + .release = aufs_release_nondir,
16165 + .fsync = aufs_fsync_nondir,
16166 + .fasync = aufs_fasync,
16167 + /* .sendpage = aufs_sendpage, */
16168 + .setfl = aufs_setfl,
16169 + .splice_write = aufs_splice_write,
16170 + .splice_read = aufs_splice_read,
16171 +#if 0 /* reserved for future use */
16172 + .aio_splice_write = aufs_aio_splice_write,
16173 + .aio_splice_read = aufs_aio_splice_read,
16175 + .fallocate = aufs_fallocate,
16176 + .copy_file_range = aufs_copy_file_range
16178 diff -urN /usr/share/empty/fs/aufs/fsctx.c linux/fs/aufs/fsctx.c
16179 --- /usr/share/empty/fs/aufs/fsctx.c 1970-01-01 01:00:00.000000000 +0100
16180 +++ linux/fs/aufs/fsctx.c 2022-10-03 07:51:14.926209260 +0200
16182 +// SPDX-License-Identifier: GPL-2.0
16184 + * Copyright (C) 2022 Junjiro R. Okajima
16186 + * This program, aufs is free software; you can redistribute it and/or modify
16187 + * it under the terms of the GNU General Public License as published by
16188 + * the Free Software Foundation; either version 2 of the License, or
16189 + * (at your option) any later version.
16191 + * This program is distributed in the hope that it will be useful,
16192 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16193 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16194 + * GNU General Public License for more details.
16196 + * You should have received a copy of the GNU General Public License
16197 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16201 + * fs context, aka new mount api
16204 +#include <linux/fs_context.h>
16207 +struct au_fsctx_opts {
16208 + aufs_bindex_t bindex;
16209 + unsigned char skipped;
16210 + struct au_opt *opt, *opt_tail;
16211 + struct super_block *sb;
16212 + struct au_sbinfo *sbinfo;
16213 + struct au_opts opts;
16216 +/* stop extra interpretation of errno in mount(8), and strange error messages */
16217 +static int cvt_err(int err)
16231 +static int au_fsctx_reconfigure(struct fs_context *fc)
16234 + unsigned int mntflags;
16235 + struct dentry *root;
16236 + struct super_block *sb;
16237 + struct inode *inode;
16238 + struct au_fsctx_opts *a = fc->fs_private;
16240 + AuDbg("fc %p\n", fc);
16244 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16246 + di_write_lock_child(root);
16247 + err = au_opts_verify(sb, fc->sb_flags, /*pending*/0);
16248 + aufs_write_unlock(root);
16251 + inode = d_inode(root);
16252 + inode_lock(inode);
16253 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16254 + if (unlikely(err))
16256 + di_write_lock_child(root);
16258 + /* au_opts_remount() may return an error */
16259 + err = au_opts_remount(sb, &a->opts);
16261 + if (au_ftest_opts(a->opts.flags, REFRESH))
16262 + au_remount_refresh(sb, au_ftest_opts(a->opts.flags,
16265 + if (au_ftest_opts(a->opts.flags, REFRESH_DYAOP)) {
16266 + mntflags = au_mntflags(sb);
16267 + do_dx = !!au_opt_test(mntflags, DIO);
16268 + au_dy_arefresh(do_dx);
16271 + au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
16272 + aufs_write_unlock(root);
16275 + inode_unlock(inode);
16276 + err = cvt_err(err);
16282 +/* ---------------------------------------------------------------------- */
16284 +static int au_fsctx_fill_super(struct super_block *sb, struct fs_context *fc)
16287 + struct au_fsctx_opts *a = fc->fs_private;
16288 + struct au_sbinfo *sbinfo = a->sbinfo;
16289 + struct dentry *root;
16290 + struct inode *inode;
16292 + sbinfo->si_sb = sb;
16293 + sb->s_fs_info = sbinfo;
16294 + kobject_get(&sbinfo->si_kobj);
16296 + __si_write_lock(sb);
16298 + au_sbilist_add(sb);
16300 + /* all timestamps always follow the ones on the branch */
16301 + sb->s_flags |= SB_NOATIME | SB_NODIRATIME;
16302 + sb->s_flags |= SB_I_VERSION; /* do we really need this? */
16303 + sb->s_op = &aufs_sop;
16304 + sb->s_d_op = &aufs_dop;
16305 + sb->s_magic = AUFS_SUPER_MAGIC;
16306 + sb->s_maxbytes = 0;
16307 + sb->s_stack_depth = 1;
16308 + au_export_init(sb);
16309 + au_xattr_init(sb);
16311 + err = au_alloc_root(sb);
16312 + if (unlikely(err)) {
16313 + si_write_unlock(sb);
16316 + root = sb->s_root;
16317 + inode = d_inode(root);
16318 + ii_write_lock_parent(inode);
16319 + aufs_write_unlock(root);
16321 + /* lock vfs_inode first, then aufs. */
16322 + inode_lock(inode);
16323 + aufs_write_lock(root);
16324 + err = au_opts_mount(sb, &a->opts);
16326 + if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
16327 + sb->s_d_op = &aufs_dop_noreval;
16328 + /* infofc(fc, "%ps", sb->s_d_op); */
16329 + pr_info("%ps\n", sb->s_d_op);
16330 + au_refresh_dop(root, /*force_reval*/0);
16331 + sbinfo->si_iop_array = aufs_iop_nogetattr;
16332 + au_refresh_iop(inode, /*force_getattr*/0);
16334 + aufs_write_unlock(root);
16335 + inode_unlock(inode);
16337 + goto out; /* success */
16340 + sb->s_root = NULL;
16343 + if (unlikely(err))
16344 + kobject_put(&sbinfo->si_kobj);
16346 + err = cvt_err(err);
16351 +static int au_fsctx_get_tree(struct fs_context *fc)
16355 + AuDbg("fc %p\n", fc);
16356 + err = get_tree_nodev(fc, au_fsctx_fill_super);
16362 +/* ---------------------------------------------------------------------- */
16364 +static void au_fsctx_dump(struct au_opts *opts)
16366 +#ifdef CONFIG_AUFS_DEBUG
16367 + /* reduce stack space */
16369 + struct au_opt_add *add;
16370 + struct au_opt_del *del;
16371 + struct au_opt_mod *mod;
16372 + struct au_opt_xino *xino;
16373 + struct au_opt_xino_itrunc *xino_itrunc;
16374 + struct au_opt_wbr_create *create;
16376 + struct au_opt *opt;
16379 + while (opt->type != Opt_tail) {
16380 + switch (opt->type) {
16382 + u.add = &opt->add;
16383 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
16384 + u.add->bindex, u.add->pathname, u.add->perm,
16385 + u.add->path.dentry);
16390 + u.del = &opt->del;
16391 + AuDbg("del {%s, %p}\n",
16392 + u.del->pathname, u.del->h_path.dentry);
16397 + u.mod = &opt->mod;
16398 + AuDbg("mod {%s, 0x%x, %p}\n",
16399 + u.mod->path, u.mod->perm, u.mod->h_root);
16402 + u.add = &opt->add;
16403 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
16404 + u.add->bindex, u.add->pathname, u.add->perm,
16405 + u.add->path.dentry);
16407 + case Opt_prepend:
16408 + u.add = &opt->add;
16409 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
16410 + u.add->bindex, u.add->pathname, u.add->perm,
16411 + u.add->path.dentry);
16415 + AuDbg("dirwh %d\n", opt->dirwh);
16417 + case Opt_rdcache:
16418 + AuDbg("rdcache %d\n", opt->rdcache);
16421 + AuDbg("rdblk %d\n", opt->rdblk);
16424 + AuDbg("rdhash %u\n", opt->rdhash);
16428 + u.xino = &opt->xino;
16429 + AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
16432 +#define au_fsctx_TF(name) \
16433 + case Opt_##name: \
16437 + AuLabel(no##name); \
16440 + /* simple true/false flag */
16441 + au_fsctx_TF(trunc_xino);
16442 + au_fsctx_TF(trunc_xib);
16443 + au_fsctx_TF(dirperm1);
16444 + au_fsctx_TF(plink);
16445 + au_fsctx_TF(shwh);
16446 + au_fsctx_TF(dio);
16447 + au_fsctx_TF(warn_perm);
16448 + au_fsctx_TF(verbose);
16449 + au_fsctx_TF(sum);
16450 + au_fsctx_TF(dirren);
16451 + au_fsctx_TF(acl);
16452 +#undef au_fsctx_TF
16454 + case Opt_trunc_xino_path:
16456 + case Opt_itrunc_xino:
16457 + u.xino_itrunc = &opt->xino_itrunc;
16458 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
16464 + case Opt_list_plink:
16465 + AuLabel(list_plink);
16468 + AuDbg("udba %d, %s\n",
16469 + opt->udba, au_optstr_udba(opt->udba));
16471 + case Opt_diropq_a:
16472 + AuLabel(diropq_a);
16474 + case Opt_diropq_w:
16475 + AuLabel(diropq_w);
16480 + case Opt_wbr_create:
16481 + u.create = &opt->wbr_create;
16482 + AuDbg("create %d, %s\n", u.create->wbr_create,
16483 + au_optstr_wbr_create(u.create->wbr_create));
16484 + switch (u.create->wbr_create) {
16485 + case AuWbrCreate_MFSV:
16487 + case AuWbrCreate_PMFSV:
16488 + AuDbg("%d sec\n", u.create->mfs_second);
16490 + case AuWbrCreate_MFSRR:
16492 + case AuWbrCreate_TDMFS:
16493 + AuDbg("%llu watermark\n",
16494 + u.create->mfsrr_watermark);
16496 + case AuWbrCreate_MFSRRV:
16498 + case AuWbrCreate_TDMFSV:
16500 + case AuWbrCreate_PMFSRRV:
16501 + AuDbg("%llu watermark, %d sec\n",
16502 + u.create->mfsrr_watermark,
16503 + u.create->mfs_second);
16507 + case Opt_wbr_copyup:
16508 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
16509 + au_optstr_wbr_copyup(opt->wbr_copyup));
16511 + case Opt_fhsm_sec:
16512 + AuDbg("fhsm_sec %u\n", opt->fhsm_second);
16516 + AuDbg("type %d\n", opt->type);
16524 +/* ---------------------------------------------------------------------- */
16527 + * For conditionally compiled mount options.
16528 + * Instead of fsparam_flag_no(), use this macro to distinguish ignore_silent.
16530 +#define au_ignore_flag(name, action) \
16531 + fsparam_flag(name, action), \
16532 + fsparam_flag("no" name, Opt_ignore_silent)
16534 +const struct fs_parameter_spec aufs_fsctx_paramspec[] = {
16535 + fsparam_string("br", Opt_br),
16537 + /* "add=%d:%s" or "ins=%d:%s" */
16538 + fsparam_string("add", Opt_add),
16539 + fsparam_string("ins", Opt_add),
16540 + fsparam_path("append", Opt_append),
16541 + fsparam_path("prepend", Opt_prepend),
16543 + fsparam_path("del", Opt_del),
16544 + /* fsparam_s32("idel", Opt_idel), */
16545 + fsparam_path("mod", Opt_mod),
16546 + /* fsparam_string("imod", Opt_imod), */
16548 + fsparam_s32("dirwh", Opt_dirwh),
16550 + fsparam_path("xino", Opt_xino),
16551 + fsparam_flag("noxino", Opt_noxino),
16552 + fsparam_flag_no("trunc_xino", Opt_trunc_xino),
16553 + /* "trunc_xino_v=%d:%d" */
16554 + /* fsparam_string("trunc_xino_v", Opt_trunc_xino_v), */
16555 + fsparam_path("trunc_xino", Opt_trunc_xino_path),
16556 + fsparam_s32("itrunc_xino", Opt_itrunc_xino),
16557 + /* fsparam_path("zxino", Opt_zxino), */
16558 + fsparam_flag_no("trunc_xib", Opt_trunc_xib),
16560 +#ifdef CONFIG_PROC_FS
16561 + fsparam_flag_no("plink", Opt_plink),
16563 + au_ignore_flag("plink", Opt_ignore),
16566 +#ifdef CONFIG_AUFS_DEBUG
16567 + fsparam_flag("list_plink", Opt_list_plink),
16570 + fsparam_string("udba", Opt_udba),
16572 + fsparam_flag_no("dio", Opt_dio),
16574 +#ifdef CONFIG_AUFS_DIRREN
16575 + fsparam_flag_no("dirren", Opt_dirren),
16577 + au_ignore_flag("dirren", Opt_ignore),
16580 +#ifdef CONFIG_AUFS_FHSM
16581 + fsparam_s32("fhsm_sec", Opt_fhsm_sec),
16583 + fsparam_s32("fhsm_sec", Opt_ignore),
16586 + /* always | a | whiteouted | w */
16587 + fsparam_string("diropq", Opt_diropq),
16589 + fsparam_flag_no("warn_perm", Opt_warn_perm),
16591 +#ifdef CONFIG_AUFS_SHWH
16592 + fsparam_flag_no("shwh", Opt_shwh),
16594 + au_ignore_flag("shwh", Opt_err),
16597 + fsparam_flag_no("dirperm1", Opt_dirperm1),
16599 + fsparam_flag_no("verbose", Opt_verbose),
16600 + fsparam_flag("v", Opt_verbose),
16601 + fsparam_flag("quiet", Opt_noverbose),
16602 + fsparam_flag("q", Opt_noverbose),
16603 + /* user-space may handle this */
16604 + fsparam_flag("silent", Opt_noverbose),
16606 + fsparam_flag_no("sum", Opt_sum),
16607 + fsparam_flag("wsum", Opt_wsum),
16609 + fsparam_s32("rdcache", Opt_rdcache),
16610 + /* "def" or s32 */
16611 + fsparam_string("rdblk", Opt_rdblk),
16612 + /* "def" or s32 */
16613 + fsparam_string("rdhash", Opt_rdhash),
16615 + fsparam_string("create", Opt_wbr_create),
16616 + fsparam_string("create_policy", Opt_wbr_create),
16617 + fsparam_string("cpup", Opt_wbr_copyup),
16618 + fsparam_string("copyup", Opt_wbr_copyup),
16619 + fsparam_string("copyup_policy", Opt_wbr_copyup),
16621 + /* generic VFS flag */
16622 +#ifdef CONFIG_FS_POSIX_ACL
16623 + fsparam_flag_no("acl", Opt_acl),
16625 + au_ignore_flag("acl", Opt_ignore),
16628 + /* internal use for the scripts */
16629 + fsparam_string("si", Opt_ignore_silent),
16631 + /* obsoleted, keep them temporary */
16632 + fsparam_flag("nodlgt", Opt_ignore_silent),
16633 + fsparam_flag("clean_plink", Opt_ignore),
16634 + fsparam_string("dirs", Opt_br),
16635 + fsparam_u32("debug", Opt_ignore),
16636 + /* "whiteout" or "all" */
16637 + fsparam_string("delete", Opt_ignore),
16638 + fsparam_string("imap", Opt_ignore),
16640 + /* temporary workaround, due to old mount(8)? */
16641 + fsparam_flag("relatime", Opt_ignore_silent),
16646 +static int au_fsctx_parse_do_add(struct fs_context *fc, struct au_opt *opt,
16647 + char *brspec, size_t speclen,
16648 + aufs_bindex_t bindex)
16653 + AuDbg("brspec %s\n", brspec);
16657 + speclen = strlen(brspec);
16658 + /* will be freed by au_fsctx_free() */
16659 + p = kmemdup_nul(brspec, speclen, GFP_NOFS);
16660 + if (unlikely(!p)) {
16661 + errorfc(fc, "failed in %s", brspec);
16664 + err = au_opt_add(opt, p, fc->sb_flags, bindex);
16671 +static int au_fsctx_parse_br(struct fs_context *fc, char *brspec)
16675 + struct au_fsctx_opts *a = fc->fs_private;
16676 + struct au_opt *opt = a->opt;
16677 + aufs_bindex_t bindex = a->bindex;
16679 + AuDbg("brspec %s\n", brspec);
16682 + while ((p = strsep(&brspec, ":")) && *p) {
16683 + err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, bindex);
16685 + if (unlikely(err))
16689 + if (unlikely(opt > a->opt_tail)) {
16695 + opt->type = Opt_tail;
16698 + a->bindex = bindex;
16705 +static int au_fsctx_parse_add(struct fs_context *fc, char *addspec)
16709 + struct au_fsctx_opts *a = fc->fs_private;
16710 + struct au_opt *opt = a->opt;
16713 + p = strchr(addspec, ':');
16714 + if (unlikely(!p)) {
16715 + errorfc(fc, "bad arg in %s", addspec);
16719 + err = kstrtoint(addspec, 0, &n);
16720 + if (unlikely(err)) {
16721 + errorfc(fc, "bad integer in %s", addspec);
16724 + AuDbg("n %d\n", n);
16725 + err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, n);
16732 +static int au_fsctx_parse_del(struct fs_context *fc, struct au_opt_del *del,
16733 + struct fs_parameter *param)
16738 + /* will be freed by au_fsctx_free() */
16739 + del->pathname = kmemdup_nul(param->string, param->size, GFP_NOFS);
16740 + if (unlikely(!del->pathname))
16742 + AuDbg("del %s\n", del->pathname);
16743 + err = vfsub_kern_path(del->pathname, AuOpt_LkupDirFlags, &del->h_path);
16744 + if (unlikely(err))
16745 + errorfc(fc, "lookup failed %s (%d)", del->pathname, err);
16752 +#if 0 /* reserved for future use */
16753 +static int au_fsctx_parse_idel(struct fs_context *fc, struct au_opt_del *del,
16754 + aufs_bindex_t bindex)
16757 + struct super_block *sb;
16758 + struct dentry *root;
16759 + struct au_fsctx_opts *a = fc->fs_private;
16765 + root = sb->s_root;
16766 + aufs_read_lock(root, AuLock_FLUSH);
16767 + if (bindex < 0 || au_sbbot(sb) < bindex) {
16768 + errorfc(fc, "out of bounds, %d", bindex);
16773 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
16774 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
16777 + aufs_read_unlock(root, !AuLock_IR);
16783 +static int au_fsctx_parse_mod(struct fs_context *fc, struct au_opt_mod *mod,
16784 + struct fs_parameter *param)
16787 + struct path path;
16791 + /* will be freed by au_fsctx_free() */
16792 + mod->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16793 + if (unlikely(!mod->path))
16797 + p = strchr(mod->path, '=');
16798 + if (unlikely(!p)) {
16799 + errorfc(fc, "no permission %s", mod->path);
16804 + err = vfsub_kern_path(mod->path, AuOpt_LkupDirFlags, &path);
16805 + if (unlikely(err)) {
16806 + errorfc(fc, "lookup failed %s (%d)", mod->path, err);
16810 + mod->perm = au_br_perm_val(p);
16811 + AuDbg("mod path %s, perm 0x%x, %s", mod->path, mod->perm, p);
16812 + mod->h_root = dget(path.dentry);
16820 +#if 0 /* reserved for future use */
16821 +static int au_fsctx_parse_imod(struct fs_context *fc, struct au_opt_mod *mod,
16826 + struct super_block *sb;
16827 + struct dentry *root;
16828 + struct au_fsctx_opts *a = fc->fs_private;
16834 + p = strchr(ibrspec, ':');
16835 + if (unlikely(!p)) {
16836 + errorfc(fc, "no index, %s", ibrspec);
16840 + err = kstrtoint(ibrspec, 0, &n);
16841 + if (unlikely(err)) {
16842 + errorfc(fc, "bad integer in %s", ibrspec);
16845 + AuDbg("n %d\n", n);
16847 + root = sb->s_root;
16848 + aufs_read_lock(root, AuLock_FLUSH);
16849 + if (n < 0 || au_sbbot(sb) < n) {
16850 + errorfc(fc, "out of bounds, %d", bindex);
16855 + mod->perm = au_br_perm_val(p);
16856 + AuDbg("mod path %s, perm 0x%x, %s\n",
16857 + mod->path, mod->perm, p);
16858 + mod->h_root = dget(au_h_dptr(root, bindex));
16861 + aufs_read_unlock(root, !AuLock_IR);
16868 +static int au_fsctx_parse_xino(struct fs_context *fc,
16869 + struct au_opt_xino *xino,
16870 + struct fs_parameter *param)
16873 + struct au_fsctx_opts *a = fc->fs_private;
16876 + /* will be freed by au_opts_free() */
16877 + xino->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16878 + if (unlikely(!xino->path))
16880 + AuDbg("path %s\n", xino->path);
16882 + xino->file = au_xino_create(a->sb, xino->path, /*silent*/0,
16884 + err = PTR_ERR(xino->file);
16885 + if (IS_ERR(xino->file)) {
16886 + xino->file = NULL;
16891 + if (unlikely(a->sb && xino->file->f_path.dentry->d_sb == a->sb)) {
16893 + errorfc(fc, "%s must be outside", xino->path);
16902 +int au_fsctx_parse_xino_itrunc_path(struct fs_context *fc,
16903 + struct au_opt_xino_itrunc *xino_itrunc,
16907 + aufs_bindex_t bbot, bindex;
16908 + struct path path;
16909 + struct dentry *root;
16910 + struct au_fsctx_opts *a = fc->fs_private;
16912 + AuDebugOn(!a->sb);
16914 + err = vfsub_kern_path(pathname, AuOpt_LkupDirFlags, &path);
16915 + if (unlikely(err)) {
16916 + errorfc(fc, "lookup failed %s (%d)", pathname, err);
16920 + xino_itrunc->bindex = -1;
16921 + root = a->sb->s_root;
16922 + aufs_read_lock(root, AuLock_FLUSH);
16923 + bbot = au_sbbot(a->sb);
16924 + for (bindex = 0; bindex <= bbot; bindex++) {
16925 + if (au_h_dptr(root, bindex) == path.dentry) {
16926 + xino_itrunc->bindex = bindex;
16930 + aufs_read_unlock(root, !AuLock_IR);
16933 + if (unlikely(xino_itrunc->bindex < 0)) {
16935 + errorfc(fc, "no such branch %s", pathname);
16943 +static int au_fsctx_parse_xino_itrunc(struct fs_context *fc,
16944 + struct au_opt_xino_itrunc *xino_itrunc,
16945 + unsigned int bindex)
16948 + aufs_bindex_t bbot;
16949 + struct super_block *sb;
16950 + struct au_fsctx_opts *a = fc->fs_private;
16956 + si_noflush_read_lock(sb);
16957 + bbot = au_sbbot(sb);
16958 + si_read_unlock(sb);
16959 + if (bindex <= bbot)
16960 + xino_itrunc->bindex = bindex;
16963 + errorfc(fc, "out of bounds, %u", bindex);
16970 +static int au_fsctx_parse_param(struct fs_context *fc, struct fs_parameter *param)
16973 + struct fs_parse_result result;
16974 + struct au_fsctx_opts *a = fc->fs_private;
16975 + struct au_opt *opt = a->opt;
16977 + AuDbg("fc %p, param {key %s, string %s}\n",
16978 + fc, param->key, param->string);
16979 + err = fs_parse(fc, aufs_fsctx_paramspec, param, &result);
16980 + if (unlikely(err < 0))
16983 + AuDbg("token %d, res{negated %d, uint64 %llu}\n",
16984 + token, result.negated, result.uint_64);
16990 + err = au_fsctx_parse_br(fc, param->string);
16993 + err = au_fsctx_parse_add(fc, param->string);
16996 + err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
16997 + /*dummy bindex*/1);
16999 + case Opt_prepend:
17000 + err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17005 + err = au_fsctx_parse_del(fc, &opt->del, param);
17007 +#if 0 /* reserved for future use */
17014 + del->pathname = "(indexed)";
17015 + err = au_opts_parse_idel(fc, &opt->del, result.uint_32);
17020 + err = au_fsctx_parse_mod(fc, &opt->mod, param);
17022 +#ifdef IMOD /* reserved for future use */
17029 + u.mod->path = "(indexed)";
17030 + err = au_opts_parse_imod(fc, &opt->mod, param->string);
17035 + err = au_fsctx_parse_xino(fc, &opt->xino, param);
17037 + case Opt_trunc_xino_path:
17039 + errorfc(fc, "no such branch %s", param->string);
17042 + err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17046 + case Opt_trunc_xino_v:
17052 + err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17056 + case Opt_itrunc_xino:
17058 + errorfc(fc, "out of bounds %s", param->string);
17061 + err = au_fsctx_parse_xino_itrunc(fc, &opt->xino_itrunc,
17067 + opt->dirwh = result.int_32;
17070 + case Opt_rdcache:
17071 + if (unlikely(result.int_32 > AUFS_RDCACHE_MAX)) {
17072 + errorfc(fc, "rdcache must be smaller than %d",
17073 + AUFS_RDCACHE_MAX);
17077 + opt->rdcache = result.int_32;
17082 + opt->rdblk = AUFS_RDBLK_DEF;
17083 + if (!strcmp(param->string, "def"))
17086 + err = kstrtoint(param->string, 0, &result.int_32);
17087 + if (unlikely(err)) {
17088 + errorfc(fc, "bad value in %s", param->key);
17092 + if (unlikely(result.int_32 < 0
17093 + || result.int_32 > KMALLOC_MAX_SIZE)) {
17094 + errorfc(fc, "bad value in %s", param->key);
17097 + if (unlikely(result.int_32 && result.int_32 < NAME_MAX)) {
17098 + errorfc(fc, "rdblk must be larger than %d", NAME_MAX);
17102 + opt->rdblk = result.int_32;
17107 + opt->rdhash = AUFS_RDHASH_DEF;
17108 + if (!strcmp(param->string, "def"))
17111 + err = kstrtoint(param->string, 0, &result.int_32);
17112 + if (unlikely(err)) {
17113 + errorfc(fc, "bad value in %s", param->key);
17116 + /* how about zero? */
17117 + if (result.int_32 < 0
17118 + || result.int_32 * sizeof(struct hlist_head)
17119 + > KMALLOC_MAX_SIZE) {
17121 + errorfc(fc, "bad integer in %s", param->key);
17124 + opt->rdhash = result.int_32;
17129 + * As other options, fs/aufs/opts.c can handle these strings by
17130 + * match_token(). But "diropq=" is deprecated now and will
17131 + * never have other value. So simple strcmp() is enough here.
17133 + if (!strcmp(param->string, "a") ||
17134 + !strcmp(param->string, "always")) {
17136 + opt->type = Opt_diropq_a;
17137 + } else if (!strcmp(param->string, "w") ||
17138 + !strcmp(param->string, "whiteouted")) {
17140 + opt->type = Opt_diropq_w;
17142 + errorfc(fc, "unknown value %s", param->string);
17146 + opt->udba = au_udba_val(param->string);
17147 + if (opt->udba >= 0)
17150 + errorf(fc, "wrong value, %s", param->string);
17153 + case Opt_wbr_create:
17154 + opt->wbr_create.wbr_create
17155 + = au_wbr_create_val(param->string, &opt->wbr_create);
17156 + if (opt->wbr_create.wbr_create >= 0)
17159 + errorf(fc, "wrong value, %s", param->key);
17162 + case Opt_wbr_copyup:
17163 + opt->wbr_copyup = au_wbr_copyup_val(param->string);
17164 + if (opt->wbr_copyup >= 0)
17167 + errorfc(fc, "wrong value, %s", param->key);
17170 + case Opt_fhsm_sec:
17171 + if (unlikely(result.int_32 < 0)) {
17172 + errorfc(fc, "bad integer in %s\n", param->key);
17177 + opt->fhsm_second = result.int_32;
17179 + warnfc(fc, "ignored %s %s", param->key, param->string);
17182 + /* simple true/false flag */
17183 +#define au_fsctx_TF(name) \
17184 + case Opt_##name: \
17186 + opt->tf = !result.negated; \
17188 + au_fsctx_TF(trunc_xino);
17189 + au_fsctx_TF(trunc_xib);
17190 + au_fsctx_TF(dirperm1);
17191 + au_fsctx_TF(plink);
17192 + au_fsctx_TF(shwh);
17193 + au_fsctx_TF(dio);
17194 + au_fsctx_TF(warn_perm);
17195 + au_fsctx_TF(verbose);
17196 + au_fsctx_TF(sum);
17197 + au_fsctx_TF(dirren);
17198 + au_fsctx_TF(acl);
17199 +#undef au_fsctx_TF
17201 + case Opt_noverbose:
17203 + opt->type = Opt_verbose;
17209 + case Opt_list_plink:
17216 + warnfc(fc, "ignored %s", param->key);
17218 + case Opt_ignore_silent:
17227 + if (unlikely(err))
17235 + case Opt_noverbose:
17240 + opt->type = token;
17244 + if (unlikely(opt > a->opt_tail)) {
17248 + opt->type = Opt_tail;
17256 + * these options accept both 'name=val' and 'name:val' form.
17257 + * some accept optional '=' in its value.
17258 + * eg. br:/br1=rw:/br2=ro and br=/br1=rw:/br2=ro
17260 +static inline unsigned int is_colonopt(char *str)
17262 +#define do_test(name) \
17263 + if (!strncmp(str, name ":", sizeof(name))) \
17264 + return sizeof(name) - 1
17268 + do_test("append");
17269 + do_test("prepend");
17271 + /* do_test("idel"); */
17273 + /* do_test("imod"); */
17279 +static int au_fsctx_parse_monolithic(struct fs_context *fc, void *data)
17284 + struct au_fsctx_opts *a = fc->fs_private;
17287 + AuDbg("str %s\n", str);
17289 + u = is_colonopt(str);
17292 + str = strchr(str, ',');
17298 + AuDbg("str %s\n", str);
17300 + err = generic_parse_monolithic(fc, str);
17302 + au_fsctx_dump(&a->opts);
17307 +/* ---------------------------------------------------------------------- */
17309 +static void au_fsctx_opts_free(struct au_opts *opts)
17311 + struct au_opt *opt;
17314 + while (opt->type != Opt_tail) {
17315 + switch (opt->type) {
17320 + case Opt_prepend:
17321 + kfree(opt->add.pathname);
17322 + path_put(&opt->add.path);
17325 + kfree(opt->del.pathname);
17328 + path_put(&opt->del.h_path);
17331 + kfree(opt->mod.path);
17334 + dput(opt->mod.h_root);
17337 + kfree(opt->xino.path);
17338 + fput(opt->xino.file);
17345 +static void au_fsctx_free(struct fs_context *fc)
17347 + struct au_fsctx_opts *a = fc->fs_private;
17349 + /* fs_type=%p, root=%pD */
17350 + AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17351 + fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17353 + kobject_put(&a->sbinfo->si_kobj);
17354 + au_fsctx_opts_free(&a->opts);
17355 + free_page((unsigned long)a->opts.opt);
17359 +static const struct fs_context_operations au_fsctx_ops = {
17360 + .free = au_fsctx_free,
17361 + .parse_param = au_fsctx_parse_param,
17362 + .parse_monolithic = au_fsctx_parse_monolithic,
17363 + .get_tree = au_fsctx_get_tree,
17364 + .reconfigure = au_fsctx_reconfigure
17366 + * nfs4 requires ->dup()? No.
17367 + * I don't know what is this ->dup() for.
17371 +int aufs_fsctx_init(struct fs_context *fc)
17374 + struct au_fsctx_opts *a;
17376 + /* fs_type=%p, root=%pD */
17377 + AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17378 + fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17380 + /* they will be freed by au_fsctx_free() */
17382 + a = kzalloc(sizeof(*a), GFP_NOFS);
17383 + if (unlikely(!a))
17386 + a->opts.opt = (void *)__get_free_page(GFP_NOFS);
17387 + if (unlikely(!a->opts.opt))
17389 + a->opt = a->opts.opt;
17390 + a->opt->type = Opt_tail;
17391 + a->opts.max_opt = PAGE_SIZE / sizeof(*a->opts.opt);
17392 + a->opt_tail = a->opt + a->opts.max_opt - 1;
17393 + a->opts.sb_flags = fc->sb_flags;
17397 + AuDebugOn(fc->purpose != FS_CONTEXT_FOR_RECONFIGURE);
17398 + a->opts.flags = AuOpts_REMOUNT;
17399 + a->sb = fc->root->d_sb;
17400 + a->sbinfo = au_sbi(a->sb);
17401 + kobject_get(&a->sbinfo->si_kobj);
17403 + a->sbinfo = au_si_alloc(a->sb);
17404 + AuDebugOn(!a->sbinfo);
17405 + err = PTR_ERR(a->sbinfo);
17406 + if (IS_ERR(a->sbinfo))
17408 + au_rw_write_unlock(&a->sbinfo->si_rwsem);
17412 + fc->fs_private = a;
17413 + fc->ops = &au_fsctx_ops;
17414 + goto out; /* success */
17417 + free_page((unsigned long)a->opts.opt);
17424 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
17425 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
17426 +++ linux/fs/aufs/fstype.h 2022-03-21 14:49:05.723299677 +0100
17428 +/* SPDX-License-Identifier: GPL-2.0 */
17430 + * Copyright (C) 2005-2021 Junjiro R. Okajima
17432 + * This program, aufs is free software; you can redistribute it and/or modify
17433 + * it under the terms of the GNU General Public License as published by
17434 + * the Free Software Foundation; either version 2 of the License, or
17435 + * (at your option) any later version.
17437 + * This program is distributed in the hope that it will be useful,
17438 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17439 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17440 + * GNU General Public License for more details.
17442 + * You should have received a copy of the GNU General Public License
17443 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17447 + * judging filesystem type
17450 +#ifndef __AUFS_FSTYPE_H__
17451 +#define __AUFS_FSTYPE_H__
17455 +#include <linux/fs.h>
17456 +#include <linux/magic.h>
17457 +#include <linux/nfs_fs.h>
17458 +#include <linux/romfs_fs.h>
17460 +static inline int au_test_aufs(struct super_block *sb)
17462 + return sb->s_magic == AUFS_SUPER_MAGIC;
17465 +static inline const char *au_sbtype(struct super_block *sb)
17467 + return sb->s_type->name;
17470 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
17472 +#if IS_ENABLED(CONFIG_ISO9660_FS)
17473 + return sb->s_magic == ISOFS_SUPER_MAGIC;
17479 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
17481 +#if IS_ENABLED(CONFIG_ROMFS_FS)
17482 + return sb->s_magic == ROMFS_MAGIC;
17488 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
17490 +#if IS_ENABLED(CONFIG_CRAMFS)
17491 + return sb->s_magic == CRAMFS_MAGIC;
17496 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
17498 +#if IS_ENABLED(CONFIG_NFS_FS)
17499 + return sb->s_magic == NFS_SUPER_MAGIC;
17505 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
17507 +#if IS_ENABLED(CONFIG_FUSE_FS)
17508 + return sb->s_magic == FUSE_SUPER_MAGIC;
17514 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
17516 +#if IS_ENABLED(CONFIG_XFS_FS)
17517 + return sb->s_magic == XFS_SB_MAGIC;
17523 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
17525 +#ifdef CONFIG_TMPFS
17526 + return sb->s_magic == TMPFS_MAGIC;
17532 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
17534 +#if IS_ENABLED(CONFIG_ECRYPT_FS)
17535 + return !strcmp(au_sbtype(sb), "ecryptfs");
17541 +static inline int au_test_ramfs(struct super_block *sb)
17543 + return sb->s_magic == RAMFS_MAGIC;
17546 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
17548 +#if IS_ENABLED(CONFIG_UBIFS_FS)
17549 + return sb->s_magic == UBIFS_SUPER_MAGIC;
17555 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
17557 +#ifdef CONFIG_PROC_FS
17558 + return sb->s_magic == PROC_SUPER_MAGIC;
17564 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
17566 +#ifdef CONFIG_SYSFS
17567 + return sb->s_magic == SYSFS_MAGIC;
17573 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
17575 +#if IS_ENABLED(CONFIG_CONFIGFS_FS)
17576 + return sb->s_magic == CONFIGFS_MAGIC;
17582 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
17584 +#if IS_ENABLED(CONFIG_MINIX_FS)
17585 + return sb->s_magic == MINIX3_SUPER_MAGIC
17586 + || sb->s_magic == MINIX2_SUPER_MAGIC
17587 + || sb->s_magic == MINIX2_SUPER_MAGIC2
17588 + || sb->s_magic == MINIX_SUPER_MAGIC
17589 + || sb->s_magic == MINIX_SUPER_MAGIC2;
17595 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
17597 +#if IS_ENABLED(CONFIG_FAT_FS)
17598 + return sb->s_magic == MSDOS_SUPER_MAGIC;
17604 +static inline int au_test_msdos(struct super_block *sb)
17606 + return au_test_fat(sb);
17609 +static inline int au_test_vfat(struct super_block *sb)
17611 + return au_test_fat(sb);
17614 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
17616 +#ifdef CONFIG_SECURITYFS
17617 + return sb->s_magic == SECURITYFS_MAGIC;
17623 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
17625 +#if IS_ENABLED(CONFIG_SQUASHFS)
17626 + return sb->s_magic == SQUASHFS_MAGIC;
17632 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
17634 +#if IS_ENABLED(CONFIG_BTRFS_FS)
17635 + return sb->s_magic == BTRFS_SUPER_MAGIC;
17641 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
17643 +#if IS_ENABLED(CONFIG_XENFS)
17644 + return sb->s_magic == XENFS_SUPER_MAGIC;
17650 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
17652 +#ifdef CONFIG_DEBUG_FS
17653 + return sb->s_magic == DEBUGFS_MAGIC;
17659 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
17661 +#if IS_ENABLED(CONFIG_NILFS)
17662 + return sb->s_magic == NILFS_SUPER_MAGIC;
17668 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
17670 +#if IS_ENABLED(CONFIG_HFSPLUS_FS)
17671 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
17677 +/* ---------------------------------------------------------------------- */
17679 + * they can't be an aufs branch.
17681 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
17684 +#ifndef CONFIG_AUFS_BR_RAMFS
17685 + au_test_ramfs(sb) ||
17687 + au_test_procfs(sb)
17688 + || au_test_sysfs(sb)
17689 + || au_test_configfs(sb)
17690 + || au_test_debugfs(sb)
17691 + || au_test_securityfs(sb)
17692 + || au_test_xenfs(sb)
17693 + || au_test_ecryptfs(sb)
17694 + /* || !strcmp(au_sbtype(sb), "unionfs") */
17695 + || au_test_aufs(sb); /* will be supported in next version */
17698 +static inline int au_test_fs_remote(struct super_block *sb)
17700 + return !au_test_tmpfs(sb)
17701 +#ifdef CONFIG_AUFS_BR_RAMFS
17702 + && !au_test_ramfs(sb)
17704 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
17707 +/* ---------------------------------------------------------------------- */
17710 + * Note: these functions (below) are created after reading ->getattr() in all
17711 + * filesystems under linux/fs. it means we have to do so in every update...
17715 + * some filesystems require getattr to refresh the inode attributes before
17717 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
17718 + * and leave the work for d_revalidate()
17720 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
17722 + return au_test_nfs(sb)
17723 + || au_test_fuse(sb)
17724 + /* || au_test_btrfs(sb) */ /* untested */
17729 + * filesystems which don't maintain i_size or i_blocks.
17731 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
17733 + return au_test_xfs(sb)
17734 + || au_test_btrfs(sb)
17735 + || au_test_ubifs(sb)
17736 + || au_test_hfsplus(sb) /* maintained, but incorrect */
17737 + /* || au_test_minix(sb) */ /* untested */
17742 + * filesystems which don't store the correct value in some of their inode
17745 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
17747 + return au_test_fs_bad_iattr_size(sb)
17748 + || au_test_fat(sb)
17749 + || au_test_msdos(sb)
17750 + || au_test_vfat(sb);
17753 +/* they don't check i_nlink in link(2) */
17754 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
17756 + return au_test_tmpfs(sb)
17757 +#ifdef CONFIG_AUFS_BR_RAMFS
17758 + || au_test_ramfs(sb)
17760 + || au_test_ubifs(sb)
17761 + || au_test_hfsplus(sb);
17765 + * filesystems which sets S_NOATIME and S_NOCMTIME.
17767 +static inline int au_test_fs_notime(struct super_block *sb)
17769 + return au_test_nfs(sb)
17770 + || au_test_fuse(sb)
17771 + || au_test_ubifs(sb)
17775 +/* temporary support for i#1 in cramfs */
17776 +static inline int au_test_fs_unique_ino(struct inode *inode)
17778 + if (au_test_cramfs(inode->i_sb))
17779 + return inode->i_ino != 1;
17783 +/* ---------------------------------------------------------------------- */
17786 + * the filesystem where the xino files placed must support i/o after unlink and
17787 + * maintain i_size and i_blocks.
17789 +static inline int au_test_fs_bad_xino(struct super_block *sb)
17791 + return au_test_fs_remote(sb)
17792 + || au_test_fs_bad_iattr_size(sb)
17793 + /* don't want unnecessary work for xino */
17794 + || au_test_aufs(sb)
17795 + || au_test_ecryptfs(sb)
17796 + || au_test_nilfs(sb);
17799 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
17801 + return au_test_tmpfs(sb)
17802 + || au_test_ramfs(sb);
17806 + * test if the @sb is real-readonly.
17808 +static inline int au_test_fs_rr(struct super_block *sb)
17810 + return au_test_squashfs(sb)
17811 + || au_test_iso9660(sb)
17812 + || au_test_cramfs(sb)
17813 + || au_test_romfs(sb);
17817 + * test if the @inode is nfs with 'noacl' option
17818 + * NFS always sets SB_POSIXACL regardless its mount option 'noacl.'
17820 +static inline int au_test_nfs_noacl(struct inode *inode)
17822 + return au_test_nfs(inode->i_sb)
17823 + /* && IS_POSIXACL(inode) */
17824 + && !nfs_server_capable(inode, NFS_CAP_ACLS);
17827 +#endif /* __KERNEL__ */
17828 +#endif /* __AUFS_FSTYPE_H__ */
17829 diff -urN /usr/share/empty/fs/aufs/hbl.h linux/fs/aufs/hbl.h
17830 --- /usr/share/empty/fs/aufs/hbl.h 1970-01-01 01:00:00.000000000 +0100
17831 +++ linux/fs/aufs/hbl.h 2022-03-21 14:49:05.723299677 +0100
17833 +/* SPDX-License-Identifier: GPL-2.0 */
17835 + * Copyright (C) 2017-2021 Junjiro R. Okajima
17837 + * This program, aufs is free software; you can redistribute it and/or modify
17838 + * it under the terms of the GNU General Public License as published by
17839 + * the Free Software Foundation; either version 2 of the License, or
17840 + * (at your option) any later version.
17842 + * This program is distributed in the hope that it will be useful,
17843 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17844 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17845 + * GNU General Public License for more details.
17847 + * You should have received a copy of the GNU General Public License
17848 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17852 + * helpers for hlist_bl.h
17855 +#ifndef __AUFS_HBL_H__
17856 +#define __AUFS_HBL_H__
17860 +#include <linux/list_bl.h>
17862 +static inline void au_hbl_add(struct hlist_bl_node *node,
17863 + struct hlist_bl_head *hbl)
17865 + hlist_bl_lock(hbl);
17866 + hlist_bl_add_head(node, hbl);
17867 + hlist_bl_unlock(hbl);
17870 +static inline void au_hbl_del(struct hlist_bl_node *node,
17871 + struct hlist_bl_head *hbl)
17873 + hlist_bl_lock(hbl);
17874 + hlist_bl_del(node);
17875 + hlist_bl_unlock(hbl);
17878 +#define au_hbl_for_each(pos, head) \
17879 + for (pos = hlist_bl_first(head); \
17883 +static inline unsigned long au_hbl_count(struct hlist_bl_head *hbl)
17885 + unsigned long cnt;
17886 + struct hlist_bl_node *pos;
17889 + hlist_bl_lock(hbl);
17890 + au_hbl_for_each(pos, hbl)
17892 + hlist_bl_unlock(hbl);
17896 +#endif /* __KERNEL__ */
17897 +#endif /* __AUFS_HBL_H__ */
17898 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
17899 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
17900 +++ linux/fs/aufs/hfsnotify.c 2022-10-03 07:51:14.926209260 +0200
17902 +// SPDX-License-Identifier: GPL-2.0
17904 + * Copyright (C) 2005-2021 Junjiro R. Okajima
17906 + * This program, aufs is free software; you can redistribute it and/or modify
17907 + * it under the terms of the GNU General Public License as published by
17908 + * the Free Software Foundation; either version 2 of the License, or
17909 + * (at your option) any later version.
17911 + * This program is distributed in the hope that it will be useful,
17912 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17913 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17914 + * GNU General Public License for more details.
17916 + * You should have received a copy of the GNU General Public License
17917 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17921 + * fsnotify for the lower directories
17926 +/* FS_IN_IGNORED is unnecessary */
17927 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
17928 + | FS_CREATE | FS_EVENT_ON_CHILD);
17929 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
17930 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
17932 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
17934 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
17936 + /* AuDbg("here\n"); */
17937 + au_cache_free_hnotify(hn);
17938 + smp_mb__before_atomic(); /* for atomic64_dec */
17939 + if (atomic64_dec_and_test(&au_hfsn_ifree))
17940 + wake_up(&au_hfsn_wq);
17943 +static int au_hfsn_alloc(struct au_hinode *hinode)
17946 + struct au_hnotify *hn;
17947 + struct super_block *sb;
17948 + struct au_branch *br;
17949 + struct fsnotify_mark *mark;
17950 + aufs_bindex_t bindex;
17952 + hn = hinode->hi_notify;
17953 + sb = hn->hn_aufs_inode->i_sb;
17954 + bindex = au_br_index(sb, hinode->hi_id);
17955 + br = au_sbr(sb, bindex);
17956 + AuDebugOn(!br->br_hfsn);
17958 + mark = &hn->hn_mark;
17959 + fsnotify_init_mark(mark, br->br_hfsn->hfsn_group);
17960 + mark->mask = AuHfsnMask;
17962 + * by udba rename or rmdir, aufs assign a new inode to the known
17963 + * h_inode, so specify 1 to allow dups.
17966 + err = fsnotify_add_inode_mark(mark, hinode->hi_inode, /*allow_dups*/1);
17972 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
17974 + struct fsnotify_mark *mark;
17975 + unsigned long long ull;
17976 + struct fsnotify_group *group;
17978 + ull = atomic64_inc_return(&au_hfsn_ifree);
17981 + mark = &hn->hn_mark;
17982 + spin_lock(&mark->lock);
17983 + group = mark->group;
17984 + fsnotify_get_group(group);
17985 + spin_unlock(&mark->lock);
17987 + fsnotify_destroy_mark(mark, group);
17988 + fsnotify_put_mark(mark);
17989 + fsnotify_put_group(group);
17992 + /* free hn by myself */
17996 +/* ---------------------------------------------------------------------- */
17998 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
18000 + struct fsnotify_mark *mark;
18002 + mark = &hinode->hi_notify->hn_mark;
18003 + spin_lock(&mark->lock);
18005 + AuDebugOn(mark->mask & AuHfsnMask);
18006 + mark->mask |= AuHfsnMask;
18008 + AuDebugOn(!(mark->mask & AuHfsnMask));
18009 + mark->mask &= ~AuHfsnMask;
18011 + spin_unlock(&mark->lock);
18012 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
18015 +/* ---------------------------------------------------------------------- */
18017 +/* #define AuDbgHnotify */
18018 +#ifdef AuDbgHnotify
18019 +static char *au_hfsn_name(u32 mask)
18021 +#ifdef CONFIG_AUFS_DEBUG
18022 +#define test_ret(flag) \
18024 + if (mask & flag) \
18027 + test_ret(FS_ACCESS);
18028 + test_ret(FS_MODIFY);
18029 + test_ret(FS_ATTRIB);
18030 + test_ret(FS_CLOSE_WRITE);
18031 + test_ret(FS_CLOSE_NOWRITE);
18032 + test_ret(FS_OPEN);
18033 + test_ret(FS_MOVED_FROM);
18034 + test_ret(FS_MOVED_TO);
18035 + test_ret(FS_CREATE);
18036 + test_ret(FS_DELETE);
18037 + test_ret(FS_DELETE_SELF);
18038 + test_ret(FS_MOVE_SELF);
18039 + test_ret(FS_UNMOUNT);
18040 + test_ret(FS_Q_OVERFLOW);
18041 + test_ret(FS_IN_IGNORED);
18042 + test_ret(FS_ISDIR);
18043 + test_ret(FS_IN_ONESHOT);
18044 + test_ret(FS_EVENT_ON_CHILD);
18053 +/* ---------------------------------------------------------------------- */
18055 +static void au_hfsn_free_group(struct fsnotify_group *group)
18057 + struct au_br_hfsnotify *hfsn = group->private;
18059 + /* AuDbg("here\n"); */
18060 + au_kfree_try_rcu(hfsn);
18063 +static int au_hfsn_handle_event(struct fsnotify_group *group,
18064 + u32 mask, const void *data, int data_type,
18065 + struct inode *dir,
18066 + const struct qstr *file_name, u32 cookie,
18067 + struct fsnotify_iter_info *iter_info)
18070 + struct au_hnotify *hnotify;
18071 + struct inode *h_dir, *h_inode;
18072 + struct fsnotify_mark *inode_mark;
18074 + AuDebugOn(!(data_type == FSNOTIFY_EVENT_INODE
18075 + || data_type == FSNOTIFY_EVENT_DENTRY));
18078 + /* if FS_UNMOUNT happens, there must be another bug */
18079 + AuDebugOn(mask & FS_UNMOUNT);
18080 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
18085 +#ifdef AuDbgHnotify
18087 + if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
18088 + || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
18089 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
18090 + h_dir->i_ino, mask, au_hfsn_name(mask),
18091 + AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
18092 + /* WARN_ON(1); */
18097 + inode_mark = fsnotify_iter_inode_mark(iter_info);
18098 + AuDebugOn(!inode_mark);
18099 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
18100 + err = au_hnotify(h_dir, hnotify, mask, file_name, h_inode);
18106 +static struct fsnotify_ops au_hfsn_ops = {
18107 + .handle_event = au_hfsn_handle_event,
18108 + .free_group_priv = au_hfsn_free_group,
18109 + .free_mark = au_hfsn_free_mark
18112 +/* ---------------------------------------------------------------------- */
18114 +static void au_hfsn_fin_br(struct au_branch *br)
18116 + struct au_br_hfsnotify *hfsn;
18118 + hfsn = br->br_hfsn;
18121 + fsnotify_put_group(hfsn->hfsn_group);
18126 +static int au_hfsn_init_br(struct au_branch *br, int perm)
18129 + struct fsnotify_group *group;
18130 + struct au_br_hfsnotify *hfsn;
18133 + br->br_hfsn = NULL;
18134 + if (!au_br_hnotifyable(perm))
18138 + hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
18139 + if (unlikely(!hfsn))
18143 + group = fsnotify_alloc_group(&au_hfsn_ops,
18144 + /*flags - not for userspace*/0);
18145 + if (IS_ERR(group)) {
18146 + err = PTR_ERR(group);
18147 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
18151 + group->private = hfsn;
18152 + hfsn->hfsn_group = group;
18153 + br->br_hfsn = hfsn;
18154 + goto out; /* success */
18157 + au_kfree_try_rcu(hfsn);
18162 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
18167 + if (!br->br_hfsn)
18168 + err = au_hfsn_init_br(br, perm);
18173 +/* ---------------------------------------------------------------------- */
18175 +static void au_hfsn_fin(void)
18177 + AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
18178 + wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
18181 +const struct au_hnotify_op au_hnotify_op = {
18182 + .ctl = au_hfsn_ctl,
18183 + .alloc = au_hfsn_alloc,
18184 + .free = au_hfsn_free,
18186 + .fin = au_hfsn_fin,
18188 + .reset_br = au_hfsn_reset_br,
18189 + .fin_br = au_hfsn_fin_br,
18190 + .init_br = au_hfsn_init_br
18192 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
18193 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
18194 +++ linux/fs/aufs/hfsplus.c 2022-03-21 14:49:05.723299677 +0100
18196 +// SPDX-License-Identifier: GPL-2.0
18198 + * Copyright (C) 2010-2021 Junjiro R. Okajima
18200 + * This program, aufs is free software; you can redistribute it and/or modify
18201 + * it under the terms of the GNU General Public License as published by
18202 + * the Free Software Foundation; either version 2 of the License, or
18203 + * (at your option) any later version.
18205 + * This program is distributed in the hope that it will be useful,
18206 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18207 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18208 + * GNU General Public License for more details.
18210 + * You should have received a copy of the GNU General Public License
18211 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18215 + * special support for filesystems which acquires an inode mutex
18216 + * at final closing a file, eg, hfsplus.
18218 + * This trick is very simple and stupid, just to open the file before really
18219 + * necessary open to tell hfsplus that this is not the final closing.
18220 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
18221 + * and au_h_open_post() after releasing it.
18226 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
18229 + struct file *h_file;
18230 + struct dentry *h_dentry;
18232 + h_dentry = au_h_dptr(dentry, bindex);
18233 + AuDebugOn(!h_dentry);
18234 + AuDebugOn(d_is_negative(h_dentry));
18237 + if (au_test_hfsplus(h_dentry->d_sb)
18238 + && d_is_reg(h_dentry))
18239 + h_file = au_h_open(dentry, bindex,
18240 + O_RDONLY | O_NOATIME | O_LARGEFILE,
18241 + /*file*/NULL, force_wr);
18245 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
18246 + struct file *h_file)
18248 + struct au_branch *br;
18252 + br = au_sbr(dentry->d_sb, bindex);
18253 + au_lcnt_dec(&br->br_nfiles);
18256 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
18257 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
18258 +++ linux/fs/aufs/hnotify.c 2022-03-21 14:49:05.723299677 +0100
18260 +// SPDX-License-Identifier: GPL-2.0
18262 + * Copyright (C) 2005-2021 Junjiro R. Okajima
18264 + * This program, aufs is free software; you can redistribute it and/or modify
18265 + * it under the terms of the GNU General Public License as published by
18266 + * the Free Software Foundation; either version 2 of the License, or
18267 + * (at your option) any later version.
18269 + * This program is distributed in the hope that it will be useful,
18270 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18271 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18272 + * GNU General Public License for more details.
18274 + * You should have received a copy of the GNU General Public License
18275 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18279 + * abstraction to notify the direct changes on lower directories
18282 +/* #include <linux/iversion.h> */
18285 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
18288 + struct au_hnotify *hn;
18291 + hn = au_cache_alloc_hnotify();
18293 + hn->hn_aufs_inode = inode;
18294 + hinode->hi_notify = hn;
18295 + err = au_hnotify_op.alloc(hinode);
18297 + if (unlikely(err)) {
18298 + hinode->hi_notify = NULL;
18299 + au_cache_free_hnotify(hn);
18301 + * The upper dir was removed by udba, but the same named
18302 + * dir left. In this case, aufs assigns a new inode
18303 + * number and set the monitor again.
18304 + * For the lower dir, the old monitor is still left.
18306 + if (err == -EEXIST)
18315 +void au_hn_free(struct au_hinode *hinode)
18317 + struct au_hnotify *hn;
18319 + hn = hinode->hi_notify;
18321 + hinode->hi_notify = NULL;
18322 + if (au_hnotify_op.free(hinode, hn))
18323 + au_cache_free_hnotify(hn);
18327 +/* ---------------------------------------------------------------------- */
18329 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
18331 + if (hinode->hi_notify)
18332 + au_hnotify_op.ctl(hinode, do_set);
18335 +void au_hn_reset(struct inode *inode, unsigned int flags)
18337 + aufs_bindex_t bindex, bbot;
18338 + struct inode *hi;
18339 + struct dentry *iwhdentry;
18341 + bbot = au_ibbot(inode);
18342 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
18343 + hi = au_h_iptr(inode, bindex);
18347 + /* inode_lock_nested(hi, AuLsc_I_CHILD); */
18348 + iwhdentry = au_hi_wh(inode, bindex);
18352 + au_set_h_iptr(inode, bindex, NULL, 0);
18353 + au_set_h_iptr(inode, bindex, au_igrab(hi),
18354 + flags & ~AuHi_XINO);
18357 + /* inode_unlock(hi); */
18361 +/* ---------------------------------------------------------------------- */
18363 +static int hn_xino(struct inode *inode, struct inode *h_inode)
18366 + aufs_bindex_t bindex, bbot, bfound, btop;
18367 + struct inode *h_i;
18370 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18371 + pr_warn("branch root dir was changed\n");
18376 + bbot = au_ibbot(inode);
18377 + btop = au_ibtop(inode);
18378 +#if 0 /* reserved for future use */
18379 + if (bindex == bbot) {
18380 + /* keep this ino in rename case */
18384 + for (bindex = btop; bindex <= bbot; bindex++)
18385 + if (au_h_iptr(inode, bindex) == h_inode) {
18392 + for (bindex = btop; bindex <= bbot; bindex++) {
18393 + h_i = au_h_iptr(inode, bindex);
18397 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
18398 + /* ignore this error */
18399 + /* bad action? */
18402 + /* children inode number will be broken */
18409 +static int hn_gen_tree(struct dentry *dentry)
18411 + int err, i, j, ndentry;
18412 + struct au_dcsub_pages dpages;
18413 + struct au_dpage *dpage;
18414 + struct dentry **dentries;
18416 + err = au_dpages_init(&dpages, GFP_NOFS);
18417 + if (unlikely(err))
18419 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
18420 + if (unlikely(err))
18423 + for (i = 0; i < dpages.ndpage; i++) {
18424 + dpage = dpages.dpages + i;
18425 + dentries = dpage->dentries;
18426 + ndentry = dpage->ndentry;
18427 + for (j = 0; j < ndentry; j++) {
18428 + struct dentry *d;
18435 + if (d_really_is_positive(d))
18436 + /* todo: reset children xino?
18437 + cached children only? */
18438 + au_iigen_dec(d_inode(d));
18443 + au_dpages_free(&dpages);
18449 + * return 0 if processed.
18451 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
18452 + const unsigned int isdir)
18455 + struct dentry *d;
18456 + struct qstr *dname;
18459 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18460 + pr_warn("branch root dir was changed\n");
18466 + AuDebugOn(!name);
18467 + au_iigen_dec(inode);
18468 + spin_lock(&inode->i_lock);
18469 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
18470 + spin_lock(&d->d_lock);
18471 + dname = &d->d_name;
18472 + if (dname->len != nlen
18473 + && memcmp(dname->name, name, nlen)) {
18474 + spin_unlock(&d->d_lock);
18479 + spin_unlock(&d->d_lock);
18482 + spin_unlock(&inode->i_lock);
18484 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
18485 + d = d_find_any_alias(inode);
18487 + au_iigen_dec(inode);
18491 + spin_lock(&d->d_lock);
18492 + dname = &d->d_name;
18493 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
18494 + spin_unlock(&d->d_lock);
18495 + err = hn_gen_tree(d);
18496 + spin_lock(&d->d_lock);
18498 + spin_unlock(&d->d_lock);
18507 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
18511 + if (IS_ROOT(dentry)) {
18512 + pr_warn("branch root dir was changed\n");
18518 + au_digen_dec(dentry);
18519 + if (d_really_is_positive(dentry))
18520 + au_iigen_dec(d_inode(dentry));
18522 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
18523 + if (d_really_is_positive(dentry))
18524 + err = hn_gen_tree(dentry);
18531 +/* ---------------------------------------------------------------------- */
18533 +/* hnotify job flags */
18534 +#define AuHnJob_XINO0 1
18535 +#define AuHnJob_GEN (1 << 1)
18536 +#define AuHnJob_DIRENT (1 << 2)
18537 +#define AuHnJob_ISDIR (1 << 3)
18538 +#define AuHnJob_TRYXINO0 (1 << 4)
18539 +#define AuHnJob_MNTPNT (1 << 5)
18540 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
18541 +#define au_fset_hnjob(flags, name) \
18542 + do { (flags) |= AuHnJob_##name; } while (0)
18543 +#define au_fclr_hnjob(flags, name) \
18544 + do { (flags) &= ~AuHnJob_##name; } while (0)
18552 +struct au_hnotify_args {
18553 + struct inode *h_dir, *dir, *h_child_inode;
18555 + unsigned int flags[AuHnLast];
18556 + unsigned int h_child_nlen;
18557 + char h_child_name[];
18560 +struct hn_job_args {
18561 + unsigned int flags;
18562 + struct inode *inode, *h_inode, *dir, *h_dir;
18563 + struct dentry *dentry;
18568 +static int hn_job(struct hn_job_args *a)
18570 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
18574 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
18575 + hn_xino(a->inode, a->h_inode); /* ignore this error */
18577 + if (au_ftest_hnjob(a->flags, TRYXINO0)
18580 + inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
18581 + if (!a->h_inode->i_nlink
18582 + && !(a->h_inode->i_state & I_LINKABLE))
18583 + hn_xino(a->inode, a->h_inode); /* ignore this error */
18584 + inode_unlock_shared(a->h_inode);
18587 + /* make the generation obsolete */
18588 + if (au_ftest_hnjob(a->flags, GEN)) {
18591 + e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
18593 + if (e && a->dentry)
18594 + hn_gen_by_name(a->dentry, isdir);
18595 + /* ignore this error */
18598 + /* make dir entries obsolete */
18599 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
18600 + struct au_vdir *vdir;
18602 + vdir = au_ivdir(a->inode);
18604 + vdir->vd_jiffy = 0;
18605 + /* IMustLock(a->inode); */
18606 + /* inode_inc_iversion(a->inode); */
18609 + /* can do nothing but warn */
18610 + if (au_ftest_hnjob(a->flags, MNTPNT)
18612 + && d_mountpoint(a->dentry))
18613 + pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
18618 +/* ---------------------------------------------------------------------- */
18620 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
18621 + struct inode *dir)
18623 + struct dentry *dentry, *d, *parent;
18624 + struct qstr *dname;
18626 + parent = d_find_any_alias(dir);
18631 + spin_lock(&parent->d_lock);
18632 + list_for_each_entry(d, &parent->d_subdirs, d_child) {
18633 + /* AuDbg("%pd\n", d); */
18634 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
18635 + dname = &d->d_name;
18636 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
18637 + goto cont_unlock;
18641 + goto cont_unlock;
18642 + if (au_dcount(d) > 0) {
18643 + dentry = dget_dlock(d);
18644 + spin_unlock(&d->d_lock);
18649 + spin_unlock(&d->d_lock);
18651 + spin_unlock(&parent->d_lock);
18655 + di_write_lock_child(dentry);
18660 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
18661 + aufs_bindex_t bindex, ino_t h_ino)
18663 + struct inode *inode;
18668 + err = au_xino_read(sb, bindex, h_ino, &ino);
18670 + inode = ilookup(sb, ino);
18674 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18675 + pr_warn("wrong root branch\n");
18681 + ii_write_lock_child(inode);
18687 +static void au_hn_bh(void *_args)
18689 + struct au_hnotify_args *a = _args;
18690 + struct super_block *sb;
18691 + aufs_bindex_t bindex, bbot, bfound;
18692 + unsigned char xino, try_iput;
18694 + struct inode *inode;
18696 + struct hn_job_args args;
18697 + struct dentry *dentry;
18698 + struct au_sbinfo *sbinfo;
18700 + AuDebugOn(!_args);
18701 + AuDebugOn(!a->h_dir);
18702 + AuDebugOn(!a->dir);
18703 + AuDebugOn(!a->mask);
18704 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
18705 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
18706 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
18711 + * do not lock a->dir->i_mutex here
18712 + * because of d_revalidate() may cause a deadlock.
18714 + sb = a->dir->i_sb;
18716 + sbinfo = au_sbi(sb);
18717 + AuDebugOn(!sbinfo);
18718 + si_write_lock(sb, AuLock_NOPLMW);
18720 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
18721 + switch (a->mask & FS_EVENTS_POSS_ON_CHILD) {
18722 + case FS_MOVED_FROM:
18723 + case FS_MOVED_TO:
18724 + AuWarn1("DIRREN with UDBA may not work correctly "
18725 + "for the direct rename(2)\n");
18728 + ii_read_lock_parent(a->dir);
18730 + bbot = au_ibbot(a->dir);
18731 + for (bindex = au_ibtop(a->dir); bindex <= bbot; bindex++)
18732 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
18736 + ii_read_unlock(a->dir);
18737 + if (unlikely(bfound < 0))
18740 + xino = !!au_opt_test(au_mntflags(sb), XINO);
18742 + if (a->h_child_inode)
18743 + h_ino = a->h_child_inode->i_ino;
18745 + if (a->h_child_nlen
18746 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
18747 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
18748 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
18751 + if (dentry && d_really_is_positive(dentry))
18752 + inode = d_inode(dentry);
18753 + if (xino && !inode && h_ino
18754 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
18755 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
18756 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
18757 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
18761 + args.flags = a->flags[AuHn_CHILD];
18762 + args.dentry = dentry;
18763 + args.inode = inode;
18764 + args.h_inode = a->h_child_inode;
18765 + args.dir = a->dir;
18766 + args.h_dir = a->h_dir;
18767 + args.h_name = a->h_child_name;
18768 + args.h_nlen = a->h_child_nlen;
18769 + err = hn_job(&args);
18771 + if (au_di(dentry))
18772 + di_write_unlock(dentry);
18775 + if (inode && try_iput) {
18776 + ii_write_unlock(inode);
18780 + ii_write_lock_parent(a->dir);
18781 + args.flags = a->flags[AuHn_PARENT];
18782 + args.dentry = NULL;
18783 + args.inode = a->dir;
18784 + args.h_inode = a->h_dir;
18786 + args.h_dir = NULL;
18787 + args.h_name = NULL;
18789 + err = hn_job(&args);
18790 + ii_write_unlock(a->dir);
18793 + iput(a->h_child_inode);
18796 + si_write_unlock(sb);
18797 + au_nwt_done(&sbinfo->si_nowait);
18801 +/* ---------------------------------------------------------------------- */
18803 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
18804 + const struct qstr *h_child_qstr, struct inode *h_child_inode)
18807 + unsigned int flags[AuHnLast], f;
18808 + unsigned char isdir, isroot, wh;
18809 + struct inode *dir;
18810 + struct au_hnotify_args *args;
18811 + char *p, *h_child_name;
18814 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
18815 + dir = igrab(hnotify->hn_aufs_inode);
18819 + isroot = (dir->i_ino == AUFS_ROOT_INO);
18821 + h_child_name = (void *)h_child_qstr->name;
18822 + len = h_child_qstr->len;
18823 + if (h_child_name) {
18824 + if (len > AUFS_WH_PFX_LEN
18825 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
18826 + h_child_name += AUFS_WH_PFX_LEN;
18827 + len -= AUFS_WH_PFX_LEN;
18833 + if (h_child_inode)
18834 + isdir = !!S_ISDIR(h_child_inode->i_mode);
18835 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
18836 + flags[AuHn_CHILD] = 0;
18838 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
18839 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
18840 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
18841 + switch (mask & ALL_FSNOTIFY_DIRENT_EVENTS) {
18842 + case FS_MOVED_FROM:
18843 + case FS_MOVED_TO:
18844 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
18845 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18848 + AuDebugOn(!h_child_name);
18853 + * aufs never be able to get this child inode.
18854 + * revalidation should be in d_revalidate()
18855 + * by checking i_nlink, i_generation or d_unhashed().
18857 + AuDebugOn(!h_child_name);
18858 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
18859 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18867 + h_child_inode = NULL;
18870 + /* iput() and kfree() will be called in au_hnotify() */
18871 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
18872 + if (unlikely(!args)) {
18873 + AuErr1("no memory\n");
18877 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
18878 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
18879 + args->mask = mask;
18881 + args->h_dir = igrab(h_dir);
18882 + if (h_child_inode)
18883 + h_child_inode = igrab(h_child_inode); /* can be NULL */
18884 + args->h_child_inode = h_child_inode;
18885 + args->h_child_nlen = len;
18887 + p = (void *)args;
18888 + p += sizeof(*args);
18889 + memcpy(p, h_child_name, len);
18893 + /* NFS fires the event for silly-renamed one from kworker */
18895 + if (!dir->i_nlink
18896 + || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
18898 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
18899 + if (unlikely(err)) {
18900 + pr_err("wkq %d\n", err);
18901 + iput(args->h_child_inode);
18902 + iput(args->h_dir);
18904 + au_kfree_rcu(args);
18911 +/* ---------------------------------------------------------------------- */
18913 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
18917 + AuDebugOn(!(udba & AuOptMask_UDBA));
18920 + if (au_hnotify_op.reset_br)
18921 + err = au_hnotify_op.reset_br(udba, br, perm);
18926 +int au_hnotify_init_br(struct au_branch *br, int perm)
18931 + if (au_hnotify_op.init_br)
18932 + err = au_hnotify_op.init_br(br, perm);
18937 +void au_hnotify_fin_br(struct au_branch *br)
18939 + if (au_hnotify_op.fin_br)
18940 + au_hnotify_op.fin_br(br);
18943 +static void au_hn_destroy_cache(void)
18945 + kmem_cache_destroy(au_cache[AuCache_HNOTIFY]);
18946 + au_cache[AuCache_HNOTIFY] = NULL;
18949 +int __init au_hnotify_init(void)
18954 + au_cache[AuCache_HNOTIFY] = AuCache(au_hnotify);
18955 + if (au_cache[AuCache_HNOTIFY]) {
18957 + if (au_hnotify_op.init)
18958 + err = au_hnotify_op.init();
18959 + if (unlikely(err))
18960 + au_hn_destroy_cache();
18966 +void au_hnotify_fin(void)
18968 + if (au_hnotify_op.fin)
18969 + au_hnotify_op.fin();
18971 + /* cf. au_cache_fin() */
18972 + if (au_cache[AuCache_HNOTIFY])
18973 + au_hn_destroy_cache();
18975 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
18976 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
18977 +++ linux/fs/aufs/iinfo.c 2022-03-21 14:49:05.726633010 +0100
18979 +// SPDX-License-Identifier: GPL-2.0
18981 + * Copyright (C) 2005-2021 Junjiro R. Okajima
18983 + * This program, aufs is free software; you can redistribute it and/or modify
18984 + * it under the terms of the GNU General Public License as published by
18985 + * the Free Software Foundation; either version 2 of the License, or
18986 + * (at your option) any later version.
18988 + * This program is distributed in the hope that it will be useful,
18989 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18990 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18991 + * GNU General Public License for more details.
18993 + * You should have received a copy of the GNU General Public License
18994 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18998 + * inode private data
19003 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
19005 + struct inode *h_inode;
19006 + struct au_hinode *hinode;
19008 + IiMustAnyLock(inode);
19010 + hinode = au_hinode(au_ii(inode), bindex);
19011 + h_inode = hinode->hi_inode;
19012 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19016 +/* todo: hard/soft set? */
19017 +void au_hiput(struct au_hinode *hinode)
19019 + au_hn_free(hinode);
19020 + dput(hinode->hi_whdentry);
19021 + iput(hinode->hi_inode);
19024 +unsigned int au_hi_flags(struct inode *inode, int isdir)
19026 + unsigned int flags;
19027 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
19030 + if (au_opt_test(mnt_flags, XINO))
19031 + au_fset_hi(flags, XINO);
19032 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
19033 + au_fset_hi(flags, HNOTIFY);
19037 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
19038 + struct inode *h_inode, unsigned int flags)
19040 + struct au_hinode *hinode;
19041 + struct inode *hi;
19042 + struct au_iinfo *iinfo = au_ii(inode);
19044 + IiMustWriteLock(inode);
19046 + hinode = au_hinode(iinfo, bindex);
19047 + hi = hinode->hi_inode;
19048 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19051 + au_hiput(hinode);
19052 + hinode->hi_inode = h_inode;
19055 + struct super_block *sb = inode->i_sb;
19056 + struct au_branch *br;
19058 + AuDebugOn(inode->i_mode
19059 + && (h_inode->i_mode & S_IFMT)
19060 + != (inode->i_mode & S_IFMT));
19061 + if (bindex == iinfo->ii_btop)
19062 + au_cpup_igen(inode, h_inode);
19063 + br = au_sbr(sb, bindex);
19064 + hinode->hi_id = br->br_id;
19065 + if (au_ftest_hi(flags, XINO)) {
19066 + err = au_xino_write(sb, bindex, h_inode->i_ino,
19068 + if (unlikely(err))
19069 + AuIOErr1("failed au_xino_write() %d\n", err);
19072 + if (au_ftest_hi(flags, HNOTIFY)
19073 + && au_br_hnotifyable(br->br_perm)) {
19074 + err = au_hn_alloc(hinode, inode);
19075 + if (unlikely(err))
19076 + AuIOErr1("au_hn_alloc() %d\n", err);
19081 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
19082 + struct dentry *h_wh)
19084 + struct au_hinode *hinode;
19086 + IiMustWriteLock(inode);
19088 + hinode = au_hinode(au_ii(inode), bindex);
19089 + AuDebugOn(hinode->hi_whdentry);
19090 + hinode->hi_whdentry = h_wh;
19093 +void au_update_iigen(struct inode *inode, int half)
19095 + struct au_iinfo *iinfo;
19096 + struct au_iigen *iigen;
19097 + unsigned int sigen;
19099 + sigen = au_sigen(inode->i_sb);
19100 + iinfo = au_ii(inode);
19101 + iigen = &iinfo->ii_generation;
19102 + spin_lock(&iigen->ig_spin);
19103 + iigen->ig_generation = sigen;
19105 + au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
19107 + au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
19108 + spin_unlock(&iigen->ig_spin);
19111 +/* it may be called at remount time, too */
19112 +void au_update_ibrange(struct inode *inode, int do_put_zero)
19114 + struct au_iinfo *iinfo;
19115 + aufs_bindex_t bindex, bbot;
19117 + AuDebugOn(au_is_bad_inode(inode));
19118 + IiMustWriteLock(inode);
19120 + iinfo = au_ii(inode);
19121 + if (do_put_zero && iinfo->ii_btop >= 0) {
19122 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19124 + struct inode *h_i;
19126 + h_i = au_hinode(iinfo, bindex)->hi_inode;
19129 + && !(h_i->i_state & I_LINKABLE))
19130 + au_set_h_iptr(inode, bindex, NULL, 0);
19134 + iinfo->ii_btop = -1;
19135 + iinfo->ii_bbot = -1;
19136 + bbot = au_sbbot(inode->i_sb);
19137 + for (bindex = 0; bindex <= bbot; bindex++)
19138 + if (au_hinode(iinfo, bindex)->hi_inode) {
19139 + iinfo->ii_btop = bindex;
19142 + if (iinfo->ii_btop >= 0)
19143 + for (bindex = bbot; bindex >= iinfo->ii_btop; bindex--)
19144 + if (au_hinode(iinfo, bindex)->hi_inode) {
19145 + iinfo->ii_bbot = bindex;
19148 + AuDebugOn(iinfo->ii_btop > iinfo->ii_bbot);
19151 +/* ---------------------------------------------------------------------- */
19153 +void au_icntnr_init_once(void *_c)
19155 + struct au_icntnr *c = _c;
19156 + struct au_iinfo *iinfo = &c->iinfo;
19158 + spin_lock_init(&iinfo->ii_generation.ig_spin);
19159 + au_rw_init(&iinfo->ii_rwsem);
19160 + inode_init_once(&c->vfs_inode);
19163 +void au_hinode_init(struct au_hinode *hinode)
19165 + hinode->hi_inode = NULL;
19166 + hinode->hi_id = -1;
19167 + au_hn_init(hinode);
19168 + hinode->hi_whdentry = NULL;
19171 +int au_iinfo_init(struct inode *inode)
19173 + struct au_iinfo *iinfo;
19174 + struct super_block *sb;
19175 + struct au_hinode *hi;
19178 + sb = inode->i_sb;
19179 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19180 + nbr = au_sbbot(sb) + 1;
19181 + if (unlikely(nbr <= 0))
19183 + hi = kmalloc_array(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
19185 + au_lcnt_inc(&au_sbi(sb)->si_ninodes);
19187 + iinfo->ii_hinode = hi;
19188 + for (i = 0; i < nbr; i++, hi++)
19189 + au_hinode_init(hi);
19191 + iinfo->ii_generation.ig_generation = au_sigen(sb);
19192 + iinfo->ii_btop = -1;
19193 + iinfo->ii_bbot = -1;
19194 + iinfo->ii_vdir = NULL;
19200 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink)
19203 + struct au_hinode *hip;
19205 + AuRwMustWriteLock(&iinfo->ii_rwsem);
19208 + hip = au_krealloc(iinfo->ii_hinode, sizeof(*hip) * nbr, GFP_NOFS,
19211 + iinfo->ii_hinode = hip;
19212 + i = iinfo->ii_bbot + 1;
19214 + for (; i < nbr; i++, hip++)
19215 + au_hinode_init(hip);
19222 +void au_iinfo_fin(struct inode *inode)
19224 + struct au_iinfo *iinfo;
19225 + struct au_hinode *hi;
19226 + struct super_block *sb;
19227 + aufs_bindex_t bindex, bbot;
19228 + const unsigned char unlinked = !inode->i_nlink;
19230 + AuDebugOn(au_is_bad_inode(inode));
19232 + sb = inode->i_sb;
19233 + au_lcnt_dec(&au_sbi(sb)->si_ninodes);
19234 + if (si_pid_test(sb))
19235 + au_xino_delete_inode(inode, unlinked);
19238 + * it is safe to hide the dependency between sbinfo and
19242 + si_noflush_read_lock(sb);
19243 + au_xino_delete_inode(inode, unlinked);
19244 + si_read_unlock(sb);
19248 + iinfo = au_ii(inode);
19249 + if (iinfo->ii_vdir)
19250 + au_vdir_free(iinfo->ii_vdir);
19252 + bindex = iinfo->ii_btop;
19253 + if (bindex >= 0) {
19254 + hi = au_hinode(iinfo, bindex);
19255 + bbot = iinfo->ii_bbot;
19256 + while (bindex++ <= bbot) {
19257 + if (hi->hi_inode)
19262 + au_kfree_rcu(iinfo->ii_hinode);
19263 + AuRwDestroy(&iinfo->ii_rwsem);
19265 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
19266 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
19267 +++ linux/fs/aufs/inode.c 2022-03-21 14:49:05.726633010 +0100
19269 +// SPDX-License-Identifier: GPL-2.0
19271 + * Copyright (C) 2005-2021 Junjiro R. Okajima
19273 + * This program, aufs is free software; you can redistribute it and/or modify
19274 + * it under the terms of the GNU General Public License as published by
19275 + * the Free Software Foundation; either version 2 of the License, or
19276 + * (at your option) any later version.
19278 + * This program is distributed in the hope that it will be useful,
19279 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19280 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19281 + * GNU General Public License for more details.
19283 + * You should have received a copy of the GNU General Public License
19284 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19288 + * inode functions
19291 +#include <linux/iversion.h>
19294 +struct inode *au_igrab(struct inode *inode)
19297 + AuDebugOn(!atomic_read(&inode->i_count));
19303 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
19305 + au_cpup_attr_all(inode, /*force*/0);
19306 + au_update_iigen(inode, /*half*/1);
19308 + inode_inc_iversion(inode);
19311 +static int au_ii_refresh(struct inode *inode, int *update)
19315 + aufs_bindex_t bindex, new_bindex;
19316 + struct super_block *sb;
19317 + struct au_iinfo *iinfo;
19318 + struct au_hinode *p, *q, tmp;
19320 + AuDebugOn(au_is_bad_inode(inode));
19321 + IiMustWriteLock(inode);
19324 + sb = inode->i_sb;
19325 + nbr = au_sbbot(sb) + 1;
19326 + type = inode->i_mode & S_IFMT;
19327 + iinfo = au_ii(inode);
19328 + err = au_hinode_realloc(iinfo, nbr, /*may_shrink*/0);
19329 + if (unlikely(err))
19332 + AuDebugOn(iinfo->ii_btop < 0);
19333 + p = au_hinode(iinfo, iinfo->ii_btop);
19334 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19336 + if (!p->hi_inode)
19339 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
19340 + new_bindex = au_br_index(sb, p->hi_id);
19341 + if (new_bindex == bindex)
19344 + if (new_bindex < 0) {
19347 + p->hi_inode = NULL;
19351 + if (new_bindex < iinfo->ii_btop)
19352 + iinfo->ii_btop = new_bindex;
19353 + if (iinfo->ii_bbot < new_bindex)
19354 + iinfo->ii_bbot = new_bindex;
19355 + /* swap two lower inode, and loop again */
19356 + q = au_hinode(iinfo, new_bindex);
19360 + if (tmp.hi_inode) {
19365 + au_update_ibrange(inode, /*do_put_zero*/0);
19366 + au_hinode_realloc(iinfo, nbr, /*may_shrink*/1); /* harmless if err */
19367 + e = au_dy_irefresh(inode);
19368 + if (unlikely(e && !err))
19376 +void au_refresh_iop(struct inode *inode, int force_getattr)
19379 + struct au_sbinfo *sbi = au_sbi(inode->i_sb);
19380 + const struct inode_operations *iop
19381 + = force_getattr ? aufs_iop : sbi->si_iop_array;
19383 + if (inode->i_op == iop)
19386 + switch (inode->i_mode & S_IFMT) {
19388 + type = AuIop_DIR;
19391 + type = AuIop_SYMLINK;
19394 + type = AuIop_OTHER;
19398 + inode->i_op = iop + type;
19399 + /* unnecessary smp_wmb() */
19402 +int au_refresh_hinode_self(struct inode *inode)
19406 + err = au_ii_refresh(inode, &update);
19408 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
19414 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
19416 + int err, e, update;
19417 + unsigned int flags;
19419 + aufs_bindex_t bindex, bbot;
19420 + unsigned char isdir;
19421 + struct au_hinode *p;
19422 + struct au_iinfo *iinfo;
19424 + err = au_ii_refresh(inode, &update);
19425 + if (unlikely(err))
19429 + iinfo = au_ii(inode);
19430 + p = au_hinode(iinfo, iinfo->ii_btop);
19431 + mode = (inode->i_mode & S_IFMT);
19432 + isdir = S_ISDIR(mode);
19433 + flags = au_hi_flags(inode, isdir);
19434 + bbot = au_dbbot(dentry);
19435 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
19436 + struct inode *h_i, *h_inode;
19437 + struct dentry *h_d;
19439 + h_d = au_h_dptr(dentry, bindex);
19440 + if (!h_d || d_is_negative(h_d))
19443 + h_inode = d_inode(h_d);
19444 + AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
19445 + if (iinfo->ii_btop <= bindex && bindex <= iinfo->ii_bbot) {
19446 + h_i = au_h_iptr(inode, bindex);
19448 + if (h_i == h_inode)
19454 + if (bindex < iinfo->ii_btop)
19455 + iinfo->ii_btop = bindex;
19456 + if (iinfo->ii_bbot < bindex)
19457 + iinfo->ii_bbot = bindex;
19458 + au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
19461 + au_update_ibrange(inode, /*do_put_zero*/0);
19462 + e = au_dy_irefresh(inode);
19463 + if (unlikely(e && !err))
19466 + au_refresh_hinode_attr(inode, update && isdir);
19473 +static int set_inode(struct inode *inode, struct dentry *dentry)
19476 + unsigned int flags;
19478 + aufs_bindex_t bindex, btop, btail;
19479 + unsigned char isdir;
19480 + struct dentry *h_dentry;
19481 + struct inode *h_inode;
19482 + struct au_iinfo *iinfo;
19483 + const struct inode_operations *iop;
19485 + IiMustWriteLock(inode);
19489 + iop = au_sbi(inode->i_sb)->si_iop_array;
19490 + btop = au_dbtop(dentry);
19491 + h_dentry = au_h_dptr(dentry, btop);
19492 + h_inode = d_inode(h_dentry);
19493 + mode = h_inode->i_mode;
19494 + switch (mode & S_IFMT) {
19496 + btail = au_dbtail(dentry);
19497 + inode->i_op = iop + AuIop_OTHER;
19498 + inode->i_fop = &aufs_file_fop;
19499 + err = au_dy_iaop(inode, btop, h_inode);
19500 + if (unlikely(err))
19505 + btail = au_dbtaildir(dentry);
19506 + inode->i_op = iop + AuIop_DIR;
19507 + inode->i_fop = &aufs_dir_fop;
19510 + btail = au_dbtail(dentry);
19511 + inode->i_op = iop + AuIop_SYMLINK;
19517 + btail = au_dbtail(dentry);
19518 + inode->i_op = iop + AuIop_OTHER;
19519 + init_special_inode(inode, mode, h_inode->i_rdev);
19522 + AuIOErr("Unknown file type 0%o\n", mode);
19527 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
19528 + flags = au_hi_flags(inode, isdir);
19529 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
19530 + && au_ftest_hi(flags, HNOTIFY)
19531 + && dentry->d_name.len > AUFS_WH_PFX_LEN
19532 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
19533 + au_fclr_hi(flags, HNOTIFY);
19534 + iinfo = au_ii(inode);
19535 + iinfo->ii_btop = btop;
19536 + iinfo->ii_bbot = btail;
19537 + for (bindex = btop; bindex <= btail; bindex++) {
19538 + h_dentry = au_h_dptr(dentry, bindex);
19540 + au_set_h_iptr(inode, bindex,
19541 + au_igrab(d_inode(h_dentry)), flags);
19543 + au_cpup_attr_all(inode, /*force*/1);
19545 + * to force calling aufs_get_acl() every time,
19546 + * do not call cache_no_acl() for aufs inode.
19554 + * successful returns with iinfo write_locked
19556 + * zero: success, matched
19557 + * plus: no error, but unmatched
19559 +static int reval_inode(struct inode *inode, struct dentry *dentry)
19562 + unsigned int gen, igflags;
19563 + aufs_bindex_t bindex, bbot;
19564 + struct inode *h_inode, *h_dinode;
19565 + struct dentry *h_dentry;
19568 + * before this function, if aufs got any iinfo lock, it must be only
19569 + * one, the parent dir.
19570 + * it can happen by UDBA and the obsoleted inode number.
19573 + if (unlikely(inode->i_ino == parent_ino(dentry)))
19577 + ii_write_lock_new_child(inode);
19578 + h_dentry = au_h_dptr(dentry, au_dbtop(dentry));
19579 + h_dinode = d_inode(h_dentry);
19580 + bbot = au_ibbot(inode);
19581 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
19582 + h_inode = au_h_iptr(inode, bindex);
19583 + if (!h_inode || h_inode != h_dinode)
19587 + gen = au_iigen(inode, &igflags);
19588 + if (gen == au_digen(dentry)
19589 + && !au_ig_ftest(igflags, HALF_REFRESHED))
19592 + /* fully refresh inode using dentry */
19593 + err = au_refresh_hinode(inode, dentry);
19595 + au_update_iigen(inode, /*half*/0);
19599 + if (unlikely(err))
19600 + ii_write_unlock(inode);
19605 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19606 + unsigned int d_type, ino_t *ino)
19609 + const int isnondir = d_type != DT_DIR;
19611 + /* prevent hardlinked inode number from race condition */
19613 + err = au_xinondir_enter(sb, bindex, h_ino, &idx);
19614 + if (unlikely(err))
19618 + err = au_xino_read(sb, bindex, h_ino, ino);
19619 + if (unlikely(err))
19620 + goto out_xinondir;
19624 + *ino = au_xino_new_ino(sb);
19625 + if (unlikely(!*ino))
19626 + goto out_xinondir;
19627 + err = au_xino_write(sb, bindex, h_ino, *ino);
19628 + if (unlikely(err))
19629 + goto out_xinondir;
19633 + if (isnondir && idx >= 0)
19634 + au_xinondir_leave(sb, bindex, h_ino, idx);
19639 +/* successful returns with iinfo write_locked */
19640 +/* todo: return with unlocked? */
19641 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
19643 + struct inode *inode, *h_inode;
19644 + struct dentry *h_dentry;
19645 + struct super_block *sb;
19646 + ino_t h_ino, ino;
19647 + int err, idx, hlinked;
19648 + aufs_bindex_t btop;
19650 + sb = dentry->d_sb;
19651 + btop = au_dbtop(dentry);
19652 + h_dentry = au_h_dptr(dentry, btop);
19653 + h_inode = d_inode(h_dentry);
19654 + h_ino = h_inode->i_ino;
19655 + hlinked = !d_is_dir(h_dentry) && h_inode->i_nlink > 1;
19659 + * stop 'race'-ing between hardlinks under different
19663 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
19664 + inode = ERR_PTR(err);
19665 + if (unlikely(err))
19669 + err = au_xino_read(sb, btop, h_ino, &ino);
19670 + inode = ERR_PTR(err);
19671 + if (unlikely(err))
19672 + goto out_xinondir;
19675 + ino = au_xino_new_ino(sb);
19676 + if (unlikely(!ino)) {
19677 + inode = ERR_PTR(-EIO);
19678 + goto out_xinondir;
19682 + AuDbg("i%lu\n", (unsigned long)ino);
19683 + inode = au_iget_locked(sb, ino);
19684 + err = PTR_ERR(inode);
19685 + if (IS_ERR(inode))
19686 + goto out_xinondir;
19688 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
19689 + if (inode->i_state & I_NEW) {
19690 + ii_write_lock_new_child(inode);
19691 + err = set_inode(inode, dentry);
19693 + unlock_new_inode(inode);
19694 + goto out_xinondir; /* success */
19698 + * iget_failed() calls iput(), but we need to call
19699 + * ii_write_unlock() after iget_failed(). so dirty hack for
19702 + atomic_inc(&inode->i_count);
19703 + iget_failed(inode);
19704 + ii_write_unlock(inode);
19705 + au_xino_write(sb, btop, h_ino, /*ino*/0);
19706 + /* ignore this error */
19708 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
19710 + * horrible race condition between lookup, readdir and copyup
19711 + * (or something).
19713 + if (hlinked && idx >= 0)
19714 + au_xinondir_leave(sb, btop, h_ino, idx);
19715 + err = reval_inode(inode, dentry);
19716 + if (unlikely(err < 0)) {
19721 + goto out; /* success */
19722 + else if (hlinked && idx >= 0) {
19723 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
19724 + if (unlikely(err)) {
19726 + inode = ERR_PTR(err);
19732 + if (unlikely(au_test_fs_unique_ino(h_inode)))
19733 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
19734 + " b%d, %s, %pd, hi%lu, i%lu.\n",
19735 + btop, au_sbtype(h_dentry->d_sb), dentry,
19736 + (unsigned long)h_ino, (unsigned long)ino);
19738 + err = au_xino_write(sb, btop, h_ino, /*ino*/0);
19741 + if (hlinked && idx >= 0)
19742 + au_xinondir_leave(sb, btop, h_ino, idx);
19748 + inode = ERR_PTR(err);
19750 + if (hlinked && idx >= 0)
19751 + au_xinondir_leave(sb, btop, h_ino, idx);
19756 +/* ---------------------------------------------------------------------- */
19758 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19759 + struct inode *inode)
19762 + struct inode *hi;
19764 + err = au_br_rdonly(au_sbr(sb, bindex));
19766 + /* pseudo-link after flushed may happen out of bounds */
19769 + && au_ibtop(inode) <= bindex
19770 + && bindex <= au_ibbot(inode)) {
19772 + * permission check is unnecessary since vfsub routine
19773 + * will be called later
19775 + hi = au_h_iptr(inode, bindex);
19777 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
19783 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19786 + if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
19788 + return inode_permission(h_userns, h_inode, mask);
19791 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19794 + if (au_test_nfs(h_inode->i_sb)
19795 + && (mask & MAY_WRITE)
19796 + && S_ISDIR(h_inode->i_mode))
19797 + mask |= MAY_READ; /* force permission check */
19798 + return au_test_h_perm(h_userns, h_inode, mask);
19800 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
19801 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
19802 +++ linux/fs/aufs/inode.h 2022-10-03 07:51:14.929542593 +0200
19804 +/* SPDX-License-Identifier: GPL-2.0 */
19806 + * Copyright (C) 2005-2021 Junjiro R. Okajima
19808 + * This program, aufs is free software; you can redistribute it and/or modify
19809 + * it under the terms of the GNU General Public License as published by
19810 + * the Free Software Foundation; either version 2 of the License, or
19811 + * (at your option) any later version.
19813 + * This program is distributed in the hope that it will be useful,
19814 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19815 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19816 + * GNU General Public License for more details.
19818 + * You should have received a copy of the GNU General Public License
19819 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19823 + * inode operations
19826 +#ifndef __AUFS_INODE_H__
19827 +#define __AUFS_INODE_H__
19831 +#include <linux/fsnotify.h>
19832 +#include "rwsem.h"
19836 +struct au_hnotify {
19837 +#ifdef CONFIG_AUFS_HNOTIFY
19838 +#ifdef CONFIG_AUFS_HFSNOTIFY
19839 + /* never use fsnotify_add_vfsmount_mark() */
19840 + struct fsnotify_mark hn_mark;
19842 + struct inode *hn_aufs_inode; /* no get/put */
19843 + struct rcu_head rcu;
19845 +} ____cacheline_aligned_in_smp;
19847 +struct au_hinode {
19848 + struct inode *hi_inode;
19849 + aufs_bindex_t hi_id;
19850 +#ifdef CONFIG_AUFS_HNOTIFY
19851 + struct au_hnotify *hi_notify;
19854 + /* reference to the copied-up whiteout with get/put */
19855 + struct dentry *hi_whdentry;
19859 +#define AuIG_HALF_REFRESHED 1
19860 +#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
19861 +#define au_ig_fset(flags, name) \
19862 + do { (flags) |= AuIG_##name; } while (0)
19863 +#define au_ig_fclr(flags, name) \
19864 + do { (flags) &= ~AuIG_##name; } while (0)
19867 + spinlock_t ig_spin;
19868 + __u32 ig_generation, ig_flags;
19873 + struct au_iigen ii_generation;
19874 + struct super_block *ii_hsb1; /* no get/put */
19876 + struct au_rwsem ii_rwsem;
19877 + aufs_bindex_t ii_btop, ii_bbot;
19879 + struct au_hinode *ii_hinode;
19880 + struct au_vdir *ii_vdir;
19883 +struct au_icntnr {
19884 + struct au_iinfo iinfo;
19885 + struct inode vfs_inode;
19886 + struct hlist_bl_node plink;
19887 + struct rcu_head rcu;
19888 +} ____cacheline_aligned_in_smp;
19890 +/* au_pin flags */
19891 +#define AuPin_DI_LOCKED 1
19892 +#define AuPin_MNT_WRITE (1 << 1)
19893 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
19894 +#define au_fset_pin(flags, name) \
19895 + do { (flags) |= AuPin_##name; } while (0)
19896 +#define au_fclr_pin(flags, name) \
19897 + do { (flags) &= ~AuPin_##name; } while (0)
19901 + struct dentry *dentry;
19902 + unsigned int udba;
19903 + unsigned char lsc_di, lsc_hi, flags;
19904 + aufs_bindex_t bindex;
19907 + struct dentry *parent;
19908 + struct au_hinode *hdir;
19909 + struct vfsmount *h_mnt;
19911 + /* temporary unlock/relock for copyup */
19912 + struct dentry *h_dentry, *h_parent;
19913 + struct au_branch *br;
19914 + struct task_struct *task;
19917 +void au_pin_hdir_unlock(struct au_pin *p);
19918 +int au_pin_hdir_lock(struct au_pin *p);
19919 +int au_pin_hdir_relock(struct au_pin *p);
19920 +void au_pin_hdir_acquire_nest(struct au_pin *p);
19921 +void au_pin_hdir_release(struct au_pin *p);
19923 +/* ---------------------------------------------------------------------- */
19925 +static inline struct au_iinfo *au_ii(struct inode *inode)
19927 + BUG_ON(is_bad_inode(inode));
19928 + return &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19931 +/* ---------------------------------------------------------------------- */
19934 +struct inode *au_igrab(struct inode *inode);
19935 +void au_refresh_iop(struct inode *inode, int force_getattr);
19936 +int au_refresh_hinode_self(struct inode *inode);
19937 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
19938 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19939 + unsigned int d_type, ino_t *ino);
19940 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
19941 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19942 + struct inode *inode);
19943 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19945 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19948 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
19949 + ino_t h_ino, unsigned int d_type, ino_t *ino)
19951 +#ifdef CONFIG_AUFS_SHWH
19952 + return au_ino(sb, bindex, h_ino, d_type, ino);
19965 +extern struct inode_operations aufs_iop[AuIop_Last], /* not const */
19966 + aufs_iop_nogetattr[AuIop_Last];
19968 +/* au_wr_dir flags */
19969 +#define AuWrDir_ADD_ENTRY 1
19970 +#define AuWrDir_ISDIR (1 << 1)
19971 +#define AuWrDir_TMPFILE (1 << 2)
19972 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
19973 +#define au_fset_wrdir(flags, name) \
19974 + do { (flags) |= AuWrDir_##name; } while (0)
19975 +#define au_fclr_wrdir(flags, name) \
19976 + do { (flags) &= ~AuWrDir_##name; } while (0)
19978 +struct au_wr_dir_args {
19979 + aufs_bindex_t force_btgt;
19980 + unsigned char flags;
19982 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
19983 + struct au_wr_dir_args *args);
19985 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
19986 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
19987 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
19988 + unsigned int udba, unsigned char flags);
19989 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
19990 + unsigned int udba, unsigned char flags) __must_check;
19991 +int au_do_pin(struct au_pin *pin) __must_check;
19992 +void au_unpin(struct au_pin *pin);
19993 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
19995 +#define AuIcpup_DID_CPUP 1
19996 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
19997 +#define au_fset_icpup(flags, name) \
19998 + do { (flags) |= AuIcpup_##name; } while (0)
19999 +#define au_fclr_icpup(flags, name) \
20000 + do { (flags) &= ~AuIcpup_##name; } while (0)
20002 +struct au_icpup_args {
20003 + unsigned char flags;
20004 + unsigned char pin_flags;
20005 + aufs_bindex_t btgt;
20006 + unsigned int udba;
20007 + struct au_pin pin;
20008 + struct path h_path;
20009 + struct inode *h_inode;
20012 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
20013 + struct au_icpup_args *a);
20015 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
20016 + struct path *h_path, int locked);
20019 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20020 + struct dentry *h_parent, int isdir);
20021 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
20022 + struct dentry *dentry, umode_t mode, dev_t dev);
20023 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
20024 + struct dentry *dentry, const char *symname);
20025 +int aufs_create(struct user_namespace *userns, struct inode *dir,
20026 + struct dentry *dentry, umode_t mode, bool want_excl);
20027 +struct vfsub_aopen_args;
20028 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
20029 + struct vfsub_aopen_args *args);
20030 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
20031 + struct dentry *dentry, umode_t mode);
20032 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
20033 + struct dentry *dentry);
20034 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
20035 + struct dentry *dentry, umode_t mode);
20038 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
20039 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
20040 + struct dentry *h_parent, int isdir);
20041 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
20042 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
20045 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
20046 +int aufs_rename(struct user_namespace *userns,
20047 + struct inode *_src_dir, struct dentry *_src_dentry,
20048 + struct inode *_dst_dir, struct dentry *_dst_dentry,
20049 + unsigned int _flags);
20052 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
20053 +void au_hiput(struct au_hinode *hinode);
20054 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
20055 + struct dentry *h_wh);
20056 +unsigned int au_hi_flags(struct inode *inode, int isdir);
20058 +/* hinode flags */
20059 +#define AuHi_XINO 1
20060 +#define AuHi_HNOTIFY (1 << 1)
20061 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
20062 +#define au_fset_hi(flags, name) \
20063 + do { (flags) |= AuHi_##name; } while (0)
20064 +#define au_fclr_hi(flags, name) \
20065 + do { (flags) &= ~AuHi_##name; } while (0)
20067 +#ifndef CONFIG_AUFS_HNOTIFY
20068 +#undef AuHi_HNOTIFY
20069 +#define AuHi_HNOTIFY 0
20072 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
20073 + struct inode *h_inode, unsigned int flags);
20075 +void au_update_iigen(struct inode *inode, int half);
20076 +void au_update_ibrange(struct inode *inode, int do_put_zero);
20078 +void au_icntnr_init_once(void *_c);
20079 +void au_hinode_init(struct au_hinode *hinode);
20080 +int au_iinfo_init(struct inode *inode);
20081 +void au_iinfo_fin(struct inode *inode);
20082 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink);
20084 +#ifdef CONFIG_PROC_FS
20086 +int au_plink_maint(struct super_block *sb, int flags);
20088 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
20089 +int au_plink_maint_enter(struct super_block *sb);
20090 +#ifdef CONFIG_AUFS_DEBUG
20091 +void au_plink_list(struct super_block *sb);
20093 +AuStubVoid(au_plink_list, struct super_block *sb)
20095 +int au_plink_test(struct inode *inode);
20096 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
20097 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
20098 + struct dentry *h_dentry);
20099 +void au_plink_put(struct super_block *sb, int verbose);
20100 +void au_plink_clean(struct super_block *sb, int verbose);
20101 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
20103 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
20104 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
20105 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
20106 +AuStubVoid(au_plink_list, struct super_block *sb);
20107 +AuStubInt0(au_plink_test, struct inode *inode);
20108 +AuStub(struct dentry *, au_plink_lkup, return NULL,
20109 + struct inode *inode, aufs_bindex_t bindex);
20110 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
20111 + struct dentry *h_dentry);
20112 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
20113 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
20114 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
20115 +#endif /* CONFIG_PROC_FS */
20117 +#ifdef CONFIG_AUFS_XATTR
20119 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
20120 + unsigned int verbose);
20121 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
20122 +void au_xattr_init(struct super_block *sb);
20124 +AuStubInt0(au_cpup_xattr, struct path *h_dst, struct path *h_src,
20125 + int ignore_flags, unsigned int verbose);
20126 +AuStubVoid(au_xattr_init, struct super_block *sb);
20129 +#ifdef CONFIG_FS_POSIX_ACL
20130 +struct posix_acl *aufs_get_acl(struct inode *inode, int type, bool rcu);
20131 +int aufs_set_acl(struct user_namespace *userns, struct inode *inode,
20132 + struct posix_acl *acl, int type);
20135 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
20141 +struct au_sxattr {
20145 + const char *name;
20151 + struct posix_acl *acl;
20156 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
20157 + struct au_sxattr *arg);
20160 +/* ---------------------------------------------------------------------- */
20162 +/* lock subclass for iinfo */
20164 + AuLsc_II_CHILD, /* child first */
20165 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
20166 + AuLsc_II_CHILD3, /* copyup dirs */
20167 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
20168 + AuLsc_II_PARENT2,
20169 + AuLsc_II_PARENT3, /* copyup dirs */
20170 + AuLsc_II_NEW_CHILD
20174 + * ii_read_lock_child, ii_write_lock_child,
20175 + * ii_read_lock_child2, ii_write_lock_child2,
20176 + * ii_read_lock_child3, ii_write_lock_child3,
20177 + * ii_read_lock_parent, ii_write_lock_parent,
20178 + * ii_read_lock_parent2, ii_write_lock_parent2,
20179 + * ii_read_lock_parent3, ii_write_lock_parent3,
20180 + * ii_read_lock_new_child, ii_write_lock_new_child,
20182 +#define AuReadLockFunc(name, lsc) \
20183 +static inline void ii_read_lock_##name(struct inode *i) \
20185 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20188 +#define AuWriteLockFunc(name, lsc) \
20189 +static inline void ii_write_lock_##name(struct inode *i) \
20191 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20194 +#define AuRWLockFuncs(name, lsc) \
20195 + AuReadLockFunc(name, lsc) \
20196 + AuWriteLockFunc(name, lsc)
20198 +AuRWLockFuncs(child, CHILD);
20199 +AuRWLockFuncs(child2, CHILD2);
20200 +AuRWLockFuncs(child3, CHILD3);
20201 +AuRWLockFuncs(parent, PARENT);
20202 +AuRWLockFuncs(parent2, PARENT2);
20203 +AuRWLockFuncs(parent3, PARENT3);
20204 +AuRWLockFuncs(new_child, NEW_CHILD);
20206 +#undef AuReadLockFunc
20207 +#undef AuWriteLockFunc
20208 +#undef AuRWLockFuncs
20210 +#define ii_read_unlock(i) au_rw_read_unlock(&au_ii(i)->ii_rwsem)
20211 +#define ii_write_unlock(i) au_rw_write_unlock(&au_ii(i)->ii_rwsem)
20212 +#define ii_downgrade_lock(i) au_rw_dgrade_lock(&au_ii(i)->ii_rwsem)
20214 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
20215 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
20216 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
20218 +/* ---------------------------------------------------------------------- */
20220 +static inline void au_icntnr_init(struct au_icntnr *c)
20222 +#ifdef CONFIG_AUFS_DEBUG
20223 + c->vfs_inode.i_mode = 0;
20227 +static inline unsigned int au_iigen(struct inode *inode, unsigned int *igflags)
20229 + unsigned int gen;
20230 + struct au_iinfo *iinfo;
20231 + struct au_iigen *iigen;
20233 + iinfo = au_ii(inode);
20234 + iigen = &iinfo->ii_generation;
20235 + spin_lock(&iigen->ig_spin);
20237 + *igflags = iigen->ig_flags;
20238 + gen = iigen->ig_generation;
20239 + spin_unlock(&iigen->ig_spin);
20244 +/* tiny test for inode number */
20245 +/* tmpfs generation is too rough */
20246 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
20248 + struct au_iinfo *iinfo;
20250 + iinfo = au_ii(inode);
20251 + AuRwMustAnyLock(&iinfo->ii_rwsem);
20252 + return !(iinfo->ii_hsb1 == h_inode->i_sb
20253 + && iinfo->ii_higen == h_inode->i_generation);
20256 +static inline void au_iigen_dec(struct inode *inode)
20258 + struct au_iinfo *iinfo;
20259 + struct au_iigen *iigen;
20261 + iinfo = au_ii(inode);
20262 + iigen = &iinfo->ii_generation;
20263 + spin_lock(&iigen->ig_spin);
20264 + iigen->ig_generation--;
20265 + spin_unlock(&iigen->ig_spin);
20268 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
20273 + if (unlikely(inode && au_iigen(inode, NULL) != sigen))
20279 +/* ---------------------------------------------------------------------- */
20281 +static inline struct au_hinode *au_hinode(struct au_iinfo *iinfo,
20282 + aufs_bindex_t bindex)
20284 + return iinfo->ii_hinode + bindex;
20287 +static inline int au_is_bad_inode(struct inode *inode)
20289 + return !!(is_bad_inode(inode) || !au_hinode(au_ii(inode), 0));
20292 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
20293 + aufs_bindex_t bindex)
20295 + IiMustAnyLock(inode);
20296 + return au_hinode(au_ii(inode), bindex)->hi_id;
20299 +static inline aufs_bindex_t au_ibtop(struct inode *inode)
20301 + IiMustAnyLock(inode);
20302 + return au_ii(inode)->ii_btop;
20305 +static inline aufs_bindex_t au_ibbot(struct inode *inode)
20307 + IiMustAnyLock(inode);
20308 + return au_ii(inode)->ii_bbot;
20311 +static inline struct au_vdir *au_ivdir(struct inode *inode)
20313 + IiMustAnyLock(inode);
20314 + return au_ii(inode)->ii_vdir;
20317 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
20319 + IiMustAnyLock(inode);
20320 + return au_hinode(au_ii(inode), bindex)->hi_whdentry;
20323 +static inline void au_set_ibtop(struct inode *inode, aufs_bindex_t bindex)
20325 + IiMustWriteLock(inode);
20326 + au_ii(inode)->ii_btop = bindex;
20329 +static inline void au_set_ibbot(struct inode *inode, aufs_bindex_t bindex)
20331 + IiMustWriteLock(inode);
20332 + au_ii(inode)->ii_bbot = bindex;
20335 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
20337 + IiMustWriteLock(inode);
20338 + au_ii(inode)->ii_vdir = vdir;
20341 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
20343 + IiMustAnyLock(inode);
20344 + return au_hinode(au_ii(inode), bindex);
20347 +/* ---------------------------------------------------------------------- */
20349 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
20352 + return pin->parent;
20356 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
20358 + if (pin && pin->hdir)
20359 + return pin->hdir->hi_inode;
20363 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
20366 + return pin->hdir;
20370 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
20373 + pin->dentry = dentry;
20376 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
20377 + unsigned char lflag)
20381 + au_fset_pin(pin->flags, DI_LOCKED);
20383 + au_fclr_pin(pin->flags, DI_LOCKED);
20387 +#if 0 /* reserved */
20388 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
20391 + dput(pin->parent);
20392 + pin->parent = dget(parent);
20397 +/* ---------------------------------------------------------------------- */
20400 +#ifdef CONFIG_AUFS_HNOTIFY
20401 +struct au_hnotify_op {
20402 + void (*ctl)(struct au_hinode *hinode, int do_set);
20403 + int (*alloc)(struct au_hinode *hinode);
20406 + * if it returns true, the caller should free hinode->hi_notify,
20407 + * otherwise ->free() frees it.
20409 + int (*free)(struct au_hinode *hinode,
20410 + struct au_hnotify *hn) __must_check;
20412 + void (*fin)(void);
20413 + int (*init)(void);
20415 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
20416 + void (*fin_br)(struct au_branch *br);
20417 + int (*init_br)(struct au_branch *br, int perm);
20421 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
20422 +void au_hn_free(struct au_hinode *hinode);
20423 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
20424 +void au_hn_reset(struct inode *inode, unsigned int flags);
20425 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
20426 + const struct qstr *h_child_qstr, struct inode *h_child_inode);
20427 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
20428 +int au_hnotify_init_br(struct au_branch *br, int perm);
20429 +void au_hnotify_fin_br(struct au_branch *br);
20430 +int __init au_hnotify_init(void);
20431 +void au_hnotify_fin(void);
20434 +extern const struct au_hnotify_op au_hnotify_op;
20437 +void au_hn_init(struct au_hinode *hinode)
20439 + hinode->hi_notify = NULL;
20442 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
20444 + return hinode->hi_notify;
20448 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
20449 + struct au_hinode *hinode __maybe_unused,
20450 + struct inode *inode __maybe_unused)
20451 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
20452 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
20453 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
20454 + int do_set __maybe_unused)
20455 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
20456 + unsigned int flags __maybe_unused)
20457 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
20458 + struct au_branch *br __maybe_unused,
20459 + int perm __maybe_unused)
20460 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
20461 + int perm __maybe_unused)
20462 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
20463 +AuStubInt0(__init au_hnotify_init, void)
20464 +AuStubVoid(au_hnotify_fin, void)
20465 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
20466 +#endif /* CONFIG_AUFS_HNOTIFY */
20468 +static inline void au_hn_suspend(struct au_hinode *hdir)
20470 + au_hn_ctl(hdir, /*do_set*/0);
20473 +static inline void au_hn_resume(struct au_hinode *hdir)
20475 + au_hn_ctl(hdir, /*do_set*/1);
20478 +static inline void au_hn_inode_lock(struct au_hinode *hdir)
20480 + inode_lock(hdir->hi_inode);
20481 + au_hn_suspend(hdir);
20484 +static inline void au_hn_inode_lock_nested(struct au_hinode *hdir,
20485 + unsigned int sc __maybe_unused)
20487 + inode_lock_nested(hdir->hi_inode, sc);
20488 + au_hn_suspend(hdir);
20491 +#if 0 /* unused */
20492 +#include "vfsub.h"
20493 +static inline void au_hn_inode_lock_shared_nested(struct au_hinode *hdir,
20496 + inode_lock_shared_nested(hdir->hi_inode, sc);
20497 + au_hn_suspend(hdir);
20501 +static inline void au_hn_inode_unlock(struct au_hinode *hdir)
20503 + au_hn_resume(hdir);
20504 + inode_unlock(hdir->hi_inode);
20507 +#endif /* __KERNEL__ */
20508 +#endif /* __AUFS_INODE_H__ */
20509 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
20510 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
20511 +++ linux/fs/aufs/ioctl.c 2022-03-21 14:49:05.726633010 +0100
20513 +// SPDX-License-Identifier: GPL-2.0
20515 + * Copyright (C) 2005-2021 Junjiro R. Okajima
20517 + * This program, aufs is free software; you can redistribute it and/or modify
20518 + * it under the terms of the GNU General Public License as published by
20519 + * the Free Software Foundation; either version 2 of the License, or
20520 + * (at your option) any later version.
20522 + * This program is distributed in the hope that it will be useful,
20523 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20524 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20525 + * GNU General Public License for more details.
20527 + * You should have received a copy of the GNU General Public License
20528 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20533 + * plink-management and readdir in userspace.
20534 + * assist the pathconf(3) wrapper library.
20536 + * File-based Hierarchical Storage Management.
20539 +#include <linux/compat.h>
20540 +#include <linux/file.h>
20543 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
20546 + aufs_bindex_t wbi, bindex, bbot;
20547 + struct file *h_file;
20548 + struct super_block *sb;
20549 + struct dentry *root;
20550 + struct au_branch *br;
20551 + struct aufs_wbr_fd wbrfd = {
20552 + .oflags = au_dir_roflags,
20555 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
20556 + | O_NOATIME | O_CLOEXEC;
20558 + AuDebugOn(wbrfd.oflags & ~valid);
20561 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
20562 + if (unlikely(err)) {
20568 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
20569 + wbrfd.oflags |= au_dir_roflags;
20570 + AuDbg("0%o\n", wbrfd.oflags);
20571 + if (unlikely(wbrfd.oflags & ~valid))
20575 + fd = get_unused_fd_flags(0);
20577 + if (unlikely(fd < 0))
20580 + h_file = ERR_PTR(-EINVAL);
20583 + sb = path->dentry->d_sb;
20584 + root = sb->s_root;
20585 + aufs_read_lock(root, AuLock_IR);
20586 + bbot = au_sbbot(sb);
20587 + if (wbrfd.brid >= 0) {
20588 + wbi = au_br_index(sb, wbrfd.brid);
20589 + if (unlikely(wbi < 0 || wbi > bbot))
20593 + h_file = ERR_PTR(-ENOENT);
20594 + br = au_sbr(sb, wbi);
20595 + if (!au_br_writable(br->br_perm)) {
20599 + bindex = wbi + 1;
20601 + for (; bindex <= bbot; bindex++) {
20602 + br = au_sbr(sb, bindex);
20603 + if (au_br_writable(br->br_perm)) {
20605 + br = au_sbr(sb, wbi);
20610 + AuDbg("wbi %d\n", wbi);
20612 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
20616 + aufs_read_unlock(root, AuLock_IR);
20617 + err = PTR_ERR(h_file);
20618 + if (IS_ERR(h_file))
20621 + au_lcnt_dec(&br->br_nfiles); /* cf. au_h_open() */
20622 + fd_install(fd, h_file);
20624 + goto out; /* success */
20627 + put_unused_fd(fd);
20633 +/* ---------------------------------------------------------------------- */
20635 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
20638 + struct dentry *dentry;
20641 + case AUFS_CTL_RDU:
20642 + case AUFS_CTL_RDU_INO:
20643 + err = au_rdu_ioctl(file, cmd, arg);
20646 + case AUFS_CTL_WBR_FD:
20647 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
20650 + case AUFS_CTL_IBUSY:
20651 + err = au_ibusy_ioctl(file, arg);
20654 + case AUFS_CTL_BRINFO:
20655 + err = au_brinfo_ioctl(file, arg);
20658 + case AUFS_CTL_FHSM_FD:
20659 + dentry = file->f_path.dentry;
20660 + if (IS_ROOT(dentry))
20661 + err = au_fhsm_fd(dentry->d_sb, arg);
20667 + /* do not call the lower */
20668 + AuDbg("0x%x\n", cmd);
20676 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
20681 + case AUFS_CTL_MVDOWN:
20682 + err = au_mvdown(file->f_path.dentry, (void __user *)arg);
20685 + case AUFS_CTL_WBR_FD:
20686 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
20690 + /* do not call the lower */
20691 + AuDbg("0x%x\n", cmd);
20699 +#ifdef CONFIG_COMPAT
20700 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
20701 + unsigned long arg)
20706 + case AUFS_CTL_RDU:
20707 + case AUFS_CTL_RDU_INO:
20708 + err = au_rdu_compat_ioctl(file, cmd, arg);
20711 + case AUFS_CTL_IBUSY:
20712 + err = au_ibusy_compat_ioctl(file, arg);
20715 + case AUFS_CTL_BRINFO:
20716 + err = au_brinfo_compat_ioctl(file, arg);
20720 + err = aufs_ioctl_dir(file, cmd, arg);
20727 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
20728 + unsigned long arg)
20730 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
20733 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
20734 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
20735 +++ linux/fs/aufs/i_op_add.c 2022-03-21 14:49:05.726633010 +0100
20737 +// SPDX-License-Identifier: GPL-2.0
20739 + * Copyright (C) 2005-2021 Junjiro R. Okajima
20741 + * This program, aufs is free software; you can redistribute it and/or modify
20742 + * it under the terms of the GNU General Public License as published by
20743 + * the Free Software Foundation; either version 2 of the License, or
20744 + * (at your option) any later version.
20746 + * This program is distributed in the hope that it will be useful,
20747 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20748 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20749 + * GNU General Public License for more details.
20751 + * You should have received a copy of the GNU General Public License
20752 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20756 + * inode operations (add entry)
20759 +#include <linux/iversion.h>
20763 + * final procedure of adding a new entry, except link(2).
20764 + * remove whiteout, instantiate, copyup the parent dir's times and size
20765 + * and update version.
20766 + * if it failed, re-create the removed whiteout.
20768 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
20769 + struct dentry *wh_dentry, struct dentry *dentry)
20772 + aufs_bindex_t bwh;
20773 + struct path h_path;
20774 + struct super_block *sb;
20775 + struct inode *inode, *h_dir;
20776 + struct dentry *wh;
20781 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
20782 + IMustLock(h_dir);
20783 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
20784 + bwh = au_dbwh(dentry);
20785 + h_path.dentry = wh_dentry;
20786 + h_path.mnt = au_sbr_mnt(sb, bindex);
20787 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
20789 + if (unlikely(err))
20793 + inode = au_new_inode(dentry, /*must_new*/1);
20794 + if (!IS_ERR(inode)) {
20795 + d_instantiate(dentry, inode);
20796 + dir = d_inode(dentry->d_parent); /* dir inode is locked */
20798 + au_dir_ts(dir, bindex);
20799 + inode_inc_iversion(dir);
20800 + au_fhsm_wrote(sb, bindex, /*force*/0);
20801 + return 0; /* success */
20804 + err = PTR_ERR(inode);
20809 + /* dir inode is locked */
20810 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
20811 + rerr = PTR_ERR(wh);
20812 + if (IS_ERR(wh)) {
20813 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
20814 + dentry, err, rerr);
20823 +static int au_d_may_add(struct dentry *dentry)
20828 + if (unlikely(d_unhashed(dentry)))
20830 + if (unlikely(d_really_is_positive(dentry)))
20836 + * simple tests for the adding inode operations.
20837 + * following the checks in vfs, plus the parent-child relationship.
20839 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20840 + struct dentry *h_parent, int isdir)
20844 + struct dentry *h_dentry;
20845 + struct inode *h_inode;
20847 + err = -ENAMETOOLONG;
20848 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20851 + h_dentry = au_h_dptr(dentry, bindex);
20852 + if (d_really_is_negative(dentry)) {
20854 + if (unlikely(d_is_positive(h_dentry)))
20857 + /* rename(2) case */
20859 + if (unlikely(d_is_negative(h_dentry)))
20861 + h_inode = d_inode(h_dentry);
20862 + if (unlikely(!h_inode->i_nlink))
20865 + h_mode = h_inode->i_mode;
20868 + if (unlikely(S_ISDIR(h_mode)))
20870 + } else if (unlikely(!S_ISDIR(h_mode))) {
20877 + /* expected parent dir is locked */
20878 + if (unlikely(h_parent != h_dentry->d_parent))
20887 + * initial procedure of adding a new entry.
20888 + * prepare writable branch and the parent dir, lock it,
20889 + * and lookup whiteout for the new entry.
20891 +static struct dentry*
20892 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
20893 + struct dentry *src_dentry, struct au_pin *pin,
20894 + struct au_wr_dir_args *wr_dir_args)
20896 + struct dentry *wh_dentry, *h_parent;
20897 + struct super_block *sb;
20898 + struct au_branch *br;
20900 + unsigned int udba;
20901 + aufs_bindex_t bcpup;
20903 + AuDbg("%pd\n", dentry);
20905 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
20907 + wh_dentry = ERR_PTR(err);
20908 + if (unlikely(err < 0))
20911 + sb = dentry->d_sb;
20912 + udba = au_opt_udba(sb);
20913 + err = au_pin(pin, dentry, bcpup, udba,
20914 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20915 + wh_dentry = ERR_PTR(err);
20916 + if (unlikely(err))
20919 + h_parent = au_pinned_h_parent(pin);
20920 + if (udba != AuOpt_UDBA_NONE
20921 + && au_dbtop(dentry) == bcpup)
20922 + err = au_may_add(dentry, bcpup, h_parent,
20923 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
20924 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20925 + err = -ENAMETOOLONG;
20926 + wh_dentry = ERR_PTR(err);
20927 + if (unlikely(err))
20930 + br = au_sbr(sb, bcpup);
20932 + struct path tmp = {
20933 + .dentry = h_parent,
20934 + .mnt = au_br_mnt(br)
20936 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
20939 + wh_dentry = NULL;
20940 + if (bcpup != au_dbwh(dentry))
20941 + goto out; /* success */
20944 + * ENAMETOOLONG here means that if we allowed create such name, then it
20945 + * would not be able to removed in the future. So we don't allow such
20946 + * name here and we don't handle ENAMETOOLONG differently here.
20948 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
20951 + if (IS_ERR(wh_dentry))
20954 + return wh_dentry;
20957 +/* ---------------------------------------------------------------------- */
20959 +enum { Mknod, Symlink, Creat };
20960 +struct simple_arg {
20967 + struct vfsub_aopen_args *aopen;
20970 + const char *symname;
20979 +static int add_simple(struct inode *dir, struct dentry *dentry,
20980 + struct simple_arg *arg)
20983 + aufs_bindex_t btop;
20984 + unsigned char created;
20985 + const unsigned char try_aopen
20986 + = (arg->type == Creat && arg->u.c.try_aopen);
20987 + struct vfsub_aopen_args *aopen = arg->u.c.aopen;
20988 + struct dentry *wh_dentry, *parent;
20989 + struct inode *h_dir;
20990 + struct super_block *sb;
20991 + struct au_branch *br;
20992 + /* to reduce stack size */
20994 + struct au_dtime dt;
20995 + struct au_pin pin;
20996 + struct path h_path;
20997 + struct au_wr_dir_args wr_dir_args;
21000 + AuDbg("%pd\n", dentry);
21004 + a = kmalloc(sizeof(*a), GFP_NOFS);
21005 + if (unlikely(!a))
21007 + a->wr_dir_args.force_btgt = -1;
21008 + a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
21010 + parent = dentry->d_parent; /* dir inode is locked */
21011 + if (!try_aopen) {
21012 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21013 + if (unlikely(err))
21016 + err = au_d_may_add(dentry);
21017 + if (unlikely(err))
21020 + di_write_lock_parent(parent);
21021 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21022 + &a->pin, &a->wr_dir_args);
21023 + err = PTR_ERR(wh_dentry);
21024 + if (IS_ERR(wh_dentry))
21027 + btop = au_dbtop(dentry);
21028 + sb = dentry->d_sb;
21029 + br = au_sbr(sb, btop);
21030 + a->h_path.dentry = au_h_dptr(dentry, btop);
21031 + a->h_path.mnt = au_br_mnt(br);
21032 + h_dir = au_pinned_h_dir(&a->pin);
21033 + switch (arg->type) {
21035 + if (!try_aopen || !h_dir->i_op->atomic_open) {
21036 + err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
21037 + arg->u.c.want_excl);
21039 + if (!err && try_aopen)
21040 + aopen->file->f_mode |= FMODE_CREATED;
21043 + err = vfsub_atomic_open(h_dir, a->h_path.dentry, aopen);
21044 + AuDbg("err %d\n", err);
21045 + AuDbgFile(aopen->file);
21046 + created = err >= 0
21047 + && !!(aopen->file->f_mode & FMODE_CREATED);
21051 + err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
21055 + err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
21062 + if (unlikely(err < 0))
21065 + err = epilog(dir, btop, wh_dentry, dentry);
21067 + goto out_unpin; /* success */
21070 + if (created /* && d_is_positive(a->h_path.dentry) */) {
21071 + /* no delegation since it is just created */
21072 + rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
21075 + AuIOErr("%pd revert failure(%d, %d)\n",
21076 + dentry, err, rerr);
21079 + au_dtime_revert(&a->dt);
21081 + if (try_aopen && h_dir->i_op->atomic_open
21082 + && (aopen->file->f_mode & FMODE_OPENED))
21083 + /* aopen->file is still opened */
21084 + au_lcnt_dec(&aopen->br->br_nfiles);
21087 + au_unpin(&a->pin);
21091 + di_write_unlock(parent);
21093 + if (unlikely(err)) {
21094 + au_update_dbtop(dentry);
21098 + aufs_read_unlock(dentry, AuLock_DW);
21105 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
21106 + struct dentry *dentry, umode_t mode, dev_t dev)
21108 + struct simple_arg arg = {
21115 + return add_simple(dir, dentry, &arg);
21118 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
21119 + struct dentry *dentry, const char *symname)
21121 + struct simple_arg arg = {
21123 + .u.s.symname = symname
21125 + return add_simple(dir, dentry, &arg);
21128 +int aufs_create(struct user_namespace *userns, struct inode *dir,
21129 + struct dentry *dentry, umode_t mode, bool want_excl)
21131 + struct simple_arg arg = {
21135 + .want_excl = want_excl
21138 + return add_simple(dir, dentry, &arg);
21141 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
21142 + struct vfsub_aopen_args *aopen_args)
21144 + struct simple_arg arg = {
21147 + .mode = aopen_args->create_mode,
21148 + .want_excl = aopen_args->open_flag & O_EXCL,
21149 + .try_aopen = true,
21150 + .aopen = aopen_args
21153 + return add_simple(dir, dentry, &arg);
21156 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
21157 + struct dentry *dentry, umode_t mode)
21160 + aufs_bindex_t bindex;
21161 + struct super_block *sb;
21162 + struct dentry *parent, *h_parent, *h_dentry;
21163 + struct inode *h_dir, *inode;
21164 + struct vfsmount *h_mnt;
21165 + struct user_namespace *h_userns;
21166 + struct au_wr_dir_args wr_dir_args = {
21167 + .force_btgt = -1,
21168 + .flags = AuWrDir_TMPFILE
21171 + /* copy-up may happen */
21175 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21176 + if (unlikely(err))
21179 + err = au_di_init(dentry);
21180 + if (unlikely(err))
21184 + parent = d_find_any_alias(dir);
21185 + AuDebugOn(!parent);
21186 + di_write_lock_parent(parent);
21187 + if (unlikely(d_inode(parent) != dir))
21190 + err = au_digen_test(parent, au_sigen(sb));
21191 + if (unlikely(err))
21194 + bindex = au_dbtop(parent);
21195 + au_set_dbtop(dentry, bindex);
21196 + au_set_dbbot(dentry, bindex);
21197 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
21199 + if (unlikely(err < 0))
21202 + err = -EOPNOTSUPP;
21203 + h_dir = au_h_iptr(dir, bindex);
21204 + if (unlikely(!h_dir->i_op->tmpfile))
21207 + h_mnt = au_sbr_mnt(sb, bindex);
21208 + err = vfsub_mnt_want_write(h_mnt);
21209 + if (unlikely(err))
21212 + h_userns = mnt_user_ns(h_mnt);
21213 + h_parent = au_h_dptr(parent, bindex);
21214 + h_dentry = vfs_tmpfile(h_userns, h_parent, mode, /*open_flag*/0);
21215 + if (IS_ERR(h_dentry)) {
21216 + err = PTR_ERR(h_dentry);
21220 + au_set_dbtop(dentry, bindex);
21221 + au_set_dbbot(dentry, bindex);
21222 + au_set_h_dptr(dentry, bindex, dget(h_dentry));
21223 + inode = au_new_inode(dentry, /*must_new*/1);
21224 + if (IS_ERR(inode)) {
21225 + err = PTR_ERR(inode);
21226 + au_set_h_dptr(dentry, bindex, NULL);
21227 + au_set_dbtop(dentry, -1);
21228 + au_set_dbbot(dentry, -1);
21230 + if (!inode->i_nlink)
21231 + set_nlink(inode, 1);
21232 + d_tmpfile(dentry, inode);
21233 + au_di(dentry)->di_tmpfile = 1;
21235 + /* update without i_mutex */
21236 + if (au_ibtop(dir) == au_dbtop(dentry))
21237 + au_cpup_attr_timesizes(dir);
21242 + vfsub_mnt_drop_write(h_mnt);
21244 + di_write_unlock(parent);
21246 + di_write_unlock(dentry);
21247 + if (unlikely(err)) {
21248 + au_di_fin(dentry);
21249 + dentry->d_fsdata = NULL;
21252 + si_read_unlock(sb);
21254 + inode_unlock(dir);
21258 +/* ---------------------------------------------------------------------- */
21260 +struct au_link_args {
21261 + aufs_bindex_t bdst, bsrc;
21262 + struct au_pin pin;
21263 + struct path h_path;
21264 + struct dentry *src_parent, *parent;
21267 +static int au_cpup_before_link(struct dentry *src_dentry,
21268 + struct au_link_args *a)
21271 + struct dentry *h_src_dentry;
21272 + struct au_cp_generic cpg = {
21273 + .dentry = src_dentry,
21278 + .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
21281 + di_read_lock_parent(a->src_parent, AuLock_IR);
21282 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
21283 + if (unlikely(err))
21286 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
21287 + err = au_pin(&a->pin, src_dentry, a->bdst,
21288 + au_opt_udba(src_dentry->d_sb),
21289 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21290 + if (unlikely(err))
21293 + err = au_sio_cpup_simple(&cpg);
21294 + au_unpin(&a->pin);
21297 + di_read_unlock(a->src_parent, AuLock_IR);
21301 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
21302 + struct au_link_args *a)
21305 + unsigned char plink;
21306 + aufs_bindex_t bbot;
21307 + struct dentry *h_src_dentry;
21308 + struct inode *h_inode, *inode, *delegated;
21309 + struct super_block *sb;
21310 + struct file *h_file;
21314 + sb = src_dentry->d_sb;
21315 + inode = d_inode(src_dentry);
21316 + if (au_ibtop(inode) <= a->bdst)
21317 + h_inode = au_h_iptr(inode, a->bdst);
21318 + if (!h_inode || !h_inode->i_nlink) {
21319 + /* copyup src_dentry as the name of dentry. */
21320 + bbot = au_dbbot(dentry);
21321 + if (bbot < a->bsrc)
21322 + au_set_dbbot(dentry, a->bsrc);
21323 + au_set_h_dptr(dentry, a->bsrc,
21324 + dget(au_h_dptr(src_dentry, a->bsrc)));
21325 + dget(a->h_path.dentry);
21326 + au_set_h_dptr(dentry, a->bdst, NULL);
21327 + AuDbg("temporary d_inode...\n");
21328 + spin_lock(&dentry->d_lock);
21329 + dentry->d_inode = d_inode(src_dentry); /* tmp */
21330 + spin_unlock(&dentry->d_lock);
21331 + h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
21332 + if (IS_ERR(h_file))
21333 + err = PTR_ERR(h_file);
21335 + struct au_cp_generic cpg = {
21336 + .dentry = dentry,
21341 + .flags = AuCpup_KEEPLINO
21343 + err = au_sio_cpup_simple(&cpg);
21344 + au_h_open_post(dentry, a->bsrc, h_file);
21346 + dput(a->h_path.dentry);
21347 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21349 + au_set_h_dptr(dentry, a->bdst,
21350 + a->h_path.dentry);
21352 + spin_lock(&dentry->d_lock);
21353 + dentry->d_inode = NULL; /* restore */
21354 + spin_unlock(&dentry->d_lock);
21355 + AuDbg("temporary d_inode...done\n");
21356 + au_set_h_dptr(dentry, a->bsrc, NULL);
21357 + au_set_dbbot(dentry, bbot);
21359 + /* the inode of src_dentry already exists on a.bdst branch */
21360 + h_src_dentry = d_find_alias(h_inode);
21361 + if (!h_src_dentry && au_plink_test(inode)) {
21363 + h_src_dentry = au_plink_lkup(inode, a->bdst);
21364 + err = PTR_ERR(h_src_dentry);
21365 + if (IS_ERR(h_src_dentry))
21368 + if (unlikely(d_is_negative(h_src_dentry))) {
21369 + dput(h_src_dentry);
21370 + h_src_dentry = NULL;
21374 + if (h_src_dentry) {
21375 + delegated = NULL;
21376 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21377 + &a->h_path, &delegated);
21378 + if (unlikely(err == -EWOULDBLOCK)) {
21379 + pr_warn("cannot retry for NFSv4 delegation"
21380 + " for an internal link\n");
21383 + dput(h_src_dentry);
21385 + AuIOErr("no dentry found for hi%lu on b%d\n",
21386 + h_inode->i_ino, a->bdst);
21391 + if (!err && !plink)
21392 + au_plink_append(inode, a->bdst, a->h_path.dentry);
21399 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
21400 + struct dentry *dentry)
21403 + struct au_dtime dt;
21404 + struct au_link_args *a;
21405 + struct dentry *wh_dentry, *h_src_dentry;
21406 + struct inode *inode, *delegated;
21407 + struct super_block *sb;
21408 + struct au_wr_dir_args wr_dir_args = {
21409 + /* .force_btgt = -1, */
21410 + .flags = AuWrDir_ADD_ENTRY
21414 + inode = d_inode(src_dentry);
21415 + IMustLock(inode);
21418 + a = kzalloc(sizeof(*a), GFP_NOFS);
21419 + if (unlikely(!a))
21422 + a->parent = dentry->d_parent; /* dir inode is locked */
21423 + err = aufs_read_and_write_lock2(dentry, src_dentry,
21424 + AuLock_NOPLM | AuLock_GEN);
21425 + if (unlikely(err))
21427 + err = au_d_linkable(src_dentry);
21428 + if (unlikely(err))
21430 + err = au_d_may_add(dentry);
21431 + if (unlikely(err))
21434 + a->src_parent = dget_parent(src_dentry);
21435 + wr_dir_args.force_btgt = au_ibtop(inode);
21437 + di_write_lock_parent(a->parent);
21438 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
21439 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
21441 + err = PTR_ERR(wh_dentry);
21442 + if (IS_ERR(wh_dentry))
21446 + sb = dentry->d_sb;
21447 + a->bdst = au_dbtop(dentry);
21448 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21449 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
21450 + a->bsrc = au_ibtop(inode);
21451 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21452 + if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
21453 + h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
21454 + if (!h_src_dentry) {
21455 + a->bsrc = au_dbtop(src_dentry);
21456 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21457 + AuDebugOn(!h_src_dentry);
21458 + } else if (IS_ERR(h_src_dentry)) {
21459 + err = PTR_ERR(h_src_dentry);
21464 + * aufs doesn't touch the credential so
21465 + * security_dentry_create_files_as() is unnecessary.
21467 + if (au_opt_test(au_mntflags(sb), PLINK)) {
21468 + if (a->bdst < a->bsrc
21469 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
21470 + err = au_cpup_or_link(src_dentry, dentry, a);
21472 + delegated = NULL;
21473 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21474 + &a->h_path, &delegated);
21475 + if (unlikely(err == -EWOULDBLOCK)) {
21476 + pr_warn("cannot retry for NFSv4 delegation"
21477 + " for an internal link\n");
21481 + dput(h_src_dentry);
21484 + * copyup src_dentry to the branch we process,
21485 + * and then link(2) to it.
21487 + dput(h_src_dentry);
21488 + if (a->bdst < a->bsrc
21489 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
21490 + au_unpin(&a->pin);
21491 + di_write_unlock(a->parent);
21492 + err = au_cpup_before_link(src_dentry, a);
21493 + di_write_lock_parent(a->parent);
21495 + err = au_pin(&a->pin, dentry, a->bdst,
21497 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21498 + if (unlikely(err))
21502 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
21504 + if (h_src_dentry && d_is_positive(h_src_dentry)) {
21505 + delegated = NULL;
21506 + err = vfsub_link(h_src_dentry,
21507 + au_pinned_h_dir(&a->pin),
21508 + &a->h_path, &delegated);
21509 + if (unlikely(err == -EWOULDBLOCK)) {
21510 + pr_warn("cannot retry"
21511 + " for NFSv4 delegation"
21512 + " for an internal link\n");
21518 + if (unlikely(err))
21522 + a->h_path.dentry = wh_dentry;
21523 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
21525 + if (unlikely(err))
21529 + au_dir_ts(dir, a->bdst);
21530 + inode_inc_iversion(dir);
21531 + inc_nlink(inode);
21532 + inode->i_ctime = dir->i_ctime;
21533 + d_instantiate(dentry, au_igrab(inode));
21534 + if (d_unhashed(a->h_path.dentry))
21535 + /* some filesystem calls d_drop() */
21537 + /* some filesystems consume an inode even hardlink */
21538 + au_fhsm_wrote(sb, a->bdst, /*force*/0);
21539 + goto out_unpin; /* success */
21542 + /* no delegation since it is just created */
21543 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
21544 + /*delegated*/NULL, /*force*/0);
21545 + if (unlikely(rerr)) {
21546 + AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
21549 + au_dtime_revert(&dt);
21551 + au_unpin(&a->pin);
21555 + di_write_unlock(a->parent);
21556 + dput(a->src_parent);
21558 + if (unlikely(err)) {
21559 + au_update_dbtop(dentry);
21562 + aufs_read_and_write_unlock2(dentry, src_dentry);
21570 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
21571 + struct dentry *dentry, umode_t mode)
21574 + aufs_bindex_t bindex;
21575 + unsigned char diropq;
21576 + struct path h_path;
21577 + struct dentry *wh_dentry, *parent, *opq_dentry;
21578 + struct inode *h_inode;
21579 + struct super_block *sb;
21581 + struct au_pin pin;
21582 + struct au_dtime dt;
21583 + } *a; /* reduce the stack usage */
21584 + struct au_wr_dir_args wr_dir_args = {
21585 + .force_btgt = -1,
21586 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
21592 + a = kmalloc(sizeof(*a), GFP_NOFS);
21593 + if (unlikely(!a))
21596 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21597 + if (unlikely(err))
21599 + err = au_d_may_add(dentry);
21600 + if (unlikely(err))
21603 + parent = dentry->d_parent; /* dir inode is locked */
21604 + di_write_lock_parent(parent);
21605 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21606 + &a->pin, &wr_dir_args);
21607 + err = PTR_ERR(wh_dentry);
21608 + if (IS_ERR(wh_dentry))
21611 + sb = dentry->d_sb;
21612 + bindex = au_dbtop(dentry);
21613 + h_path.dentry = au_h_dptr(dentry, bindex);
21614 + h_path.mnt = au_sbr_mnt(sb, bindex);
21615 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
21616 + if (unlikely(err))
21619 + /* make the dir opaque */
21621 + h_inode = d_inode(h_path.dentry);
21623 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
21624 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
21625 + opq_dentry = au_diropq_create(dentry, bindex);
21626 + inode_unlock(h_inode);
21627 + err = PTR_ERR(opq_dentry);
21628 + if (IS_ERR(opq_dentry))
21630 + dput(opq_dentry);
21634 + err = epilog(dir, bindex, wh_dentry, dentry);
21637 + goto out_unpin; /* success */
21642 + AuLabel(revert opq);
21643 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
21644 + rerr = au_diropq_remove(dentry, bindex);
21645 + inode_unlock(h_inode);
21647 + AuIOErr("%pd reverting diropq failed(%d, %d)\n",
21648 + dentry, err, rerr);
21654 + AuLabel(revert dir);
21655 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
21657 + AuIOErr("%pd reverting dir failed(%d, %d)\n",
21658 + dentry, err, rerr);
21661 + au_dtime_revert(&a->dt);
21663 + au_unpin(&a->pin);
21666 + di_write_unlock(parent);
21668 + if (unlikely(err)) {
21669 + au_update_dbtop(dentry);
21672 + aufs_read_unlock(dentry, AuLock_DW);
21678 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
21679 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
21680 +++ linux/fs/aufs/i_op.c 2022-03-21 14:49:05.726633010 +0100
21682 +// SPDX-License-Identifier: GPL-2.0
21684 + * Copyright (C) 2005-2021 Junjiro R. Okajima
21686 + * This program, aufs is free software; you can redistribute it and/or modify
21687 + * it under the terms of the GNU General Public License as published by
21688 + * the Free Software Foundation; either version 2 of the License, or
21689 + * (at your option) any later version.
21691 + * This program is distributed in the hope that it will be useful,
21692 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21693 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21694 + * GNU General Public License for more details.
21696 + * You should have received a copy of the GNU General Public License
21697 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21701 + * inode operations (except add/del/rename)
21704 +#include <linux/device_cgroup.h>
21705 +#include <linux/fs_stack.h>
21706 +#include <linux/iversion.h>
21707 +#include <linux/security.h>
21710 +static int h_permission(struct inode *h_inode, int mask,
21711 + struct path *h_path, int brperm)
21714 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21715 + struct user_namespace *h_userns;
21718 + if (write_mask && IS_IMMUTABLE(h_inode))
21722 + if (((mask & MAY_EXEC)
21723 + && S_ISREG(h_inode->i_mode)
21724 + && (path_noexec(h_path)
21725 + || !(h_inode->i_mode & 0111))))
21729 + * - skip the lower fs test in the case of write to ro branch.
21730 + * - nfs dir permission write check is optimized, but a policy for
21731 + * link/rename requires a real check.
21732 + * - nfs always sets SB_POSIXACL regardless its mount option 'noacl.'
21733 + * in this case, generic_permission() returns -EOPNOTSUPP.
21735 + h_userns = mnt_user_ns(h_path->mnt);
21736 + if ((write_mask && !au_br_writable(brperm))
21737 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
21738 + && write_mask && !(mask & MAY_READ))
21739 + || !h_inode->i_op->permission) {
21740 + /* AuLabel(generic_permission); */
21741 + /* AuDbg("get_acl %ps\n", h_inode->i_op->get_acl); */
21742 + err = generic_permission(h_userns, h_inode, mask);
21743 + if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
21744 + err = h_inode->i_op->permission(h_userns, h_inode,
21748 + /* AuLabel(h_inode->permission); */
21749 + err = h_inode->i_op->permission(h_userns, h_inode, mask);
21754 + err = devcgroup_inode_permission(h_inode, mask);
21756 + err = security_inode_permission(h_inode, mask);
21762 +static int aufs_permission(struct user_namespace *userns, struct inode *inode,
21766 + aufs_bindex_t bindex, bbot;
21767 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
21768 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21769 + struct inode *h_inode;
21770 + struct super_block *sb;
21771 + struct au_branch *br;
21773 + /* todo: support rcu-walk? */
21774 + if (mask & MAY_NOT_BLOCK)
21777 + sb = inode->i_sb;
21778 + si_read_lock(sb, AuLock_FLUSH);
21779 + ii_read_lock_child(inode);
21780 +#if 0 /* reserved for future use */
21782 + * This test may be rather 'too much' since the test is essentially done
21783 + * in the aufs_lookup(). Theoretically it is possible that the inode
21784 + * generation doesn't match to the superblock's here. But it isn't a
21785 + * big deal I suppose.
21787 + err = au_iigen_test(inode, au_sigen(sb));
21788 + if (unlikely(err))
21794 + || au_opt_test(au_mntflags(sb), DIRPERM1)) {
21795 + err = au_busy_or_stale();
21796 + h_inode = au_h_iptr(inode, au_ibtop(inode));
21797 + if (unlikely(!h_inode
21798 + || (h_inode->i_mode & S_IFMT)
21799 + != (inode->i_mode & S_IFMT)))
21803 + bindex = au_ibtop(inode);
21804 + br = au_sbr(sb, bindex);
21805 + err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
21808 + && !special_file(h_inode->i_mode)) {
21809 + /* test whether the upper writable branch exists */
21811 + for (; bindex >= 0; bindex--)
21812 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
21820 + /* non-write to dir */
21822 + bbot = au_ibbot(inode);
21823 + for (bindex = au_ibtop(inode); !err && bindex <= bbot; bindex++) {
21824 + h_inode = au_h_iptr(inode, bindex);
21826 + err = au_busy_or_stale();
21827 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
21830 + br = au_sbr(sb, bindex);
21831 + err = h_permission(h_inode, mask, &br->br_path,
21837 + ii_read_unlock(inode);
21838 + si_read_unlock(sb);
21842 +/* ---------------------------------------------------------------------- */
21844 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
21845 + unsigned int flags)
21847 + struct dentry *ret, *parent;
21848 + struct inode *inode;
21849 + struct super_block *sb;
21850 + int err, npositive;
21854 + /* todo: support rcu-walk? */
21855 + ret = ERR_PTR(-ECHILD);
21856 + if (flags & LOOKUP_RCU)
21859 + ret = ERR_PTR(-ENAMETOOLONG);
21860 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
21864 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21865 + ret = ERR_PTR(err);
21866 + if (unlikely(err))
21869 + err = au_di_init(dentry);
21870 + ret = ERR_PTR(err);
21871 + if (unlikely(err))
21875 + npositive = 0; /* suppress a warning */
21876 + parent = dentry->d_parent; /* dir inode is locked */
21877 + di_read_lock_parent(parent, AuLock_IR);
21878 + err = au_alive_dir(parent);
21880 + err = au_digen_test(parent, au_sigen(sb));
21882 + /* regardless LOOKUP_CREATE, always ALLOW_NEG */
21883 + npositive = au_lkup_dentry(dentry, au_dbtop(parent),
21884 + AuLkup_ALLOW_NEG);
21887 + di_read_unlock(parent, AuLock_IR);
21888 + ret = ERR_PTR(err);
21889 + if (unlikely(err < 0))
21893 + inode = au_new_inode(dentry, /*must_new*/0);
21894 + if (IS_ERR(inode)) {
21895 + ret = (void *)inode;
21902 + atomic_inc(&inode->i_count);
21903 + ret = d_splice_alias(inode, dentry);
21904 +#if 0 /* reserved for future use */
21905 + if (unlikely(d_need_lookup(dentry))) {
21906 + spin_lock(&dentry->d_lock);
21907 + dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
21908 + spin_unlock(&dentry->d_lock);
21912 + if (!IS_ERR(ret)) {
21914 + if (ret && ret != dentry)
21915 + ii_write_unlock(inode);
21917 + ii_write_unlock(inode);
21924 + di_write_unlock(dentry);
21926 + si_read_unlock(sb);
21931 +/* ---------------------------------------------------------------------- */
21934 + * very dirty and complicated aufs ->atomic_open().
21935 + * aufs_atomic_open()
21936 + * + au_aopen_or_create()
21938 + * + vfsub_atomic_open()
21939 + * + branch fs ->atomic_open()
21940 + * may call the actual 'open' for h_file
21941 + * + inc br_nfiles only if opened
21942 + * + au_aopen_no_open() or au_aopen_do_open()
21944 + * au_aopen_do_open()
21945 + * + finish_open()
21946 + * + au_do_aopen()
21947 + * + au_do_open() the body of all 'open'
21948 + * + au_do_open_nondir()
21949 + * set the passed h_file
21951 + * au_aopen_no_open()
21952 + * + finish_no_open()
21955 +struct aopen_node {
21956 + struct hlist_bl_node hblist;
21957 + struct file *file, *h_file;
21960 +static int au_do_aopen(struct inode *inode, struct file *file)
21962 + struct hlist_bl_head *aopen;
21963 + struct hlist_bl_node *pos;
21964 + struct aopen_node *node;
21965 + struct au_do_open_args args = {
21967 + .open = au_do_open_nondir
21970 + aopen = &au_sbi(inode->i_sb)->si_aopen;
21971 + hlist_bl_lock(aopen);
21972 + hlist_bl_for_each_entry(node, pos, aopen, hblist)
21973 + if (node->file == file) {
21974 + args.h_file = node->h_file;
21977 + hlist_bl_unlock(aopen);
21978 + /* AuDebugOn(!args.h_file); */
21980 + return au_do_open(file, &args);
21983 +static int au_aopen_do_open(struct file *file, struct dentry *dentry,
21984 + struct aopen_node *aopen_node)
21987 + struct hlist_bl_head *aopen;
21990 + aopen = &au_sbi(dentry->d_sb)->si_aopen;
21991 + au_hbl_add(&aopen_node->hblist, aopen);
21992 + err = finish_open(file, dentry, au_do_aopen);
21993 + au_hbl_del(&aopen_node->hblist, aopen);
21994 + /* AuDbgFile(file); */
21995 + AuDbg("%pd%s%s\n", dentry,
21996 + (file->f_mode & FMODE_CREATED) ? " created" : "",
21997 + (file->f_mode & FMODE_OPENED) ? " opened" : "");
22003 +static int au_aopen_no_open(struct file *file, struct dentry *dentry)
22009 + err = finish_no_open(file, dentry);
22015 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
22016 + struct file *file, unsigned int open_flag,
22017 + umode_t create_mode)
22019 + int err, did_open;
22020 + unsigned int lkup_flags;
22021 + aufs_bindex_t bindex;
22022 + struct super_block *sb;
22023 + struct dentry *parent, *d;
22024 + struct vfsub_aopen_args args = {
22025 + .open_flag = open_flag,
22026 + .create_mode = create_mode
22028 + struct aopen_node aopen_node = {
22033 + AuDbg("open_flag 0%o\n", open_flag);
22034 + AuDbgDentry(dentry);
22037 + if (!au_di(dentry)) {
22038 + lkup_flags = LOOKUP_OPEN;
22039 + if (open_flag & O_CREAT)
22040 + lkup_flags |= LOOKUP_CREATE;
22041 + d = aufs_lookup(dir, dentry, lkup_flags);
22043 + err = PTR_ERR(d);
22048 + * obsoleted dentry found.
22049 + * another error will be returned later.
22055 + AuDbgDentry(dentry);
22058 + if (d_is_positive(dentry)
22059 + || d_unhashed(dentry)
22060 + || d_unlinked(dentry)
22061 + || !(open_flag & O_CREAT)) {
22062 + err = au_aopen_no_open(file, dentry);
22063 + goto out; /* success */
22066 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
22067 + if (unlikely(err))
22070 + sb = dentry->d_sb;
22071 + parent = dentry->d_parent; /* dir is locked */
22072 + di_write_lock_parent(parent);
22073 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
22074 + if (unlikely(err < 0))
22077 + AuDbgDentry(dentry);
22078 + if (d_is_positive(dentry)) {
22079 + err = au_aopen_no_open(file, dentry);
22080 + goto out_parent; /* success */
22083 + args.file = alloc_empty_file(file->f_flags, current_cred());
22084 + err = PTR_ERR(args.file);
22085 + if (IS_ERR(args.file))
22088 + bindex = au_dbtop(dentry);
22089 + err = au_aopen_or_create(dir, dentry, &args);
22091 + AuDbgFile(args.file);
22092 + file->f_mode = args.file->f_mode & ~FMODE_OPENED;
22093 + did_open = !!(args.file->f_mode & FMODE_OPENED);
22096 + args.file = NULL;
22098 + di_write_unlock(parent);
22099 + di_write_unlock(dentry);
22100 + if (unlikely(err < 0)) {
22107 + err = au_aopen_no_open(file, dentry);
22109 + aopen_node.h_file = args.file;
22110 + err = au_aopen_do_open(file, dentry, &aopen_node);
22112 + if (unlikely(err < 0)) {
22116 + au_lcnt_dec(&args.br->br_nfiles);
22118 + goto out_sb; /* success */
22121 + di_write_unlock(parent);
22122 + di_write_unlock(dentry);
22124 + si_read_unlock(sb);
22132 +/* ---------------------------------------------------------------------- */
22134 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
22135 + const unsigned char add_entry, aufs_bindex_t bcpup,
22136 + aufs_bindex_t btop)
22139 + struct dentry *h_parent;
22140 + struct inode *h_dir;
22143 + IMustLock(d_inode(parent));
22145 + di_write_lock_parent(parent);
22148 + if (!au_h_dptr(parent, bcpup)) {
22149 + if (btop > bcpup)
22150 + err = au_cpup_dirs(dentry, bcpup);
22151 + else if (btop < bcpup)
22152 + err = au_cpdown_dirs(dentry, bcpup);
22156 + if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
22157 + h_parent = au_h_dptr(parent, bcpup);
22158 + h_dir = d_inode(h_parent);
22159 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
22160 + err = au_lkup_neg(dentry, bcpup, /*wh*/0);
22161 + /* todo: no unlock here */
22162 + inode_unlock_shared(h_dir);
22164 + AuDbg("bcpup %d\n", bcpup);
22166 + if (d_really_is_negative(dentry))
22167 + au_set_h_dptr(dentry, btop, NULL);
22168 + au_update_dbrange(dentry, /*do_put_zero*/0);
22173 + di_write_unlock(parent);
22175 + err = bcpup; /* success */
22182 + * decide the branch and the parent dir where we will create a new entry.
22183 + * returns new bindex or an error.
22184 + * copyup the parent dir if needed.
22186 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
22187 + struct au_wr_dir_args *args)
22190 + unsigned int flags;
22191 + aufs_bindex_t bcpup, btop, src_btop;
22192 + const unsigned char add_entry
22193 + = au_ftest_wrdir(args->flags, ADD_ENTRY)
22194 + | au_ftest_wrdir(args->flags, TMPFILE);
22195 + struct super_block *sb;
22196 + struct dentry *parent;
22197 + struct au_sbinfo *sbinfo;
22199 + sb = dentry->d_sb;
22200 + sbinfo = au_sbi(sb);
22201 + parent = dget_parent(dentry);
22202 + btop = au_dbtop(dentry);
22204 + if (args->force_btgt < 0) {
22205 + if (src_dentry) {
22206 + src_btop = au_dbtop(src_dentry);
22207 + if (src_btop < btop)
22208 + bcpup = src_btop;
22209 + } else if (add_entry) {
22211 + if (au_ftest_wrdir(args->flags, ISDIR))
22212 + au_fset_wbr(flags, DIR);
22213 + err = AuWbrCreate(sbinfo, dentry, flags);
22217 + if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
22219 + err = AuWbrCopyup(sbinfo, dentry);
22221 + if (!IS_ROOT(dentry)) {
22222 + di_read_lock_parent(parent, !AuLock_IR);
22223 + err = AuWbrCopyup(sbinfo, dentry);
22224 + di_read_unlock(parent, !AuLock_IR);
22226 + err = AuWbrCopyup(sbinfo, dentry);
22229 + if (unlikely(err < 0))
22233 + bcpup = args->force_btgt;
22234 + AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
22237 + AuDbg("btop %d, bcpup %d\n", btop, bcpup);
22239 + if (bcpup == btop)
22240 + goto out; /* success */
22242 + /* copyup the new parent into the branch we process */
22243 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, btop);
22245 + if (d_really_is_negative(dentry)) {
22246 + au_set_h_dptr(dentry, btop, NULL);
22247 + au_set_dbtop(dentry, bcpup);
22248 + au_set_dbbot(dentry, bcpup);
22250 + AuDebugOn(add_entry
22251 + && !au_ftest_wrdir(args->flags, TMPFILE)
22252 + && !au_h_dptr(dentry, bcpup));
22260 +/* ---------------------------------------------------------------------- */
22262 +void au_pin_hdir_unlock(struct au_pin *p)
22265 + au_hn_inode_unlock(p->hdir);
22268 +int au_pin_hdir_lock(struct au_pin *p)
22276 + /* even if an error happens later, keep this lock */
22277 + au_hn_inode_lock_nested(p->hdir, p->lsc_hi);
22280 + if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
22285 + err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
22286 + p->h_parent, p->br);
22292 +int au_pin_hdir_relock(struct au_pin *p)
22295 + struct inode *h_i;
22296 + struct dentry *h_d[] = {
22301 + err = au_pin_hdir_lock(p);
22302 + if (unlikely(err))
22305 + for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
22308 + if (d_is_positive(h_d[i])) {
22309 + h_i = d_inode(h_d[i]);
22310 + err = !h_i->i_nlink;
22318 +static void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
22320 + atomic_long_set(&p->hdir->hi_inode->i_rwsem.owner, (long)task);
22323 +void au_pin_hdir_acquire_nest(struct au_pin *p)
22326 + rwsem_acquire_nest(&p->hdir->hi_inode->i_rwsem.dep_map,
22327 + p->lsc_hi, 0, NULL, _RET_IP_);
22328 + au_pin_hdir_set_owner(p, current);
22332 +void au_pin_hdir_release(struct au_pin *p)
22335 + au_pin_hdir_set_owner(p, p->task);
22336 + rwsem_release(&p->hdir->hi_inode->i_rwsem.dep_map, _RET_IP_);
22340 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
22342 + if (pin && pin->parent)
22343 + return au_h_dptr(pin->parent, pin->bindex);
22347 +void au_unpin(struct au_pin *p)
22350 + au_pin_hdir_unlock(p);
22351 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
22352 + vfsub_mnt_drop_write(p->h_mnt);
22356 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22357 + di_read_unlock(p->parent, AuLock_IR);
22358 + iput(p->hdir->hi_inode);
22360 + p->parent = NULL;
22363 + /* do not clear p->task */
22366 +int au_do_pin(struct au_pin *p)
22369 + struct super_block *sb;
22370 + struct inode *h_dir;
22373 + sb = p->dentry->d_sb;
22374 + p->br = au_sbr(sb, p->bindex);
22375 + if (IS_ROOT(p->dentry)) {
22376 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
22377 + p->h_mnt = au_br_mnt(p->br);
22378 + err = vfsub_mnt_want_write(p->h_mnt);
22379 + if (unlikely(err)) {
22380 + au_fclr_pin(p->flags, MNT_WRITE);
22387 + p->h_dentry = NULL;
22388 + if (p->bindex <= au_dbbot(p->dentry))
22389 + p->h_dentry = au_h_dptr(p->dentry, p->bindex);
22391 + p->parent = dget_parent(p->dentry);
22392 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22393 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
22396 + p->h_parent = au_h_dptr(p->parent, p->bindex);
22397 + p->hdir = au_hi(d_inode(p->parent), p->bindex);
22399 + h_dir = p->hdir->hi_inode;
22403 + * if DI_LOCKED is not set, then p->parent may be different
22404 + * and h_parent can be NULL.
22406 + if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
22408 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22409 + di_read_unlock(p->parent, AuLock_IR);
22411 + p->parent = NULL;
22415 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
22416 + p->h_mnt = au_br_mnt(p->br);
22417 + err = vfsub_mnt_want_write(p->h_mnt);
22418 + if (unlikely(err)) {
22419 + au_fclr_pin(p->flags, MNT_WRITE);
22420 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22421 + di_read_unlock(p->parent, AuLock_IR);
22423 + p->parent = NULL;
22429 + err = au_pin_hdir_lock(p);
22431 + goto out; /* success */
22436 + pr_err("err %d\n", err);
22437 + err = au_busy_or_stale();
22442 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
22443 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
22444 + unsigned int udba, unsigned char flags)
22446 + p->dentry = dentry;
22448 + p->lsc_di = lsc_di;
22449 + p->lsc_hi = lsc_hi;
22450 + p->flags = flags;
22451 + p->bindex = bindex;
22453 + p->parent = NULL;
22457 + p->h_dentry = NULL;
22458 + p->h_parent = NULL;
22460 + p->task = current;
22463 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
22464 + unsigned int udba, unsigned char flags)
22466 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
22468 + return au_do_pin(pin);
22471 +/* ---------------------------------------------------------------------- */
22474 + * ->setattr() and ->getattr() are called in various cases.
22475 + * chmod, stat: dentry is revalidated.
22476 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
22478 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
22480 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
22481 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
22484 + struct dentry *parent;
22487 + if (au_digen_test(dentry, sigen)) {
22488 + parent = dget_parent(dentry);
22489 + di_read_lock_parent(parent, AuLock_IR);
22490 + err = au_refresh_dentry(dentry, parent);
22491 + di_read_unlock(parent, AuLock_IR);
22499 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
22500 + struct au_icpup_args *a)
22504 + aufs_bindex_t btop, ibtop;
22505 + struct dentry *hi_wh, *parent;
22506 + struct inode *inode;
22507 + struct au_wr_dir_args wr_dir_args = {
22508 + .force_btgt = -1,
22512 + if (d_is_dir(dentry))
22513 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
22514 + /* plink or hi_wh() case */
22515 + btop = au_dbtop(dentry);
22516 + inode = d_inode(dentry);
22517 + ibtop = au_ibtop(inode);
22518 + if (btop != ibtop && !au_test_ro(inode->i_sb, ibtop, inode))
22519 + wr_dir_args.force_btgt = ibtop;
22520 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
22521 + if (unlikely(err < 0))
22525 + au_fset_icpup(a->flags, DID_CPUP);
22528 + a->pin_flags = AuPin_MNT_WRITE;
22530 + if (!IS_ROOT(dentry)) {
22531 + au_fset_pin(a->pin_flags, DI_LOCKED);
22532 + parent = dget_parent(dentry);
22533 + di_write_lock_parent(parent);
22536 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
22537 + if (unlikely(err))
22541 + a->h_path.dentry = au_h_dptr(dentry, btop);
22542 + a->h_inode = d_inode(a->h_path.dentry);
22543 + if (ia && (ia->ia_valid & ATTR_SIZE)) {
22544 + inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
22545 + if (ia->ia_size < i_size_read(a->h_inode))
22546 + sz = ia->ia_size;
22547 + inode_unlock_shared(a->h_inode);
22551 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
22552 + hi_wh = au_hi_wh(inode, a->btgt);
22554 + struct au_cp_generic cpg = {
22555 + .dentry = dentry,
22561 + err = au_sio_cpup_wh(&cpg, /*file*/NULL);
22562 + if (unlikely(err))
22564 + hi_wh = au_hi_wh(inode, a->btgt);
22565 + /* todo: revalidate hi_wh? */
22570 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
22571 + di_downgrade_lock(parent, AuLock_IR);
22575 + if (!au_ftest_icpup(a->flags, DID_CPUP))
22576 + goto out; /* success */
22578 + if (!d_unhashed(dentry)) {
22579 + struct au_cp_generic cpg = {
22580 + .dentry = dentry,
22585 + .flags = AuCpup_DTIME | AuCpup_HOPEN
22587 + err = au_sio_cpup_simple(&cpg);
22589 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22590 + } else if (!hi_wh)
22591 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22593 + a->h_path.dentry = hi_wh; /* do not dget here */
22596 + a->h_inode = d_inode(a->h_path.dentry);
22598 + goto out; /* success */
22599 + au_unpin(&a->pin);
22602 + di_write_unlock(parent);
22607 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22611 +static int aufs_setattr(struct user_namespace *userns, struct dentry *dentry,
22612 + struct iattr *ia)
22615 + struct inode *inode, *delegated;
22616 + struct super_block *sb;
22617 + struct file *file;
22618 + struct au_icpup_args *a;
22619 + struct user_namespace *h_userns;
22621 + inode = d_inode(dentry);
22622 + IMustLock(inode);
22624 + err = setattr_prepare(userns, dentry, ia);
22625 + if (unlikely(err))
22629 + a = kzalloc(sizeof(*a), GFP_NOFS);
22630 + if (unlikely(!a))
22633 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
22634 + ia->ia_valid &= ~ATTR_MODE;
22637 + sb = dentry->d_sb;
22638 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22639 + if (unlikely(err))
22642 + if (ia->ia_valid & ATTR_FILE) {
22643 + /* currently ftruncate(2) only */
22644 + AuDebugOn(!d_is_reg(dentry));
22645 + file = ia->ia_file;
22646 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1,
22648 + if (unlikely(err))
22650 + ia->ia_file = au_hf_top(file);
22651 + a->udba = AuOpt_UDBA_NONE;
22653 + /* fchmod() doesn't pass ia_file */
22654 + a->udba = au_opt_udba(sb);
22655 + di_write_lock_child(dentry);
22656 + /* no d_unlinked(), to set UDBA_NONE for root */
22657 + if (d_unhashed(dentry))
22658 + a->udba = AuOpt_UDBA_NONE;
22659 + if (a->udba != AuOpt_UDBA_NONE) {
22660 + AuDebugOn(IS_ROOT(dentry));
22661 + err = au_reval_for_attr(dentry, au_sigen(sb));
22662 + if (unlikely(err))
22667 + err = au_pin_and_icpup(dentry, ia, a);
22668 + if (unlikely(err < 0))
22670 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
22671 + ia->ia_file = NULL;
22672 + ia->ia_valid &= ~ATTR_FILE;
22675 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
22676 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
22677 + == (ATTR_MODE | ATTR_CTIME)) {
22678 + err = security_path_chmod(&a->h_path, ia->ia_mode);
22679 + if (unlikely(err))
22681 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
22682 + && (ia->ia_valid & ATTR_CTIME)) {
22683 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
22684 + if (unlikely(err))
22688 + if (ia->ia_valid & ATTR_SIZE) {
22691 + if (ia->ia_size < i_size_read(inode))
22693 + truncate_setsize(inode, ia->ia_size);
22696 + if (ia->ia_valid & ATTR_FILE)
22698 + inode_unlock(a->h_inode);
22699 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
22700 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22702 + delegated = NULL;
22704 + err = vfsub_notify_change(&a->h_path, ia, &delegated);
22706 + err = break_deleg_wait(&delegated);
22714 + * regardless aufs 'acl' option setting.
22715 + * why don't all acl-aware fs call this func from their ->setattr()?
22717 + if (!err && (ia->ia_valid & ATTR_MODE)) {
22718 + h_userns = mnt_user_ns(a->h_path.mnt);
22719 + err = vfsub_acl_chmod(h_userns, a->h_inode, ia->ia_mode);
22722 + au_cpup_attr_changeable(inode);
22725 + inode_unlock(a->h_inode);
22726 + au_unpin(&a->pin);
22727 + if (unlikely(err))
22728 + au_update_dbtop(dentry);
22730 + di_write_unlock(dentry);
22732 + fi_write_unlock(file);
22733 + ia->ia_file = file;
22734 + ia->ia_valid |= ATTR_FILE;
22737 + si_read_unlock(sb);
22745 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
22746 +static int au_h_path_to_set_attr(struct dentry *dentry,
22747 + struct au_icpup_args *a, struct path *h_path)
22750 + struct super_block *sb;
22752 + sb = dentry->d_sb;
22753 + a->udba = au_opt_udba(sb);
22754 + /* no d_unlinked(), to set UDBA_NONE for root */
22755 + if (d_unhashed(dentry))
22756 + a->udba = AuOpt_UDBA_NONE;
22757 + if (a->udba != AuOpt_UDBA_NONE) {
22758 + AuDebugOn(IS_ROOT(dentry));
22759 + err = au_reval_for_attr(dentry, au_sigen(sb));
22760 + if (unlikely(err))
22763 + err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
22764 + if (unlikely(err < 0))
22767 + h_path->dentry = a->h_path.dentry;
22768 + h_path->mnt = au_sbr_mnt(sb, a->btgt);
22774 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
22775 + struct au_sxattr *arg)
22778 + struct path h_path;
22779 + struct super_block *sb;
22780 + struct au_icpup_args *a;
22781 + struct inode *h_inode;
22782 + struct user_namespace *h_userns;
22784 + IMustLock(inode);
22787 + a = kzalloc(sizeof(*a), GFP_NOFS);
22788 + if (unlikely(!a))
22791 + sb = dentry->d_sb;
22792 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22793 + if (unlikely(err))
22796 + h_path.dentry = NULL; /* silence gcc */
22797 + di_write_lock_child(dentry);
22798 + err = au_h_path_to_set_attr(dentry, a, &h_path);
22799 + if (unlikely(err))
22801 + h_userns = mnt_user_ns(h_path.mnt);
22803 + inode_unlock(a->h_inode);
22804 + switch (arg->type) {
22805 + case AU_XATTR_SET:
22806 + AuDebugOn(d_is_negative(h_path.dentry));
22807 + err = vfsub_setxattr(h_userns, h_path.dentry,
22808 + arg->u.set.name, arg->u.set.value,
22809 + arg->u.set.size, arg->u.set.flags);
22812 + err = -EOPNOTSUPP;
22813 + h_inode = d_inode(h_path.dentry);
22814 + if (h_inode->i_op->set_acl) {
22815 + /* this will call posix_acl_update_mode */
22816 + err = h_inode->i_op->set_acl(h_userns, h_inode,
22817 + arg->u.acl_set.acl,
22818 + arg->u.acl_set.type);
22823 + au_cpup_attr_timesizes(inode);
22825 + au_unpin(&a->pin);
22826 + if (unlikely(err))
22827 + au_update_dbtop(dentry);
22830 + di_write_unlock(dentry);
22831 + si_read_unlock(sb);
22840 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
22841 + unsigned int nlink)
22845 + inode->i_mode = st->mode;
22846 + /* don't i_[ug]id_write() here */
22847 + inode->i_uid = st->uid;
22848 + inode->i_gid = st->gid;
22849 + inode->i_atime = st->atime;
22850 + inode->i_mtime = st->mtime;
22851 + inode->i_ctime = st->ctime;
22853 + au_cpup_attr_nlink(inode, /*force*/0);
22854 + if (S_ISDIR(inode->i_mode)) {
22855 + n = inode->i_nlink;
22858 + smp_mb(); /* for i_nlink */
22859 + /* 0 can happen */
22860 + set_nlink(inode, n);
22863 + spin_lock(&inode->i_lock);
22864 + inode->i_blocks = st->blocks;
22865 + i_size_write(inode, st->size);
22866 + spin_unlock(&inode->i_lock);
22870 + * common routine for aufs_getattr() and au_getxattr().
22871 + * returns zero or negative (an error).
22872 + * @dentry will be read-locked in success.
22874 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
22875 + struct path *h_path, int locked)
22878 + unsigned int mnt_flags, sigen;
22879 + unsigned char udba_none;
22880 + aufs_bindex_t bindex;
22881 + struct super_block *sb, *h_sb;
22883 + h_path->mnt = NULL;
22884 + h_path->dentry = NULL;
22887 + sb = dentry->d_sb;
22888 + mnt_flags = au_mntflags(sb);
22889 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
22891 + if (unlikely(locked))
22892 + goto body; /* skip locking dinfo */
22894 + /* support fstat(2) */
22895 + if (!d_unlinked(dentry) && !udba_none) {
22896 + sigen = au_sigen(sb);
22897 + err = au_digen_test(dentry, sigen);
22899 + di_read_lock_child(dentry, AuLock_IR);
22900 + err = au_dbrange_test(dentry);
22901 + if (unlikely(err)) {
22902 + di_read_unlock(dentry, AuLock_IR);
22906 + AuDebugOn(IS_ROOT(dentry));
22907 + di_write_lock_child(dentry);
22908 + err = au_dbrange_test(dentry);
22910 + err = au_reval_for_attr(dentry, sigen);
22912 + di_downgrade_lock(dentry, AuLock_IR);
22914 + di_write_unlock(dentry);
22919 + di_read_lock_child(dentry, AuLock_IR);
22923 + inode = d_inode(dentry);
22924 + if (unlikely(!inode))
22927 + bindex = au_ibtop(inode);
22928 + h_path->mnt = au_sbr_mnt(sb, bindex);
22929 + h_sb = h_path->mnt->mnt_sb;
22931 + && !au_test_fs_bad_iattr(h_sb)
22933 + goto out; /* success */
22935 + if (au_dbtop(dentry) == bindex)
22936 + h_path->dentry = au_h_dptr(dentry, bindex);
22937 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
22938 + h_path->dentry = au_plink_lkup(inode, bindex);
22939 + if (IS_ERR(h_path->dentry))
22940 + /* pretending success */
22941 + h_path->dentry = NULL;
22943 + dput(h_path->dentry);
22950 +static int aufs_getattr(struct user_namespace *userns, const struct path *path,
22951 + struct kstat *st, u32 request, unsigned int query)
22954 + unsigned char positive;
22955 + struct path h_path;
22956 + struct dentry *dentry;
22957 + struct inode *inode;
22958 + struct super_block *sb;
22960 + dentry = path->dentry;
22961 + inode = d_inode(dentry);
22962 + sb = dentry->d_sb;
22963 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22964 + if (unlikely(err))
22966 + err = au_h_path_getattr(dentry, /*inode*/NULL, /*force*/0, &h_path,
22968 + if (unlikely(err))
22970 + if (unlikely(!h_path.dentry))
22971 + /* illegally overlapped or something */
22972 + goto out_fill; /* pretending success */
22974 + positive = d_is_positive(h_path.dentry);
22976 + /* no vfsub version */
22977 + err = vfs_getattr(&h_path, st, request, query);
22980 + au_refresh_iattr(inode, st,
22981 + d_inode(h_path.dentry)->i_nlink);
22982 + goto out_fill; /* success */
22988 + generic_fillattr(userns, inode, st);
22990 + di_read_unlock(dentry, AuLock_IR);
22992 + si_read_unlock(sb);
22998 +/* ---------------------------------------------------------------------- */
23000 +static const char *aufs_get_link(struct dentry *dentry, struct inode *inode,
23001 + struct delayed_call *done)
23004 + struct dentry *h_dentry;
23005 + struct inode *h_inode;
23007 + aufs_bindex_t bindex;
23009 + ret = NULL; /* suppress a warning */
23014 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
23015 + if (unlikely(err))
23018 + err = au_d_hashed_positive(dentry);
23019 + if (unlikely(err))
23023 + inode = d_inode(dentry);
23024 + bindex = au_ibtop(inode);
23025 + h_inode = au_h_iptr(inode, bindex);
23026 + if (unlikely(!h_inode->i_op->get_link))
23031 + if (au_dbtop(dentry) <= bindex) {
23032 + h_dentry = au_h_dptr(dentry, bindex);
23037 + h_dentry = d_find_any_alias(h_inode);
23038 + if (IS_ERR(h_dentry)) {
23039 + err = PTR_ERR(h_dentry);
23043 + if (unlikely(!h_dentry))
23047 + AuDbg("%ps\n", h_inode->i_op->get_link);
23048 + AuDbgDentry(h_dentry);
23049 + ret = vfs_get_link(h_dentry, done);
23052 + err = PTR_ERR(ret);
23055 + aufs_read_unlock(dentry, AuLock_IR);
23057 + if (unlikely(err))
23058 + ret = ERR_PTR(err);
23059 + AuTraceErrPtr(ret);
23063 +/* ---------------------------------------------------------------------- */
23065 +static int au_is_special(struct inode *inode)
23067 + return (inode->i_mode & (S_IFBLK | S_IFCHR | S_IFIFO | S_IFSOCK));
23070 +static int aufs_update_time(struct inode *inode, struct timespec64 *ts,
23074 + aufs_bindex_t bindex;
23075 + struct super_block *sb;
23076 + struct inode *h_inode;
23077 + struct vfsmount *h_mnt;
23079 + sb = inode->i_sb;
23080 + WARN_ONCE((flags & S_ATIME) && !IS_NOATIME(inode),
23081 + "unexpected s_flags 0x%lx", sb->s_flags);
23083 + /* mmap_sem might be acquired already, cf. aufs_mmap() */
23085 + si_read_lock(sb, AuLock_FLUSH);
23086 + ii_write_lock_child(inode);
23089 + bindex = au_ibtop(inode);
23090 + h_inode = au_h_iptr(inode, bindex);
23091 + if (!au_test_ro(sb, bindex, inode)) {
23092 + h_mnt = au_sbr_mnt(sb, bindex);
23093 + err = vfsub_mnt_want_write(h_mnt);
23095 + err = vfsub_update_time(h_inode, ts, flags);
23096 + vfsub_mnt_drop_write(h_mnt);
23098 + } else if (au_is_special(h_inode)) {
23100 + * Never copy-up here.
23101 + * These special files may already be opened and used for
23102 + * communicating. If we copied it up, then the communication
23103 + * would be corrupted.
23105 + AuWarn1("timestamps for i%lu are ignored "
23106 + "since it is on readonly branch (hi%lu).\n",
23107 + inode->i_ino, h_inode->i_ino);
23108 + } else if (flags & ~S_ATIME) {
23110 + AuIOErr1("unexpected flags 0x%x\n", flags);
23115 + au_cpup_attr_timesizes(inode);
23116 + ii_write_unlock(inode);
23117 + si_read_unlock(sb);
23120 + if (!err && (flags & S_VERSION))
23121 + inode_inc_iversion(inode);
23126 +/* ---------------------------------------------------------------------- */
23128 +/* no getattr version will be set by module.c:aufs_init() */
23129 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
23131 + [AuIop_SYMLINK] = {
23132 + .permission = aufs_permission,
23133 +#ifdef CONFIG_FS_POSIX_ACL
23134 + .get_acl = aufs_get_acl,
23135 + .set_acl = aufs_set_acl, /* unsupport for symlink? */
23138 + .setattr = aufs_setattr,
23139 + .getattr = aufs_getattr,
23141 +#ifdef CONFIG_AUFS_XATTR
23142 + .listxattr = aufs_listxattr,
23145 + .get_link = aufs_get_link,
23147 + /* .update_time = aufs_update_time */
23150 + .create = aufs_create,
23151 + .lookup = aufs_lookup,
23152 + .link = aufs_link,
23153 + .unlink = aufs_unlink,
23154 + .symlink = aufs_symlink,
23155 + .mkdir = aufs_mkdir,
23156 + .rmdir = aufs_rmdir,
23157 + .mknod = aufs_mknod,
23158 + .rename = aufs_rename,
23160 + .permission = aufs_permission,
23161 +#ifdef CONFIG_FS_POSIX_ACL
23162 + .get_acl = aufs_get_acl,
23163 + .set_acl = aufs_set_acl,
23166 + .setattr = aufs_setattr,
23167 + .getattr = aufs_getattr,
23169 +#ifdef CONFIG_AUFS_XATTR
23170 + .listxattr = aufs_listxattr,
23173 + .update_time = aufs_update_time,
23174 + .atomic_open = aufs_atomic_open,
23175 + .tmpfile = aufs_tmpfile
23177 + [AuIop_OTHER] = {
23178 + .permission = aufs_permission,
23179 +#ifdef CONFIG_FS_POSIX_ACL
23180 + .get_acl = aufs_get_acl,
23181 + .set_acl = aufs_set_acl,
23184 + .setattr = aufs_setattr,
23185 + .getattr = aufs_getattr,
23187 +#ifdef CONFIG_AUFS_XATTR
23188 + .listxattr = aufs_listxattr,
23191 + .update_time = aufs_update_time
23194 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
23195 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
23196 +++ linux/fs/aufs/i_op_del.c 2022-03-21 14:49:05.726633010 +0100
23198 +// SPDX-License-Identifier: GPL-2.0
23200 + * Copyright (C) 2005-2021 Junjiro R. Okajima
23202 + * This program, aufs is free software; you can redistribute it and/or modify
23203 + * it under the terms of the GNU General Public License as published by
23204 + * the Free Software Foundation; either version 2 of the License, or
23205 + * (at your option) any later version.
23207 + * This program is distributed in the hope that it will be useful,
23208 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23209 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23210 + * GNU General Public License for more details.
23212 + * You should have received a copy of the GNU General Public License
23213 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23217 + * inode operations (del entry)
23220 +#include <linux/iversion.h>
23224 + * decide if a new whiteout for @dentry is necessary or not.
23225 + * when it is necessary, prepare the parent dir for the upper branch whose
23226 + * branch index is @bcpup for creation. the actual creation of the whiteout will
23227 + * be done by caller.
23229 + * 0: wh is unnecessary
23230 + * plus: wh is necessary
23233 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
23235 + int need_wh, err;
23236 + aufs_bindex_t btop;
23237 + struct super_block *sb;
23239 + sb = dentry->d_sb;
23240 + btop = au_dbtop(dentry);
23241 + if (*bcpup < 0) {
23243 + if (au_test_ro(sb, btop, d_inode(dentry))) {
23244 + err = AuWbrCopyup(au_sbi(sb), dentry);
23246 + if (unlikely(err < 0))
23250 + AuDebugOn(btop < *bcpup
23251 + || au_test_ro(sb, *bcpup, d_inode(dentry)));
23252 + AuDbg("bcpup %d, btop %d\n", *bcpup, btop);
23254 + if (*bcpup != btop) {
23255 + err = au_cpup_dirs(dentry, *bcpup);
23256 + if (unlikely(err))
23260 + struct au_dinfo *dinfo, *tmp;
23262 + need_wh = -ENOMEM;
23263 + dinfo = au_di(dentry);
23264 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
23266 + au_di_cp(tmp, dinfo);
23267 + au_di_swap(tmp, dinfo);
23268 + /* returns the number of positive dentries */
23269 + need_wh = au_lkup_dentry(dentry, btop + 1,
23270 + /* AuLkup_IGNORE_PERM */ 0);
23271 + au_di_swap(tmp, dinfo);
23272 + au_rw_write_unlock(&tmp->di_rwsem);
23276 + AuDbg("need_wh %d\n", need_wh);
23284 + * simple tests for the del-entry operations.
23285 + * following the checks in vfs, plus the parent-child relationship.
23287 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
23288 + struct dentry *h_parent, int isdir)
23292 + struct dentry *h_dentry, *h_latest;
23293 + struct inode *h_inode;
23294 + struct path h_ppath;
23295 + struct super_block *sb;
23296 + struct au_branch *br;
23297 + struct user_namespace *h_userns;
23299 + h_dentry = au_h_dptr(dentry, bindex);
23300 + if (d_really_is_positive(dentry)) {
23302 + if (unlikely(d_is_negative(h_dentry)))
23304 + h_inode = d_inode(h_dentry);
23305 + if (unlikely(!h_inode->i_nlink))
23308 + h_mode = h_inode->i_mode;
23311 + if (unlikely(S_ISDIR(h_mode)))
23313 + } else if (unlikely(!S_ISDIR(h_mode))) {
23318 + /* rename(2) case */
23320 + if (unlikely(d_is_positive(h_dentry)))
23325 + /* expected parent dir is locked */
23326 + if (unlikely(h_parent != h_dentry->d_parent))
23331 + * rmdir a dir may break the consistency on some filesystem.
23332 + * let's try heavy test.
23335 + sb = dentry->d_sb;
23336 + br = au_sbr(sb, bindex);
23337 + h_userns = au_br_userns(br);
23338 + if (unlikely(!au_opt_test(au_mntflags(sb), DIRPERM1)
23339 + && au_test_h_perm(h_userns, d_inode(h_parent),
23340 + MAY_EXEC | MAY_WRITE)))
23343 + h_ppath.dentry = h_parent;
23344 + h_ppath.mnt = au_br_mnt(br);
23345 + h_latest = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
23347 + if (IS_ERR(h_latest))
23349 + if (h_latest == h_dentry)
23358 + * decide the branch where we operate for @dentry. the branch index will be set
23359 + * @rbcpup. after deciding it, 'pin' it and store the timestamps of the parent
23360 + * dir for reverting.
23361 + * when a new whiteout is necessary, create it.
23363 +static struct dentry*
23364 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
23365 + struct au_dtime *dt, struct au_pin *pin)
23367 + struct dentry *wh_dentry;
23368 + struct super_block *sb;
23369 + struct path h_path;
23370 + int err, need_wh;
23371 + unsigned int udba;
23372 + aufs_bindex_t bcpup;
23374 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
23375 + wh_dentry = ERR_PTR(need_wh);
23376 + if (unlikely(need_wh < 0))
23379 + sb = dentry->d_sb;
23380 + udba = au_opt_udba(sb);
23382 + err = au_pin(pin, dentry, bcpup, udba,
23383 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23384 + wh_dentry = ERR_PTR(err);
23385 + if (unlikely(err))
23388 + h_path.dentry = au_pinned_h_parent(pin);
23389 + if (udba != AuOpt_UDBA_NONE
23390 + && au_dbtop(dentry) == bcpup) {
23391 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
23392 + wh_dentry = ERR_PTR(err);
23393 + if (unlikely(err))
23397 + h_path.mnt = au_sbr_mnt(sb, bcpup);
23398 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
23399 + wh_dentry = NULL;
23401 + goto out; /* success, no need to create whiteout */
23403 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
23404 + if (IS_ERR(wh_dentry))
23407 + /* returns with the parent is locked and wh_dentry is dget-ed */
23408 + goto out; /* success */
23413 + return wh_dentry;
23417 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
23418 + * in order to be revertible and save time for removing many child whiteouts
23420 + * returns 1 when there are too many child whiteout and caller should remove
23421 + * them asynchronously. returns 0 when the number of children is enough small to
23422 + * remove now or the branch fs is a remote fs.
23423 + * otherwise return an error.
23425 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
23426 + struct au_nhash *whlist, struct inode *dir)
23428 + int rmdir_later, err, dirwh;
23429 + struct dentry *h_dentry;
23430 + struct super_block *sb;
23431 + struct inode *inode;
23433 + sb = dentry->d_sb;
23434 + SiMustAnyLock(sb);
23435 + h_dentry = au_h_dptr(dentry, bindex);
23436 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
23437 + if (unlikely(err))
23440 + /* stop monitoring */
23441 + inode = d_inode(dentry);
23442 + au_hn_free(au_hi(inode, bindex));
23444 + if (!au_test_fs_remote(h_dentry->d_sb)) {
23445 + dirwh = au_sbi(sb)->si_dirwh;
23446 + rmdir_later = (dirwh <= 1);
23447 + if (!rmdir_later)
23448 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
23451 + return rmdir_later;
23454 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
23455 + if (unlikely(err)) {
23456 + AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
23457 + h_dentry, bindex, err);
23467 + * final procedure for deleting a entry.
23468 + * maintain dentry and iattr.
23470 +static void epilog(struct inode *dir, struct dentry *dentry,
23471 + aufs_bindex_t bindex)
23473 + struct inode *inode;
23475 + inode = d_inode(dentry);
23477 + inode->i_ctime = dir->i_ctime;
23479 + au_dir_ts(dir, bindex);
23480 + inode_inc_iversion(dir);
23484 + * when an error happened, remove the created whiteout and revert everything.
23486 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
23487 + aufs_bindex_t bwh, struct dentry *wh_dentry,
23488 + struct dentry *dentry, struct au_dtime *dt)
23491 + struct path h_path = {
23492 + .dentry = wh_dentry,
23493 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
23496 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
23498 + au_set_dbwh(dentry, bwh);
23499 + au_dtime_revert(dt);
23503 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
23507 +/* ---------------------------------------------------------------------- */
23509 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
23512 + aufs_bindex_t bwh, bindex, btop;
23513 + struct inode *inode, *h_dir, *delegated;
23514 + struct dentry *parent, *wh_dentry;
23515 + /* to reduce stack size */
23517 + struct au_dtime dt;
23518 + struct au_pin pin;
23519 + struct path h_path;
23525 + a = kmalloc(sizeof(*a), GFP_NOFS);
23526 + if (unlikely(!a))
23529 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
23530 + if (unlikely(err))
23532 + err = au_d_hashed_positive(dentry);
23533 + if (unlikely(err))
23535 + inode = d_inode(dentry);
23536 + IMustLock(inode);
23538 + if (unlikely(d_is_dir(dentry)))
23539 + goto out_unlock; /* possible? */
23541 + btop = au_dbtop(dentry);
23542 + bwh = au_dbwh(dentry);
23544 + parent = dentry->d_parent; /* dir inode is locked */
23545 + di_write_lock_parent(parent);
23546 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
23548 + err = PTR_ERR(wh_dentry);
23549 + if (IS_ERR(wh_dentry))
23552 + a->h_path.mnt = au_sbr_mnt(dentry->d_sb, btop);
23553 + a->h_path.dentry = au_h_dptr(dentry, btop);
23554 + dget(a->h_path.dentry);
23555 + if (bindex == btop) {
23556 + h_dir = au_pinned_h_dir(&a->pin);
23557 + delegated = NULL;
23558 + err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
23559 + if (unlikely(err == -EWOULDBLOCK)) {
23560 + pr_warn("cannot retry for NFSv4 delegation"
23561 + " for an internal unlink\n");
23565 + /* dir inode is locked */
23566 + h_dir = d_inode(wh_dentry->d_parent);
23567 + IMustLock(h_dir);
23572 + vfsub_drop_nlink(inode);
23573 + epilog(dir, dentry, bindex);
23575 + /* update target timestamps */
23576 + if (bindex == btop) {
23577 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
23579 + inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
23581 + /* todo: this timestamp may be reverted later */
23582 + inode->i_ctime = h_dir->i_ctime;
23583 + goto out_unpin; /* success */
23590 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23597 + au_unpin(&a->pin);
23599 + dput(a->h_path.dentry);
23601 + di_write_unlock(parent);
23603 + aufs_read_unlock(dentry, AuLock_DW);
23610 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
23612 + int err, rmdir_later;
23613 + aufs_bindex_t bwh, bindex, btop;
23614 + struct inode *inode;
23615 + struct dentry *parent, *wh_dentry, *h_dentry;
23616 + struct au_whtmp_rmdir *args;
23617 + /* to reduce stack size */
23619 + struct au_dtime dt;
23620 + struct au_pin pin;
23626 + a = kmalloc(sizeof(*a), GFP_NOFS);
23627 + if (unlikely(!a))
23630 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
23631 + if (unlikely(err))
23633 + err = au_alive_dir(dentry);
23634 + if (unlikely(err))
23636 + inode = d_inode(dentry);
23637 + IMustLock(inode);
23639 + if (unlikely(!d_is_dir(dentry)))
23640 + goto out_unlock; /* possible? */
23643 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
23644 + if (unlikely(!args))
23647 + parent = dentry->d_parent; /* dir inode is locked */
23648 + di_write_lock_parent(parent);
23649 + err = au_test_empty(dentry, &args->whlist);
23650 + if (unlikely(err))
23653 + btop = au_dbtop(dentry);
23654 + bwh = au_dbwh(dentry);
23656 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
23658 + err = PTR_ERR(wh_dentry);
23659 + if (IS_ERR(wh_dentry))
23662 + h_dentry = au_h_dptr(dentry, btop);
23665 + if (bindex == btop) {
23666 + err = renwh_and_rmdir(dentry, btop, &args->whlist, dir);
23668 + rmdir_later = err;
23672 + /* stop monitoring */
23673 + au_hn_free(au_hi(inode, btop));
23675 + /* dir inode is locked */
23676 + IMustLock(d_inode(wh_dentry->d_parent));
23681 + vfsub_dead_dir(inode);
23682 + au_set_dbdiropq(dentry, -1);
23683 + epilog(dir, dentry, bindex);
23685 + if (rmdir_later) {
23686 + au_whtmp_kick_rmdir(dir, btop, h_dentry, args);
23690 + goto out_unpin; /* success */
23698 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23705 + au_unpin(&a->pin);
23709 + di_write_unlock(parent);
23711 + au_whtmp_rmdir_free(args);
23713 + aufs_read_unlock(dentry, AuLock_DW);
23720 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
23721 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
23722 +++ linux/fs/aufs/i_op_ren.c 2022-03-21 14:49:05.726633010 +0100
23724 +// SPDX-License-Identifier: GPL-2.0
23726 + * Copyright (C) 2005-2021 Junjiro R. Okajima
23728 + * This program, aufs is free software; you can redistribute it and/or modify
23729 + * it under the terms of the GNU General Public License as published by
23730 + * the Free Software Foundation; either version 2 of the License, or
23731 + * (at your option) any later version.
23733 + * This program is distributed in the hope that it will be useful,
23734 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23735 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23736 + * GNU General Public License for more details.
23738 + * You should have received a copy of the GNU General Public License
23739 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23743 + * inode operation (rename entry)
23744 + * todo: this is crazy monster
23747 +#include <linux/iversion.h>
23750 +enum { AuSRC, AuDST, AuSrcDst };
23751 +enum { AuPARENT, AuCHILD, AuParentChild };
23753 +#define AuRen_ISDIR_SRC 1
23754 +#define AuRen_ISDIR_DST (1 << 1)
23755 +#define AuRen_ISSAMEDIR (1 << 2)
23756 +#define AuRen_WHSRC (1 << 3)
23757 +#define AuRen_WHDST (1 << 4)
23758 +#define AuRen_MNT_WRITE (1 << 5)
23759 +#define AuRen_DT_DSTDIR (1 << 6)
23760 +#define AuRen_DIROPQ_SRC (1 << 7)
23761 +#define AuRen_DIROPQ_DST (1 << 8)
23762 +#define AuRen_DIRREN (1 << 9)
23763 +#define AuRen_DROPPED_SRC (1 << 10)
23764 +#define AuRen_DROPPED_DST (1 << 11)
23765 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
23766 +#define au_fset_ren(flags, name) \
23767 + do { (flags) |= AuRen_##name; } while (0)
23768 +#define au_fclr_ren(flags, name) \
23769 + do { (flags) &= ~AuRen_##name; } while (0)
23771 +#ifndef CONFIG_AUFS_DIRREN
23772 +#undef AuRen_DIRREN
23773 +#define AuRen_DIRREN 0
23776 +struct au_ren_args {
23778 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
23780 + struct inode *dir, *inode;
23781 + struct au_hinode *hdir, *hinode;
23782 + struct au_dtime dt[AuParentChild];
23783 + aufs_bindex_t btop, bdiropq;
23786 +#define src_dentry sd[AuSRC].dentry
23787 +#define src_dir sd[AuSRC].dir
23788 +#define src_inode sd[AuSRC].inode
23789 +#define src_h_dentry sd[AuSRC].h_dentry
23790 +#define src_parent sd[AuSRC].parent
23791 +#define src_h_parent sd[AuSRC].h_parent
23792 +#define src_wh_dentry sd[AuSRC].wh_dentry
23793 +#define src_hdir sd[AuSRC].hdir
23794 +#define src_hinode sd[AuSRC].hinode
23795 +#define src_h_dir sd[AuSRC].hdir->hi_inode
23796 +#define src_dt sd[AuSRC].dt
23797 +#define src_btop sd[AuSRC].btop
23798 +#define src_bdiropq sd[AuSRC].bdiropq
23800 +#define dst_dentry sd[AuDST].dentry
23801 +#define dst_dir sd[AuDST].dir
23802 +#define dst_inode sd[AuDST].inode
23803 +#define dst_h_dentry sd[AuDST].h_dentry
23804 +#define dst_parent sd[AuDST].parent
23805 +#define dst_h_parent sd[AuDST].h_parent
23806 +#define dst_wh_dentry sd[AuDST].wh_dentry
23807 +#define dst_hdir sd[AuDST].hdir
23808 +#define dst_hinode sd[AuDST].hinode
23809 +#define dst_h_dir sd[AuDST].hdir->hi_inode
23810 +#define dst_dt sd[AuDST].dt
23811 +#define dst_btop sd[AuDST].btop
23812 +#define dst_bdiropq sd[AuDST].bdiropq
23814 + struct dentry *h_trap;
23815 + struct au_branch *br;
23816 + struct path h_path;
23817 + struct au_nhash whlist;
23818 + aufs_bindex_t btgt, src_bwh;
23821 + unsigned short auren_flags;
23822 + unsigned char flags; /* syscall parameter */
23823 + unsigned char exchange;
23826 + struct au_whtmp_rmdir *thargs;
23827 + struct dentry *h_dst;
23828 + struct au_hinode *h_root;
23831 +/* ---------------------------------------------------------------------- */
23834 + * functions for reverting.
23835 + * when an error happened in a single rename systemcall, we should revert
23836 + * everything as if nothing happened.
23837 + * we don't need to revert the copied-up/down the parent dir since they are
23841 +#define RevertFailure(fmt, ...) do { \
23842 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
23843 + ##__VA_ARGS__, err, rerr); \
23847 +static void au_ren_do_rev_diropq(int err, struct au_ren_args *a, int idx)
23850 + struct dentry *d;
23851 +#define src_or_dst(member) a->sd[idx].member
23853 + d = src_or_dst(dentry); /* {src,dst}_dentry */
23854 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
23855 + rerr = au_diropq_remove(d, a->btgt);
23856 + au_hn_inode_unlock(src_or_dst(hinode));
23857 + au_set_dbdiropq(d, src_or_dst(bdiropq));
23859 + RevertFailure("remove diropq %pd", d);
23861 +#undef src_or_dst_
23864 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
23866 + if (au_ftest_ren(a->auren_flags, DIROPQ_SRC))
23867 + au_ren_do_rev_diropq(err, a, AuSRC);
23868 + if (au_ftest_ren(a->auren_flags, DIROPQ_DST))
23869 + au_ren_do_rev_diropq(err, a, AuDST);
23872 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
23875 + struct inode *delegated;
23876 + struct path h_ppath = {
23877 + .dentry = a->src_h_parent,
23878 + .mnt = a->h_path.mnt
23881 + a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name, &h_ppath);
23882 + rerr = PTR_ERR(a->h_path.dentry);
23883 + if (IS_ERR(a->h_path.dentry)) {
23884 + RevertFailure("lkup one %pd", a->src_dentry);
23888 + delegated = NULL;
23889 + rerr = vfsub_rename(a->dst_h_dir,
23890 + au_h_dptr(a->src_dentry, a->btgt),
23891 + a->src_h_dir, &a->h_path, &delegated, a->flags);
23892 + if (unlikely(rerr == -EWOULDBLOCK)) {
23893 + pr_warn("cannot retry for NFSv4 delegation"
23894 + " for an internal rename\n");
23897 + d_drop(a->h_path.dentry);
23898 + dput(a->h_path.dentry);
23899 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
23901 + RevertFailure("rename %pd", a->src_dentry);
23904 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
23907 + struct inode *delegated;
23908 + struct path h_ppath = {
23909 + .dentry = a->dst_h_parent,
23910 + .mnt = a->h_path.mnt
23913 + a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name, &h_ppath);
23914 + rerr = PTR_ERR(a->h_path.dentry);
23915 + if (IS_ERR(a->h_path.dentry)) {
23916 + RevertFailure("lkup one %pd", a->dst_dentry);
23919 + if (d_is_positive(a->h_path.dentry)) {
23920 + d_drop(a->h_path.dentry);
23921 + dput(a->h_path.dentry);
23925 + delegated = NULL;
23926 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
23927 + &delegated, a->flags);
23928 + if (unlikely(rerr == -EWOULDBLOCK)) {
23929 + pr_warn("cannot retry for NFSv4 delegation"
23930 + " for an internal rename\n");
23933 + d_drop(a->h_path.dentry);
23934 + dput(a->h_path.dentry);
23936 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
23938 + RevertFailure("rename %pd", a->h_dst);
23941 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
23945 + a->h_path.dentry = a->src_wh_dentry;
23946 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
23947 + au_set_dbwh(a->src_dentry, a->src_bwh);
23949 + RevertFailure("unlink %pd", a->src_wh_dentry);
23951 +#undef RevertFailure
23953 +/* ---------------------------------------------------------------------- */
23956 + * when we have to copyup the renaming entry, do it with the rename-target name
23957 + * in order to minimize the cost (the later actual rename is unnecessary).
23958 + * otherwise rename it on the target branch.
23960 +static int au_ren_or_cpup(struct au_ren_args *a)
23963 + struct dentry *d;
23964 + struct inode *delegated;
23966 + d = a->src_dentry;
23967 + if (au_dbtop(d) == a->btgt) {
23968 + a->h_path.dentry = a->dst_h_dentry;
23969 + AuDebugOn(au_dbtop(d) != a->btgt);
23970 + delegated = NULL;
23971 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
23972 + a->dst_h_dir, &a->h_path, &delegated,
23974 + if (unlikely(err == -EWOULDBLOCK)) {
23975 + pr_warn("cannot retry for NFSv4 delegation"
23976 + " for an internal rename\n");
23982 + if (!err && a->h_dst)
23983 + /* it will be set to dinfo later */
23989 +/* cf. aufs_rmdir() */
23990 +static int au_ren_del_whtmp(struct au_ren_args *a)
23993 + struct inode *dir;
23995 + dir = a->dst_dir;
23996 + SiMustAnyLock(dir->i_sb);
23997 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
23998 + au_sbi(dir->i_sb)->si_dirwh)
23999 + || au_test_fs_remote(a->h_dst->d_sb)) {
24000 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
24001 + if (unlikely(err))
24002 + pr_warn("failed removing whtmp dir %pd (%d), "
24003 + "ignored.\n", a->h_dst, err);
24005 + au_nhash_wh_free(&a->thargs->whlist);
24006 + a->thargs->whlist = a->whlist;
24007 + a->whlist.nh_num = 0;
24008 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
24010 + a->thargs = NULL;
24016 +/* make it 'opaque' dir. */
24017 +static int au_ren_do_diropq(struct au_ren_args *a, int idx)
24020 + struct dentry *d, *diropq;
24021 +#define src_or_dst(member) a->sd[idx].member
24024 + d = src_or_dst(dentry); /* {src,dst}_dentry */
24025 + src_or_dst(bdiropq) = au_dbdiropq(d);
24026 + src_or_dst(hinode) = au_hi(src_or_dst(inode), a->btgt);
24027 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
24028 + diropq = au_diropq_create(d, a->btgt);
24029 + au_hn_inode_unlock(src_or_dst(hinode));
24030 + if (IS_ERR(diropq))
24031 + err = PTR_ERR(diropq);
24035 +#undef src_or_dst_
24039 +static int au_ren_diropq(struct au_ren_args *a)
24042 + unsigned char always;
24043 + struct dentry *d;
24046 + d = a->dst_dentry; /* already renamed on the branch */
24047 + always = !!au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ);
24048 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24049 + && !au_ftest_ren(a->auren_flags, DIRREN)
24050 + && a->btgt != au_dbdiropq(a->src_dentry)
24051 + && (a->dst_wh_dentry
24052 + || a->btgt <= au_dbdiropq(d)
24053 + /* hide the lower to keep xino */
24054 + /* the lowers may not be a dir, but we hide them anyway */
24055 + || a->btgt < au_dbbot(d)
24058 + err = au_ren_do_diropq(a, AuSRC);
24059 + if (unlikely(err))
24061 + au_fset_ren(a->auren_flags, DIROPQ_SRC);
24063 + if (!a->exchange)
24064 + goto out; /* success */
24066 + d = a->src_dentry; /* already renamed on the branch */
24067 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24068 + && a->btgt != au_dbdiropq(a->dst_dentry)
24069 + && (a->btgt < au_dbdiropq(d)
24070 + || a->btgt < au_dbbot(d)
24072 + AuDbgDentry(a->src_dentry);
24073 + AuDbgDentry(a->dst_dentry);
24074 + err = au_ren_do_diropq(a, AuDST);
24075 + if (unlikely(err))
24076 + goto out_rev_src;
24077 + au_fset_ren(a->auren_flags, DIROPQ_DST);
24079 + goto out; /* success */
24082 + AuDbg("err %d, reverting src\n", err);
24083 + au_ren_rev_diropq(err, a);
24088 +static int do_rename(struct au_ren_args *a)
24091 + struct dentry *d, *h_d;
24093 + if (!a->exchange) {
24094 + /* prepare workqueue args for asynchronous rmdir */
24095 + h_d = a->dst_h_dentry;
24096 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24097 + /* && !au_ftest_ren(a->auren_flags, DIRREN) */
24098 + && d_is_positive(h_d)) {
24100 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb,
24102 + if (unlikely(!a->thargs))
24104 + a->h_dst = dget(h_d);
24107 + /* create whiteout for src_dentry */
24108 + if (au_ftest_ren(a->auren_flags, WHSRC)) {
24109 + a->src_bwh = au_dbwh(a->src_dentry);
24110 + AuDebugOn(a->src_bwh >= 0);
24111 + a->src_wh_dentry = au_wh_create(a->src_dentry, a->btgt,
24112 + a->src_h_parent);
24113 + err = PTR_ERR(a->src_wh_dentry);
24114 + if (IS_ERR(a->src_wh_dentry))
24118 + /* lookup whiteout for dentry */
24119 + if (au_ftest_ren(a->auren_flags, WHDST)) {
24120 + h_d = au_wh_lkup(a->dst_h_parent,
24121 + &a->dst_dentry->d_name, a->br);
24122 + err = PTR_ERR(h_d);
24125 + if (d_is_negative(h_d))
24128 + a->dst_wh_dentry = h_d;
24131 + /* rename dentry to tmpwh */
24133 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
24134 + if (unlikely(err))
24137 + d = a->dst_dentry;
24138 + au_set_h_dptr(d, a->btgt, NULL);
24139 + err = au_lkup_neg(d, a->btgt, /*wh*/0);
24140 + if (unlikely(err))
24142 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
24146 + BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_btop != a->btgt);
24147 +#if 0 /* debugging */
24148 + BUG_ON(!au_ftest_ren(a->auren_flags, DIRREN)
24149 + && d_is_positive(a->dst_h_dentry)
24150 + && a->src_btop != a->btgt);
24153 + /* rename by vfs_rename or cpup */
24154 + err = au_ren_or_cpup(a);
24155 + if (unlikely(err))
24156 + /* leave the copied-up one */
24159 + /* make dir opaque */
24160 + err = au_ren_diropq(a);
24161 + if (unlikely(err))
24164 + /* update target timestamps */
24165 + if (a->exchange) {
24166 + AuDebugOn(au_dbtop(a->dst_dentry) != a->btgt);
24167 + a->h_path.dentry = au_h_dptr(a->dst_dentry, a->btgt);
24168 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24169 + a->dst_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24171 + AuDebugOn(au_dbtop(a->src_dentry) != a->btgt);
24172 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
24173 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24174 + a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24176 + if (!a->exchange) {
24177 + /* remove whiteout for dentry */
24178 + if (a->dst_wh_dentry) {
24179 + a->h_path.dentry = a->dst_wh_dentry;
24180 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
24182 + if (unlikely(err))
24186 + /* remove whtmp */
24188 + au_ren_del_whtmp(a); /* ignore this error */
24190 + au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
24193 + goto out_success;
24196 + au_ren_rev_diropq(err, a);
24198 + au_ren_rev_rename(err, a);
24202 + au_ren_rev_whtmp(err, a);
24204 + dput(a->dst_wh_dentry);
24205 + a->dst_wh_dentry = NULL;
24207 + if (a->src_wh_dentry)
24208 + au_ren_rev_whsrc(err, a);
24210 + dput(a->src_wh_dentry);
24211 + dput(a->dst_wh_dentry);
24215 + au_whtmp_rmdir_free(a->thargs);
24216 + a->thargs = NULL;
24222 +/* ---------------------------------------------------------------------- */
24225 + * test if @dentry dir can be rename destination or not.
24226 + * success means, it is a logically empty dir.
24228 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
24230 + return au_test_empty(dentry, whlist);
24234 + * test if @a->src_dentry dir can be rename source or not.
24235 + * if it can, return 0.
24237 + * - it is a logically empty dir.
24238 + * - or, it exists on writable branch and has no children including whiteouts
24239 + * on the lower branch unless DIRREN is on.
24241 +static int may_rename_srcdir(struct au_ren_args *a)
24244 + unsigned int rdhash;
24245 + aufs_bindex_t btop, btgt;
24246 + struct dentry *dentry;
24247 + struct super_block *sb;
24248 + struct au_sbinfo *sbinfo;
24250 + dentry = a->src_dentry;
24251 + sb = dentry->d_sb;
24252 + sbinfo = au_sbi(sb);
24253 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
24254 + au_fset_ren(a->auren_flags, DIRREN);
24257 + btop = au_dbtop(dentry);
24258 + if (btop != btgt) {
24259 + struct au_nhash whlist;
24261 + SiMustAnyLock(sb);
24262 + rdhash = sbinfo->si_rdhash;
24264 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
24266 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
24267 + if (unlikely(err))
24269 + err = au_test_empty(dentry, &whlist);
24270 + au_nhash_wh_free(&whlist);
24274 + if (btop == au_dbtaildir(dentry))
24275 + return 0; /* success */
24277 + err = au_test_empty_lower(dentry);
24280 + if (err == -ENOTEMPTY) {
24281 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24284 + AuWarn1("renaming dir who has child(ren) on multiple "
24285 + "branches, is not supported\n");
24292 +/* side effect: sets whlist and h_dentry */
24293 +static int au_ren_may_dir(struct au_ren_args *a)
24296 + unsigned int rdhash;
24297 + struct dentry *d;
24299 + d = a->dst_dentry;
24300 + SiMustAnyLock(d->d_sb);
24303 + if (au_ftest_ren(a->auren_flags, ISDIR_DST) && a->dst_inode) {
24304 + rdhash = au_sbi(d->d_sb)->si_rdhash;
24306 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
24307 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
24308 + if (unlikely(err))
24311 + if (!a->exchange) {
24312 + au_set_dbtop(d, a->dst_btop);
24313 + err = may_rename_dstdir(d, &a->whlist);
24314 + au_set_dbtop(d, a->btgt);
24316 + err = may_rename_srcdir(a);
24318 + a->dst_h_dentry = au_h_dptr(d, au_dbtop(d));
24319 + if (unlikely(err))
24322 + d = a->src_dentry;
24323 + a->src_h_dentry = au_h_dptr(d, au_dbtop(d));
24324 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24325 + err = may_rename_srcdir(a);
24326 + if (unlikely(err)) {
24327 + au_nhash_wh_free(&a->whlist);
24328 + a->whlist.nh_num = 0;
24335 +/* ---------------------------------------------------------------------- */
24338 + * simple tests for rename.
24339 + * following the checks in vfs, plus the parent-child relationship.
24341 +static int au_may_ren(struct au_ren_args *a)
24344 + struct inode *h_inode;
24346 + if (a->src_btop == a->btgt) {
24347 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
24348 + au_ftest_ren(a->auren_flags, ISDIR_SRC));
24349 + if (unlikely(err))
24352 + if (unlikely(a->src_h_dentry == a->h_trap))
24357 + if (a->dst_btop != a->btgt)
24360 + err = -ENOTEMPTY;
24361 + if (unlikely(a->dst_h_dentry == a->h_trap))
24365 + isdir = !!au_ftest_ren(a->auren_flags, ISDIR_DST);
24366 + if (d_really_is_negative(a->dst_dentry)) {
24367 + if (d_is_negative(a->dst_h_dentry))
24368 + err = au_may_add(a->dst_dentry, a->btgt,
24369 + a->dst_h_parent, isdir);
24371 + if (unlikely(d_is_negative(a->dst_h_dentry)))
24373 + h_inode = d_inode(a->dst_h_dentry);
24374 + if (h_inode->i_nlink)
24375 + err = au_may_del(a->dst_dentry, a->btgt,
24376 + a->dst_h_parent, isdir);
24380 + if (unlikely(err == -ENOENT || err == -EEXIST))
24386 +/* ---------------------------------------------------------------------- */
24391 + * - src_dir and dir by lock_rename()
24392 + * - inode if exists
24395 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
24397 + * + di_write_lock2_child()
24398 + * + di_write_lock_child()
24399 + * + ii_write_lock_child()
24400 + * + di_write_lock_child2()
24401 + * + ii_write_lock_child2()
24402 + * + src_parent and parent
24403 + * + di_write_lock_parent()
24404 + * + ii_write_lock_parent()
24405 + * + di_write_lock_parent2()
24406 + * + ii_write_lock_parent2()
24407 + * + lower src_dir and dir by vfsub_lock_rename()
24408 + * + verify the every relationships between child and parent. if any
24409 + * of them failed, unlock all and return -EBUSY.
24411 +static void au_ren_unlock(struct au_ren_args *a)
24413 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
24414 + a->dst_h_parent, a->dst_hdir);
24415 + if (au_ftest_ren(a->auren_flags, DIRREN)
24417 + au_hn_inode_unlock(a->h_root);
24418 + if (au_ftest_ren(a->auren_flags, MNT_WRITE))
24419 + vfsub_mnt_drop_write(au_br_mnt(a->br));
24422 +static int au_ren_lock(struct au_ren_args *a)
24425 + unsigned int udba;
24428 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
24429 + a->src_hdir = au_hi(a->src_dir, a->btgt);
24430 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
24431 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
24433 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
24434 + if (unlikely(err))
24436 + au_fset_ren(a->auren_flags, MNT_WRITE);
24437 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24438 + struct dentry *root;
24439 + struct inode *dir;
24442 + * sbinfo is already locked, so this ii_read_lock is
24443 + * unnecessary. but our debugging feature checks it.
24445 + root = a->src_inode->i_sb->s_root;
24446 + if (root != a->src_parent && root != a->dst_parent) {
24447 + dir = d_inode(root);
24448 + ii_read_lock_parent3(dir);
24449 + a->h_root = au_hi(dir, a->btgt);
24450 + ii_read_unlock(dir);
24451 + au_hn_inode_lock_nested(a->h_root, AuLsc_I_PARENT3);
24454 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
24455 + a->dst_h_parent, a->dst_hdir);
24456 + udba = au_opt_udba(a->src_dentry->d_sb);
24457 + if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
24458 + || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
24459 + err = au_busy_or_stale();
24460 + if (!err && au_dbtop(a->src_dentry) == a->btgt)
24461 + err = au_h_verify(a->src_h_dentry, udba,
24462 + d_inode(a->src_h_parent), a->src_h_parent,
24464 + if (!err && au_dbtop(a->dst_dentry) == a->btgt)
24465 + err = au_h_verify(a->dst_h_dentry, udba,
24466 + d_inode(a->dst_h_parent), a->dst_h_parent,
24469 + goto out; /* success */
24471 + err = au_busy_or_stale();
24472 + au_ren_unlock(a);
24478 +/* ---------------------------------------------------------------------- */
24480 +static void au_ren_refresh_dir(struct au_ren_args *a)
24482 + struct inode *dir;
24484 + dir = a->dst_dir;
24485 + inode_inc_iversion(dir);
24486 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24487 + /* is this updating defined in POSIX? */
24488 + au_cpup_attr_timesizes(a->src_inode);
24489 + au_cpup_attr_nlink(dir, /*force*/1);
24491 + au_dir_ts(dir, a->btgt);
24493 + if (a->exchange) {
24494 + dir = a->src_dir;
24495 + inode_inc_iversion(dir);
24496 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24497 + /* is this updating defined in POSIX? */
24498 + au_cpup_attr_timesizes(a->dst_inode);
24499 + au_cpup_attr_nlink(dir, /*force*/1);
24501 + au_dir_ts(dir, a->btgt);
24504 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24507 + dir = a->src_dir;
24508 + inode_inc_iversion(dir);
24509 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC))
24510 + au_cpup_attr_nlink(dir, /*force*/1);
24511 + au_dir_ts(dir, a->btgt);
24514 +static void au_ren_refresh(struct au_ren_args *a)
24516 + aufs_bindex_t bbot, bindex;
24517 + struct dentry *d, *h_d;
24518 + struct inode *i, *h_i;
24519 + struct super_block *sb;
24521 + d = a->dst_dentry;
24524 + /* already dget-ed by au_ren_or_cpup() */
24525 + au_set_h_dptr(d, a->btgt, a->h_dst);
24527 + i = a->dst_inode;
24529 + if (!a->exchange) {
24530 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST))
24531 + vfsub_drop_nlink(i);
24533 + vfsub_dead_dir(i);
24534 + au_cpup_attr_timesizes(i);
24536 + au_update_dbrange(d, /*do_put_zero*/1);
24538 + au_cpup_attr_nlink(i, /*force*/1);
24541 + for (bindex = au_dbtop(d); bindex < bbot; bindex++)
24542 + au_set_h_dptr(d, bindex, NULL);
24543 + bbot = au_dbbot(d);
24544 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++)
24545 + au_set_h_dptr(d, bindex, NULL);
24546 + au_update_dbrange(d, /*do_put_zero*/0);
24550 + || au_ftest_ren(a->auren_flags, DIRREN)) {
24551 + d_drop(a->src_dentry);
24552 + if (au_ftest_ren(a->auren_flags, DIRREN))
24553 + au_set_dbwh(a->src_dentry, -1);
24557 + d = a->src_dentry;
24558 + au_set_dbwh(d, -1);
24559 + bbot = au_dbbot(d);
24560 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24561 + h_d = au_h_dptr(d, bindex);
24563 + au_set_h_dptr(d, bindex, NULL);
24565 + au_set_dbbot(d, a->btgt);
24568 + i = a->src_inode;
24569 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
24570 + return; /* success */
24572 + bbot = au_ibbot(i);
24573 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24574 + h_i = au_h_iptr(i, bindex);
24576 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
24577 + /* ignore this error */
24578 + au_set_h_iptr(i, bindex, NULL, 0);
24581 + au_set_ibbot(i, a->btgt);
24584 +/* ---------------------------------------------------------------------- */
24586 +/* mainly for link(2) and rename(2) */
24587 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
24589 + aufs_bindex_t bdiropq, bwh;
24590 + struct dentry *parent;
24591 + struct au_branch *br;
24593 + parent = dentry->d_parent;
24594 + IMustLock(d_inode(parent)); /* dir is locked */
24596 + bdiropq = au_dbdiropq(parent);
24597 + bwh = au_dbwh(dentry);
24598 + br = au_sbr(dentry->d_sb, btgt);
24599 + if (au_br_rdonly(br)
24600 + || (0 <= bdiropq && bdiropq < btgt)
24601 + || (0 <= bwh && bwh < btgt))
24604 + AuDbg("btgt %d\n", btgt);
24608 +/* sets src_btop, dst_btop and btgt */
24609 +static int au_ren_wbr(struct au_ren_args *a)
24612 + struct au_wr_dir_args wr_dir_args = {
24613 + /* .force_btgt = -1, */
24614 + .flags = AuWrDir_ADD_ENTRY
24617 + a->src_btop = au_dbtop(a->src_dentry);
24618 + a->dst_btop = au_dbtop(a->dst_dentry);
24619 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24620 + || au_ftest_ren(a->auren_flags, ISDIR_DST))
24621 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
24622 + wr_dir_args.force_btgt = a->src_btop;
24623 + if (a->dst_inode && a->dst_btop < a->src_btop)
24624 + wr_dir_args.force_btgt = a->dst_btop;
24625 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
24626 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
24629 + au_update_dbtop(a->dst_dentry);
24634 +static void au_ren_dt(struct au_ren_args *a)
24636 + a->h_path.dentry = a->src_h_parent;
24637 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
24638 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR)) {
24639 + a->h_path.dentry = a->dst_h_parent;
24640 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
24643 + au_fclr_ren(a->auren_flags, DT_DSTDIR);
24644 + if (!au_ftest_ren(a->auren_flags, ISDIR_SRC)
24648 + a->h_path.dentry = a->src_h_dentry;
24649 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
24650 + if (d_is_positive(a->dst_h_dentry)) {
24651 + au_fset_ren(a->auren_flags, DT_DSTDIR);
24652 + a->h_path.dentry = a->dst_h_dentry;
24653 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
24657 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
24659 + struct dentry *h_d;
24660 + struct inode *h_inode;
24662 + au_dtime_revert(a->src_dt + AuPARENT);
24663 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR))
24664 + au_dtime_revert(a->dst_dt + AuPARENT);
24666 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC) && err != -EIO) {
24667 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
24668 + h_inode = d_inode(h_d);
24669 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
24670 + au_dtime_revert(a->src_dt + AuCHILD);
24671 + inode_unlock(h_inode);
24673 + if (au_ftest_ren(a->auren_flags, DT_DSTDIR)) {
24674 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
24675 + h_inode = d_inode(h_d);
24676 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
24677 + au_dtime_revert(a->dst_dt + AuCHILD);
24678 + inode_unlock(h_inode);
24683 +/* ---------------------------------------------------------------------- */
24685 +int aufs_rename(struct user_namespace *userns,
24686 + struct inode *_src_dir, struct dentry *_src_dentry,
24687 + struct inode *_dst_dir, struct dentry *_dst_dentry,
24688 + unsigned int _flags)
24690 + int err, lock_flags;
24692 + /* reduce stack space */
24693 + struct au_ren_args *a;
24694 + struct au_pin pin;
24696 + AuDbg("%pd, %pd, 0x%x\n", _src_dentry, _dst_dentry, _flags);
24697 + IMustLock(_src_dir);
24698 + IMustLock(_dst_dir);
24701 + if (unlikely(_flags & RENAME_WHITEOUT))
24705 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
24706 + a = kzalloc(sizeof(*a), GFP_NOFS);
24707 + if (unlikely(!a))
24710 + a->flags = _flags;
24711 + BUILD_BUG_ON(sizeof(a->exchange) == sizeof(u8)
24712 + && RENAME_EXCHANGE > U8_MAX);
24713 + a->exchange = _flags & RENAME_EXCHANGE;
24714 + a->src_dir = _src_dir;
24715 + a->src_dentry = _src_dentry;
24716 + a->src_inode = NULL;
24717 + if (d_really_is_positive(a->src_dentry))
24718 + a->src_inode = d_inode(a->src_dentry);
24719 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
24720 + a->dst_dir = _dst_dir;
24721 + a->dst_dentry = _dst_dentry;
24722 + a->dst_inode = NULL;
24723 + if (d_really_is_positive(a->dst_dentry))
24724 + a->dst_inode = d_inode(a->dst_dentry);
24725 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
24726 + if (a->dst_inode) {
24728 + * if EXCHANGE && src is non-dir && dst is dir,
24729 + * dst is not locked.
24731 + /* IMustLock(a->dst_inode); */
24732 + au_igrab(a->dst_inode);
24736 + lock_flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
24737 + if (d_is_dir(a->src_dentry)) {
24738 + au_fset_ren(a->auren_flags, ISDIR_SRC);
24739 + if (unlikely(!a->exchange
24740 + && d_really_is_positive(a->dst_dentry)
24741 + && !d_is_dir(a->dst_dentry)))
24743 + lock_flags |= AuLock_DIRS;
24745 + if (a->dst_inode && d_is_dir(a->dst_dentry)) {
24746 + au_fset_ren(a->auren_flags, ISDIR_DST);
24747 + if (unlikely(!a->exchange
24748 + && d_really_is_positive(a->src_dentry)
24749 + && !d_is_dir(a->src_dentry)))
24751 + lock_flags |= AuLock_DIRS;
24753 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
24755 + if (unlikely(err))
24758 + err = au_d_hashed_positive(a->src_dentry);
24759 + if (unlikely(err))
24762 + if (a->dst_inode) {
24764 + * If it is a dir, VFS unhash it before this
24765 + * function. It means we cannot rely upon d_unhashed().
24767 + if (unlikely(!a->dst_inode->i_nlink))
24769 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24770 + err = au_d_hashed_positive(a->dst_dentry);
24771 + if (unlikely(err && !a->exchange))
24773 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
24775 + } else if (unlikely(d_unhashed(a->dst_dentry)))
24779 + * is it possible?
24780 + * yes, it happened (in linux-3.3-rcN) but I don't know why.
24781 + * there may exist a problem somewhere else.
24784 + if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
24787 + au_fset_ren(a->auren_flags, ISSAMEDIR); /* temporary */
24788 + di_write_lock_parent(a->dst_parent);
24790 + /* which branch we process */
24791 + err = au_ren_wbr(a);
24792 + if (unlikely(err < 0))
24794 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
24795 + a->h_path.mnt = au_br_mnt(a->br);
24797 + /* are they available to be renamed */
24798 + err = au_ren_may_dir(a);
24799 + if (unlikely(err))
24800 + goto out_children;
24802 + /* prepare the writable parent dir on the same branch */
24803 + if (a->dst_btop == a->btgt) {
24804 + au_fset_ren(a->auren_flags, WHDST);
24806 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
24807 + if (unlikely(err))
24808 + goto out_children;
24812 + if (!a->exchange) {
24813 + if (a->src_dir != a->dst_dir) {
24815 + * this temporary unlock is safe,
24816 + * because both dir->i_mutex are locked.
24818 + di_write_unlock(a->dst_parent);
24819 + di_write_lock_parent(a->src_parent);
24820 + err = au_wr_dir_need_wh(a->src_dentry,
24821 + au_ftest_ren(a->auren_flags,
24824 + di_write_unlock(a->src_parent);
24825 + di_write_lock2_parent(a->src_parent, a->dst_parent,
24827 + au_fclr_ren(a->auren_flags, ISSAMEDIR);
24829 + err = au_wr_dir_need_wh(a->src_dentry,
24830 + au_ftest_ren(a->auren_flags,
24834 + if (unlikely(err < 0))
24835 + goto out_children;
24837 + au_fset_ren(a->auren_flags, WHSRC);
24840 + if (a->src_btop != a->btgt) {
24841 + err = au_pin(&pin, a->src_dentry, a->btgt,
24842 + au_opt_udba(a->src_dentry->d_sb),
24843 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24845 + struct au_cp_generic cpg = {
24846 + .dentry = a->src_dentry,
24848 + .bsrc = a->src_btop,
24851 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24853 + AuDebugOn(au_dbtop(a->src_dentry) != a->src_btop);
24854 + err = au_sio_cpup_simple(&cpg);
24857 + if (unlikely(err))
24858 + goto out_children;
24859 + a->src_btop = a->btgt;
24860 + a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
24861 + if (!a->exchange)
24862 + au_fset_ren(a->auren_flags, WHSRC);
24866 + if (a->exchange && a->dst_inode
24867 + && a->dst_btop != a->btgt) {
24868 + err = au_pin(&pin, a->dst_dentry, a->btgt,
24869 + au_opt_udba(a->dst_dentry->d_sb),
24870 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24872 + struct au_cp_generic cpg = {
24873 + .dentry = a->dst_dentry,
24875 + .bsrc = a->dst_btop,
24878 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24880 + err = au_sio_cpup_simple(&cpg);
24883 + if (unlikely(err))
24884 + goto out_children;
24885 + a->dst_btop = a->btgt;
24886 + a->dst_h_dentry = au_h_dptr(a->dst_dentry, a->btgt);
24889 + /* lock them all */
24890 + err = au_ren_lock(a);
24891 + if (unlikely(err))
24892 + /* leave the copied-up one */
24893 + goto out_children;
24895 + if (!a->exchange) {
24896 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
24897 + err = au_may_ren(a);
24898 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
24899 + err = -ENAMETOOLONG;
24900 + if (unlikely(err))
24904 + /* store timestamps to be revertible */
24907 + /* store dirren info */
24908 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24909 + err = au_dr_rename(a->src_dentry, a->btgt,
24910 + &a->dst_dentry->d_name, &rev);
24912 + if (unlikely(err))
24917 + err = do_rename(a);
24918 + if (unlikely(err))
24921 + if (au_ftest_ren(a->auren_flags, DIRREN))
24922 + au_dr_rename_fin(a->src_dentry, a->btgt, rev);
24924 + /* update dir attributes */
24925 + au_ren_refresh_dir(a);
24927 + /* dput/iput all lower dentries */
24928 + au_ren_refresh(a);
24930 + goto out_hdir; /* success */
24933 + if (au_ftest_ren(a->auren_flags, DIRREN))
24934 + au_dr_rename_rev(a->src_dentry, a->btgt, rev);
24936 + au_ren_rev_dt(err, a);
24938 + au_ren_unlock(a);
24940 + au_nhash_wh_free(&a->whlist);
24941 + if (err && a->dst_inode && a->dst_btop != a->btgt) {
24942 + AuDbg("btop %d, btgt %d\n", a->dst_btop, a->btgt);
24943 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
24944 + au_set_dbtop(a->dst_dentry, a->dst_btop);
24948 + if (d_unhashed(a->src_dentry))
24949 + au_fset_ren(a->auren_flags, DROPPED_SRC);
24950 + if (d_unhashed(a->dst_dentry))
24951 + au_fset_ren(a->auren_flags, DROPPED_DST);
24952 + if (!a->exchange)
24953 + d_move(a->src_dentry, a->dst_dentry);
24955 + d_exchange(a->src_dentry, a->dst_dentry);
24956 + if (au_ftest_ren(a->auren_flags, DROPPED_DST))
24957 + d_drop(a->dst_dentry);
24959 + if (au_ftest_ren(a->auren_flags, DROPPED_SRC))
24960 + d_drop(a->src_dentry);
24962 + au_update_dbtop(a->dst_dentry);
24963 + if (!a->dst_inode)
24964 + d_drop(a->dst_dentry);
24966 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24967 + di_write_unlock(a->dst_parent);
24969 + di_write_unlock2(a->src_parent, a->dst_parent);
24971 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
24973 + iput(a->dst_inode);
24975 + au_whtmp_rmdir_free(a->thargs);
24981 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
24982 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
24983 +++ linux/fs/aufs/Kconfig 2022-10-03 07:51:14.926209260 +0200
24985 +# SPDX-License-Identifier: GPL-2.0
24987 + tristate "Aufs (Advanced multi layered unification filesystem) support"
24989 + Aufs is a stackable unification filesystem such as Unionfs,
24990 + which unifies several directories and provides a merged single
24992 + In the early days, aufs was entirely re-designed and
24993 + re-implemented Unionfs Version 1.x series. Introducing many
24994 + original ideas, approaches and improvements, it becomes totally
24995 + different from Unionfs while keeping the basic features.
24999 + prompt "Maximum number of branches"
25000 + default AUFS_BRANCH_MAX_127
25002 + Specifies the maximum number of branches (or member directories)
25003 + in a single aufs. The larger value consumes more system
25004 + resources and has a minor impact to performance.
25005 +config AUFS_BRANCH_MAX_127
25008 + Specifies the maximum number of branches (or member directories)
25009 + in a single aufs. The larger value consumes more system
25010 + resources and has a minor impact to performance.
25011 +config AUFS_BRANCH_MAX_511
25014 + Specifies the maximum number of branches (or member directories)
25015 + in a single aufs. The larger value consumes more system
25016 + resources and has a minor impact to performance.
25017 +config AUFS_BRANCH_MAX_1023
25020 + Specifies the maximum number of branches (or member directories)
25021 + in a single aufs. The larger value consumes more system
25022 + resources and has a minor impact to performance.
25023 +config AUFS_BRANCH_MAX_32767
25026 + Specifies the maximum number of branches (or member directories)
25027 + in a single aufs. The larger value consumes more system
25028 + resources and has a minor impact to performance.
25031 +config AUFS_SBILIST
25033 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
25036 + Automatic configuration for internal use.
25037 + When aufs supports Magic SysRq or /proc, enabled automatically.
25039 +config AUFS_HNOTIFY
25040 + bool "Detect direct branch access (bypassing aufs)"
25042 + If you want to modify files on branches directly, eg. bypassing aufs,
25043 + and want aufs to detect the changes of them fully, then enable this
25044 + option and use 'udba=notify' mount option.
25045 + Currently there is only one available configuration, "fsnotify".
25046 + It will have a negative impact to the performance.
25047 + See detail in aufs.5.
25050 + prompt "method" if AUFS_HNOTIFY
25051 + default AUFS_HFSNOTIFY
25052 +config AUFS_HFSNOTIFY
25057 +config AUFS_EXPORT
25058 + bool "NFS-exportable aufs"
25059 + depends on EXPORTFS
25061 + If you want to export your mounted aufs via NFS, then enable this
25062 + option. There are several requirements for this configuration.
25063 + See detail in aufs.5.
25065 +config AUFS_INO_T_64
25067 + depends on AUFS_EXPORT
25068 + depends on 64BIT && !(ALPHA || S390)
25071 + Automatic configuration for internal use.
25072 + /* typedef unsigned long/int __kernel_ino_t */
25073 + /* alpha and s390x are int */
25076 + bool "support for XATTR/EA (including Security Labels)"
25078 + If your branch fs supports XATTR/EA and you want to make them
25079 + available in aufs too, then enable this opsion and specify the
25080 + branch attributes for EA.
25081 + See detail in aufs.5.
25084 + bool "File-based Hierarchical Storage Management"
25086 + Hierarchical Storage Management (or HSM) is a well-known feature
25087 + in the storage world. Aufs provides this feature as file-based.
25088 + with multiple branches.
25089 + These multiple branches are prioritized, ie. the topmost one
25090 + should be the fastest drive and be used heavily.
25093 + bool "Readdir in userspace"
25095 + Aufs has two methods to provide a merged view for a directory,
25096 + by a user-space library and by kernel-space natively. The latter
25097 + is always enabled but sometimes large and slow.
25098 + If you enable this option, install the library in aufs2-util
25099 + package, and set some environment variables for your readdir(3),
25100 + then the work will be handled in user-space which generally
25101 + shows better performance in most cases.
25102 + See detail in aufs.5.
25104 +config AUFS_DIRREN
25105 + bool "Workaround for rename(2)-ing a directory"
25107 + By default, aufs returns EXDEV error in renameing a dir who has
25108 + his child on the lower branch, since it is a bad idea to issue
25109 + rename(2) internally for every lower branch. But user may not
25110 + accept this behaviour. So here is a workaround to allow such
25111 + rename(2) and store some extra information on the writable
25112 + branch. Obviously this costs high (and I don't like it).
25113 + To use this feature, you need to enable this configuration AND
25114 + to specify the mount option `dirren.'
25115 + See details in aufs.5 and the design documents.
25118 + bool "Show whiteouts"
25120 + If you want to make the whiteouts in aufs visible, then enable
25121 + this option and specify 'shwh' mount option. Although it may
25122 + sounds like philosophy or something, but in technically it
25123 + simply shows the name of whiteout with keeping its behaviour.
25125 +config AUFS_BR_RAMFS
25126 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
25128 + If you want to use ramfs as an aufs branch fs, then enable this
25129 + option. Generally tmpfs is recommended.
25130 + Aufs prohibited them to be a branch fs by default, because
25131 + initramfs becomes unusable after switch_root or something
25132 + generally. If you sets initramfs as an aufs branch and boot your
25133 + system by switch_root, you will meet a problem easily since the
25134 + files in initramfs may be inaccessible.
25135 + Unless you are going to use ramfs as an aufs branch fs without
25136 + switch_root or something, leave it N.
25138 +config AUFS_BR_FUSE
25139 + bool "Fuse fs as an aufs branch"
25140 + depends on FUSE_FS
25143 + If you want to use fuse-based userspace filesystem as an aufs
25144 + branch fs, then enable this option.
25145 + It implements the internal poll(2) operation which is
25146 + implemented by fuse only (curretnly).
25151 + Automatic configuration for internal use.
25153 +config AUFS_BR_HFSPLUS
25154 + bool "Hfsplus as an aufs branch"
25155 + depends on HFSPLUS_FS
25158 + If you want to use hfsplus fs as an aufs branch fs, then enable
25159 + this option. This option introduces a small overhead at
25160 + copying-up a file on hfsplus.
25162 +config AUFS_BDEV_LOOP
25164 + depends on BLK_DEV_LOOP
25167 + Automatic configuration for internal use.
25168 + Convert =[ym] into =y.
25171 + bool "Debug aufs"
25173 + Enable this to compile aufs internal debug code.
25174 + It will have a negative impact to the performance.
25176 +config AUFS_MAGIC_SYSRQ
25178 + depends on AUFS_DEBUG && MAGIC_SYSRQ
25181 + Automatic configuration for internal use.
25182 + When aufs supports Magic SysRq, enabled automatically.
25184 diff -urN /usr/share/empty/fs/aufs/lcnt.h linux/fs/aufs/lcnt.h
25185 --- /usr/share/empty/fs/aufs/lcnt.h 1970-01-01 01:00:00.000000000 +0100
25186 +++ linux/fs/aufs/lcnt.h 2022-03-21 14:49:05.726633010 +0100
25188 +/* SPDX-License-Identifier: GPL-2.0 */
25190 + * Copyright (C) 2018-2021 Junjiro R. Okajima
25192 + * This program, aufs is free software; you can redistribute it and/or modify
25193 + * it under the terms of the GNU General Public License as published by
25194 + * the Free Software Foundation; either version 2 of the License, or
25195 + * (at your option) any later version.
25197 + * This program is distributed in the hope that it will be useful,
25198 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25199 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25200 + * GNU General Public License for more details.
25202 + * You should have received a copy of the GNU General Public License
25203 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25207 + * simple long counter wrapper
25210 +#ifndef __AUFS_LCNT_H__
25211 +#define __AUFS_LCNT_H__
25215 +#include "debug.h"
25217 +#define AuLCntATOMIC 1
25218 +#define AuLCntPCPUCNT 2
25220 + * why does percpu_refcount require extra synchronize_rcu()s in
25221 + * au_br_do_free()
25223 +#define AuLCntPCPUREF 3
25225 +/* #define AuLCntChosen AuLCntATOMIC */
25226 +#define AuLCntChosen AuLCntPCPUCNT
25227 +/* #define AuLCntChosen AuLCntPCPUREF */
25229 +#if AuLCntChosen == AuLCntATOMIC
25230 +#include <linux/atomic.h>
25232 +typedef atomic_long_t au_lcnt_t;
25234 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25236 + atomic_long_set(cnt, 0);
25240 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25245 +static inline void au_lcnt_fin(au_lcnt_t *cnt __maybe_unused,
25246 + int do_sync __maybe_unused)
25251 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25253 + atomic_long_inc(cnt);
25256 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25258 + atomic_long_dec(cnt);
25261 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25263 + return atomic_long_read(cnt);
25267 +#if AuLCntChosen == AuLCntPCPUCNT
25268 +#include <linux/percpu_counter.h>
25270 +typedef struct percpu_counter au_lcnt_t;
25272 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25274 + return percpu_counter_init(cnt, 0, GFP_NOFS);
25277 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25282 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync __maybe_unused)
25284 + percpu_counter_destroy(cnt);
25287 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25289 + percpu_counter_inc(cnt);
25292 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25294 + percpu_counter_dec(cnt);
25297 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25301 + n = percpu_counter_sum(cnt);
25303 + if (LONG_MAX != LLONG_MAX
25305 + AuWarn1("%s\n", "wrap-around");
25311 +#if AuLCntChosen == AuLCntPCPUREF
25312 +#include <linux/percpu-refcount.h>
25314 +typedef struct percpu_ref au_lcnt_t;
25316 +static inline int au_lcnt_init(au_lcnt_t *cnt, percpu_ref_func_t *release)
25319 + release = percpu_ref_exit;
25320 + return percpu_ref_init(cnt, release, /*percpu mode*/0, GFP_NOFS);
25323 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25325 + synchronize_rcu();
25328 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync)
25330 + percpu_ref_kill(cnt);
25332 + au_lcnt_wait_for_fin(cnt);
25335 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25337 + percpu_ref_get(cnt);
25340 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25342 + percpu_ref_put(cnt);
25346 + * avoid calling this func as possible.
25348 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev)
25352 + percpu_ref_switch_to_atomic_sync(cnt);
25353 + l = atomic_long_read(&cnt->count);
25355 + percpu_ref_switch_to_percpu(cnt);
25357 + /* percpu_ref is initialized by 1 instead of 0 */
25362 +#ifdef CONFIG_AUFS_DEBUG
25363 +#define AuLCntZero(val) do { \
25366 + AuDbg("%s = %ld\n", #val, l); \
25369 +#define AuLCntZero(val) do {} while (0)
25372 +#endif /* __KERNEL__ */
25373 +#endif /* __AUFS_LCNT_H__ */
25374 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
25375 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
25376 +++ linux/fs/aufs/loop.c 2022-03-21 14:49:05.726633010 +0100
25378 +// SPDX-License-Identifier: GPL-2.0
25380 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25382 + * This program, aufs is free software; you can redistribute it and/or modify
25383 + * it under the terms of the GNU General Public License as published by
25384 + * the Free Software Foundation; either version 2 of the License, or
25385 + * (at your option) any later version.
25387 + * This program is distributed in the hope that it will be useful,
25388 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25389 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25390 + * GNU General Public License for more details.
25392 + * You should have received a copy of the GNU General Public License
25393 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25397 + * support for loopback block device as a branch
25402 +/* added into drivers/block/loop.c */
25403 +static struct file *(*backing_file_func)(struct super_block *sb);
25406 + * test if two lower dentries have overlapping branches.
25408 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
25410 + struct super_block *h_sb;
25411 + struct file *backing_file;
25413 + if (unlikely(!backing_file_func)) {
25414 + /* don't load "loop" module here */
25415 + backing_file_func = symbol_get(loop_backing_file);
25416 + if (unlikely(!backing_file_func))
25417 + /* "loop" module is not loaded */
25421 + h_sb = h_adding->d_sb;
25422 + backing_file = backing_file_func(h_sb);
25423 + if (!backing_file)
25426 + h_adding = backing_file->f_path.dentry;
25428 + * h_adding can be local NFS.
25429 + * in this case aufs cannot detect the loop.
25431 + if (unlikely(h_adding->d_sb == sb))
25433 + return !!au_test_subdir(h_adding, sb->s_root);
25436 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
25437 +int au_test_loopback_kthread(void)
25440 + struct task_struct *tsk = current;
25441 + char c, comm[sizeof(tsk->comm)];
25444 + if (tsk->flags & PF_KTHREAD) {
25445 + get_task_comm(comm, tsk);
25447 + ret = ('0' <= c && c <= '9'
25448 + && !strncmp(comm, "loop", 4));
25454 +/* ---------------------------------------------------------------------- */
25456 +#define au_warn_loopback_step 16
25457 +static int au_warn_loopback_nelem = au_warn_loopback_step;
25458 +static unsigned long *au_warn_loopback_array;
25460 +void au_warn_loopback(struct super_block *h_sb)
25462 + int i, new_nelem;
25463 + unsigned long *a, magic;
25464 + static DEFINE_SPINLOCK(spin);
25466 + magic = h_sb->s_magic;
25467 + spin_lock(&spin);
25468 + a = au_warn_loopback_array;
25469 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
25470 + if (a[i] == magic) {
25471 + spin_unlock(&spin);
25475 + /* h_sb is new to us, print it */
25476 + if (i < au_warn_loopback_nelem) {
25481 + /* expand the array */
25482 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
25483 + a = au_kzrealloc(au_warn_loopback_array,
25484 + au_warn_loopback_nelem * sizeof(unsigned long),
25485 + new_nelem * sizeof(unsigned long), GFP_ATOMIC,
25486 + /*may_shrink*/0);
25488 + au_warn_loopback_nelem = new_nelem;
25489 + au_warn_loopback_array = a;
25494 + spin_unlock(&spin);
25495 + AuWarn1("realloc failed, ignored\n");
25499 + spin_unlock(&spin);
25500 + pr_warn("you may want to try another patch for loopback file "
25501 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
25504 +int au_loopback_init(void)
25507 + struct super_block *sb __maybe_unused;
25509 + BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(*au_warn_loopback_array));
25512 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
25513 + sizeof(unsigned long), GFP_NOFS);
25514 + if (unlikely(!au_warn_loopback_array))
25520 +void au_loopback_fin(void)
25522 + if (backing_file_func)
25523 + symbol_put(loop_backing_file);
25524 + au_kfree_try_rcu(au_warn_loopback_array);
25526 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
25527 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
25528 +++ linux/fs/aufs/loop.h 2022-03-21 14:49:05.726633010 +0100
25530 +/* SPDX-License-Identifier: GPL-2.0 */
25532 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25534 + * This program, aufs is free software; you can redistribute it and/or modify
25535 + * it under the terms of the GNU General Public License as published by
25536 + * the Free Software Foundation; either version 2 of the License, or
25537 + * (at your option) any later version.
25539 + * This program is distributed in the hope that it will be useful,
25540 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25541 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25542 + * GNU General Public License for more details.
25544 + * You should have received a copy of the GNU General Public License
25545 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25549 + * support for loopback mount as a branch
25552 +#ifndef __AUFS_LOOP_H__
25553 +#define __AUFS_LOOP_H__
25558 +struct super_block;
25560 +#ifdef CONFIG_AUFS_BDEV_LOOP
25561 +/* drivers/block/loop.c */
25562 +struct file *loop_backing_file(struct super_block *sb);
25565 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
25566 +int au_test_loopback_kthread(void);
25567 +void au_warn_loopback(struct super_block *h_sb);
25569 +int au_loopback_init(void);
25570 +void au_loopback_fin(void);
25572 +AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
25574 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
25575 + struct dentry *h_adding)
25576 +AuStubInt0(au_test_loopback_kthread, void)
25577 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
25579 +AuStubInt0(au_loopback_init, void)
25580 +AuStubVoid(au_loopback_fin, void)
25581 +#endif /* BLK_DEV_LOOP */
25583 +#endif /* __KERNEL__ */
25584 +#endif /* __AUFS_LOOP_H__ */
25585 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
25586 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
25587 +++ linux/fs/aufs/magic.mk 2022-03-21 14:49:05.726633010 +0100
25589 +# SPDX-License-Identifier: GPL-2.0
25591 +# defined in ${srctree}/fs/fuse/inode.c
25593 +ifdef CONFIG_FUSE_FS
25594 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
25597 +# defined in ${srctree}/fs/xfs/xfs_sb.h
25599 +ifdef CONFIG_XFS_FS
25600 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
25603 +# defined in ${srctree}/fs/configfs/mount.c
25605 +ifdef CONFIG_CONFIGFS_FS
25606 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
25609 +# defined in ${srctree}/fs/ubifs/ubifs.h
25611 +ifdef CONFIG_UBIFS_FS
25612 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
25615 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
25617 +ifdef CONFIG_HFSPLUS_FS
25618 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
25620 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
25621 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
25622 +++ linux/fs/aufs/Makefile 2022-03-21 14:49:05.716633010 +0100
25624 +# SPDX-License-Identifier: GPL-2.0
25626 +include ${src}/magic.mk
25627 +ifeq (${CONFIG_AUFS_FS},m)
25628 +include ${src}/conf.mk
25630 +-include ${src}/priv_def.mk
25632 +# cf. include/linux/kernel.h
25634 +ccflags-y += -DDEBUG
25635 +# sparse requires the full pathname
25637 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
25639 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
25642 +obj-$(CONFIG_AUFS_FS) += aufs.o
25643 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o fsctx.o \
25644 + wkq.o vfsub.o dcsub.o \
25645 + cpup.o whout.o wbr_policy.o \
25646 + dinfo.o dentry.o \
25648 + finfo.o file.o f_op.o \
25650 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
25654 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
25655 +aufs-$(CONFIG_SYSFS) += sysfs.o
25656 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
25657 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
25658 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
25659 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
25660 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
25661 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
25662 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
25663 +aufs-$(CONFIG_AUFS_DIRREN) += dirren.o
25664 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
25665 +aufs-$(CONFIG_AUFS_POLL) += poll.o
25666 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
25667 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
25668 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
25669 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
25670 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
25671 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
25672 +++ linux/fs/aufs/module.c 2022-03-21 14:49:05.726633010 +0100
25674 +// SPDX-License-Identifier: GPL-2.0
25676 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25678 + * This program, aufs is free software; you can redistribute it and/or modify
25679 + * it under the terms of the GNU General Public License as published by
25680 + * the Free Software Foundation; either version 2 of the License, or
25681 + * (at your option) any later version.
25683 + * This program is distributed in the hope that it will be useful,
25684 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25685 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25686 + * GNU General Public License for more details.
25688 + * You should have received a copy of the GNU General Public License
25689 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25693 + * module global variables and operations
25696 +#include <linux/module.h>
25697 +#include <linux/seq_file.h>
25700 +/* shrinkable realloc */
25701 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink)
25709 +#if 0 /* unused */
25716 + AuDebugOn(!new_sz);
25719 + diff = au_kmidx_sub(sz, new_sz);
25725 + /* expand or SLOB */
25726 + p = krealloc(p, new_sz, gfp);
25727 + else if (new_sz < sz && may_shrink) {
25731 + q = kmalloc(new_sz, gfp);
25734 + memcpy(q, p, new_sz);
25735 + au_kfree_try_rcu(p);
25746 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
25749 + p = au_krealloc(p, new_sz, gfp, may_shrink);
25750 + if (p && new_sz > nused)
25751 + memset(p + nused, 0, new_sz - nused);
25755 +/* ---------------------------------------------------------------------- */
25759 +struct kmem_cache *au_cache[AuCache_Last];
25761 +static void au_cache_fin(void)
25766 + * Make sure all delayed rcu free inodes are flushed before we
25771 + /* excluding AuCache_HNOTIFY */
25772 + BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
25773 + for (i = 0; i < AuCache_HNOTIFY; i++) {
25774 + kmem_cache_destroy(au_cache[i]);
25775 + au_cache[i] = NULL;
25779 +static int __init au_cache_init(void)
25781 + au_cache[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
25782 + if (au_cache[AuCache_DINFO])
25783 + /* SLAB_DESTROY_BY_RCU */
25784 + au_cache[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
25785 + au_icntnr_init_once);
25786 + if (au_cache[AuCache_ICNTNR])
25787 + au_cache[AuCache_FINFO] = AuCacheCtor(au_finfo,
25788 + au_fi_init_once);
25789 + if (au_cache[AuCache_FINFO])
25790 + au_cache[AuCache_VDIR] = AuCache(au_vdir);
25791 + if (au_cache[AuCache_VDIR])
25792 + au_cache[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
25793 + if (au_cache[AuCache_DEHSTR])
25800 +/* ---------------------------------------------------------------------- */
25802 +int au_dir_roflags;
25804 +#ifdef CONFIG_AUFS_SBILIST
25806 + * iterate_supers_type() doesn't protect us from
25807 + * remounting (branch management)
25809 +struct hlist_bl_head au_sbilist;
25813 + * functions for module interface.
25815 +MODULE_LICENSE("GPL");
25816 +/* MODULE_LICENSE("GPL v2"); */
25817 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
25818 +MODULE_DESCRIPTION(AUFS_NAME
25819 + " -- Advanced multi layered unification filesystem");
25820 +MODULE_VERSION(AUFS_VERSION);
25821 +MODULE_ALIAS_FS(AUFS_NAME);
25823 +/* this module parameter has no meaning when SYSFS is disabled */
25824 +int sysaufs_brs = 1;
25825 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
25826 +module_param_named(brs, sysaufs_brs, int, 0444);
25828 +/* this module parameter has no meaning when USER_NS is disabled */
25830 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
25831 +module_param_named(allow_userns, au_userns, bool, 0444);
25833 +/* ---------------------------------------------------------------------- */
25835 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
25837 +int au_seq_path(struct seq_file *seq, struct path *path)
25841 + err = seq_path(seq, path, au_esc_chars);
25850 +/* ---------------------------------------------------------------------- */
25852 +static int __init aufs_init(void)
25857 + p = au_esc_chars;
25858 + for (i = 1; i <= ' '; i++)
25864 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
25866 + memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
25867 + for (i = 0; i < AuIop_Last; i++)
25868 + aufs_iop_nogetattr[i].getattr = NULL;
25870 + memset(au_cache, 0, sizeof(au_cache)); /* including hnotify */
25872 + au_sbilist_init();
25873 + sysaufs_brs_init();
25876 + err = sysaufs_init();
25877 + if (unlikely(err))
25879 + err = dbgaufs_init();
25880 + if (unlikely(err))
25881 + goto out_sysaufs;
25882 + err = au_procfs_init();
25883 + if (unlikely(err))
25884 + goto out_dbgaufs;
25885 + err = au_wkq_init();
25886 + if (unlikely(err))
25888 + err = au_loopback_init();
25889 + if (unlikely(err))
25891 + err = au_hnotify_init();
25892 + if (unlikely(err))
25893 + goto out_loopback;
25894 + err = au_sysrq_init();
25895 + if (unlikely(err))
25897 + err = au_cache_init();
25898 + if (unlikely(err))
25901 + aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
25902 + err = register_filesystem(&aufs_fs_type);
25903 + if (unlikely(err))
25906 + /* since we define pr_fmt, call printk directly */
25907 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
25908 + goto out; /* success */
25915 + au_hnotify_fin();
25917 + au_loopback_fin();
25931 +static void __exit aufs_exit(void)
25933 + unregister_filesystem(&aufs_fs_type);
25936 + au_hnotify_fin();
25937 + au_loopback_fin();
25945 +module_init(aufs_init);
25946 +module_exit(aufs_exit);
25947 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
25948 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
25949 +++ linux/fs/aufs/module.h 2022-05-23 09:44:37.729839166 +0200
25951 +/* SPDX-License-Identifier: GPL-2.0 */
25953 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25955 + * This program, aufs is free software; you can redistribute it and/or modify
25956 + * it under the terms of the GNU General Public License as published by
25957 + * the Free Software Foundation; either version 2 of the License, or
25958 + * (at your option) any later version.
25960 + * This program is distributed in the hope that it will be useful,
25961 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25962 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25963 + * GNU General Public License for more details.
25965 + * You should have received a copy of the GNU General Public License
25966 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25970 + * module initialization and module-global
25973 +#ifndef __AUFS_MODULE_H__
25974 +#define __AUFS_MODULE_H__
25978 +#include <linux/slab.h>
25979 +#include "debug.h"
25980 +#include "dentry.h"
25983 +#include "inode.h"
25988 +/* module parameters */
25989 +extern int sysaufs_brs;
25990 +extern bool au_userns;
25992 +/* ---------------------------------------------------------------------- */
25994 +extern int au_dir_roflags;
25996 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink);
25997 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
26001 + * Comparing the size of the object with sizeof(struct rcu_head)
26002 + * case 1: object is always larger
26003 + * --> au_kfree_rcu() or au_kfree_do_rcu()
26004 + * case 2: object is always smaller
26005 + * --> au_kfree_small()
26006 + * case 3: object can be any size
26007 + * --> au_kfree_try_rcu()
26010 +static inline void au_kfree_do_rcu(const void *p)
26013 + struct rcu_head rcu;
26014 + } *a = (void *)p;
26016 + kfree_rcu(a, rcu);
26019 +#define au_kfree_rcu(_p) do { \
26020 + typeof(_p) p = (_p); \
26021 + BUILD_BUG_ON(sizeof(*p) < sizeof(struct rcu_head)); \
26023 + au_kfree_do_rcu(p); \
26026 +#define au_kfree_do_sz_test(sz) (sz >= sizeof(struct rcu_head))
26027 +#define au_kfree_sz_test(p) (p && au_kfree_do_sz_test(ksize(p)))
26029 +static inline void au_kfree_try_rcu(const void *p)
26033 + if (au_kfree_sz_test(p))
26034 + au_kfree_do_rcu(p);
26039 +static inline void au_kfree_small(const void *p)
26043 + AuDebugOn(au_kfree_sz_test(p));
26047 +static inline int au_kmidx_sub(size_t sz, size_t new_sz)
26049 +#ifndef CONFIG_SLOB
26050 + return __kmalloc_index(sz, false) - __kmalloc_index(new_sz, false);
26052 + return -1; /* SLOB is untested */
26056 +int au_seq_path(struct seq_file *seq, struct path *path);
26058 +#ifdef CONFIG_PROC_FS
26060 +int __init au_procfs_init(void);
26061 +void au_procfs_fin(void);
26063 +AuStubInt0(au_procfs_init, void);
26064 +AuStubVoid(au_procfs_fin, void);
26067 +/* ---------------------------------------------------------------------- */
26076 + AuCache_HNOTIFY, /* must be last */
26080 +extern struct kmem_cache *au_cache[AuCache_Last];
26082 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
26083 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
26084 +#define AuCacheCtor(type, ctor) \
26085 + kmem_cache_create(#type, sizeof(struct type), \
26086 + __alignof__(struct type), AuCacheFlags, ctor)
26088 +#define AuCacheFuncAlloc(name, index) \
26089 + static inline struct au_##name *au_cache_alloc_##name(void) \
26090 + { return kmem_cache_alloc(au_cache[AuCache_##index], GFP_NOFS); }
26092 +#define AuCacheFuncs(name, index) \
26093 + static inline void au_cache_free_##name##_norcu(struct au_##name *p) \
26094 + { kmem_cache_free(au_cache[AuCache_##index], p); } \
26096 + static inline void au_cache_free_##name##_rcu_cb(struct rcu_head *rcu) \
26097 + { void *p = rcu; \
26098 + p -= offsetof(struct au_##name, rcu); \
26099 + kmem_cache_free(au_cache[AuCache_##index], p); } \
26100 + static inline void au_cache_free_##name##_rcu(struct au_##name *p) \
26101 + { BUILD_BUG_ON(sizeof(struct au_##name) < sizeof(struct rcu_head)); \
26102 + call_rcu(&p->rcu, au_cache_free_##name##_rcu_cb); } \
26104 + static inline void au_cache_free_##name(struct au_##name *p) \
26105 + { /* au_cache_free_##name##_norcu(p); */ \
26106 + au_cache_free_##name##_rcu(p); }
26108 +AuCacheFuncs(dinfo, DINFO);
26109 +AuCacheFuncAlloc(dinfo, DINFO);
26111 +AuCacheFuncs(icntnr, ICNTNR);
26112 +static inline struct au_icntnr *au_cache_alloc_icntnr(struct super_block *sb)
26113 +{ return alloc_inode_sb(sb, au_cache[AuCache_ICNTNR], GFP_NOFS); }
26115 +AuCacheFuncs(finfo, FINFO);
26116 +AuCacheFuncAlloc(finfo, FINFO);
26118 +AuCacheFuncs(vdir, VDIR);
26119 +AuCacheFuncAlloc(vdir, VDIR);
26121 +AuCacheFuncs(vdir_dehstr, DEHSTR);
26122 +AuCacheFuncAlloc(vdir_dehstr, DEHSTR);
26124 +#ifdef CONFIG_AUFS_HNOTIFY
26125 +AuCacheFuncs(hnotify, HNOTIFY);
26126 +AuCacheFuncAlloc(hnotify, HNOTIFY);
26129 +#endif /* __KERNEL__ */
26130 +#endif /* __AUFS_MODULE_H__ */
26131 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
26132 --- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
26133 +++ linux/fs/aufs/mvdown.c 2022-03-21 14:49:05.729966343 +0100
26135 +// SPDX-License-Identifier: GPL-2.0
26137 + * Copyright (C) 2011-2021 Junjiro R. Okajima
26139 + * This program, aufs is free software; you can redistribute it and/or modify
26140 + * it under the terms of the GNU General Public License as published by
26141 + * the Free Software Foundation; either version 2 of the License, or
26142 + * (at your option) any later version.
26144 + * This program is distributed in the hope that it will be useful,
26145 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26146 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26147 + * GNU General Public License for more details.
26149 + * You should have received a copy of the GNU General Public License
26150 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26154 + * move-down, opposite of copy-up
26159 +struct au_mvd_args {
26161 + struct super_block *h_sb;
26162 + struct dentry *h_parent;
26163 + struct au_hinode *hdir;
26164 + struct inode *h_dir, *h_inode;
26165 + struct au_pin pin;
26166 + } info[AUFS_MVDOWN_NARRAY];
26168 + struct aufs_mvdown mvdown;
26169 + struct dentry *dentry, *parent;
26170 + struct inode *inode, *dir;
26171 + struct super_block *sb;
26172 + aufs_bindex_t bopq, bwh, bfound;
26173 + unsigned char rename_lock;
26176 +#define mvd_errno mvdown.au_errno
26177 +#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
26178 +#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
26179 +#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
26180 +#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
26182 +#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
26183 +#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
26184 +#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
26185 +#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
26186 +#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
26187 +#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
26189 +#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
26190 +#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
26191 +#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
26192 +#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
26193 +#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
26194 +#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
26196 +#define AU_MVD_PR(flag, ...) do { \
26198 + pr_err(__VA_ARGS__); \
26201 +static int find_lower_writable(struct au_mvd_args *a)
26203 + struct super_block *sb;
26204 + aufs_bindex_t bindex, bbot;
26205 + struct au_branch *br;
26208 + bindex = a->mvd_bsrc;
26209 + bbot = au_sbbot(sb);
26210 + if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
26211 + for (bindex++; bindex <= bbot; bindex++) {
26212 + br = au_sbr(sb, bindex);
26213 + if (au_br_fhsm(br->br_perm)
26214 + && !sb_rdonly(au_br_sb(br)))
26217 + else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
26218 + for (bindex++; bindex <= bbot; bindex++) {
26219 + br = au_sbr(sb, bindex);
26220 + if (!au_br_rdonly(br))
26224 + for (bindex++; bindex <= bbot; bindex++) {
26225 + br = au_sbr(sb, bindex);
26226 + if (!sb_rdonly(au_br_sb(br))) {
26227 + if (au_br_rdonly(br))
26229 + |= AUFS_MVDOWN_ROLOWER_R;
26237 +/* make the parent dir on bdst */
26238 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
26243 + a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
26244 + a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
26245 + a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
26246 + a->mvd_h_dst_parent = NULL;
26247 + if (au_dbbot(a->parent) >= a->mvd_bdst)
26248 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26249 + if (!a->mvd_h_dst_parent) {
26250 + err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
26251 + if (unlikely(err)) {
26252 + AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
26255 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26263 +/* lock them all */
26264 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
26267 + struct dentry *h_trap;
26269 + a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
26270 + a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
26271 + err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
26272 + au_opt_udba(a->sb),
26273 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26275 + if (unlikely(err)) {
26276 + AU_MVD_PR(dmsg, "pin_dst failed\n");
26280 + if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
26281 + a->rename_lock = 0;
26282 + au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26283 + AuLsc_DI_PARENT, AuLsc_I_PARENT3,
26284 + au_opt_udba(a->sb),
26285 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26286 + err = au_do_pin(&a->mvd_pin_src);
26288 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26289 + if (unlikely(err)) {
26290 + AU_MVD_PR(dmsg, "pin_src failed\n");
26293 + goto out; /* success */
26296 + a->rename_lock = 1;
26297 + au_pin_hdir_unlock(&a->mvd_pin_dst);
26298 + err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26299 + au_opt_udba(a->sb),
26300 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26302 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26303 + if (unlikely(err)) {
26304 + AU_MVD_PR(dmsg, "pin_src failed\n");
26305 + au_pin_hdir_lock(&a->mvd_pin_dst);
26308 + au_pin_hdir_unlock(&a->mvd_pin_src);
26309 + h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26310 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26312 + err = (h_trap != a->mvd_h_src_parent);
26314 + err = (h_trap != a->mvd_h_dst_parent);
26316 + BUG_ON(err); /* it should never happen */
26317 + if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
26320 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26321 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26322 + au_pin_hdir_lock(&a->mvd_pin_src);
26323 + au_unpin(&a->mvd_pin_src);
26324 + au_pin_hdir_lock(&a->mvd_pin_dst);
26327 + goto out; /* success */
26330 + au_unpin(&a->mvd_pin_dst);
26336 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
26338 + if (!a->rename_lock)
26339 + au_unpin(&a->mvd_pin_src);
26341 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26342 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26343 + au_pin_hdir_lock(&a->mvd_pin_src);
26344 + au_unpin(&a->mvd_pin_src);
26345 + au_pin_hdir_lock(&a->mvd_pin_dst);
26347 + au_unpin(&a->mvd_pin_dst);
26350 +/* copy-down the file */
26351 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
26354 + struct au_cp_generic cpg = {
26355 + .dentry = a->dentry,
26356 + .bdst = a->mvd_bdst,
26357 + .bsrc = a->mvd_bsrc,
26359 + .pin = &a->mvd_pin_dst,
26360 + .flags = AuCpup_DTIME | AuCpup_HOPEN
26363 + AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
26364 + if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26365 + au_fset_cpup(cpg.flags, OVERWRITE);
26366 + if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
26367 + au_fset_cpup(cpg.flags, RWDST);
26368 + err = au_sio_cpdown_simple(&cpg);
26369 + if (unlikely(err))
26370 + AU_MVD_PR(dmsg, "cpdown failed\n");
26377 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
26380 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
26383 + struct path h_path;
26384 + struct au_branch *br;
26385 + struct inode *delegated;
26387 + br = au_sbr(a->sb, a->mvd_bdst);
26388 + h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
26389 + err = PTR_ERR(h_path.dentry);
26390 + if (IS_ERR(h_path.dentry)) {
26391 + AU_MVD_PR(dmsg, "wh_lkup failed\n");
26396 + if (d_is_positive(h_path.dentry)) {
26397 + h_path.mnt = au_br_mnt(br);
26398 + delegated = NULL;
26399 + err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
26400 + &delegated, /*force*/0);
26401 + if (unlikely(err == -EWOULDBLOCK)) {
26402 + pr_warn("cannot retry for NFSv4 delegation"
26403 + " for an internal unlink\n");
26406 + if (unlikely(err))
26407 + AU_MVD_PR(dmsg, "wh_unlink failed\n");
26409 + dput(h_path.dentry);
26417 + * unlink the topmost h_dentry
26419 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
26422 + struct path h_path;
26423 + struct inode *delegated;
26425 + h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
26426 + h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
26427 + delegated = NULL;
26428 + err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
26429 + if (unlikely(err == -EWOULDBLOCK)) {
26430 + pr_warn("cannot retry for NFSv4 delegation"
26431 + " for an internal unlink\n");
26434 + if (unlikely(err))
26435 + AU_MVD_PR(dmsg, "unlink failed\n");
26441 +/* Since mvdown succeeded, we ignore an error of this function */
26442 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
26445 + struct au_branch *br;
26447 + a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
26448 + br = au_sbr(a->sb, a->mvd_bsrc);
26449 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
26451 + br = au_sbr(a->sb, a->mvd_bdst);
26452 + a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
26453 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
26456 + a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
26458 + AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
26462 + * copy-down the file and unlink the bsrc file.
26463 + * - unlink the bdst whout if exist
26464 + * - copy-down the file (with whtmp name and rename)
26465 + * - unlink the bsrc file
26467 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
26471 + err = au_do_mkdir(dmsg, a);
26473 + err = au_do_lock(dmsg, a);
26474 + if (unlikely(err))
26478 + * do not revert the activities we made on bdst since they should be
26479 + * harmless in aufs.
26482 + err = au_do_cpdown(dmsg, a);
26484 + err = au_do_unlink_wh(dmsg, a);
26485 + if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
26486 + err = au_do_unlink(dmsg, a);
26487 + if (unlikely(err))
26490 + AuDbg("%pd2, 0x%x, %d --> %d\n",
26491 + a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
26492 + if (find_lower_writable(a) < 0)
26493 + a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
26495 + if (a->mvdown.flags & AUFS_MVDOWN_STFS)
26496 + au_do_stfs(dmsg, a);
26498 + /* maintain internal array */
26499 + if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
26500 + au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
26501 + au_set_dbtop(a->dentry, a->mvd_bdst);
26502 + au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
26503 + au_set_ibtop(a->inode, a->mvd_bdst);
26505 + /* hide the lower */
26506 + au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
26507 + au_set_dbbot(a->dentry, a->mvd_bsrc);
26508 + au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
26509 + au_set_ibbot(a->inode, a->mvd_bsrc);
26511 + if (au_dbbot(a->dentry) < a->mvd_bdst)
26512 + au_set_dbbot(a->dentry, a->mvd_bdst);
26513 + if (au_ibbot(a->inode) < a->mvd_bdst)
26514 + au_set_ibbot(a->inode, a->mvd_bdst);
26517 + au_do_unlock(dmsg, a);
26523 +/* ---------------------------------------------------------------------- */
26525 +/* make sure the file is idle */
26526 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
26528 + int err, plinked;
26531 + plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
26532 + if (au_dbtop(a->dentry) == a->mvd_bsrc
26533 + && au_dcount(a->dentry) == 1
26534 + && atomic_read(&a->inode->i_count) == 1
26535 + /* && a->mvd_h_src_inode->i_nlink == 1 */
26536 + && (!plinked || !au_plink_test(a->inode))
26537 + && a->inode->i_nlink == 1)
26542 + "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
26543 + a->mvd_bsrc, au_dbtop(a->dentry), au_dcount(a->dentry),
26544 + atomic_read(&a->inode->i_count), a->inode->i_nlink,
26545 + a->mvd_h_src_inode->i_nlink,
26546 + plinked, plinked ? au_plink_test(a->inode) : 0);
26553 +/* make sure the parent dir is fine */
26554 +static int au_mvd_args_parent(const unsigned char dmsg,
26555 + struct au_mvd_args *a)
26558 + aufs_bindex_t bindex;
26561 + if (unlikely(au_alive_dir(a->parent))) {
26563 + AU_MVD_PR(dmsg, "parent dir is dead\n");
26567 + a->bopq = au_dbdiropq(a->parent);
26568 + bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
26569 + AuDbg("b%d\n", bindex);
26570 + if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
26571 + || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
26573 + a->mvd_errno = EAU_MVDOWN_OPAQUE;
26574 + AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
26575 + a->bopq, a->mvd_bdst);
26583 +static int au_mvd_args_intermediate(const unsigned char dmsg,
26584 + struct au_mvd_args *a)
26587 + struct au_dinfo *dinfo, *tmp;
26589 + /* lookup the next lower positive entry */
26591 + tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
26592 + if (unlikely(!tmp))
26597 + dinfo = au_di(a->dentry);
26598 + au_di_cp(tmp, dinfo);
26599 + au_di_swap(tmp, dinfo);
26601 + /* returns the number of positive dentries */
26602 + err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1,
26603 + /* AuLkup_IGNORE_PERM */ 0);
26605 + a->bwh = au_dbwh(a->dentry);
26606 + else if (err > 0)
26607 + a->bfound = au_dbtop(a->dentry);
26609 + au_di_swap(tmp, dinfo);
26610 + au_rw_write_unlock(&tmp->di_rwsem);
26612 + if (unlikely(err < 0))
26613 + AU_MVD_PR(dmsg, "failed look-up lower\n");
26616 + * here, we have these cases.
26618 + * no positive dentry under bsrc. there are more sub-cases.
26620 + * there no whiteout, we can safely move-down.
26623 + * bsrc < bwh && bwh < bdst
26624 + * there is a whiteout on RO branch. cannot proceed.
26626 + * there is a whiteout on the RW target branch. it should
26629 + * there is a whiteout somewhere unrelated branch.
26630 + * -1 < bfound && bfound <= bsrc
26633 + * found, but it is on RO branch between bsrc and bdst. cannot
26636 + * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
26639 + * found, after we create the file on bdst, it will be hidden.
26642 + AuDebugOn(a->bfound == -1
26644 + && a->bwh <= a->mvd_bsrc);
26645 + AuDebugOn(-1 < a->bfound
26646 + && a->bfound <= a->mvd_bsrc);
26649 + if (a->bfound == -1
26650 + && a->mvd_bsrc < a->bwh
26652 + && a->bwh < a->mvd_bdst) {
26653 + a->mvd_errno = EAU_MVDOWN_WHITEOUT;
26654 + AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
26655 + a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
26657 + } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
26658 + a->mvd_errno = EAU_MVDOWN_UPPER;
26659 + AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
26660 + a->mvd_bdst, a->bfound);
26664 + err = 0; /* success */
26671 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
26676 + if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26677 + && a->bfound == a->mvd_bdst)
26683 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
26686 + struct au_branch *br;
26689 + if (unlikely(S_ISDIR(a->inode->i_mode)))
26693 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
26694 + a->mvd_bsrc = au_ibtop(a->inode);
26696 + a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
26697 + if (unlikely(a->mvd_bsrc < 0
26698 + || (a->mvd_bsrc < au_dbtop(a->dentry)
26699 + || au_dbbot(a->dentry) < a->mvd_bsrc
26700 + || !au_h_dptr(a->dentry, a->mvd_bsrc))
26701 + || (a->mvd_bsrc < au_ibtop(a->inode)
26702 + || au_ibbot(a->inode) < a->mvd_bsrc
26703 + || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
26704 + a->mvd_errno = EAU_MVDOWN_NOUPPER;
26705 + AU_MVD_PR(dmsg, "no upper\n");
26709 + if (unlikely(a->mvd_bsrc == au_sbbot(a->sb))) {
26710 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
26711 + AU_MVD_PR(dmsg, "on the bottom\n");
26714 + a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
26715 + br = au_sbr(a->sb, a->mvd_bsrc);
26716 + err = au_br_rdonly(br);
26717 + if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
26718 + if (unlikely(err))
26720 + } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
26721 + || IS_APPEND(a->mvd_h_src_inode))) {
26723 + a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
26729 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
26730 + a->mvd_bdst = find_lower_writable(a);
26731 + if (unlikely(a->mvd_bdst < 0)) {
26732 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
26733 + AU_MVD_PR(dmsg, "no writable lower branch\n");
26737 + a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
26738 + if (unlikely(a->mvd_bdst < 0
26739 + || au_sbbot(a->sb) < a->mvd_bdst)) {
26740 + a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
26741 + AU_MVD_PR(dmsg, "no lower brid\n");
26746 + err = au_mvd_args_busy(dmsg, a);
26748 + err = au_mvd_args_parent(dmsg, a);
26750 + err = au_mvd_args_intermediate(dmsg, a);
26752 + err = au_mvd_args_exist(dmsg, a);
26754 + AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
26761 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
26764 + unsigned char dmsg;
26765 + struct au_mvd_args *args;
26766 + struct inode *inode;
26768 + inode = d_inode(dentry);
26770 + if (unlikely(!capable(CAP_SYS_ADMIN)))
26774 + args = kmalloc(sizeof(*args), GFP_NOFS);
26775 + if (unlikely(!args))
26778 + err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
26780 + /* VERIFY_WRITE */
26781 + err = !access_ok(uarg, sizeof(*uarg));
26782 + if (unlikely(err)) {
26787 + AuDbg("flags 0x%x\n", args->mvdown.flags);
26788 + args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
26789 + args->mvdown.au_errno = 0;
26790 + args->dentry = dentry;
26791 + args->inode = inode;
26792 + args->sb = dentry->d_sb;
26795 + dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
26796 + args->parent = dget_parent(dentry);
26797 + args->dir = d_inode(args->parent);
26798 + inode_lock_nested(args->dir, I_MUTEX_PARENT);
26799 + dput(args->parent);
26800 + if (unlikely(args->parent != dentry->d_parent)) {
26801 + AU_MVD_PR(dmsg, "parent dir is moved\n");
26805 + inode_lock_nested(inode, I_MUTEX_CHILD);
26806 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
26807 + if (unlikely(err))
26810 + di_write_lock_parent(args->parent);
26811 + err = au_mvd_args(dmsg, args);
26812 + if (unlikely(err))
26815 + err = au_do_mvdown(dmsg, args);
26816 + if (unlikely(err))
26819 + au_cpup_attr_timesizes(args->dir);
26820 + au_cpup_attr_timesizes(inode);
26821 + if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
26822 + au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
26823 + /* au_digen_dec(dentry); */
26826 + di_write_unlock(args->parent);
26827 + aufs_read_unlock(dentry, AuLock_DW);
26829 + inode_unlock(inode);
26831 + inode_unlock(args->dir);
26833 + e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
26836 + au_kfree_rcu(args);
26841 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
26842 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
26843 +++ linux/fs/aufs/opts.c 2022-03-21 14:49:05.729966343 +0100
26845 +// SPDX-License-Identifier: GPL-2.0
26847 + * Copyright (C) 2005-2021 Junjiro R. Okajima
26849 + * This program, aufs is free software; you can redistribute it and/or modify
26850 + * it under the terms of the GNU General Public License as published by
26851 + * the Free Software Foundation; either version 2 of the License, or
26852 + * (at your option) any later version.
26854 + * This program is distributed in the hope that it will be useful,
26855 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26856 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26857 + * GNU General Public License for more details.
26859 + * You should have received a copy of the GNU General Public License
26860 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26864 + * mount options/flags
26867 +#include <linux/types.h> /* a distribution requires */
26868 +#include <linux/parser.h>
26871 +/* ---------------------------------------------------------------------- */
26873 +static const char *au_parser_pattern(int val, match_table_t tbl)
26875 + struct match_token *p;
26878 + while (p->pattern) {
26879 + if (p->token == val)
26880 + return p->pattern;
26887 +static const char *au_optstr(int *val, match_table_t tbl)
26889 + struct match_token *p;
26896 + while (p->pattern) {
26898 + && (v & p->token) == p->token) {
26899 + *val &= ~p->token;
26900 + return p->pattern;
26909 +/* ---------------------------------------------------------------------- */
26911 +static match_table_t brperm = {
26912 + {AuBrPerm_RO, AUFS_BRPERM_RO},
26913 + {AuBrPerm_RR, AUFS_BRPERM_RR},
26914 + {AuBrPerm_RW, AUFS_BRPERM_RW},
26918 +static match_table_t brattr = {
26920 + {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
26921 + {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
26922 + /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
26923 + {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
26924 +#ifdef CONFIG_AUFS_FHSM
26925 + {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
26927 +#ifdef CONFIG_AUFS_XATTR
26928 + {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
26929 + {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
26930 + {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
26931 + {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
26932 + {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
26933 + {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
26936 + /* ro/rr branch */
26937 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
26940 + {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
26941 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
26946 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
26953 + p = strchr(str, '+');
26956 + v = match_token(str, table, args);
26958 + if (v & AuBrAttr_CMOO_Mask)
26959 + attr &= ~AuBrAttr_CMOO_Mask;
26964 + pr_warn("ignored branch attribute %s\n", str);
26974 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
26982 + p = au_optstr(&perm, brattr);
26985 + memcpy(q, p, sz + 1);
26991 + p = au_optstr(&perm, brattr);
26995 + memcpy(q, p, sz + 1);
27001 + return q - str->a;
27004 +int au_br_perm_val(char *perm)
27006 + int val, bad, sz;
27008 + substring_t args[MAX_OPT_ARGS];
27009 + au_br_perm_str_t attr;
27011 + p = strchr(perm, '+');
27014 + val = match_token(perm, brperm, args);
27018 + pr_warn("ignored branch permission %s\n", perm);
27019 + val = AuBrPerm_RO;
27025 + val |= br_attr_val(p + 1, brattr, args);
27028 + switch (val & AuBrPerm_Mask) {
27029 + case AuBrPerm_RO:
27030 + case AuBrPerm_RR:
27031 + bad = val & AuBrWAttr_Mask;
27032 + val &= ~AuBrWAttr_Mask;
27034 + case AuBrPerm_RW:
27035 + bad = val & AuBrRAttr_Mask;
27036 + val &= ~AuBrRAttr_Mask;
27041 + * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
27042 + * does not treat it as an error, just warning.
27043 + * this is a tiny guard for the user operation.
27045 + if (val & AuBrAttr_UNPIN) {
27046 + bad |= AuBrAttr_UNPIN;
27047 + val &= ~AuBrAttr_UNPIN;
27050 + if (unlikely(bad)) {
27051 + sz = au_do_optstr_br_attr(&attr, bad);
27053 + pr_warn("ignored branch attribute %s\n", attr.a);
27060 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
27062 + au_br_perm_str_t attr;
27068 + p = au_optstr(&perm, brperm);
27069 + AuDebugOn(!p || !*p);
27071 + memcpy(q, p, sz + 1);
27074 + sz = au_do_optstr_br_attr(&attr, perm);
27077 + memcpy(q, attr.a, sz + 1);
27080 + AuDebugOn(strlen(str->a) >= sizeof(str->a));
27083 +/* ---------------------------------------------------------------------- */
27085 +static match_table_t udbalevel = {
27086 + {AuOpt_UDBA_REVAL, "reval"},
27087 + {AuOpt_UDBA_NONE, "none"},
27088 +#ifdef CONFIG_AUFS_HNOTIFY
27089 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
27090 +#ifdef CONFIG_AUFS_HFSNOTIFY
27091 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
27097 +int au_udba_val(char *str)
27099 + substring_t args[MAX_OPT_ARGS];
27101 + return match_token(str, udbalevel, args);
27104 +const char *au_optstr_udba(int udba)
27106 + return au_parser_pattern(udba, udbalevel);
27109 +/* ---------------------------------------------------------------------- */
27111 +static match_table_t au_wbr_create_policy = {
27112 + {AuWbrCreate_TDP, "tdp"},
27113 + {AuWbrCreate_TDP, "top-down-parent"},
27114 + {AuWbrCreate_RR, "rr"},
27115 + {AuWbrCreate_RR, "round-robin"},
27116 + {AuWbrCreate_MFS, "mfs"},
27117 + {AuWbrCreate_MFS, "most-free-space"},
27118 + {AuWbrCreate_MFSV, "mfs:%d"},
27119 + {AuWbrCreate_MFSV, "most-free-space:%d"},
27121 + /* top-down regardless the parent, and then mfs */
27122 + {AuWbrCreate_TDMFS, "tdmfs:%d"},
27123 + {AuWbrCreate_TDMFSV, "tdmfs:%d:%d"},
27125 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
27126 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
27127 + {AuWbrCreate_PMFS, "pmfs"},
27128 + {AuWbrCreate_PMFSV, "pmfs:%d"},
27129 + {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
27130 + {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
27135 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
27136 + struct au_opt_wbr_create *create)
27139 + unsigned long long ull;
27142 + if (!match_u64(arg, &ull))
27143 + create->mfsrr_watermark = ull;
27145 + pr_err("bad integer in %s\n", str);
27152 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
27153 + struct au_opt_wbr_create *create)
27158 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
27159 + create->mfs_second = n;
27161 + pr_err("bad integer in %s\n", str);
27168 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
27171 + substring_t args[MAX_OPT_ARGS];
27173 + err = match_token(str, au_wbr_create_policy, args);
27174 + create->wbr_create = err;
27176 + case AuWbrCreate_MFSRRV:
27177 + case AuWbrCreate_TDMFSV:
27178 + case AuWbrCreate_PMFSRRV:
27179 + e = au_wbr_mfs_wmark(&args[0], str, create);
27181 + e = au_wbr_mfs_sec(&args[1], str, create);
27185 + case AuWbrCreate_MFSRR:
27186 + case AuWbrCreate_TDMFS:
27187 + case AuWbrCreate_PMFSRR:
27188 + e = au_wbr_mfs_wmark(&args[0], str, create);
27189 + if (unlikely(e)) {
27194 + case AuWbrCreate_MFS:
27195 + case AuWbrCreate_PMFS:
27196 + create->mfs_second = AUFS_MFS_DEF_SEC;
27198 + case AuWbrCreate_MFSV:
27199 + case AuWbrCreate_PMFSV:
27200 + e = au_wbr_mfs_sec(&args[0], str, create);
27209 +const char *au_optstr_wbr_create(int wbr_create)
27211 + return au_parser_pattern(wbr_create, au_wbr_create_policy);
27214 +static match_table_t au_wbr_copyup_policy = {
27215 + {AuWbrCopyup_TDP, "tdp"},
27216 + {AuWbrCopyup_TDP, "top-down-parent"},
27217 + {AuWbrCopyup_BUP, "bup"},
27218 + {AuWbrCopyup_BUP, "bottom-up-parent"},
27219 + {AuWbrCopyup_BU, "bu"},
27220 + {AuWbrCopyup_BU, "bottom-up"},
27224 +int au_wbr_copyup_val(char *str)
27226 + substring_t args[MAX_OPT_ARGS];
27228 + return match_token(str, au_wbr_copyup_policy, args);
27231 +const char *au_optstr_wbr_copyup(int wbr_copyup)
27233 + return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
27236 +/* ---------------------------------------------------------------------- */
27238 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
27239 + aufs_bindex_t bindex)
27242 + struct au_opt_add *add = &opt->add;
27245 + add->bindex = bindex;
27246 + add->perm = AuBrPerm_RO;
27247 + add->pathname = opt_str;
27248 + p = strchr(opt_str, '=');
27252 + add->perm = au_br_perm_val(p);
27255 + err = vfsub_kern_path(add->pathname, AuOpt_LkupDirFlags, &add->path);
27258 + add->perm = AuBrPerm_RO;
27259 + if (au_test_fs_rr(add->path.dentry->d_sb))
27260 + add->perm = AuBrPerm_RR;
27261 + else if (!bindex && !(sb_flags & SB_RDONLY))
27262 + add->perm = AuBrPerm_RW;
27264 + opt->type = Opt_add;
27267 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
27274 +static int au_opt_wbr_create(struct super_block *sb,
27275 + struct au_opt_wbr_create *create)
27278 + struct au_sbinfo *sbinfo;
27280 + SiMustWriteLock(sb);
27282 + err = 1; /* handled */
27283 + sbinfo = au_sbi(sb);
27284 + if (sbinfo->si_wbr_create_ops->fin) {
27285 + err = sbinfo->si_wbr_create_ops->fin(sb);
27290 + sbinfo->si_wbr_create = create->wbr_create;
27291 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
27292 + switch (create->wbr_create) {
27293 + case AuWbrCreate_MFSRRV:
27294 + case AuWbrCreate_MFSRR:
27295 + case AuWbrCreate_TDMFS:
27296 + case AuWbrCreate_TDMFSV:
27297 + case AuWbrCreate_PMFSRR:
27298 + case AuWbrCreate_PMFSRRV:
27299 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
27301 + case AuWbrCreate_MFS:
27302 + case AuWbrCreate_MFSV:
27303 + case AuWbrCreate_PMFS:
27304 + case AuWbrCreate_PMFSV:
27305 + sbinfo->si_wbr_mfs.mfs_expire
27306 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
27310 + if (sbinfo->si_wbr_create_ops->init)
27311 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
27318 + * plus: processed without an error
27319 + * zero: unprocessed
27321 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
27322 + struct au_opts *opts)
27325 + struct au_sbinfo *sbinfo;
27327 + SiMustWriteLock(sb);
27329 + err = 1; /* handled */
27330 + sbinfo = au_sbi(sb);
27331 + switch (opt->type) {
27333 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27334 + sbinfo->si_mntflags |= opt->udba;
27335 + opts->given_udba |= opt->udba;
27340 + au_opt_set(sbinfo->si_mntflags, PLINK);
27342 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27343 + au_plink_put(sb, /*verbose*/1);
27344 + au_opt_clr(sbinfo->si_mntflags, PLINK);
27347 + case Opt_list_plink:
27348 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27349 + au_plink_list(sb);
27354 + au_opt_set(sbinfo->si_mntflags, DIO);
27355 + au_fset_opts(opts->flags, REFRESH_DYAOP);
27357 + au_opt_clr(sbinfo->si_mntflags, DIO);
27358 + au_fset_opts(opts->flags, REFRESH_DYAOP);
27362 + case Opt_fhsm_sec:
27363 + au_fhsm_set(sbinfo, opt->fhsm_second);
27366 + case Opt_diropq_a:
27367 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27369 + case Opt_diropq_w:
27370 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27373 + case Opt_warn_perm:
27375 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
27377 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
27380 + case Opt_verbose:
27382 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
27384 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
27389 + au_opt_set(sbinfo->si_mntflags, SUM);
27391 + au_opt_clr(sbinfo->si_mntflags, SUM);
27392 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
27396 + au_opt_clr(sbinfo->si_mntflags, SUM);
27397 + au_opt_set(sbinfo->si_mntflags, SUM_W);
27400 + case Opt_wbr_create:
27401 + err = au_opt_wbr_create(sb, &opt->wbr_create);
27403 + case Opt_wbr_copyup:
27404 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
27405 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
27409 + sbinfo->si_dirwh = opt->dirwh;
27412 + case Opt_rdcache:
27413 + sbinfo->si_rdcache
27414 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
27417 + sbinfo->si_rdblk = opt->rdblk;
27420 + sbinfo->si_rdhash = opt->rdhash;
27425 + au_opt_set(sbinfo->si_mntflags, SHWH);
27427 + au_opt_clr(sbinfo->si_mntflags, SHWH);
27430 + case Opt_dirperm1:
27432 + au_opt_set(sbinfo->si_mntflags, DIRPERM1);
27434 + au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
27437 + case Opt_trunc_xino:
27439 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
27441 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
27444 + case Opt_trunc_xino_path:
27445 + case Opt_itrunc_xino:
27446 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex,
27452 + case Opt_trunc_xib:
27454 + au_fset_opts(opts->flags, TRUNC_XIB);
27456 + au_fclr_opts(opts->flags, TRUNC_XIB);
27462 + if (!au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27463 + err = au_dr_opt_set(sb);
27468 + au_opt_set(sbinfo->si_mntflags, DIRREN);
27470 + if (au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27471 + err = au_dr_opt_clr(sb, au_ftest_opts(opts->flags,
27477 + au_opt_clr(sbinfo->si_mntflags, DIRREN);
27483 + sb->s_flags |= SB_POSIXACL;
27485 + sb->s_flags &= ~SB_POSIXACL;
27497 + * returns tri-state.
27498 + * plus: processed without an error
27499 + * zero: unprocessed
27502 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
27503 + struct au_opts *opts)
27505 + int err, do_refresh;
27508 + switch (opt->type) {
27510 + opt->add.bindex = au_sbbot(sb) + 1;
27511 + if (opt->add.bindex < 0)
27512 + opt->add.bindex = 0;
27514 + /* Always goto add, not fallthrough */
27515 + case Opt_prepend:
27516 + opt->add.bindex = 0;
27518 + add: /* indented label */
27520 + err = au_br_add(sb, &opt->add,
27521 + au_ftest_opts(opts->flags, REMOUNT));
27524 + au_fset_opts(opts->flags, REFRESH);
27530 + err = au_br_del(sb, &opt->del,
27531 + au_ftest_opts(opts->flags, REMOUNT));
27534 + au_fset_opts(opts->flags, TRUNC_XIB);
27535 + au_fset_opts(opts->flags, REFRESH);
27541 + err = au_br_mod(sb, &opt->mod,
27542 + au_ftest_opts(opts->flags, REMOUNT),
27547 + au_fset_opts(opts->flags, REFRESH);
27554 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
27555 + struct au_opt_xino **opt_xino,
27556 + struct au_opts *opts)
27561 + switch (opt->type) {
27563 + err = au_xino_set(sb, &opt->xino,
27564 + !!au_ftest_opts(opts->flags, REMOUNT));
27566 + *opt_xino = &opt->xino;
27570 + *opt_xino = (void *)-1;
27577 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
27578 + unsigned int pending)
27581 + aufs_bindex_t bindex, bbot;
27582 + unsigned char do_plink, skip, do_free, can_no_dreval;
27583 + struct au_branch *br;
27584 + struct au_wbr *wbr;
27585 + struct dentry *root, *dentry;
27586 + struct inode *dir, *h_dir;
27587 + struct au_sbinfo *sbinfo;
27588 + struct au_hinode *hdir;
27590 + SiMustAnyLock(sb);
27592 + sbinfo = au_sbi(sb);
27593 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
27595 + if (!(sb_flags & SB_RDONLY)) {
27596 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
27597 + pr_warn("first branch should be rw\n");
27598 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
27599 + pr_warn_once("shwh should be used with ro\n");
27602 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
27603 + && !au_opt_test(sbinfo->si_mntflags, XINO))
27604 + pr_warn_once("udba=*notify requires xino\n");
27606 + if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
27607 + pr_warn_once("dirperm1 breaks the protection"
27608 + " by the permission bits on the lower branch\n");
27612 + root = sb->s_root;
27613 + dir = d_inode(root);
27614 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
27615 + can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
27617 + bbot = au_sbbot(sb);
27618 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
27620 + h_dir = au_h_iptr(dir, bindex);
27621 + br = au_sbr(sb, bindex);
27623 + if ((br->br_perm & AuBrAttr_ICEX)
27624 + && !h_dir->i_op->listxattr)
27625 + br->br_perm &= ~AuBrAttr_ICEX;
27626 +#if 0 /* untested */
27627 + if ((br->br_perm & AuBrAttr_ICEX_SEC)
27628 + && (au_br_sb(br)->s_flags & SB_NOSEC))
27629 + br->br_perm &= ~AuBrAttr_ICEX_SEC;
27633 + wbr = br->br_wbr;
27635 + wbr_wh_read_lock(wbr);
27637 + if (!au_br_writable(br->br_perm)) {
27640 + || (!wbr->wbr_whbase
27641 + && !wbr->wbr_plink
27642 + && !wbr->wbr_orph));
27643 + } else if (!au_br_wh_linkable(br->br_perm)) {
27644 + /* skip = (!br->br_whbase && !br->br_orph); */
27645 + skip = (!wbr || !wbr->wbr_whbase);
27646 + if (skip && wbr) {
27648 + skip = !!wbr->wbr_plink;
27650 + skip = !wbr->wbr_plink;
27653 + /* skip = (br->br_whbase && br->br_ohph); */
27654 + skip = (wbr && wbr->wbr_whbase);
27657 + skip = !!wbr->wbr_plink;
27659 + skip = !wbr->wbr_plink;
27663 + wbr_wh_read_unlock(wbr);
27665 + if (can_no_dreval) {
27666 + dentry = br->br_path.dentry;
27667 + spin_lock(&dentry->d_lock);
27668 + if (dentry->d_flags &
27669 + (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
27670 + can_no_dreval = 0;
27671 + spin_unlock(&dentry->d_lock);
27674 + if (au_br_fhsm(br->br_perm)) {
27676 + AuDebugOn(!br->br_fhsm);
27682 + hdir = au_hi(dir, bindex);
27683 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
27685 + wbr_wh_write_lock(wbr);
27686 + err = au_wh_init(br, sb);
27688 + wbr_wh_write_unlock(wbr);
27689 + au_hn_inode_unlock(hdir);
27691 + if (!err && do_free) {
27692 + au_kfree_rcu(wbr);
27693 + br->br_wbr = NULL;
27697 + if (can_no_dreval)
27698 + au_fset_si(sbinfo, NO_DREVAL);
27700 + au_fclr_si(sbinfo, NO_DREVAL);
27703 + au_fset_si(sbinfo, FHSM);
27704 + for (bindex = bbot; bindex >= 0; bindex--) {
27705 + br = au_sbr(sb, bindex);
27706 + if (au_br_fhsm(br->br_perm)) {
27707 + au_fhsm_set_bottom(sb, bindex);
27712 + au_fclr_si(sbinfo, FHSM);
27713 + au_fhsm_set_bottom(sb, -1);
27719 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
27722 + unsigned int tmp;
27723 + aufs_bindex_t bindex, bbot;
27724 + struct au_opt *opt;
27725 + struct au_opt_xino *opt_xino, xino;
27726 + struct au_sbinfo *sbinfo;
27727 + struct au_branch *br;
27728 + struct inode *dir;
27730 + SiMustWriteLock(sb);
27735 + while (err >= 0 && opt->type != Opt_tail)
27736 + err = au_opt_simple(sb, opt++, opts);
27739 + else if (unlikely(err < 0))
27742 + /* disable xino and udba temporary */
27743 + sbinfo = au_sbi(sb);
27744 + tmp = sbinfo->si_mntflags;
27745 + au_opt_clr(sbinfo->si_mntflags, XINO);
27746 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
27749 + while (err >= 0 && opt->type != Opt_tail)
27750 + err = au_opt_br(sb, opt++, opts);
27753 + else if (unlikely(err < 0))
27756 + bbot = au_sbbot(sb);
27757 + if (unlikely(bbot < 0)) {
27759 + pr_err("no branches\n");
27763 + if (au_opt_test(tmp, XINO))
27764 + au_opt_set(sbinfo->si_mntflags, XINO);
27766 + while (!err && opt->type != Opt_tail)
27767 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
27768 + if (unlikely(err))
27771 + err = au_opts_verify(sb, sb->s_flags, tmp);
27772 + if (unlikely(err))
27775 + /* restore xino */
27776 + if (au_opt_test(tmp, XINO) && !opt_xino) {
27777 + xino.file = au_xino_def(sb);
27778 + err = PTR_ERR(xino.file);
27779 + if (IS_ERR(xino.file))
27782 + err = au_xino_set(sb, &xino, /*remount*/0);
27784 + if (unlikely(err))
27788 + /* restore udba */
27789 + tmp &= AuOptMask_UDBA;
27790 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27791 + sbinfo->si_mntflags |= tmp;
27792 + bbot = au_sbbot(sb);
27793 + for (bindex = 0; bindex <= bbot; bindex++) {
27794 + br = au_sbr(sb, bindex);
27795 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
27796 + if (unlikely(err))
27797 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
27799 + /* go on even if err */
27801 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
27802 + dir = d_inode(sb->s_root);
27803 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
27810 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
27813 + unsigned char no_dreval;
27814 + struct inode *dir;
27815 + struct au_opt_xino *opt_xino;
27816 + struct au_opt *opt;
27817 + struct au_sbinfo *sbinfo;
27819 + SiMustWriteLock(sb);
27821 + err = au_dr_opt_flush(sb);
27822 + if (unlikely(err))
27824 + au_fset_opts(opts->flags, DR_FLUSHED);
27826 + dir = d_inode(sb->s_root);
27827 + sbinfo = au_sbi(sb);
27830 + while (err >= 0 && opt->type != Opt_tail) {
27831 + err = au_opt_simple(sb, opt, opts);
27833 + err = au_opt_br(sb, opt, opts);
27835 + err = au_opt_xino(sb, opt, &opt_xino, opts);
27841 + /* go on even err */
27843 + no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
27844 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
27845 + if (unlikely(rerr && !err))
27848 + if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
27849 + au_fset_opts(opts->flags, REFRESH_IDOP);
27851 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
27852 + rerr = au_xib_trunc(sb);
27853 + if (unlikely(rerr && !err))
27857 + /* will be handled by the caller */
27858 + if (!au_ftest_opts(opts->flags, REFRESH)
27859 + && (opts->given_udba
27860 + || au_opt_test(sbinfo->si_mntflags, XINO)
27861 + || au_ftest_opts(opts->flags, REFRESH_IDOP)
27863 + au_fset_opts(opts->flags, REFRESH);
27865 + AuDbg("status 0x%x\n", opts->flags);
27871 +/* ---------------------------------------------------------------------- */
27873 +unsigned int au_opt_udba(struct super_block *sb)
27875 + return au_mntflags(sb) & AuOptMask_UDBA;
27877 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
27878 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
27879 +++ linux/fs/aufs/opts.h 2022-03-21 14:49:05.729966343 +0100
27881 +/* SPDX-License-Identifier: GPL-2.0 */
27883 + * Copyright (C) 2005-2021 Junjiro R. Okajima
27885 + * This program, aufs is free software; you can redistribute it and/or modify
27886 + * it under the terms of the GNU General Public License as published by
27887 + * the Free Software Foundation; either version 2 of the License, or
27888 + * (at your option) any later version.
27890 + * This program is distributed in the hope that it will be useful,
27891 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27892 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27893 + * GNU General Public License for more details.
27895 + * You should have received a copy of the GNU General Public License
27896 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27900 + * mount options/flags
27903 +#ifndef __AUFS_OPTS_H__
27904 +#define __AUFS_OPTS_H__
27908 +#include <linux/fs_parser.h>
27909 +#include <linux/namei.h>
27910 +#include <linux/path.h>
27914 + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
27915 + Opt_idel, Opt_imod,
27916 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
27917 + Opt_xino, Opt_noxino,
27918 + Opt_trunc_xino, Opt_trunc_xino_v,
27919 + Opt_trunc_xino_path, Opt_itrunc_xino,
27922 + Opt_plink, Opt_list_plink,
27925 + Opt_diropq, Opt_diropq_a, Opt_diropq_w,
27927 + Opt_wbr_copyup, Opt_wbr_create,
27929 + Opt_verbose, Opt_noverbose,
27930 + Opt_sum, Opt_wsum,
27934 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
27937 +/* ---------------------------------------------------------------------- */
27940 +#define AuOpt_XINO 1 /* external inode number bitmap
27941 + and translation table */
27942 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
27943 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
27944 +#define AuOpt_UDBA_REVAL (1 << 3)
27945 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
27946 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
27947 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
27948 +#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
27950 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
27951 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
27952 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
27953 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
27954 +#define AuOpt_VERBOSE (1 << 13) /* print the cause of error */
27955 +#define AuOpt_DIO (1 << 14) /* direct io */
27956 +#define AuOpt_DIRREN (1 << 15) /* directory rename */
27958 +#ifndef CONFIG_AUFS_HNOTIFY
27959 +#undef AuOpt_UDBA_HNOTIFY
27960 +#define AuOpt_UDBA_HNOTIFY 0
27962 +#ifndef CONFIG_AUFS_DIRREN
27963 +#undef AuOpt_DIRREN
27964 +#define AuOpt_DIRREN 0
27966 +#ifndef CONFIG_AUFS_SHWH
27968 +#define AuOpt_SHWH 0
27971 +#define AuOpt_Def (AuOpt_XINO \
27972 + | AuOpt_UDBA_REVAL \
27974 + /* | AuOpt_DIRPERM1 */ \
27975 + | AuOpt_WARN_PERM)
27976 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
27977 + | AuOpt_UDBA_REVAL \
27978 + | AuOpt_UDBA_HNOTIFY)
27980 +#define AuOpt_LkupDirFlags (LOOKUP_FOLLOW | LOOKUP_DIRECTORY)
27982 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
27983 +#define au_opt_set(flags, name) do { \
27984 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
27985 + ((flags) |= AuOpt_##name); \
27987 +#define au_opt_set_udba(flags, name) do { \
27988 + (flags) &= ~AuOptMask_UDBA; \
27989 + ((flags) |= AuOpt_##name); \
27991 +#define au_opt_clr(flags, name) do { \
27992 + ((flags) &= ~AuOpt_##name); \
27995 +static inline unsigned int au_opts_plink(unsigned int mntflags)
27997 +#ifdef CONFIG_PROC_FS
28000 + return mntflags & ~AuOpt_PLINK;
28004 +/* ---------------------------------------------------------------------- */
28006 +/* policies to select one among multiple writable branches */
28008 + AuWbrCreate_TDP, /* top down parent */
28009 + AuWbrCreate_RR, /* round robin */
28010 + AuWbrCreate_MFS, /* most free space */
28011 + AuWbrCreate_MFSV, /* mfs with seconds */
28012 + AuWbrCreate_MFSRR, /* mfs then rr */
28013 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
28014 + AuWbrCreate_TDMFS, /* top down regardless parent and mfs */
28015 + AuWbrCreate_TDMFSV, /* top down regardless parent and mfs */
28016 + AuWbrCreate_PMFS, /* parent and mfs */
28017 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
28018 + AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
28019 + AuWbrCreate_PMFSRRV, /* plus seconds */
28021 + AuWbrCreate_Def = AuWbrCreate_TDP
28025 + AuWbrCopyup_TDP, /* top down parent */
28026 + AuWbrCopyup_BUP, /* bottom up parent */
28027 + AuWbrCopyup_BU, /* bottom up */
28029 + AuWbrCopyup_Def = AuWbrCopyup_TDP
28032 +/* ---------------------------------------------------------------------- */
28036 +struct au_opt_add {
28037 + aufs_bindex_t bindex;
28040 + struct path path;
28043 +struct au_opt_del {
28045 + struct path h_path;
28048 +struct au_opt_mod {
28051 + struct dentry *h_root;
28054 +struct au_opt_xino {
28056 + struct file *file;
28059 +struct au_opt_xino_itrunc {
28060 + aufs_bindex_t bindex;
28063 +struct au_opt_wbr_create {
28066 + unsigned long long mfsrr_watermark;
28072 + struct au_opt_xino xino;
28073 + struct au_opt_xino_itrunc xino_itrunc;
28074 + struct au_opt_add add;
28075 + struct au_opt_del del;
28076 + struct au_opt_mod mod;
28079 + unsigned int rdblk;
28080 + unsigned int rdhash;
28082 + struct au_opt_wbr_create wbr_create;
28084 + unsigned int fhsm_second;
28085 + bool tf; /* generic flag, true or false */
28090 +#define AuOpts_REMOUNT 1
28091 +#define AuOpts_REFRESH (1 << 1)
28092 +#define AuOpts_TRUNC_XIB (1 << 2)
28093 +#define AuOpts_REFRESH_DYAOP (1 << 3)
28094 +#define AuOpts_REFRESH_IDOP (1 << 4)
28095 +#define AuOpts_DR_FLUSHED (1 << 5)
28096 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
28097 +#define au_fset_opts(flags, name) \
28098 + do { (flags) |= AuOpts_##name; } while (0)
28099 +#define au_fclr_opts(flags, name) \
28100 + do { (flags) &= ~AuOpts_##name; } while (0)
28102 +#ifndef CONFIG_AUFS_DIRREN
28103 +#undef AuOpts_DR_FLUSHED
28104 +#define AuOpts_DR_FLUSHED 0
28108 + struct au_opt *opt;
28111 + unsigned int given_udba;
28112 + unsigned int flags;
28113 + unsigned long sb_flags;
28116 +/* ---------------------------------------------------------------------- */
28119 +int au_br_perm_val(char *perm);
28120 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
28121 +int au_udba_val(char *str);
28122 +const char *au_optstr_udba(int udba);
28123 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create);
28124 +const char *au_optstr_wbr_create(int wbr_create);
28125 +int au_wbr_copyup_val(char *str);
28126 +const char *au_optstr_wbr_copyup(int wbr_copyup);
28128 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
28129 + aufs_bindex_t bindex);
28130 +struct super_block;
28131 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
28132 + unsigned int pending);
28133 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
28134 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
28136 +unsigned int au_opt_udba(struct super_block *sb);
28139 +int aufs_fsctx_init(struct fs_context *fc);
28140 +extern const struct fs_parameter_spec aufs_fsctx_paramspec[];
28142 +#endif /* __KERNEL__ */
28143 +#endif /* __AUFS_OPTS_H__ */
28144 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
28145 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
28146 +++ linux/fs/aufs/plink.c 2022-03-21 14:49:05.729966343 +0100
28148 +// SPDX-License-Identifier: GPL-2.0
28150 + * Copyright (C) 2005-2021 Junjiro R. Okajima
28152 + * This program, aufs is free software; you can redistribute it and/or modify
28153 + * it under the terms of the GNU General Public License as published by
28154 + * the Free Software Foundation; either version 2 of the License, or
28155 + * (at your option) any later version.
28157 + * This program is distributed in the hope that it will be useful,
28158 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28159 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28160 + * GNU General Public License for more details.
28162 + * You should have received a copy of the GNU General Public License
28163 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28173 + * the pseudo-link maintenance mode.
28174 + * during a user process maintains the pseudo-links,
28175 + * prohibit adding a new plink and branch manipulation.
28179 + * For entry functions which will handle plink, and i_mutex is already held
28181 + * They cannot wait and should return an error at once.
28182 + * Callers has to check the error.
28184 + * For entry functions which will handle plink, but i_mutex is not held
28186 + * They can wait the plink maintenance mode to finish.
28188 + * They behave like F_SETLK and F_SETLKW.
28189 + * If the caller never handle plink, then both flags are unnecessary.
28192 +int au_plink_maint(struct super_block *sb, int flags)
28196 + struct task_struct *parent, *prev;
28197 + struct au_sbinfo *sbi;
28199 + SiMustAnyLock(sb);
28202 + if (!au_opt_test(au_mntflags(sb), PLINK))
28205 + sbi = au_sbi(sb);
28206 + pid = sbi->si_plink_maint_pid;
28207 + if (!pid || pid == current->pid)
28210 + /* todo: it highly depends upon /sbin/mount.aufs */
28212 + parent = current;
28216 + parent = rcu_dereference(parent->real_parent);
28217 + if (parent == prev)
28219 + ppid = task_pid_vnr(parent);
28220 + if (pid == ppid) {
28221 + rcu_read_unlock();
28226 + rcu_read_unlock();
28228 + if (au_ftest_lock(flags, NOPLMW)) {
28229 + /* if there is no i_mutex lock in VFS, we don't need to wait */
28230 + /* AuDebugOn(!lockdep_depth(current)); */
28231 + while (sbi->si_plink_maint_pid) {
28232 + si_read_unlock(sb);
28233 + /* gave up wake_up_bit() */
28234 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
28236 + if (au_ftest_lock(flags, FLUSH))
28237 + au_nwt_flush(&sbi->si_nowait);
28238 + si_noflush_read_lock(sb);
28240 + } else if (au_ftest_lock(flags, NOPLM)) {
28241 + AuDbg("ppid %d, pid %d\n", ppid, pid);
28249 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
28251 + spin_lock(&sbinfo->si_plink_maint_lock);
28252 + sbinfo->si_plink_maint_pid = 0;
28253 + spin_unlock(&sbinfo->si_plink_maint_lock);
28254 + wake_up_all(&sbinfo->si_plink_wq);
28257 +int au_plink_maint_enter(struct super_block *sb)
28260 + struct au_sbinfo *sbinfo;
28263 + sbinfo = au_sbi(sb);
28264 + /* make sure i am the only one in this fs */
28265 + si_write_lock(sb, AuLock_FLUSH);
28266 + if (au_opt_test(au_mntflags(sb), PLINK)) {
28267 + spin_lock(&sbinfo->si_plink_maint_lock);
28268 + if (!sbinfo->si_plink_maint_pid)
28269 + sbinfo->si_plink_maint_pid = current->pid;
28272 + spin_unlock(&sbinfo->si_plink_maint_lock);
28274 + si_write_unlock(sb);
28279 +/* ---------------------------------------------------------------------- */
28281 +#ifdef CONFIG_AUFS_DEBUG
28282 +void au_plink_list(struct super_block *sb)
28285 + struct au_sbinfo *sbinfo;
28286 + struct hlist_bl_head *hbl;
28287 + struct hlist_bl_node *pos;
28288 + struct au_icntnr *icntnr;
28290 + SiMustAnyLock(sb);
28292 + sbinfo = au_sbi(sb);
28293 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28294 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28296 + for (i = 0; i < AuPlink_NHASH; i++) {
28297 + hbl = sbinfo->si_plink + i;
28298 + hlist_bl_lock(hbl);
28299 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28300 + AuDbg("%lu\n", icntnr->vfs_inode.i_ino);
28301 + hlist_bl_unlock(hbl);
28306 +/* is the inode pseudo-linked? */
28307 +int au_plink_test(struct inode *inode)
28310 + struct au_sbinfo *sbinfo;
28311 + struct hlist_bl_head *hbl;
28312 + struct hlist_bl_node *pos;
28313 + struct au_icntnr *icntnr;
28315 + sbinfo = au_sbi(inode->i_sb);
28316 + AuRwMustAnyLock(&sbinfo->si_rwsem);
28317 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
28318 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28321 + i = au_plink_hash(inode->i_ino);
28322 + hbl = sbinfo->si_plink + i;
28323 + hlist_bl_lock(hbl);
28324 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28325 + if (&icntnr->vfs_inode == inode) {
28329 + hlist_bl_unlock(hbl);
28333 +/* ---------------------------------------------------------------------- */
28336 + * generate a name for plink.
28337 + * the file will be stored under AUFS_WH_PLINKDIR.
28339 +/* 20 is max digits length of ulong 64 */
28340 +#define PLINK_NAME_LEN ((20 + 1) * 2)
28342 +static int plink_name(char *name, int len, struct inode *inode,
28343 + aufs_bindex_t bindex)
28346 + struct inode *h_inode;
28348 + h_inode = au_h_iptr(inode, bindex);
28349 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
28353 +struct au_do_plink_lkup_args {
28354 + struct dentry **errp;
28355 + struct qstr *tgtname;
28356 + struct path *h_ppath;
28359 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
28360 + struct path *h_ppath)
28362 + struct dentry *h_dentry;
28363 + struct inode *h_inode;
28365 + h_inode = d_inode(h_ppath->dentry);
28366 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2);
28367 + h_dentry = vfsub_lkup_one(tgtname, h_ppath);
28368 + inode_unlock_shared(h_inode);
28373 +static void au_call_do_plink_lkup(void *args)
28375 + struct au_do_plink_lkup_args *a = args;
28376 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_ppath);
28379 +/* lookup the plink-ed @inode under the branch at @bindex */
28380 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
28382 + struct dentry *h_dentry;
28383 + struct au_branch *br;
28384 + struct path h_ppath;
28386 + char a[PLINK_NAME_LEN];
28387 + struct qstr tgtname = QSTR_INIT(a, 0);
28389 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28391 + br = au_sbr(inode->i_sb, bindex);
28392 + h_ppath.dentry = br->br_wbr->wbr_plink;
28393 + h_ppath.mnt = au_br_mnt(br);
28394 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28396 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28397 + struct au_do_plink_lkup_args args = {
28398 + .errp = &h_dentry,
28399 + .tgtname = &tgtname,
28400 + .h_ppath = &h_ppath
28403 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
28404 + if (unlikely(wkq_err))
28405 + h_dentry = ERR_PTR(wkq_err);
28407 + h_dentry = au_do_plink_lkup(&tgtname, &h_ppath);
28412 +/* create a pseudo-link */
28413 +static int do_whplink(struct qstr *tgt, struct path *h_ppath,
28414 + struct dentry *h_dentry)
28417 + struct path h_path;
28418 + struct inode *h_dir, *delegated;
28420 + h_dir = d_inode(h_ppath->dentry);
28421 + inode_lock_nested(h_dir, AuLsc_I_CHILD2);
28422 + h_path.mnt = h_ppath->mnt;
28424 + h_path.dentry = vfsub_lkup_one(tgt, h_ppath);
28425 + err = PTR_ERR(h_path.dentry);
28426 + if (IS_ERR(h_path.dentry))
28430 + /* wh.plink dir is not monitored */
28431 + /* todo: is it really safe? */
28432 + if (d_is_positive(h_path.dentry)
28433 + && d_inode(h_path.dentry) != d_inode(h_dentry)) {
28434 + delegated = NULL;
28435 + err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
28436 + if (unlikely(err == -EWOULDBLOCK)) {
28437 + pr_warn("cannot retry for NFSv4 delegation"
28438 + " for an internal unlink\n");
28441 + dput(h_path.dentry);
28442 + h_path.dentry = NULL;
28446 + if (!err && d_is_negative(h_path.dentry)) {
28447 + delegated = NULL;
28448 + err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
28449 + if (unlikely(err == -EWOULDBLOCK)) {
28450 + pr_warn("cannot retry for NFSv4 delegation"
28451 + " for an internal link\n");
28455 + dput(h_path.dentry);
28458 + inode_unlock(h_dir);
28462 +struct do_whplink_args {
28464 + struct qstr *tgt;
28465 + struct path *h_ppath;
28466 + struct dentry *h_dentry;
28469 +static void call_do_whplink(void *args)
28471 + struct do_whplink_args *a = args;
28472 + *a->errp = do_whplink(a->tgt, a->h_ppath, a->h_dentry);
28475 +static int whplink(struct dentry *h_dentry, struct inode *inode,
28476 + aufs_bindex_t bindex)
28478 + int err, wkq_err;
28479 + struct au_branch *br;
28480 + struct au_wbr *wbr;
28481 + struct path h_ppath;
28482 + char a[PLINK_NAME_LEN];
28483 + struct qstr tgtname = QSTR_INIT(a, 0);
28485 + br = au_sbr(inode->i_sb, bindex);
28486 + wbr = br->br_wbr;
28487 + h_ppath.dentry = wbr->wbr_plink;
28488 + h_ppath.mnt = au_br_mnt(br);
28489 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28491 + /* always superio. */
28492 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28493 + struct do_whplink_args args = {
28496 + .h_ppath = &h_ppath,
28497 + .h_dentry = h_dentry
28499 + wkq_err = au_wkq_wait(call_do_whplink, &args);
28500 + if (unlikely(wkq_err))
28503 + err = do_whplink(&tgtname, &h_ppath, h_dentry);
28509 + * create a new pseudo-link for @h_dentry on @bindex.
28510 + * the linked inode is held in aufs @inode.
28512 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
28513 + struct dentry *h_dentry)
28515 + struct super_block *sb;
28516 + struct au_sbinfo *sbinfo;
28517 + struct hlist_bl_head *hbl;
28518 + struct hlist_bl_node *pos;
28519 + struct au_icntnr *icntnr;
28520 + int found, err, cnt, i;
28522 + sb = inode->i_sb;
28523 + sbinfo = au_sbi(sb);
28524 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28525 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28527 + found = au_plink_test(inode);
28531 + i = au_plink_hash(inode->i_ino);
28532 + hbl = sbinfo->si_plink + i;
28535 + hlist_bl_lock(hbl);
28536 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink) {
28537 + if (&icntnr->vfs_inode == inode) {
28543 + icntnr = container_of(inode, struct au_icntnr, vfs_inode);
28544 + hlist_bl_add_head(&icntnr->plink, hbl);
28546 + hlist_bl_unlock(hbl);
28548 + cnt = au_hbl_count(hbl);
28549 +#define msg "unexpectedly unbalanced or too many pseudo-links"
28550 + if (cnt > AUFS_PLINK_WARN)
28551 + AuWarn1(msg ", %d\n", cnt);
28553 + err = whplink(h_dentry, inode, bindex);
28554 + if (unlikely(err)) {
28555 + pr_warn("err %d, damaged pseudo link.\n", err);
28556 + au_hbl_del(&icntnr->plink, hbl);
28557 + iput(&icntnr->vfs_inode);
28560 + iput(&icntnr->vfs_inode);
28563 +/* free all plinks */
28564 +void au_plink_put(struct super_block *sb, int verbose)
28567 + struct au_sbinfo *sbinfo;
28568 + struct hlist_bl_head *hbl;
28569 + struct hlist_bl_node *pos, *tmp;
28570 + struct au_icntnr *icntnr;
28572 + SiMustWriteLock(sb);
28574 + sbinfo = au_sbi(sb);
28575 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28576 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28578 + /* no spin_lock since sbinfo is write-locked */
28580 + for (i = 0; i < AuPlink_NHASH; i++) {
28581 + hbl = sbinfo->si_plink + i;
28582 + if (!warned && verbose && !hlist_bl_empty(hbl)) {
28583 + pr_warn("pseudo-link is not flushed");
28586 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink)
28587 + iput(&icntnr->vfs_inode);
28588 + INIT_HLIST_BL_HEAD(hbl);
28592 +void au_plink_clean(struct super_block *sb, int verbose)
28594 + struct dentry *root;
28596 + root = sb->s_root;
28597 + aufs_write_lock(root);
28598 + if (au_opt_test(au_mntflags(sb), PLINK))
28599 + au_plink_put(sb, verbose);
28600 + aufs_write_unlock(root);
28603 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
28606 + aufs_bindex_t btop, bbot, bindex;
28609 + btop = au_ibtop(inode);
28610 + bbot = au_ibbot(inode);
28612 + for (bindex = btop; bindex <= bbot; bindex++) {
28613 + if (!au_h_iptr(inode, bindex)
28614 + || au_ii_br_id(inode, bindex) != br_id)
28616 + au_set_h_iptr(inode, bindex, NULL, 0);
28621 + for (bindex = btop; bindex <= bbot; bindex++)
28622 + if (au_h_iptr(inode, bindex)) {
28632 +/* free the plinks on a branch specified by @br_id */
28633 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
28635 + struct au_sbinfo *sbinfo;
28636 + struct hlist_bl_head *hbl;
28637 + struct hlist_bl_node *pos, *tmp;
28638 + struct au_icntnr *icntnr;
28639 + struct inode *inode;
28642 + SiMustWriteLock(sb);
28644 + sbinfo = au_sbi(sb);
28645 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28646 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28648 + /* no bit_lock since sbinfo is write-locked */
28649 + for (i = 0; i < AuPlink_NHASH; i++) {
28650 + hbl = sbinfo->si_plink + i;
28651 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) {
28652 + inode = au_igrab(&icntnr->vfs_inode);
28653 + ii_write_lock_child(inode);
28654 + do_put = au_plink_do_half_refresh(inode, br_id);
28656 + hlist_bl_del(&icntnr->plink);
28659 + ii_write_unlock(inode);
28664 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
28665 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
28666 +++ linux/fs/aufs/poll.c 2022-03-21 14:49:05.729966343 +0100
28668 +// SPDX-License-Identifier: GPL-2.0
28670 + * Copyright (C) 2005-2021 Junjiro R. Okajima
28672 + * This program, aufs is free software; you can redistribute it and/or modify
28673 + * it under the terms of the GNU General Public License as published by
28674 + * the Free Software Foundation; either version 2 of the License, or
28675 + * (at your option) any later version.
28677 + * This program is distributed in the hope that it will be useful,
28678 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28679 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28680 + * GNU General Public License for more details.
28682 + * You should have received a copy of the GNU General Public License
28683 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28688 + * There is only one filesystem which implements ->poll operation, currently.
28693 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt)
28696 + struct file *h_file;
28697 + struct super_block *sb;
28699 + /* We should pretend an error happened. */
28700 + mask = EPOLLERR /* | EPOLLIN | EPOLLOUT */;
28701 + sb = file->f_path.dentry->d_sb;
28702 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
28704 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
28705 + if (IS_ERR(h_file)) {
28706 + AuDbg("h_file %ld\n", PTR_ERR(h_file));
28710 + mask = vfs_poll(h_file, pt);
28711 + fput(h_file); /* instead of au_read_post() */
28714 + si_read_unlock(sb);
28715 + if (mask & EPOLLERR)
28716 + AuDbg("mask 0x%x\n", mask);
28719 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
28720 --- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
28721 +++ linux/fs/aufs/posix_acl.c 2022-03-21 14:49:05.729966343 +0100
28723 +// SPDX-License-Identifier: GPL-2.0
28725 + * Copyright (C) 2014-2021 Junjiro R. Okajima
28727 + * This program, aufs is free software; you can redistribute it and/or modify
28728 + * it under the terms of the GNU General Public License as published by
28729 + * the Free Software Foundation; either version 2 of the License, or
28730 + * (at your option) any later version.
28732 + * This program is distributed in the hope that it will be useful,
28733 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28734 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28735 + * GNU General Public License for more details.
28737 + * You should have received a copy of the GNU General Public License
28738 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28742 + * posix acl operations
28745 +#include <linux/fs.h>
28748 +struct posix_acl *aufs_get_acl(struct inode *inode, int type, bool rcu)
28750 + struct posix_acl *acl;
28752 + aufs_bindex_t bindex;
28753 + struct inode *h_inode;
28754 + struct super_block *sb;
28756 + acl = ERR_PTR(-ECHILD);
28761 + sb = inode->i_sb;
28762 + si_read_lock(sb, AuLock_FLUSH);
28763 + ii_read_lock_child(inode);
28764 + if (!(sb->s_flags & SB_POSIXACL))
28767 + bindex = au_ibtop(inode);
28768 + h_inode = au_h_iptr(inode, bindex);
28769 + if (unlikely(!h_inode
28770 + || ((h_inode->i_mode & S_IFMT)
28771 + != (inode->i_mode & S_IFMT)))) {
28772 + err = au_busy_or_stale();
28773 + acl = ERR_PTR(err);
28777 + /* always topmost only */
28778 + acl = get_acl(h_inode, type);
28780 + forget_cached_acl(inode, type);
28782 + set_cached_acl(inode, type, acl);
28785 + ii_read_unlock(inode);
28786 + si_read_unlock(sb);
28789 + AuTraceErrPtr(acl);
28793 +int aufs_set_acl(struct user_namespace *userns, struct inode *inode,
28794 + struct posix_acl *acl, int type)
28798 + struct dentry *dentry;
28799 + struct au_sxattr arg = {
28800 + .type = AU_ACL_SET,
28807 + IMustLock(inode);
28809 + if (inode->i_ino == AUFS_ROOT_INO)
28810 + dentry = dget(inode->i_sb->s_root);
28812 + dentry = d_find_alias(inode);
28814 + dentry = d_find_any_alias(inode);
28816 + pr_warn("cannot handle this inode, "
28817 + "please report to aufs-users ML\n");
28823 + ssz = au_sxattr(dentry, inode, &arg);
28824 + /* forget even it if succeeds since the branch might set differently */
28825 + forget_cached_acl(inode, type);
28834 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
28835 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
28836 +++ linux/fs/aufs/procfs.c 2022-03-21 14:49:05.729966343 +0100
28838 +// SPDX-License-Identifier: GPL-2.0
28840 + * Copyright (C) 2010-2021 Junjiro R. Okajima
28842 + * This program, aufs is free software; you can redistribute it and/or modify
28843 + * it under the terms of the GNU General Public License as published by
28844 + * the Free Software Foundation; either version 2 of the License, or
28845 + * (at your option) any later version.
28847 + * This program is distributed in the hope that it will be useful,
28848 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28849 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28850 + * GNU General Public License for more details.
28852 + * You should have received a copy of the GNU General Public License
28853 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28857 + * procfs interfaces
28860 +#include <linux/proc_fs.h>
28863 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
28865 + struct au_sbinfo *sbinfo;
28867 + sbinfo = file->private_data;
28869 + au_plink_maint_leave(sbinfo);
28870 + kobject_put(&sbinfo->si_kobj);
28876 +static void au_procfs_plm_write_clean(struct file *file)
28878 + struct au_sbinfo *sbinfo;
28880 + sbinfo = file->private_data;
28882 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
28885 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
28888 + struct super_block *sb;
28889 + struct au_sbinfo *sbinfo;
28890 + struct hlist_bl_node *pos;
28893 + if (unlikely(file->private_data))
28897 + /* don't use au_sbilist_lock() here */
28898 + hlist_bl_lock(&au_sbilist);
28899 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
28900 + if (id == sysaufs_si_id(sbinfo)) {
28901 + if (kobject_get_unless_zero(&sbinfo->si_kobj))
28902 + sb = sbinfo->si_sb;
28905 + hlist_bl_unlock(&au_sbilist);
28908 + if (unlikely(!sb))
28911 + err = au_plink_maint_enter(sb);
28913 + /* keep kobject_get() */
28914 + file->private_data = sbinfo;
28916 + kobject_put(&sbinfo->si_kobj);
28922 + * Accept a valid "si=xxxx" only.
28923 + * Once it is accepted successfully, accept "clean" too.
28925 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
28926 + size_t count, loff_t *ppos)
28929 + unsigned long id;
28930 + /* last newline is allowed */
28931 + char buf[3 + sizeof(unsigned long) * 2 + 1];
28934 + if (unlikely(!capable(CAP_SYS_ADMIN)))
28938 + if (unlikely(count > sizeof(buf)))
28941 + err = copy_from_user(buf, ubuf, count);
28942 + if (unlikely(err)) {
28949 + if (!strcmp("clean", buf)) {
28950 + au_procfs_plm_write_clean(file);
28951 + goto out_success;
28952 + } else if (unlikely(strncmp("si=", buf, 3)))
28955 + err = kstrtoul(buf + 3, 16, &id);
28956 + if (unlikely(err))
28959 + err = au_procfs_plm_write_si(file, id);
28960 + if (unlikely(err))
28964 + err = count; /* success */
28969 +static const struct proc_ops au_procfs_plm_op = {
28970 + .proc_write = au_procfs_plm_write,
28971 + .proc_release = au_procfs_plm_release
28974 +/* ---------------------------------------------------------------------- */
28976 +static struct proc_dir_entry *au_procfs_dir;
28978 +void au_procfs_fin(void)
28980 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
28981 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
28984 +int __init au_procfs_init(void)
28987 + struct proc_dir_entry *entry;
28990 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
28991 + if (unlikely(!au_procfs_dir))
28994 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | 0200,
28995 + au_procfs_dir, &au_procfs_plm_op);
28996 + if (unlikely(!entry))
29000 + goto out; /* success */
29004 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29008 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
29009 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
29010 +++ linux/fs/aufs/rdu.c 2022-03-21 14:49:05.729966343 +0100
29012 +// SPDX-License-Identifier: GPL-2.0
29014 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29016 + * This program, aufs is free software; you can redistribute it and/or modify
29017 + * it under the terms of the GNU General Public License as published by
29018 + * the Free Software Foundation; either version 2 of the License, or
29019 + * (at your option) any later version.
29021 + * This program is distributed in the hope that it will be useful,
29022 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29023 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29024 + * GNU General Public License for more details.
29026 + * You should have received a copy of the GNU General Public License
29027 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29031 + * readdir in userspace.
29034 +#include <linux/compat.h>
29035 +#include <linux/fs_stack.h>
29036 +#include <linux/security.h>
29039 +/* bits for struct aufs_rdu.flags */
29040 +#define AuRdu_CALLED 1
29041 +#define AuRdu_CONT (1 << 1)
29042 +#define AuRdu_FULL (1 << 2)
29043 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
29044 +#define au_fset_rdu(flags, name) \
29045 + do { (flags) |= AuRdu_##name; } while (0)
29046 +#define au_fclr_rdu(flags, name) \
29047 + do { (flags) &= ~AuRdu_##name; } while (0)
29049 +struct au_rdu_arg {
29050 + struct dir_context ctx;
29051 + struct aufs_rdu *rdu;
29052 + union au_rdu_ent_ul ent;
29053 + unsigned long end;
29055 + struct super_block *sb;
29059 +static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
29060 + loff_t offset, u64 h_ino, unsigned int d_type)
29063 + struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
29064 + struct aufs_rdu *rdu = arg->rdu;
29065 + struct au_rdu_ent ent;
29069 + au_fset_rdu(rdu->cookie.flags, CALLED);
29070 + len = au_rdu_len(nlen);
29071 + if (arg->ent.ul + len < arg->end) {
29073 + ent.bindex = rdu->cookie.bindex;
29074 + ent.type = d_type;
29076 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
29077 + ent.type = DT_UNKNOWN;
29079 + /* unnecessary to support mmap_sem since this is a dir */
29081 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
29083 + if (copy_to_user(arg->ent.e->name, name, nlen))
29085 + /* the terminating NULL */
29086 + if (__put_user(0, arg->ent.e->name + nlen))
29089 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
29090 + arg->ent.ul += len;
29094 + au_fset_rdu(rdu->cookie.flags, FULL);
29096 + rdu->tail = arg->ent;
29100 + /* AuTraceErr(err); */
29104 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
29108 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
29110 + /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
29111 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
29113 + if (unlikely(offset != cookie->h_pos))
29119 + au_fclr_rdu(cookie->flags, CALLED);
29121 + err = vfsub_iterate_dir(h_file, &arg->ctx);
29125 + && au_ftest_rdu(cookie->flags, CALLED)
29126 + && !au_ftest_rdu(cookie->flags, FULL));
29127 + cookie->h_pos = h_file->f_pos;
29134 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
29137 + aufs_bindex_t bbot;
29138 + struct au_rdu_arg arg = {
29140 + .actor = au_rdu_fill
29143 + struct dentry *dentry;
29144 + struct inode *inode;
29145 + struct file *h_file;
29146 + struct au_rdu_cookie *cookie = &rdu->cookie;
29148 + /* VERIFY_WRITE */
29149 + err = !access_ok(rdu->ent.e, rdu->sz);
29150 + if (unlikely(err)) {
29156 + rdu->tail = rdu->ent;
29159 + arg.ent = rdu->ent;
29160 + arg.end = arg.ent.ul;
29161 + arg.end += rdu->sz;
29164 + if (unlikely(!file->f_op->iterate && !file->f_op->iterate_shared))
29167 + err = security_file_permission(file, MAY_READ);
29169 + if (unlikely(err))
29172 + dentry = file->f_path.dentry;
29173 + inode = d_inode(dentry);
29174 + inode_lock_shared(inode);
29176 + arg.sb = inode->i_sb;
29177 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
29178 + if (unlikely(err))
29180 + err = au_alive_dir(dentry);
29181 + if (unlikely(err))
29183 + /* todo: reval? */
29184 + fi_read_lock(file);
29187 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
29188 + && cookie->generation != au_figen(file)))
29193 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
29195 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
29197 + bbot = au_fbtop(file);
29198 + if (cookie->bindex < bbot)
29199 + cookie->bindex = bbot;
29200 + bbot = au_fbbot_dir(file);
29201 + /* AuDbg("b%d, b%d\n", cookie->bindex, bbot); */
29202 + for (; !err && cookie->bindex <= bbot;
29203 + cookie->bindex++, cookie->h_pos = 0) {
29204 + h_file = au_hf_dir(file, cookie->bindex);
29208 + au_fclr_rdu(cookie->flags, FULL);
29209 + err = au_rdu_do(h_file, &arg);
29211 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
29214 + AuDbg("rent %llu\n", rdu->rent);
29216 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
29217 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
29218 + au_fset_rdu(cookie->flags, CONT);
29219 + cookie->generation = au_figen(file);
29222 + ii_read_lock_child(inode);
29223 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibtop(inode)));
29224 + ii_read_unlock(inode);
29227 + fi_read_unlock(file);
29229 + si_read_unlock(arg.sb);
29231 + inode_unlock_shared(inode);
29237 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
29241 + unsigned long long nent;
29242 + union au_rdu_ent_ul *u;
29243 + struct au_rdu_ent ent;
29244 + struct super_block *sb;
29247 + nent = rdu->nent;
29249 + sb = file->f_path.dentry->d_sb;
29250 + si_read_lock(sb, AuLock_FLUSH);
29251 + while (nent-- > 0) {
29252 + /* unnecessary to support mmap_sem since this is a dir */
29253 + err = copy_from_user(&ent, u->e, sizeof(ent));
29255 + /* VERIFY_WRITE */
29256 + err = !access_ok(&u->e->ino, sizeof(ino));
29257 + if (unlikely(err)) {
29263 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
29265 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
29267 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
29269 + if (unlikely(err)) {
29274 + err = __put_user(ino, &u->e->ino);
29275 + if (unlikely(err)) {
29280 + u->ul += au_rdu_len(ent.nlen);
29282 + si_read_unlock(sb);
29287 +/* ---------------------------------------------------------------------- */
29289 +static int au_rdu_verify(struct aufs_rdu *rdu)
29291 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
29292 + "%llu, b%d, 0x%x, g%u}\n",
29293 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
29295 + rdu->rent, rdu->shwh, rdu->full,
29296 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
29297 + rdu->cookie.generation);
29299 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
29303 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
29307 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29310 + struct aufs_rdu rdu;
29311 + void __user *p = (void __user *)arg;
29313 + err = copy_from_user(&rdu, p, sizeof(rdu));
29314 + if (unlikely(err)) {
29319 + err = au_rdu_verify(&rdu);
29320 + if (unlikely(err))
29324 + case AUFS_CTL_RDU:
29325 + err = au_rdu(file, &rdu);
29326 + if (unlikely(err))
29329 + e = copy_to_user(p, &rdu, sizeof(rdu));
29330 + if (unlikely(e)) {
29335 + case AUFS_CTL_RDU_INO:
29336 + err = au_rdu_ino(file, &rdu);
29340 + /* err = -ENOTTY; */
29349 +#ifdef CONFIG_COMPAT
29350 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29353 + struct aufs_rdu rdu;
29354 + void __user *p = compat_ptr(arg);
29356 + /* todo: get_user()? */
29357 + err = copy_from_user(&rdu, p, sizeof(rdu));
29358 + if (unlikely(err)) {
29363 + rdu.ent.e = compat_ptr(rdu.ent.ul);
29364 + err = au_rdu_verify(&rdu);
29365 + if (unlikely(err))
29369 + case AUFS_CTL_RDU:
29370 + err = au_rdu(file, &rdu);
29371 + if (unlikely(err))
29374 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
29375 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
29376 + e = copy_to_user(p, &rdu, sizeof(rdu));
29377 + if (unlikely(e)) {
29382 + case AUFS_CTL_RDU_INO:
29383 + err = au_rdu_ino(file, &rdu);
29387 + /* err = -ENOTTY; */
29396 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
29397 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
29398 +++ linux/fs/aufs/rwsem.h 2022-03-21 14:49:05.729966343 +0100
29400 +/* SPDX-License-Identifier: GPL-2.0 */
29402 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29404 + * This program, aufs is free software; you can redistribute it and/or modify
29405 + * it under the terms of the GNU General Public License as published by
29406 + * the Free Software Foundation; either version 2 of the License, or
29407 + * (at your option) any later version.
29409 + * This program is distributed in the hope that it will be useful,
29410 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29411 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29412 + * GNU General Public License for more details.
29414 + * You should have received a copy of the GNU General Public License
29415 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29419 + * simple read-write semaphore wrappers
29422 +#ifndef __AUFS_RWSEM_H__
29423 +#define __AUFS_RWSEM_H__
29427 +#include "debug.h"
29429 +/* in the future, the name 'au_rwsem' will be totally gone */
29430 +#define au_rwsem rw_semaphore
29432 +/* to debug easier, do not make them inlined functions */
29433 +#define AuRwMustNoWaiters(rw) AuDebugOn(rwsem_is_contended(rw))
29435 +#ifdef CONFIG_LOCKDEP
29436 +/* rwsem_is_locked() is unusable */
29437 +#define AuRwMustReadLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29438 + && !lockdep_recursing(current) \
29440 + && !lockdep_is_held_type(rw, 1))
29441 +#define AuRwMustWriteLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29442 + && !lockdep_recursing(current) \
29444 + && !lockdep_is_held_type(rw, 0))
29445 +#define AuRwMustAnyLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29446 + && !lockdep_recursing(current) \
29448 + && !lockdep_is_held(rw))
29449 +#define AuRwDestroy(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29450 + && !lockdep_recursing(current) \
29452 + && lockdep_is_held(rw))
29454 +#define AuRwMustReadLock(rw) do {} while (0)
29455 +#define AuRwMustWriteLock(rw) do {} while (0)
29456 +#define AuRwMustAnyLock(rw) do {} while (0)
29457 +#define AuRwDestroy(rw) do {} while (0)
29460 +#define au_rw_init(rw) init_rwsem(rw)
29462 +#define au_rw_init_wlock(rw) do { \
29463 + au_rw_init(rw); \
29464 + down_write(rw); \
29467 +#define au_rw_init_wlock_nested(rw, lsc) do { \
29468 + au_rw_init(rw); \
29469 + down_write_nested(rw, lsc); \
29472 +#define au_rw_read_lock(rw) down_read(rw)
29473 +#define au_rw_read_lock_nested(rw, lsc) down_read_nested(rw, lsc)
29474 +#define au_rw_read_unlock(rw) up_read(rw)
29475 +#define au_rw_dgrade_lock(rw) downgrade_write(rw)
29476 +#define au_rw_write_lock(rw) down_write(rw)
29477 +#define au_rw_write_lock_nested(rw, lsc) down_write_nested(rw, lsc)
29478 +#define au_rw_write_unlock(rw) up_write(rw)
29479 +/* why is not _nested version defined? */
29480 +#define au_rw_read_trylock(rw) down_read_trylock(rw)
29481 +#define au_rw_write_trylock(rw) down_write_trylock(rw)
29483 +#endif /* __KERNEL__ */
29484 +#endif /* __AUFS_RWSEM_H__ */
29485 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
29486 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
29487 +++ linux/fs/aufs/sbinfo.c 2022-03-21 14:49:05.729966343 +0100
29489 +// SPDX-License-Identifier: GPL-2.0
29491 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29493 + * This program, aufs is free software; you can redistribute it and/or modify
29494 + * it under the terms of the GNU General Public License as published by
29495 + * the Free Software Foundation; either version 2 of the License, or
29496 + * (at your option) any later version.
29498 + * This program is distributed in the hope that it will be useful,
29499 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29500 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29501 + * GNU General Public License for more details.
29503 + * You should have received a copy of the GNU General Public License
29504 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29508 + * superblock private data
29511 +#include <linux/iversion.h>
29515 + * they are necessary regardless sysfs is disabled.
29517 +void au_si_free(struct kobject *kobj)
29520 + struct au_sbinfo *sbinfo;
29521 + char *locked __maybe_unused; /* debug only */
29523 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
29524 + for (i = 0; i < AuPlink_NHASH; i++)
29525 + AuDebugOn(!hlist_bl_empty(sbinfo->si_plink + i));
29526 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
29528 + AuLCntZero(au_lcnt_read(&sbinfo->si_ninodes, /*do_rev*/0));
29529 + au_lcnt_fin(&sbinfo->si_ninodes, /*do_sync*/0);
29530 + AuLCntZero(au_lcnt_read(&sbinfo->si_nfiles, /*do_rev*/0));
29531 + au_lcnt_fin(&sbinfo->si_nfiles, /*do_sync*/0);
29533 + dbgaufs_si_fin(sbinfo);
29534 + au_rw_write_lock(&sbinfo->si_rwsem);
29535 + au_br_free(sbinfo);
29536 + au_rw_write_unlock(&sbinfo->si_rwsem);
29538 + au_kfree_try_rcu(sbinfo->si_branch);
29539 + mutex_destroy(&sbinfo->si_xib_mtx);
29540 + AuRwDestroy(&sbinfo->si_rwsem);
29542 + au_lcnt_wait_for_fin(&sbinfo->si_ninodes);
29543 + /* si_nfiles is waited too */
29544 + au_kfree_rcu(sbinfo);
29547 +struct au_sbinfo *au_si_alloc(struct super_block *sb)
29549 + struct au_sbinfo *sbinfo;
29553 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
29554 + if (unlikely(!sbinfo))
29557 + /* will be reallocated separately */
29558 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
29559 + if (unlikely(!sbinfo->si_branch))
29562 + err = sysaufs_si_init(sbinfo);
29564 + dbgaufs_si_null(sbinfo);
29565 + err = dbgaufs_si_init(sbinfo);
29566 + if (unlikely(err))
29567 + kobject_put(&sbinfo->si_kobj);
29569 + if (unlikely(err))
29572 + au_nwt_init(&sbinfo->si_nowait);
29573 + au_rw_init_wlock(&sbinfo->si_rwsem);
29575 + au_lcnt_init(&sbinfo->si_ninodes, /*release*/NULL);
29576 + au_lcnt_init(&sbinfo->si_nfiles, /*release*/NULL);
29578 + sbinfo->si_bbot = -1;
29579 + sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
29581 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
29582 + sbinfo->si_wbr_create = AuWbrCreate_Def;
29583 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
29584 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
29586 + au_fhsm_init(sbinfo);
29588 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
29590 + sbinfo->si_xino_jiffy = jiffies;
29591 + sbinfo->si_xino_expire
29592 + = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
29593 + mutex_init(&sbinfo->si_xib_mtx);
29594 + /* leave si_xib_last_pindex and si_xib_next_bit */
29596 + INIT_HLIST_BL_HEAD(&sbinfo->si_aopen);
29598 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
29599 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
29600 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
29601 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
29603 + for (i = 0; i < AuPlink_NHASH; i++)
29604 + INIT_HLIST_BL_HEAD(sbinfo->si_plink + i);
29605 + init_waitqueue_head(&sbinfo->si_plink_wq);
29606 + spin_lock_init(&sbinfo->si_plink_maint_lock);
29608 + INIT_HLIST_BL_HEAD(&sbinfo->si_files);
29610 + /* with getattr by default */
29611 + sbinfo->si_iop_array = aufs_iop;
29613 + /* leave other members for sysaufs and si_mnt. */
29614 + sbinfo->si_sb = sb;
29616 + sb->s_fs_info = sbinfo;
29619 + return sbinfo; /* success */
29622 + au_kfree_try_rcu(sbinfo->si_branch);
29624 + au_kfree_rcu(sbinfo);
29626 + return ERR_PTR(err);
29629 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink)
29632 + struct au_branch **brp;
29634 + AuRwMustWriteLock(&sbinfo->si_rwsem);
29637 + sz = sizeof(*brp) * (sbinfo->si_bbot + 1);
29638 + if (unlikely(!sz))
29639 + sz = sizeof(*brp);
29640 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS,
29643 + sbinfo->si_branch = brp;
29650 +/* ---------------------------------------------------------------------- */
29652 +unsigned int au_sigen_inc(struct super_block *sb)
29654 + unsigned int gen;
29655 + struct inode *inode;
29657 + SiMustWriteLock(sb);
29659 + gen = ++au_sbi(sb)->si_generation;
29660 + au_update_digen(sb->s_root);
29661 + inode = d_inode(sb->s_root);
29662 + au_update_iigen(inode, /*half*/0);
29663 + inode_inc_iversion(inode);
29667 +aufs_bindex_t au_new_br_id(struct super_block *sb)
29669 + aufs_bindex_t br_id;
29671 + struct au_sbinfo *sbinfo;
29673 + SiMustWriteLock(sb);
29675 + sbinfo = au_sbi(sb);
29676 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
29677 + br_id = ++sbinfo->si_last_br_id;
29678 + AuDebugOn(br_id < 0);
29679 + if (br_id && au_br_index(sb, br_id) < 0)
29686 +/* ---------------------------------------------------------------------- */
29688 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
29689 +int si_read_lock(struct super_block *sb, int flags)
29694 + if (au_ftest_lock(flags, FLUSH))
29695 + au_nwt_flush(&au_sbi(sb)->si_nowait);
29697 + si_noflush_read_lock(sb);
29698 + err = au_plink_maint(sb, flags);
29699 + if (unlikely(err))
29700 + si_read_unlock(sb);
29705 +int si_write_lock(struct super_block *sb, int flags)
29709 + if (au_ftest_lock(flags, FLUSH))
29710 + au_nwt_flush(&au_sbi(sb)->si_nowait);
29712 + si_noflush_write_lock(sb);
29713 + err = au_plink_maint(sb, flags);
29714 + if (unlikely(err))
29715 + si_write_unlock(sb);
29720 +/* dentry and super_block lock. call at entry point */
29721 +int aufs_read_lock(struct dentry *dentry, int flags)
29724 + struct super_block *sb;
29726 + sb = dentry->d_sb;
29727 + err = si_read_lock(sb, flags);
29728 + if (unlikely(err))
29731 + if (au_ftest_lock(flags, DW))
29732 + di_write_lock_child(dentry);
29734 + di_read_lock_child(dentry, flags);
29736 + if (au_ftest_lock(flags, GEN)) {
29737 + err = au_digen_test(dentry, au_sigen(sb));
29738 + if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
29739 + AuDebugOn(!err && au_dbrange_test(dentry));
29741 + err = au_dbrange_test(dentry);
29742 + if (unlikely(err))
29743 + aufs_read_unlock(dentry, flags);
29750 +void aufs_read_unlock(struct dentry *dentry, int flags)
29752 + if (au_ftest_lock(flags, DW))
29753 + di_write_unlock(dentry);
29755 + di_read_unlock(dentry, flags);
29756 + si_read_unlock(dentry->d_sb);
29759 +void aufs_write_lock(struct dentry *dentry)
29761 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
29762 + di_write_lock_child(dentry);
29765 +void aufs_write_unlock(struct dentry *dentry)
29767 + di_write_unlock(dentry);
29768 + si_write_unlock(dentry->d_sb);
29771 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
29774 + unsigned int sigen;
29775 + struct super_block *sb;
29778 + err = si_read_lock(sb, flags);
29779 + if (unlikely(err))
29782 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
29784 + if (au_ftest_lock(flags, GEN)) {
29785 + sigen = au_sigen(sb);
29786 + err = au_digen_test(d1, sigen);
29787 + AuDebugOn(!err && au_dbrange_test(d1));
29789 + err = au_digen_test(d2, sigen);
29790 + AuDebugOn(!err && au_dbrange_test(d2));
29792 + if (unlikely(err))
29793 + aufs_read_and_write_unlock2(d1, d2);
29800 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
29802 + di_write_unlock2(d1, d2);
29803 + si_read_unlock(d1->d_sb);
29805 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
29806 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
29807 +++ linux/fs/aufs/super.c 2022-05-23 09:44:37.729839166 +0200
29809 +// SPDX-License-Identifier: GPL-2.0
29811 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29813 + * This program, aufs is free software; you can redistribute it and/or modify
29814 + * it under the terms of the GNU General Public License as published by
29815 + * the Free Software Foundation; either version 2 of the License, or
29816 + * (at your option) any later version.
29818 + * This program is distributed in the hope that it will be useful,
29819 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29820 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29821 + * GNU General Public License for more details.
29823 + * You should have received a copy of the GNU General Public License
29824 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29828 + * mount and super_block operations
29831 +#include <linux/iversion.h>
29832 +#include <linux/mm.h>
29833 +#include <linux/seq_file.h>
29834 +#include <linux/statfs.h>
29835 +#include <linux/vmalloc.h>
29839 + * super_operations
29841 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
29843 + struct au_icntnr *c;
29845 + c = au_cache_alloc_icntnr(sb);
29847 + au_icntnr_init(c);
29848 + inode_set_iversion(&c->vfs_inode, 1); /* sigen(sb); */
29849 + c->iinfo.ii_hinode = NULL;
29850 + return &c->vfs_inode;
29855 +static void aufs_destroy_inode(struct inode *inode)
29857 + if (!au_is_bad_inode(inode))
29858 + au_iinfo_fin(inode);
29861 +static void aufs_free_inode(struct inode *inode)
29863 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
29866 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
29868 + struct inode *inode;
29871 + inode = iget_locked(sb, ino);
29872 + if (unlikely(!inode)) {
29873 + inode = ERR_PTR(-ENOMEM);
29876 + if (!(inode->i_state & I_NEW))
29879 + err = au_xigen_new(inode);
29881 + err = au_iinfo_init(inode);
29883 + inode_inc_iversion(inode);
29885 + iget_failed(inode);
29886 + inode = ERR_PTR(err);
29890 + /* never return NULL */
29891 + AuDebugOn(!inode);
29892 + AuTraceErrPtr(inode);
29896 +/* lock free root dinfo */
29897 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
29900 + aufs_bindex_t bindex, bbot;
29901 + struct path path;
29902 + struct au_hdentry *hdp;
29903 + struct au_branch *br;
29904 + au_br_perm_str_t perm;
29907 + bbot = au_sbbot(sb);
29909 + hdp = au_hdentry(au_di(sb->s_root), bindex);
29910 + for (; !err && bindex <= bbot; bindex++, hdp++) {
29911 + br = au_sbr(sb, bindex);
29912 + path.mnt = au_br_mnt(br);
29913 + path.dentry = hdp->hd_dentry;
29914 + err = au_seq_path(seq, &path);
29916 + au_optstr_br_perm(&perm, br->br_perm);
29917 + seq_printf(seq, "=%s", perm.a);
29918 + if (bindex != bbot)
29919 + seq_putc(seq, ':');
29922 + if (unlikely(err || seq_has_overflowed(seq)))
29928 +static void au_gen_fmt(char *fmt, int len __maybe_unused, const char *pat,
29929 + const char *append)
29934 + while (*pat != ':')
29937 + strcpy(p, append);
29938 + AuDebugOn(strlen(fmt) >= len);
29941 +static void au_show_wbr_create(struct seq_file *m, int v,
29942 + struct au_sbinfo *sbinfo)
29946 + struct au_wbr_mfs *mfs;
29948 + AuRwMustAnyLock(&sbinfo->si_rwsem);
29950 + seq_puts(m, ",create=");
29951 + pat = au_optstr_wbr_create(v);
29952 + mfs = &sbinfo->si_wbr_mfs;
29954 + case AuWbrCreate_TDP:
29955 + case AuWbrCreate_RR:
29956 + case AuWbrCreate_MFS:
29957 + case AuWbrCreate_PMFS:
29958 + seq_puts(m, pat);
29960 + case AuWbrCreate_MFSRR:
29961 + case AuWbrCreate_TDMFS:
29962 + case AuWbrCreate_PMFSRR:
29963 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu");
29964 + seq_printf(m, fmt, mfs->mfsrr_watermark);
29966 + case AuWbrCreate_MFSV:
29967 + case AuWbrCreate_PMFSV:
29968 + au_gen_fmt(fmt, sizeof(fmt), pat, "%lu");
29969 + seq_printf(m, fmt,
29970 + jiffies_to_msecs(mfs->mfs_expire)
29973 + case AuWbrCreate_MFSRRV:
29974 + case AuWbrCreate_TDMFSV:
29975 + case AuWbrCreate_PMFSRRV:
29976 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu:%lu");
29977 + seq_printf(m, fmt, mfs->mfsrr_watermark,
29978 + jiffies_to_msecs(mfs->mfs_expire) / MSEC_PER_SEC);
29985 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
29987 +#ifdef CONFIG_SYSFS
29991 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
29992 + aufs_bindex_t bindex, brid;
29993 + struct qstr *name;
29995 + struct dentry *d, *h_root;
29996 + struct au_branch *br;
29998 + AuRwMustAnyLock(&sbinfo->si_rwsem);
30001 + f = au_sbi(sb)->si_xib;
30005 + /* stop printing the default xino path on the first writable branch */
30007 + bindex = au_xi_root(sb, f->f_path.dentry);
30008 + if (bindex >= 0) {
30009 + br = au_sbr_sb(sb, bindex);
30010 + h_root = au_br_dentry(br);
30013 + d = f->f_path.dentry;
30014 + name = &d->d_name;
30015 + /* safe ->d_parent because the file is unlinked */
30016 + if (d->d_parent == h_root
30017 + && name->len == len
30018 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
30021 + seq_puts(seq, ",xino=");
30022 + err = au_xino_path(seq, f);
30029 +/* seq_file will re-call me in case of too long string */
30030 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
30033 + unsigned int mnt_flags, v;
30034 + struct super_block *sb;
30035 + struct au_sbinfo *sbinfo;
30037 +#define AuBool(name, str) do { \
30038 + v = au_opt_test(mnt_flags, name); \
30039 + if (v != au_opt_test(AuOpt_Def, name)) \
30040 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
30043 +#define AuStr(name, str) do { \
30044 + v = mnt_flags & AuOptMask_##name; \
30045 + if (v != (AuOpt_Def & AuOptMask_##name)) \
30046 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
30049 +#define AuUInt(name, str, val) do { \
30050 + if (val != AUFS_##name##_DEF) \
30051 + seq_printf(m, "," #str "=%u", val); \
30054 + sb = dentry->d_sb;
30055 + if (sb->s_flags & SB_POSIXACL)
30056 + seq_puts(m, ",acl");
30057 +#if 0 /* reserved for future use */
30058 + if (sb->s_flags & SB_I_VERSION)
30059 + seq_puts(m, ",i_version");
30062 + /* lock free root dinfo */
30063 + si_noflush_read_lock(sb);
30064 + sbinfo = au_sbi(sb);
30065 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
30067 + mnt_flags = au_mntflags(sb);
30068 + if (au_opt_test(mnt_flags, XINO)) {
30069 + err = au_show_xino(m, sb);
30070 + if (unlikely(err))
30073 + seq_puts(m, ",noxino");
30075 + AuBool(TRUNC_XINO, trunc_xino);
30076 + AuStr(UDBA, udba);
30077 + AuBool(SHWH, shwh);
30078 + AuBool(PLINK, plink);
30079 + AuBool(DIO, dio);
30080 + AuBool(DIRPERM1, dirperm1);
30082 + v = sbinfo->si_wbr_create;
30083 + if (v != AuWbrCreate_Def)
30084 + au_show_wbr_create(m, v, sbinfo);
30086 + v = sbinfo->si_wbr_copyup;
30087 + if (v != AuWbrCopyup_Def)
30088 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
30090 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
30091 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
30092 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
30094 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
30096 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
30097 + AuUInt(RDCACHE, rdcache, v);
30099 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
30100 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
30102 + au_fhsm_show(m, sbinfo);
30104 + AuBool(DIRREN, dirren);
30105 + AuBool(SUM, sum);
30106 + /* AuBool(SUM_W, wsum); */
30107 + AuBool(WARN_PERM, warn_perm);
30108 + AuBool(VERBOSE, verbose);
30111 + /* be sure to print "br:" last */
30112 + if (!sysaufs_brs) {
30113 + seq_puts(m, ",br:");
30114 + au_show_brs(m, sb);
30116 + si_read_unlock(sb);
30124 +/* ---------------------------------------------------------------------- */
30126 +/* sum mode which returns the summation for statfs(2) */
30128 +static u64 au_add_till_max(u64 a, u64 b)
30136 + return ULLONG_MAX;
30139 +static u64 au_mul_till_max(u64 a, long mul)
30147 + return ULLONG_MAX;
30150 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
30153 + long bsize, factor;
30154 + u64 blocks, bfree, bavail, files, ffree;
30155 + aufs_bindex_t bbot, bindex, i;
30156 + unsigned char shared;
30157 + struct path h_path;
30158 + struct super_block *h_sb;
30161 + bsize = LONG_MAX;
30167 + bbot = au_sbbot(sb);
30168 + for (bindex = 0; bindex <= bbot; bindex++) {
30169 + h_path.mnt = au_sbr_mnt(sb, bindex);
30170 + h_sb = h_path.mnt->mnt_sb;
30172 + for (i = 0; !shared && i < bindex; i++)
30173 + shared = (au_sbr_sb(sb, i) == h_sb);
30177 + /* sb->s_root for NFS is unreliable */
30178 + h_path.dentry = h_path.mnt->mnt_root;
30179 + err = vfs_statfs(&h_path, buf);
30180 + if (unlikely(err))
30183 + if (bsize > buf->f_bsize) {
30185 + * we will reduce bsize, so we have to expand blocks
30186 + * etc. to match them again
30188 + factor = (bsize / buf->f_bsize);
30189 + blocks = au_mul_till_max(blocks, factor);
30190 + bfree = au_mul_till_max(bfree, factor);
30191 + bavail = au_mul_till_max(bavail, factor);
30192 + bsize = buf->f_bsize;
30195 + factor = (buf->f_bsize / bsize);
30196 + blocks = au_add_till_max(blocks,
30197 + au_mul_till_max(buf->f_blocks, factor));
30198 + bfree = au_add_till_max(bfree,
30199 + au_mul_till_max(buf->f_bfree, factor));
30200 + bavail = au_add_till_max(bavail,
30201 + au_mul_till_max(buf->f_bavail, factor));
30202 + files = au_add_till_max(files, buf->f_files);
30203 + ffree = au_add_till_max(ffree, buf->f_ffree);
30206 + buf->f_bsize = bsize;
30207 + buf->f_blocks = blocks;
30208 + buf->f_bfree = bfree;
30209 + buf->f_bavail = bavail;
30210 + buf->f_files = files;
30211 + buf->f_ffree = ffree;
30212 + buf->f_frsize = 0;
30218 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
30221 + struct path h_path;
30222 + struct super_block *sb;
30224 + /* lock free root dinfo */
30225 + sb = dentry->d_sb;
30226 + si_noflush_read_lock(sb);
30227 + if (!au_opt_test(au_mntflags(sb), SUM)) {
30228 + /* sb->s_root for NFS is unreliable */
30229 + h_path.mnt = au_sbr_mnt(sb, 0);
30230 + h_path.dentry = h_path.mnt->mnt_root;
30231 + err = vfs_statfs(&h_path, buf);
30233 + err = au_statfs_sum(sb, buf);
30234 + si_read_unlock(sb);
30237 + buf->f_type = AUFS_SUPER_MAGIC;
30238 + buf->f_namelen = AUFS_MAX_NAMELEN;
30239 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
30241 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
30246 +/* ---------------------------------------------------------------------- */
30248 +static int aufs_sync_fs(struct super_block *sb, int wait)
30251 + aufs_bindex_t bbot, bindex;
30252 + struct au_branch *br;
30253 + struct super_block *h_sb;
30256 + si_noflush_read_lock(sb);
30257 + bbot = au_sbbot(sb);
30258 + for (bindex = 0; bindex <= bbot; bindex++) {
30259 + br = au_sbr(sb, bindex);
30260 + if (!au_br_writable(br->br_perm))
30263 + h_sb = au_sbr_sb(sb, bindex);
30264 + e = vfsub_sync_filesystem(h_sb);
30265 + if (unlikely(e && !err))
30267 + /* go on even if an error happens */
30269 + si_read_unlock(sb);
30274 +/* ---------------------------------------------------------------------- */
30276 +/* final actions when unmounting a file system */
30277 +static void aufs_put_super(struct super_block *sb)
30279 + struct au_sbinfo *sbinfo;
30281 + sbinfo = au_sbi(sb);
30283 + kobject_put(&sbinfo->si_kobj);
30286 +/* ---------------------------------------------------------------------- */
30288 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30289 + struct super_block *sb, void *arg)
30292 + unsigned long long n, sz;
30299 + if (*hint > ULLONG_MAX / sizeof(array)) {
30300 + array = ERR_PTR(-EMFILE);
30301 + pr_err("hint %llu\n", *hint);
30305 + sz = sizeof(array) * *hint;
30306 + array = kzalloc(sz, GFP_NOFS);
30307 + if (unlikely(!array))
30308 + array = vzalloc(sz);
30309 + if (unlikely(!array)) {
30310 + array = ERR_PTR(-ENOMEM);
30314 + n = cb(sb, array, *hint, arg);
30315 + AuDebugOn(n > *hint);
30322 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
30323 + unsigned long long max __maybe_unused,
30326 + unsigned long long n;
30327 + struct inode **p, *inode;
30328 + struct list_head *head;
30333 + spin_lock(&sb->s_inode_list_lock);
30334 + list_for_each_entry(inode, head, i_sb_list) {
30335 + if (!au_is_bad_inode(inode)
30336 + && au_ii(inode)->ii_btop >= 0) {
30337 + spin_lock(&inode->i_lock);
30338 + if (atomic_read(&inode->i_count)) {
30342 + AuDebugOn(n > max);
30344 + spin_unlock(&inode->i_lock);
30347 + spin_unlock(&sb->s_inode_list_lock);
30352 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
30354 + struct au_sbinfo *sbi;
30356 + sbi = au_sbi(sb);
30357 + *max = au_lcnt_read(&sbi->si_ninodes, /*do_rev*/1);
30358 + return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
30361 +void au_iarray_free(struct inode **a, unsigned long long max)
30363 + unsigned long long ull;
30365 + for (ull = 0; ull < max; ull++)
30370 +/* ---------------------------------------------------------------------- */
30373 + * refresh dentry and inode at remount time.
30375 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
30376 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
30377 + struct dentry *parent)
30381 + di_write_lock_child(dentry);
30382 + di_read_lock_parent(parent, AuLock_IR);
30383 + err = au_refresh_dentry(dentry, parent);
30384 + if (!err && dir_flags)
30385 + au_hn_reset(d_inode(dentry), dir_flags);
30386 + di_read_unlock(parent, AuLock_IR);
30387 + di_write_unlock(dentry);
30392 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
30393 + struct au_sbinfo *sbinfo,
30394 + const unsigned int dir_flags, unsigned int do_idop)
30397 + struct dentry *parent;
30400 + parent = dget_parent(dentry);
30401 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
30402 + if (d_really_is_positive(dentry)) {
30403 + if (!d_is_dir(dentry))
30404 + err = au_do_refresh(dentry, /*dir_flags*/0,
30407 + err = au_do_refresh(dentry, dir_flags, parent);
30408 + if (unlikely(err))
30409 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
30412 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
30413 + AuDbgDentry(dentry);
30419 + au_refresh_dop(dentry, /*force_reval*/0);
30421 + au_refresh_dop(dentry, /*force_reval*/1);
30427 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
30429 + int err, i, j, ndentry, e;
30430 + unsigned int sigen;
30431 + struct au_dcsub_pages dpages;
30432 + struct au_dpage *dpage;
30433 + struct dentry **dentries, *d;
30434 + struct au_sbinfo *sbinfo;
30435 + struct dentry *root = sb->s_root;
30436 + const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
30439 + au_refresh_dop(root, /*force_reval*/0);
30441 + err = au_dpages_init(&dpages, GFP_NOFS);
30442 + if (unlikely(err))
30444 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
30445 + if (unlikely(err))
30448 + sigen = au_sigen(sb);
30449 + sbinfo = au_sbi(sb);
30450 + for (i = 0; i < dpages.ndpage; i++) {
30451 + dpage = dpages.dpages + i;
30452 + dentries = dpage->dentries;
30453 + ndentry = dpage->ndentry;
30454 + for (j = 0; j < ndentry; j++) {
30456 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
30458 + if (unlikely(e && !err))
30460 + /* go on even err */
30465 + au_dpages_free(&dpages);
30470 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
30473 + unsigned int sigen;
30474 + unsigned long long max, ull;
30475 + struct inode *inode, **array;
30477 + array = au_iarray_alloc(sb, &max);
30478 + err = PTR_ERR(array);
30479 + if (IS_ERR(array))
30483 + sigen = au_sigen(sb);
30484 + for (ull = 0; ull < max; ull++) {
30485 + inode = array[ull];
30486 + if (unlikely(!inode))
30490 + ii_write_lock_child(inode);
30491 + if (au_iigen(inode, NULL) != sigen) {
30492 + e = au_refresh_hinode_self(inode);
30493 + if (unlikely(e)) {
30494 + au_refresh_iop(inode, /*force_getattr*/1);
30495 + pr_err("error %d, i%lu\n", e, inode->i_ino);
30498 + /* go on even if err */
30501 + if (!e && do_idop)
30502 + au_refresh_iop(inode, /*force_getattr*/0);
30503 + ii_write_unlock(inode);
30506 + au_iarray_free(array, max);
30512 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
30515 + unsigned int udba;
30516 + aufs_bindex_t bindex, bbot;
30517 + struct dentry *root;
30518 + struct inode *inode;
30519 + struct au_branch *br;
30520 + struct au_sbinfo *sbi;
30522 + au_sigen_inc(sb);
30523 + sbi = au_sbi(sb);
30524 + au_fclr_si(sbi, FAILED_REFRESH_DIR);
30526 + root = sb->s_root;
30527 + DiMustNoWaiters(root);
30528 + inode = d_inode(root);
30529 + IiMustNoWaiters(inode);
30531 + udba = au_opt_udba(sb);
30532 + bbot = au_sbbot(sb);
30533 + for (bindex = 0; bindex <= bbot; bindex++) {
30534 + br = au_sbr(sb, bindex);
30535 + err = au_hnotify_reset_br(udba, br, br->br_perm);
30536 + if (unlikely(err))
30537 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
30539 + /* go on even if err */
30541 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
30544 + if (au_ftest_si(sbi, NO_DREVAL)) {
30545 + AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
30546 + sb->s_d_op = &aufs_dop_noreval;
30547 + AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
30548 + sbi->si_iop_array = aufs_iop_nogetattr;
30550 + AuDebugOn(sb->s_d_op == &aufs_dop);
30551 + sb->s_d_op = &aufs_dop;
30552 + AuDebugOn(sbi->si_iop_array == aufs_iop);
30553 + sbi->si_iop_array = aufs_iop;
30555 + pr_info("reset to %ps and %ps\n",
30556 + sb->s_d_op, sbi->si_iop_array);
30559 + di_write_unlock(root);
30560 + err = au_refresh_d(sb, do_idop);
30561 + e = au_refresh_i(sb, do_idop);
30562 + if (unlikely(e && !err))
30564 + /* aufs_write_lock() calls ..._child() */
30565 + di_write_lock_child(root);
30567 + au_cpup_attr_all(inode, /*force*/1);
30569 + if (unlikely(err))
30570 + AuIOErr("refresh failed, ignored, %d\n", err);
30573 +const struct super_operations aufs_sop = {
30574 + .alloc_inode = aufs_alloc_inode,
30575 + .destroy_inode = aufs_destroy_inode,
30576 + .free_inode = aufs_free_inode,
30577 + /* always deleting, no clearing */
30578 + .drop_inode = generic_delete_inode,
30579 + .show_options = aufs_show_options,
30580 + .statfs = aufs_statfs,
30581 + .put_super = aufs_put_super,
30582 + .sync_fs = aufs_sync_fs
30585 +/* ---------------------------------------------------------------------- */
30587 +int au_alloc_root(struct super_block *sb)
30590 + struct inode *inode;
30591 + struct dentry *root;
30594 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
30595 + err = PTR_ERR(inode);
30596 + if (IS_ERR(inode))
30599 + inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
30600 + inode->i_fop = &aufs_dir_fop;
30601 + inode->i_mode = S_IFDIR;
30602 + set_nlink(inode, 2);
30603 + unlock_new_inode(inode);
30605 + root = d_make_root(inode);
30606 + if (unlikely(!root))
30608 + err = PTR_ERR(root);
30609 + if (IS_ERR(root))
30612 + err = au_di_init(root);
30614 + sb->s_root = root;
30615 + return 0; /* success */
30623 +/* ---------------------------------------------------------------------- */
30625 +static void aufs_kill_sb(struct super_block *sb)
30627 + struct au_sbinfo *sbinfo;
30628 + struct dentry *root;
30630 + sbinfo = au_sbi(sb);
30634 + au_sbilist_del(sb);
30636 + root = sb->s_root;
30638 + aufs_write_lock(root);
30640 + __si_write_lock(sb);
30643 + if (sbinfo->si_wbr_create_ops->fin)
30644 + sbinfo->si_wbr_create_ops->fin(sb);
30645 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
30646 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
30647 + au_remount_refresh(sb, /*do_idop*/0);
30649 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
30650 + au_plink_put(sb, /*verbose*/1);
30653 + au_dr_opt_flush(sb);
30656 + aufs_write_unlock(root);
30658 + __si_write_unlock(sb);
30660 + sbinfo->si_sb = NULL;
30661 + au_nwt_flush(&sbinfo->si_nowait);
30664 + kill_anon_super(sb);
30667 +struct file_system_type aufs_fs_type = {
30668 + .name = AUFS_FSTYPE,
30669 + /* a race between rename and others */
30670 + .fs_flags = FS_RENAME_DOES_D_MOVE
30672 + /*| FS_ALLOW_IDMAP*/
30674 + .init_fs_context = aufs_fsctx_init,
30675 + .parameters = aufs_fsctx_paramspec,
30676 + .kill_sb = aufs_kill_sb,
30677 + /* no need to __module_get() and module_put(). */
30678 + .owner = THIS_MODULE,
30680 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
30681 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
30682 +++ linux/fs/aufs/super.h 2022-03-21 14:49:05.729966343 +0100
30684 +/* SPDX-License-Identifier: GPL-2.0 */
30686 + * Copyright (C) 2005-2021 Junjiro R. Okajima
30688 + * This program, aufs is free software; you can redistribute it and/or modify
30689 + * it under the terms of the GNU General Public License as published by
30690 + * the Free Software Foundation; either version 2 of the License, or
30691 + * (at your option) any later version.
30693 + * This program is distributed in the hope that it will be useful,
30694 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30695 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30696 + * GNU General Public License for more details.
30698 + * You should have received a copy of the GNU General Public License
30699 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30703 + * super_block operations
30706 +#ifndef __AUFS_SUPER_H__
30707 +#define __AUFS_SUPER_H__
30711 +#include <linux/fs.h>
30712 +#include <linux/kobject.h>
30715 +#include "rwsem.h"
30718 +/* policies to select one among multiple writable branches */
30719 +struct au_wbr_copyup_operations {
30720 + int (*copyup)(struct dentry *dentry);
30723 +#define AuWbr_DIR 1 /* target is a dir */
30724 +#define AuWbr_PARENT (1 << 1) /* always require a parent */
30726 +#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
30727 +#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
30728 +#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
30730 +struct au_wbr_create_operations {
30731 + int (*create)(struct dentry *dentry, unsigned int flags);
30732 + int (*init)(struct super_block *sb);
30733 + int (*fin)(struct super_block *sb);
30736 +struct au_wbr_mfs {
30737 + struct mutex mfs_lock; /* protect this structure */
30738 + unsigned long mfs_jiffy;
30739 + unsigned long mfs_expire;
30740 + aufs_bindex_t mfs_bindex;
30742 + unsigned long long mfsrr_bytes;
30743 + unsigned long long mfsrr_watermark;
30746 +#define AuPlink_NHASH 100
30747 +static inline int au_plink_hash(ino_t ino)
30749 + return ino % AuPlink_NHASH;
30752 +/* File-based Hierarchical Storage Management */
30754 +#ifdef CONFIG_AUFS_FHSM
30755 + /* allow only one process who can receive the notification */
30756 + spinlock_t fhsm_spin;
30758 + wait_queue_head_t fhsm_wqh;
30759 + atomic_t fhsm_readable;
30761 + /* these are protected by si_rwsem */
30762 + unsigned long fhsm_expire;
30763 + aufs_bindex_t fhsm_bottom;
30768 +struct au_sbinfo {
30769 + /* nowait tasks in the system-wide workqueue */
30770 + struct au_nowait_tasks si_nowait;
30773 + * tried sb->s_umount, but failed due to the dependency between i_mutex.
30774 + * rwsem for au_sbinfo is necessary.
30776 + struct au_rwsem si_rwsem;
30779 + * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
30782 + au_lcnt_t si_ninodes, si_nfiles;
30784 + /* branch management */
30785 + unsigned int si_generation;
30787 + /* see AuSi_ flags */
30788 + unsigned char au_si_status;
30790 + aufs_bindex_t si_bbot;
30792 + /* dirty trick to keep br_id plus */
30793 + unsigned int si_last_br_id :
30794 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
30795 + struct au_branch **si_branch;
30797 + /* policy to select a writable branch */
30798 + unsigned char si_wbr_copyup;
30799 + unsigned char si_wbr_create;
30800 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
30801 + struct au_wbr_create_operations *si_wbr_create_ops;
30803 + /* round robin */
30804 + atomic_t si_wbr_rr_next;
30806 + /* most free space */
30807 + struct au_wbr_mfs si_wbr_mfs;
30809 + /* File-based Hierarchical Storage Management */
30810 + struct au_fhsm si_fhsm;
30812 + /* mount flags */
30813 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
30814 + unsigned int si_mntflags;
30816 + /* external inode number (bitmap and translation table) */
30817 + loff_t si_ximaxent; /* max entries in a xino */
30819 + struct file *si_xib;
30820 + struct mutex si_xib_mtx; /* protect xib members */
30821 + unsigned long *si_xib_buf;
30822 + unsigned long si_xib_last_pindex;
30823 + int si_xib_next_bit;
30825 + unsigned long si_xino_jiffy;
30826 + unsigned long si_xino_expire;
30827 + /* reserved for future use */
30828 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
30830 +#ifdef CONFIG_AUFS_EXPORT
30831 + /* i_generation */
30832 + /* todo: make xigen file an array to support many inode numbers */
30833 + struct file *si_xigen;
30834 + atomic_t si_xigen_next;
30837 + /* dirty trick to support atomic_open */
30838 + struct hlist_bl_head si_aopen;
30840 + /* vdir parameters */
30841 + unsigned long si_rdcache; /* max cache time in jiffies */
30842 + unsigned int si_rdblk; /* deblk size */
30843 + unsigned int si_rdhash; /* hash size */
30846 + * If the number of whiteouts are larger than si_dirwh, leave all of
30847 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
30848 + * future fsck.aufs or kernel thread will remove them later.
30849 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
30851 + unsigned int si_dirwh;
30853 + /* pseudo_link list */
30854 + struct hlist_bl_head si_plink[AuPlink_NHASH];
30855 + wait_queue_head_t si_plink_wq;
30856 + spinlock_t si_plink_maint_lock;
30857 + pid_t si_plink_maint_pid;
30860 + struct hlist_bl_head si_files;
30862 + /* with/without getattr, brother of sb->s_d_op */
30863 + const struct inode_operations *si_iop_array;
30866 + * sysfs and lifetime management.
30867 + * this is not a small structure and it may be a waste of memory in case
30868 + * of sysfs is disabled, particularly when many aufs-es are mounted.
30869 + * but using sysfs is majority.
30871 + struct kobject si_kobj;
30872 +#ifdef CONFIG_DEBUG_FS
30873 + struct dentry *si_dbgaufs;
30874 + struct dentry *si_dbgaufs_plink;
30875 + struct dentry *si_dbgaufs_xib;
30876 +#ifdef CONFIG_AUFS_EXPORT
30877 + struct dentry *si_dbgaufs_xigen;
30881 +#ifdef CONFIG_AUFS_SBILIST
30882 + struct hlist_bl_node si_list;
30885 + /* dirty, necessary for unmounting, sysfs and sysrq */
30886 + struct super_block *si_sb;
30889 +/* sbinfo status flags */
30891 + * set true when refresh_dirs() failed at remount time.
30892 + * then try refreshing dirs at access time again.
30893 + * if it is false, refreshing dirs at access time is unnecessary
30895 +#define AuSi_FAILED_REFRESH_DIR 1
30896 +#define AuSi_FHSM (1 << 1) /* fhsm is active now */
30897 +#define AuSi_NO_DREVAL (1 << 2) /* disable all d_revalidate */
30899 +#ifndef CONFIG_AUFS_FHSM
30901 +#define AuSi_FHSM 0
30904 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
30905 + unsigned int flag)
30907 + AuRwMustAnyLock(&sbi->si_rwsem);
30908 + return sbi->au_si_status & flag;
30910 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
30911 +#define au_fset_si(sbinfo, name) do { \
30912 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30913 + (sbinfo)->au_si_status |= AuSi_##name; \
30915 +#define au_fclr_si(sbinfo, name) do { \
30916 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30917 + (sbinfo)->au_si_status &= ~AuSi_##name; \
30920 +/* ---------------------------------------------------------------------- */
30922 +/* policy to select one among writable branches */
30923 +#define AuWbrCopyup(sbinfo, ...) \
30924 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
30925 +#define AuWbrCreate(sbinfo, ...) \
30926 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
30928 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
30929 +#define AuLock_DW 1 /* write-lock dentry */
30930 +#define AuLock_IR (1 << 1) /* read-lock inode */
30931 +#define AuLock_IW (1 << 2) /* write-lock inode */
30932 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
30933 +#define AuLock_DIRS (1 << 4) /* target is a pair of dirs */
30934 + /* except RENAME_EXCHANGE */
30935 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
30936 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
30937 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
30938 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
30939 +#define au_fset_lock(flags, name) \
30940 + do { (flags) |= AuLock_##name; } while (0)
30941 +#define au_fclr_lock(flags, name) \
30942 + do { (flags) &= ~AuLock_##name; } while (0)
30944 +/* ---------------------------------------------------------------------- */
30947 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
30949 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
30950 + unsigned long long max, void *arg);
30951 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30952 + struct super_block *sb, void *arg);
30953 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
30954 +void au_iarray_free(struct inode **a, unsigned long long max);
30956 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop);
30957 +extern const struct super_operations aufs_sop;
30958 +int au_alloc_root(struct super_block *sb);
30959 +extern struct file_system_type aufs_fs_type;
30962 +void au_si_free(struct kobject *kobj);
30963 +struct au_sbinfo *au_si_alloc(struct super_block *sb);
30964 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink);
30966 +unsigned int au_sigen_inc(struct super_block *sb);
30967 +aufs_bindex_t au_new_br_id(struct super_block *sb);
30969 +int si_read_lock(struct super_block *sb, int flags);
30970 +int si_write_lock(struct super_block *sb, int flags);
30971 +int aufs_read_lock(struct dentry *dentry, int flags);
30972 +void aufs_read_unlock(struct dentry *dentry, int flags);
30973 +void aufs_write_lock(struct dentry *dentry);
30974 +void aufs_write_unlock(struct dentry *dentry);
30975 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
30976 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
30978 +/* wbr_policy.c */
30979 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
30980 +extern struct au_wbr_create_operations au_wbr_create_ops[];
30981 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
30982 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
30983 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop);
30986 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
30988 +#ifdef CONFIG_AUFS_FHSM
30991 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
30995 + spin_lock(&fhsm->fhsm_spin);
30996 + pid = fhsm->fhsm_pid;
30997 + spin_unlock(&fhsm->fhsm_spin);
31002 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
31003 +void au_fhsm_wrote_all(struct super_block *sb, int force);
31004 +int au_fhsm_fd(struct super_block *sb, int oflags);
31005 +int au_fhsm_br_alloc(struct au_branch *br);
31006 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
31007 +void au_fhsm_fin(struct super_block *sb);
31008 +void au_fhsm_init(struct au_sbinfo *sbinfo);
31009 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
31010 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
31012 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
31014 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
31015 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
31016 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
31017 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
31018 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
31019 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
31020 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
31021 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
31022 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
31025 +/* ---------------------------------------------------------------------- */
31027 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
31029 + return sb->s_fs_info;
31032 +/* ---------------------------------------------------------------------- */
31034 +#ifdef CONFIG_AUFS_EXPORT
31035 +int au_test_nfsd(void);
31036 +void au_export_init(struct super_block *sb);
31037 +void au_xigen_inc(struct inode *inode);
31038 +int au_xigen_new(struct inode *inode);
31039 +int au_xigen_set(struct super_block *sb, struct path *path);
31040 +void au_xigen_clr(struct super_block *sb);
31042 +static inline int au_busy_or_stale(void)
31044 + if (!au_test_nfsd())
31049 +AuStubInt0(au_test_nfsd, void)
31050 +AuStubVoid(au_export_init, struct super_block *sb)
31051 +AuStubVoid(au_xigen_inc, struct inode *inode)
31052 +AuStubInt0(au_xigen_new, struct inode *inode)
31053 +AuStubInt0(au_xigen_set, struct super_block *sb, struct path *path)
31054 +AuStubVoid(au_xigen_clr, struct super_block *sb)
31055 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
31056 +#endif /* CONFIG_AUFS_EXPORT */
31058 +/* ---------------------------------------------------------------------- */
31060 +#ifdef CONFIG_AUFS_SBILIST
31062 +extern struct hlist_bl_head au_sbilist;
31064 +static inline void au_sbilist_init(void)
31066 + INIT_HLIST_BL_HEAD(&au_sbilist);
31069 +static inline void au_sbilist_add(struct super_block *sb)
31071 + au_hbl_add(&au_sbi(sb)->si_list, &au_sbilist);
31074 +static inline void au_sbilist_del(struct super_block *sb)
31076 + au_hbl_del(&au_sbi(sb)->si_list, &au_sbilist);
31079 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
31080 +static inline void au_sbilist_lock(void)
31082 + hlist_bl_lock(&au_sbilist);
31085 +static inline void au_sbilist_unlock(void)
31087 + hlist_bl_unlock(&au_sbilist);
31089 +#define AuGFP_SBILIST GFP_ATOMIC
31091 +AuStubVoid(au_sbilist_lock, void)
31092 +AuStubVoid(au_sbilist_unlock, void)
31093 +#define AuGFP_SBILIST GFP_NOFS
31094 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
31096 +AuStubVoid(au_sbilist_init, void)
31097 +AuStubVoid(au_sbilist_add, struct super_block *sb)
31098 +AuStubVoid(au_sbilist_del, struct super_block *sb)
31099 +AuStubVoid(au_sbilist_lock, void)
31100 +AuStubVoid(au_sbilist_unlock, void)
31101 +#define AuGFP_SBILIST GFP_NOFS
31104 +/* ---------------------------------------------------------------------- */
31106 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
31109 + * This function is a dynamic '__init' function actually,
31110 + * so the tiny check for si_rwsem is unnecessary.
31112 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
31113 +#ifdef CONFIG_DEBUG_FS
31114 + sbinfo->si_dbgaufs = NULL;
31115 + sbinfo->si_dbgaufs_plink = NULL;
31116 + sbinfo->si_dbgaufs_xib = NULL;
31117 +#ifdef CONFIG_AUFS_EXPORT
31118 + sbinfo->si_dbgaufs_xigen = NULL;
31123 +/* ---------------------------------------------------------------------- */
31125 +/* current->atomic_flags */
31126 +/* this value should never corrupt the ones defined in linux/sched.h */
31127 +#define PFA_AUFS 0x10
31129 +TASK_PFA_TEST(AUFS, test_aufs) /* task_test_aufs */
31130 +TASK_PFA_SET(AUFS, aufs) /* task_set_aufs */
31131 +TASK_PFA_CLEAR(AUFS, aufs) /* task_clear_aufs */
31133 +static inline int si_pid_test(struct super_block *sb)
31135 + return !!task_test_aufs(current);
31138 +static inline void si_pid_clr(struct super_block *sb)
31140 + AuDebugOn(!task_test_aufs(current));
31141 + task_clear_aufs(current);
31144 +static inline void si_pid_set(struct super_block *sb)
31146 + AuDebugOn(task_test_aufs(current));
31147 + task_set_aufs(current);
31150 +/* ---------------------------------------------------------------------- */
31152 +/* lock superblock. mainly for entry point functions */
31153 +#define __si_read_lock(sb) au_rw_read_lock(&au_sbi(sb)->si_rwsem)
31154 +#define __si_write_lock(sb) au_rw_write_lock(&au_sbi(sb)->si_rwsem)
31155 +#define __si_read_trylock(sb) au_rw_read_trylock(&au_sbi(sb)->si_rwsem)
31156 +#define __si_write_trylock(sb) au_rw_write_trylock(&au_sbi(sb)->si_rwsem)
31158 +#define __si_read_trylock_nested(sb) \
31159 + au_rw_read_trylock_nested(&au_sbi(sb)->si_rwsem)
31160 +#define __si_write_trylock_nested(sb) \
31161 + au_rw_write_trylock_nested(&au_sbi(sb)->si_rwsem)
31164 +#define __si_read_unlock(sb) au_rw_read_unlock(&au_sbi(sb)->si_rwsem)
31165 +#define __si_write_unlock(sb) au_rw_write_unlock(&au_sbi(sb)->si_rwsem)
31166 +#define __si_downgrade_lock(sb) au_rw_dgrade_lock(&au_sbi(sb)->si_rwsem)
31168 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
31169 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
31170 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
31172 +static inline void si_noflush_read_lock(struct super_block *sb)
31174 + __si_read_lock(sb);
31178 +static inline int si_noflush_read_trylock(struct super_block *sb)
31182 + locked = __si_read_trylock(sb);
31188 +static inline void si_noflush_write_lock(struct super_block *sb)
31190 + __si_write_lock(sb);
31194 +static inline int si_noflush_write_trylock(struct super_block *sb)
31198 + locked = __si_write_trylock(sb);
31204 +#if 0 /* reserved */
31205 +static inline int si_read_trylock(struct super_block *sb, int flags)
31207 + if (au_ftest_lock(flags, FLUSH))
31208 + au_nwt_flush(&au_sbi(sb)->si_nowait);
31209 + return si_noflush_read_trylock(sb);
31213 +static inline void si_read_unlock(struct super_block *sb)
31216 + __si_read_unlock(sb);
31219 +#if 0 /* reserved */
31220 +static inline int si_write_trylock(struct super_block *sb, int flags)
31222 + if (au_ftest_lock(flags, FLUSH))
31223 + au_nwt_flush(&au_sbi(sb)->si_nowait);
31224 + return si_noflush_write_trylock(sb);
31228 +static inline void si_write_unlock(struct super_block *sb)
31231 + __si_write_unlock(sb);
31234 +#if 0 /* reserved */
31235 +static inline void si_downgrade_lock(struct super_block *sb)
31237 + __si_downgrade_lock(sb);
31241 +/* ---------------------------------------------------------------------- */
31243 +static inline aufs_bindex_t au_sbbot(struct super_block *sb)
31245 + SiMustAnyLock(sb);
31246 + return au_sbi(sb)->si_bbot;
31249 +static inline unsigned int au_mntflags(struct super_block *sb)
31251 + SiMustAnyLock(sb);
31252 + return au_sbi(sb)->si_mntflags;
31255 +static inline unsigned int au_sigen(struct super_block *sb)
31257 + SiMustAnyLock(sb);
31258 + return au_sbi(sb)->si_generation;
31261 +static inline struct au_branch *au_sbr(struct super_block *sb,
31262 + aufs_bindex_t bindex)
31264 + SiMustAnyLock(sb);
31265 + return au_sbi(sb)->si_branch[0 + bindex];
31268 +static inline loff_t au_xi_maxent(struct super_block *sb)
31270 + SiMustAnyLock(sb);
31271 + return au_sbi(sb)->si_ximaxent;
31274 +#endif /* __KERNEL__ */
31275 +#endif /* __AUFS_SUPER_H__ */
31276 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
31277 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
31278 +++ linux/fs/aufs/sysaufs.c 2022-05-23 09:44:37.729839166 +0200
31280 +// SPDX-License-Identifier: GPL-2.0
31282 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31284 + * This program, aufs is free software; you can redistribute it and/or modify
31285 + * it under the terms of the GNU General Public License as published by
31286 + * the Free Software Foundation; either version 2 of the License, or
31287 + * (at your option) any later version.
31289 + * This program is distributed in the hope that it will be useful,
31290 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31291 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31292 + * GNU General Public License for more details.
31294 + * You should have received a copy of the GNU General Public License
31295 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31299 + * sysfs interface and lifetime management
31300 + * they are necessary regardless sysfs is disabled.
31303 +#include <linux/random.h>
31306 +unsigned long sysaufs_si_mask;
31307 +struct kset *sysaufs_kset;
31309 +#define AuSiAttr(_name) { \
31310 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
31311 + .show = sysaufs_si_##_name, \
31314 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
31315 +struct attribute *sysaufs_si_attrs[] = {
31316 + &sysaufs_si_attr_xi_path.attr,
31319 +ATTRIBUTE_GROUPS(sysaufs_si);
31321 +static const struct sysfs_ops au_sbi_ops = {
31322 + .show = sysaufs_si_show
31325 +static struct kobj_type au_sbi_ktype = {
31326 + .release = au_si_free,
31327 + .sysfs_ops = &au_sbi_ops,
31328 + .default_groups = sysaufs_si_groups
31331 +/* ---------------------------------------------------------------------- */
31333 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
31337 + sbinfo->si_kobj.kset = sysaufs_kset;
31338 + /* cf. sysaufs_name() */
31339 + err = kobject_init_and_add
31340 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
31341 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
31346 +void sysaufs_fin(void)
31348 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31349 + kset_unregister(sysaufs_kset);
31352 +int __init sysaufs_init(void)
31357 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
31358 + } while (!sysaufs_si_mask);
31361 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
31362 + if (unlikely(!sysaufs_kset))
31364 + err = PTR_ERR(sysaufs_kset);
31365 + if (IS_ERR(sysaufs_kset))
31367 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31368 + if (unlikely(err))
31369 + kset_unregister(sysaufs_kset);
31374 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
31375 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
31376 +++ linux/fs/aufs/sysaufs.h 2022-03-21 14:49:05.729966343 +0100
31378 +/* SPDX-License-Identifier: GPL-2.0 */
31380 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31382 + * This program, aufs is free software; you can redistribute it and/or modify
31383 + * it under the terms of the GNU General Public License as published by
31384 + * the Free Software Foundation; either version 2 of the License, or
31385 + * (at your option) any later version.
31387 + * This program is distributed in the hope that it will be useful,
31388 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31389 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31390 + * GNU General Public License for more details.
31392 + * You should have received a copy of the GNU General Public License
31393 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31397 + * sysfs interface and mount lifetime management
31400 +#ifndef __SYSAUFS_H__
31401 +#define __SYSAUFS_H__
31405 +#include <linux/sysfs.h>
31406 +#include "module.h"
31408 +struct super_block;
31411 +struct sysaufs_si_attr {
31412 + struct attribute attr;
31413 + int (*show)(struct seq_file *seq, struct super_block *sb);
31416 +/* ---------------------------------------------------------------------- */
31419 +extern unsigned long sysaufs_si_mask;
31420 +extern struct kset *sysaufs_kset;
31421 +extern struct attribute *sysaufs_si_attrs[];
31422 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
31423 +int __init sysaufs_init(void);
31424 +void sysaufs_fin(void);
31426 +/* ---------------------------------------------------------------------- */
31428 +/* some people doesn't like to show a pointer in kernel */
31429 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
31431 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
31434 +#define SysaufsSiNamePrefix "si_"
31435 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
31436 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
31438 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
31439 + sysaufs_si_id(sbinfo));
31443 +#ifdef CONFIG_SYSFS
31445 +extern struct attribute_group *sysaufs_attr_group;
31447 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
31448 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31450 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
31451 +#ifdef CONFIG_COMPAT
31452 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
31455 +void sysaufs_br_init(struct au_branch *br);
31456 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
31457 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
31459 +#define sysaufs_brs_init() do {} while (0)
31462 +#define sysaufs_attr_group NULL
31464 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
31465 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
31466 + struct attribute *attr, char *buf)
31467 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
31468 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
31469 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
31471 +static inline void sysaufs_brs_init(void)
31476 +#endif /* CONFIG_SYSFS */
31478 +#endif /* __KERNEL__ */
31479 +#endif /* __SYSAUFS_H__ */
31480 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
31481 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
31482 +++ linux/fs/aufs/sysfs.c 2022-03-21 14:49:05.729966343 +0100
31484 +// SPDX-License-Identifier: GPL-2.0
31486 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31488 + * This program, aufs is free software; you can redistribute it and/or modify
31489 + * it under the terms of the GNU General Public License as published by
31490 + * the Free Software Foundation; either version 2 of the License, or
31491 + * (at your option) any later version.
31493 + * This program is distributed in the hope that it will be useful,
31494 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31495 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31496 + * GNU General Public License for more details.
31498 + * You should have received a copy of the GNU General Public License
31499 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31503 + * sysfs interface
31506 +#include <linux/compat.h>
31507 +#include <linux/seq_file.h>
31510 +#ifdef CONFIG_AUFS_FS_MODULE
31511 +/* this entry violates the "one line per file" policy of sysfs */
31512 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
31516 + static char *conf =
31517 +/* this file is generated at compiling */
31518 +#include "conf.str"
31521 + err = snprintf(buf, PAGE_SIZE, conf);
31522 + if (unlikely(err >= PAGE_SIZE))
31527 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
31530 +static struct attribute *au_attr[] = {
31531 +#ifdef CONFIG_AUFS_FS_MODULE
31532 + &au_config_attr.attr,
31534 + NULL, /* need to NULL terminate the list of attributes */
31537 +static struct attribute_group sysaufs_attr_group_body = {
31541 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
31543 +/* ---------------------------------------------------------------------- */
31545 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
31549 + SiMustAnyLock(sb);
31552 + if (au_opt_test(au_mntflags(sb), XINO)) {
31553 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
31554 + seq_putc(seq, '\n');
31560 + * the lifetime of branch is independent from the entry under sysfs.
31561 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
31564 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
31565 + aufs_bindex_t bindex, int idx)
31568 + struct path path;
31569 + struct dentry *root;
31570 + struct au_branch *br;
31571 + au_br_perm_str_t perm;
31573 + AuDbg("b%d\n", bindex);
31576 + root = sb->s_root;
31577 + di_read_lock_parent(root, !AuLock_IR);
31578 + br = au_sbr(sb, bindex);
31581 + case AuBrSysfs_BR:
31582 + path.mnt = au_br_mnt(br);
31583 + path.dentry = au_h_dptr(root, bindex);
31584 + err = au_seq_path(seq, &path);
31586 + au_optstr_br_perm(&perm, br->br_perm);
31587 + seq_printf(seq, "=%s\n", perm.a);
31590 + case AuBrSysfs_BRID:
31591 + seq_printf(seq, "%d\n", br->br_id);
31594 + di_read_unlock(root, !AuLock_IR);
31595 + if (unlikely(err || seq_has_overflowed(seq)))
31601 +/* ---------------------------------------------------------------------- */
31603 +static struct seq_file *au_seq(char *p, ssize_t len)
31605 + struct seq_file *seq;
31607 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
31609 + /* mutex_init(&seq.lock); */
31612 + return seq; /* success */
31615 + seq = ERR_PTR(-ENOMEM);
31619 +#define SysaufsBr_PREFIX "br"
31620 +#define SysaufsBrid_PREFIX "brid"
31622 +/* todo: file size may exceed PAGE_SIZE */
31623 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31629 + aufs_bindex_t bbot;
31630 + struct au_sbinfo *sbinfo;
31631 + struct super_block *sb;
31632 + struct seq_file *seq;
31634 + struct attribute **cattr;
31636 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
31637 + sb = sbinfo->si_sb;
31640 + * prevent a race condition between sysfs and aufs.
31641 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
31642 + * prohibits maintaining the sysfs entries.
31643 + * hew we acquire read lock after sysfs_get_active_two().
31644 + * on the other hand, the remount process may maintain the sysfs/aufs
31645 + * entries after acquiring write lock.
31646 + * it can cause a deadlock.
31647 + * simply we gave up processing read here.
31650 + if (unlikely(!si_noflush_read_trylock(sb)))
31653 + seq = au_seq(buf, PAGE_SIZE);
31654 + err = PTR_ERR(seq);
31658 + name = (void *)attr->name;
31659 + cattr = sysaufs_si_attrs;
31661 + if (!strcmp(name, (*cattr)->name)) {
31662 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
31669 + if (!strncmp(name, SysaufsBrid_PREFIX,
31670 + sizeof(SysaufsBrid_PREFIX) - 1)) {
31671 + idx = AuBrSysfs_BRID;
31672 + name += sizeof(SysaufsBrid_PREFIX) - 1;
31673 + } else if (!strncmp(name, SysaufsBr_PREFIX,
31674 + sizeof(SysaufsBr_PREFIX) - 1)) {
31675 + idx = AuBrSysfs_BR;
31676 + name += sizeof(SysaufsBr_PREFIX) - 1;
31680 + err = kstrtol(name, 10, &l);
31682 + bbot = au_sbbot(sb);
31684 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
31691 + err = seq->count;
31692 + /* sysfs limit */
31693 + if (unlikely(err == PAGE_SIZE))
31696 + au_kfree_rcu(seq);
31698 + si_read_unlock(sb);
31703 +/* ---------------------------------------------------------------------- */
31705 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
31709 + aufs_bindex_t bindex, bbot;
31712 + struct seq_file *seq;
31713 + struct au_branch *br;
31715 + si_read_lock(sb, AuLock_FLUSH);
31716 + bbot = au_sbbot(sb);
31722 + buf = (void *)__get_free_page(GFP_NOFS);
31723 + if (unlikely(!buf))
31726 + seq = au_seq(buf, PAGE_SIZE);
31727 + err = PTR_ERR(seq);
31731 + sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
31732 + for (bindex = 0; bindex <= bbot; bindex++, arg++) {
31733 + /* VERIFY_WRITE */
31734 + err = !access_ok(arg, sizeof(*arg));
31735 + if (unlikely(err))
31738 + br = au_sbr(sb, bindex);
31739 + brid = br->br_id;
31740 + BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
31741 + err = __put_user(brid, &arg->id);
31742 + if (unlikely(err))
31745 + BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
31746 + err = __put_user(br->br_perm, &arg->perm);
31747 + if (unlikely(err))
31750 + err = au_seq_path(seq, &br->br_path);
31751 + if (unlikely(err))
31753 + seq_putc(seq, '\0');
31754 + if (!seq_has_overflowed(seq)) {
31755 + err = copy_to_user(arg->path, seq->buf, seq->count);
31757 + if (unlikely(err))
31764 + if (unlikely(err))
31768 + au_kfree_rcu(seq);
31770 + free_page((unsigned long)buf);
31772 + si_read_unlock(sb);
31776 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
31778 + return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
31781 +#ifdef CONFIG_COMPAT
31782 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
31784 + return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
31788 +/* ---------------------------------------------------------------------- */
31790 +void sysaufs_br_init(struct au_branch *br)
31793 + struct au_brsysfs *br_sysfs;
31794 + struct attribute *attr;
31796 + br_sysfs = br->br_sysfs;
31797 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31798 + attr = &br_sysfs->attr;
31799 + sysfs_attr_init(attr);
31800 + attr->name = br_sysfs->name;
31801 + attr->mode = 0444;
31806 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
31808 + struct au_branch *br;
31809 + struct kobject *kobj;
31810 + struct au_brsysfs *br_sysfs;
31812 + aufs_bindex_t bbot;
31814 + if (!sysaufs_brs)
31817 + kobj = &au_sbi(sb)->si_kobj;
31818 + bbot = au_sbbot(sb);
31819 + for (; bindex <= bbot; bindex++) {
31820 + br = au_sbr(sb, bindex);
31821 + br_sysfs = br->br_sysfs;
31822 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31823 + sysfs_remove_file(kobj, &br_sysfs->attr);
31829 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
31832 + aufs_bindex_t bbot;
31833 + struct kobject *kobj;
31834 + struct au_branch *br;
31835 + struct au_brsysfs *br_sysfs;
31837 + if (!sysaufs_brs)
31840 + kobj = &au_sbi(sb)->si_kobj;
31841 + bbot = au_sbbot(sb);
31842 + for (; bindex <= bbot; bindex++) {
31843 + br = au_sbr(sb, bindex);
31844 + br_sysfs = br->br_sysfs;
31845 + snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
31846 + SysaufsBr_PREFIX "%d", bindex);
31847 + snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
31848 + SysaufsBrid_PREFIX "%d", bindex);
31849 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31850 + err = sysfs_create_file(kobj, &br_sysfs->attr);
31851 + if (unlikely(err))
31852 + pr_warn("failed %s under sysfs(%d)\n",
31853 + br_sysfs->name, err);
31858 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
31859 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
31860 +++ linux/fs/aufs/sysrq.c 2022-03-21 14:49:05.729966343 +0100
31862 +// SPDX-License-Identifier: GPL-2.0
31864 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31866 + * This program, aufs is free software; you can redistribute it and/or modify
31867 + * it under the terms of the GNU General Public License as published by
31868 + * the Free Software Foundation; either version 2 of the License, or
31869 + * (at your option) any later version.
31871 + * This program is distributed in the hope that it will be useful,
31872 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31873 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31874 + * GNU General Public License for more details.
31876 + * You should have received a copy of the GNU General Public License
31877 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31881 + * magic sysrq handler
31884 +/* #include <linux/sysrq.h> */
31885 +#include <linux/writeback.h>
31888 +/* ---------------------------------------------------------------------- */
31890 +static void sysrq_sb(struct super_block *sb)
31893 + struct au_sbinfo *sbinfo;
31894 + struct file *file;
31895 + struct hlist_bl_head *files;
31896 + struct hlist_bl_node *pos;
31897 + struct au_finfo *finfo;
31900 + plevel = au_plevel;
31901 + au_plevel = KERN_WARNING;
31903 + /* since we define pr_fmt, call printk directly */
31904 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
31906 + sbinfo = au_sbi(sb);
31907 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
31908 + pr("superblock\n");
31911 +#if 0 /* reserved */
31913 + int err, i, j, ndentry;
31914 + struct au_dcsub_pages dpages;
31915 + struct au_dpage *dpage;
31917 + err = au_dpages_init(&dpages, GFP_ATOMIC);
31918 + if (unlikely(err))
31920 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
31922 + for (i = 0; i < dpages.ndpage; i++) {
31923 + dpage = dpages.dpages + i;
31924 + ndentry = dpage->ndentry;
31925 + for (j = 0; j < ndentry; j++)
31926 + au_dpri_dentry(dpage->dentries[j]);
31928 + au_dpages_free(&dpages);
31932 + pr("isolated inode\n");
31933 + spin_lock(&sb->s_inode_list_lock);
31934 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
31935 + spin_lock(&i->i_lock);
31936 + if (hlist_empty(&i->i_dentry))
31937 + au_dpri_inode(i);
31938 + spin_unlock(&i->i_lock);
31940 + spin_unlock(&sb->s_inode_list_lock);
31943 + files = &au_sbi(sb)->si_files;
31944 + hlist_bl_lock(files);
31945 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
31948 + file = finfo->fi_file;
31949 + mode = file_inode(file)->i_mode;
31950 + if (!special_file(mode))
31951 + au_dpri_file(file);
31953 + hlist_bl_unlock(files);
31957 + au_plevel = plevel;
31960 +/* ---------------------------------------------------------------------- */
31962 +/* module parameter */
31963 +static char *aufs_sysrq_key = "a";
31964 +module_param_named(sysrq, aufs_sysrq_key, charp, 0444);
31965 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
31967 +static void au_sysrq(int key __maybe_unused)
31969 + struct au_sbinfo *sbinfo;
31970 + struct hlist_bl_node *pos;
31973 + au_sbilist_lock();
31974 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
31975 + sysrq_sb(sbinfo->si_sb);
31976 + au_sbilist_unlock();
31980 +static struct sysrq_key_op au_sysrq_op = {
31981 + .handler = au_sysrq,
31982 + .help_msg = "Aufs",
31983 + .action_msg = "Aufs",
31984 + .enable_mask = SYSRQ_ENABLE_DUMP
31987 +/* ---------------------------------------------------------------------- */
31989 +int __init au_sysrq_init(void)
31995 + key = *aufs_sysrq_key;
31996 + if ('a' <= key && key <= 'z')
31997 + err = register_sysrq_key(key, &au_sysrq_op);
31998 + if (unlikely(err))
31999 + pr_err("err %d, sysrq=%c\n", err, key);
32003 +void au_sysrq_fin(void)
32007 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
32008 + if (unlikely(err))
32009 + pr_err("err %d (ignored)\n", err);
32011 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
32012 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
32013 +++ linux/fs/aufs/vdir.c 2022-03-21 14:49:05.733299677 +0100
32015 +// SPDX-License-Identifier: GPL-2.0
32017 + * Copyright (C) 2005-2021 Junjiro R. Okajima
32019 + * This program, aufs is free software; you can redistribute it and/or modify
32020 + * it under the terms of the GNU General Public License as published by
32021 + * the Free Software Foundation; either version 2 of the License, or
32022 + * (at your option) any later version.
32024 + * This program is distributed in the hope that it will be useful,
32025 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32026 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32027 + * GNU General Public License for more details.
32029 + * You should have received a copy of the GNU General Public License
32030 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32034 + * virtual or vertical directory
32037 +#include <linux/iversion.h>
32040 +static unsigned int calc_size(int nlen)
32042 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
32045 +static int set_deblk_end(union au_vdir_deblk_p *p,
32046 + union au_vdir_deblk_p *deblk_end)
32048 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
32049 + p->de->de_str.len = 0;
32053 + return -1; /* error */
32056 +/* returns true or false */
32057 +static int is_deblk_end(union au_vdir_deblk_p *p,
32058 + union au_vdir_deblk_p *deblk_end)
32060 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
32061 + return !p->de->de_str.len;
32065 +static unsigned char *last_deblk(struct au_vdir *vdir)
32067 + return vdir->vd_deblk[vdir->vd_nblk - 1];
32070 +/* ---------------------------------------------------------------------- */
32072 +/* estimate the appropriate size for name hash table */
32073 +unsigned int au_rdhash_est(loff_t sz)
32081 + if (sz < AUFS_RDHASH_DEF)
32082 + n = AUFS_RDHASH_DEF;
32083 + /* pr_info("n %u\n", n); */
32088 + * the allocated memory has to be freed by
32089 + * au_nhash_wh_free() or au_nhash_de_free().
32091 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
32093 + struct hlist_head *head;
32097 + sz = sizeof(*nhash->nh_head) * num_hash;
32098 + head = kmalloc(sz, gfp);
32100 + nhash->nh_num = num_hash;
32101 + nhash->nh_head = head;
32102 + for (u = 0; u < num_hash; u++)
32103 + INIT_HLIST_HEAD(head++);
32104 + return 0; /* success */
32110 +static void nhash_count(struct hlist_head *head)
32112 +#if 0 /* debugging */
32114 + struct hlist_node *pos;
32117 + hlist_for_each(pos, head)
32119 + pr_info("%lu\n", n);
32123 +static void au_nhash_wh_do_free(struct hlist_head *head)
32125 + struct au_vdir_wh *pos;
32126 + struct hlist_node *node;
32128 + hlist_for_each_entry_safe(pos, node, head, wh_hash)
32129 + au_kfree_rcu(pos);
32132 +static void au_nhash_de_do_free(struct hlist_head *head)
32134 + struct au_vdir_dehstr *pos;
32135 + struct hlist_node *node;
32137 + hlist_for_each_entry_safe(pos, node, head, hash)
32138 + au_cache_free_vdir_dehstr(pos);
32141 +static void au_nhash_do_free(struct au_nhash *nhash,
32142 + void (*free)(struct hlist_head *head))
32145 + struct hlist_head *head;
32147 + n = nhash->nh_num;
32151 + head = nhash->nh_head;
32152 + while (n-- > 0) {
32153 + nhash_count(head);
32156 + au_kfree_try_rcu(nhash->nh_head);
32159 +void au_nhash_wh_free(struct au_nhash *whlist)
32161 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
32164 +static void au_nhash_de_free(struct au_nhash *delist)
32166 + au_nhash_do_free(delist, au_nhash_de_do_free);
32169 +/* ---------------------------------------------------------------------- */
32171 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
32175 + unsigned int u, n;
32176 + struct hlist_head *head;
32177 + struct au_vdir_wh *pos;
32180 + n = whlist->nh_num;
32181 + head = whlist->nh_head;
32182 + for (u = 0; u < n; u++, head++)
32183 + hlist_for_each_entry(pos, head, wh_hash)
32184 + if (pos->wh_bindex == btgt && ++num > limit)
32189 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
32190 + unsigned char *name,
32191 + unsigned int len)
32194 + /* const unsigned int magic_bit = 12; */
32196 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
32203 + /* v = hash_long(v, magic_bit); */
32204 + v %= nhash->nh_num;
32205 + return nhash->nh_head + v;
32208 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
32211 + return str->len == nlen && !memcmp(str->name, name, nlen);
32214 +/* returns found or not */
32215 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
32217 + struct hlist_head *head;
32218 + struct au_vdir_wh *pos;
32219 + struct au_vdir_destr *str;
32221 + head = au_name_hash(whlist, name, nlen);
32222 + hlist_for_each_entry(pos, head, wh_hash) {
32223 + str = &pos->wh_str;
32224 + AuDbg("%.*s\n", str->len, str->name);
32225 + if (au_nhash_test_name(str, name, nlen))
32231 +/* returns found(true) or not */
32232 +static int test_known(struct au_nhash *delist, char *name, int nlen)
32234 + struct hlist_head *head;
32235 + struct au_vdir_dehstr *pos;
32236 + struct au_vdir_destr *str;
32238 + head = au_name_hash(delist, name, nlen);
32239 + hlist_for_each_entry(pos, head, hash) {
32241 + AuDbg("%.*s\n", str->len, str->name);
32242 + if (au_nhash_test_name(str, name, nlen))
32248 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
32249 + unsigned char d_type)
32251 +#ifdef CONFIG_AUFS_SHWH
32252 + wh->wh_ino = ino;
32253 + wh->wh_type = d_type;
32257 +/* ---------------------------------------------------------------------- */
32259 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
32260 + unsigned int d_type, aufs_bindex_t bindex,
32261 + unsigned char shwh)
32264 + struct au_vdir_destr *str;
32265 + struct au_vdir_wh *wh;
32267 + AuDbg("%.*s\n", nlen, name);
32268 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
32271 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
32272 + if (unlikely(!wh))
32276 + wh->wh_bindex = bindex;
32278 + au_shwh_init_wh(wh, ino, d_type);
32279 + str = &wh->wh_str;
32281 + memcpy(str->name, name, nlen);
32282 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
32289 +static int append_deblk(struct au_vdir *vdir)
32292 + unsigned long ul;
32293 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
32294 + union au_vdir_deblk_p p, deblk_end;
32295 + unsigned char **o;
32298 + o = au_krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
32299 + GFP_NOFS, /*may_shrink*/0);
32300 + if (unlikely(!o))
32303 + vdir->vd_deblk = o;
32304 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
32306 + ul = vdir->vd_nblk++;
32307 + vdir->vd_deblk[ul] = p.deblk;
32308 + vdir->vd_last.ul = ul;
32309 + vdir->vd_last.p.deblk = p.deblk;
32310 + deblk_end.deblk = p.deblk + deblk_sz;
32311 + err = set_deblk_end(&p, &deblk_end);
32318 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
32319 + unsigned int d_type, struct au_nhash *delist)
32323 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
32324 + union au_vdir_deblk_p p, *room, deblk_end;
32325 + struct au_vdir_dehstr *dehstr;
32327 + p.deblk = last_deblk(vdir);
32328 + deblk_end.deblk = p.deblk + deblk_sz;
32329 + room = &vdir->vd_last.p;
32330 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
32331 + || !is_deblk_end(room, &deblk_end));
32333 + sz = calc_size(nlen);
32334 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
32335 + err = append_deblk(vdir);
32336 + if (unlikely(err))
32339 + p.deblk = last_deblk(vdir);
32340 + deblk_end.deblk = p.deblk + deblk_sz;
32342 + AuDebugOn(room->deblk != p.deblk);
32346 + dehstr = au_cache_alloc_vdir_dehstr();
32347 + if (unlikely(!dehstr))
32350 + dehstr->str = &room->de->de_str;
32351 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
32352 + room->de->de_ino = ino;
32353 + room->de->de_type = d_type;
32354 + room->de->de_str.len = nlen;
32355 + memcpy(room->de->de_str.name, name, nlen);
32358 + room->deblk += sz;
32359 + if (unlikely(set_deblk_end(room, &deblk_end)))
32360 + err = append_deblk(vdir);
32367 +/* ---------------------------------------------------------------------- */
32369 +void au_vdir_free(struct au_vdir *vdir)
32371 + unsigned char **deblk;
32373 + deblk = vdir->vd_deblk;
32374 + while (vdir->vd_nblk--)
32375 + au_kfree_try_rcu(*deblk++);
32376 + au_kfree_try_rcu(vdir->vd_deblk);
32377 + au_cache_free_vdir(vdir);
32380 +static struct au_vdir *alloc_vdir(struct file *file)
32382 + struct au_vdir *vdir;
32383 + struct super_block *sb;
32386 + sb = file->f_path.dentry->d_sb;
32387 + SiMustAnyLock(sb);
32390 + vdir = au_cache_alloc_vdir();
32391 + if (unlikely(!vdir))
32394 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
32395 + if (unlikely(!vdir->vd_deblk))
32398 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
32399 + if (!vdir->vd_deblk_sz) {
32400 + /* estimate the appropriate size for deblk */
32401 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
32402 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
32404 + vdir->vd_nblk = 0;
32405 + vdir->vd_version = 0;
32406 + vdir->vd_jiffy = 0;
32407 + err = append_deblk(vdir);
32409 + return vdir; /* success */
32411 + au_kfree_try_rcu(vdir->vd_deblk);
32414 + au_cache_free_vdir(vdir);
32416 + vdir = ERR_PTR(err);
32420 +static int reinit_vdir(struct au_vdir *vdir)
32423 + union au_vdir_deblk_p p, deblk_end;
32425 + while (vdir->vd_nblk > 1) {
32426 + au_kfree_try_rcu(vdir->vd_deblk[vdir->vd_nblk - 1]);
32427 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
32430 + p.deblk = vdir->vd_deblk[0];
32431 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
32432 + err = set_deblk_end(&p, &deblk_end);
32433 + /* keep vd_dblk_sz */
32434 + vdir->vd_last.ul = 0;
32435 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32436 + vdir->vd_version = 0;
32437 + vdir->vd_jiffy = 0;
32442 +/* ---------------------------------------------------------------------- */
32444 +#define AuFillVdir_CALLED 1
32445 +#define AuFillVdir_WHABLE (1 << 1)
32446 +#define AuFillVdir_SHWH (1 << 2)
32447 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
32448 +#define au_fset_fillvdir(flags, name) \
32449 + do { (flags) |= AuFillVdir_##name; } while (0)
32450 +#define au_fclr_fillvdir(flags, name) \
32451 + do { (flags) &= ~AuFillVdir_##name; } while (0)
32453 +#ifndef CONFIG_AUFS_SHWH
32454 +#undef AuFillVdir_SHWH
32455 +#define AuFillVdir_SHWH 0
32458 +struct fillvdir_arg {
32459 + struct dir_context ctx;
32460 + struct file *file;
32461 + struct au_vdir *vdir;
32462 + struct au_nhash delist;
32463 + struct au_nhash whlist;
32464 + aufs_bindex_t bindex;
32465 + unsigned int flags;
32469 +static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
32470 + loff_t offset __maybe_unused, u64 h_ino,
32471 + unsigned int d_type)
32473 + struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
32474 + char *name = (void *)__name;
32475 + struct super_block *sb;
32477 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
32480 + sb = arg->file->f_path.dentry->d_sb;
32481 + au_fset_fillvdir(arg->flags, CALLED);
32483 + if (nlen <= AUFS_WH_PFX_LEN
32484 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
32485 + if (test_known(&arg->delist, name, nlen)
32486 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
32487 + goto out; /* already exists or whiteouted */
32489 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
32491 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
32492 + d_type = DT_UNKNOWN;
32493 + arg->err = append_de(arg->vdir, name, nlen, ino,
32494 + d_type, &arg->delist);
32496 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
32497 + name += AUFS_WH_PFX_LEN;
32498 + nlen -= AUFS_WH_PFX_LEN;
32499 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
32500 + goto out; /* already whiteouted */
32502 + ino = 0; /* just to suppress a warning */
32504 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
32507 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
32508 + d_type = DT_UNKNOWN;
32509 + arg->err = au_nhash_append_wh
32510 + (&arg->whlist, name, nlen, ino, d_type,
32511 + arg->bindex, shwh);
32517 + arg->vdir->vd_jiffy = jiffies;
32519 + AuTraceErr(arg->err);
32523 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
32524 + struct au_nhash *whlist, struct au_nhash *delist)
32526 +#ifdef CONFIG_AUFS_SHWH
32528 + unsigned int nh, u;
32529 + struct hlist_head *head;
32530 + struct au_vdir_wh *pos;
32531 + struct hlist_node *n;
32533 + struct au_vdir_destr *destr;
32535 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
32538 + o = p = (void *)__get_free_page(GFP_NOFS);
32539 + if (unlikely(!p))
32543 + nh = whlist->nh_num;
32544 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32545 + p += AUFS_WH_PFX_LEN;
32546 + for (u = 0; u < nh; u++) {
32547 + head = whlist->nh_head + u;
32548 + hlist_for_each_entry_safe(pos, n, head, wh_hash) {
32549 + destr = &pos->wh_str;
32550 + memcpy(p, destr->name, destr->len);
32551 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
32552 + pos->wh_ino, pos->wh_type, delist);
32553 + if (unlikely(err))
32558 + free_page((unsigned long)o);
32568 +static int au_do_read_vdir(struct fillvdir_arg *arg)
32571 + unsigned int rdhash;
32573 + aufs_bindex_t bbot, bindex, btop;
32574 + unsigned char shwh;
32575 + struct file *hf, *file;
32576 + struct super_block *sb;
32578 + file = arg->file;
32579 + sb = file->f_path.dentry->d_sb;
32580 + SiMustAnyLock(sb);
32582 + rdhash = au_sbi(sb)->si_rdhash;
32584 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
32585 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
32586 + if (unlikely(err))
32588 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
32589 + if (unlikely(err))
32595 + if (au_opt_test(au_mntflags(sb), SHWH)) {
32597 + au_fset_fillvdir(arg->flags, SHWH);
32599 + btop = au_fbtop(file);
32600 + bbot = au_fbbot_dir(file);
32601 + for (bindex = btop; !err && bindex <= bbot; bindex++) {
32602 + hf = au_hf_dir(file, bindex);
32606 + offset = vfsub_llseek(hf, 0, SEEK_SET);
32608 + if (unlikely(offset))
32611 + arg->bindex = bindex;
32612 + au_fclr_fillvdir(arg->flags, WHABLE);
32614 + || (bindex != bbot
32615 + && au_br_whable(au_sbr_perm(sb, bindex))))
32616 + au_fset_fillvdir(arg->flags, WHABLE);
32619 + au_fclr_fillvdir(arg->flags, CALLED);
32621 + err = vfsub_iterate_dir(hf, &arg->ctx);
32624 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
32627 + * dir_relax() may be good for concurrency, but aufs should not
32628 + * use it since it will cause a lockdep problem.
32632 + if (!err && shwh)
32633 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
32635 + au_nhash_wh_free(&arg->whlist);
32638 + au_nhash_de_free(&arg->delist);
32643 +static int read_vdir(struct file *file, int may_read)
32646 + unsigned long expire;
32647 + unsigned char do_read;
32648 + struct fillvdir_arg arg = {
32650 + .actor = fillvdir
32653 + struct inode *inode;
32654 + struct au_vdir *vdir, *allocated;
32657 + inode = file_inode(file);
32658 + IMustLock(inode);
32659 + IiMustWriteLock(inode);
32660 + SiMustAnyLock(inode->i_sb);
32662 + allocated = NULL;
32664 + expire = au_sbi(inode->i_sb)->si_rdcache;
32665 + vdir = au_ivdir(inode);
32668 + vdir = alloc_vdir(file);
32669 + err = PTR_ERR(vdir);
32670 + if (IS_ERR(vdir))
32673 + allocated = vdir;
32674 + } else if (may_read
32675 + && (!inode_eq_iversion(inode, vdir->vd_version)
32676 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
32678 + err = reinit_vdir(vdir);
32679 + if (unlikely(err))
32684 + return 0; /* success */
32688 + err = au_do_read_vdir(&arg);
32690 + /* file->f_pos = 0; */ /* todo: ctx->pos? */
32691 + vdir->vd_version = inode_query_iversion(inode);
32692 + vdir->vd_last.ul = 0;
32693 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32695 + au_set_ivdir(inode, allocated);
32696 + } else if (allocated)
32697 + au_vdir_free(allocated);
32703 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
32706 + unsigned long ul, n;
32707 + const unsigned int deblk_sz = src->vd_deblk_sz;
32709 + AuDebugOn(tgt->vd_nblk != 1);
32712 + if (tgt->vd_nblk < src->vd_nblk) {
32713 + unsigned char **p;
32715 + p = au_krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
32716 + GFP_NOFS, /*may_shrink*/0);
32717 + if (unlikely(!p))
32719 + tgt->vd_deblk = p;
32722 + if (tgt->vd_deblk_sz != deblk_sz) {
32723 + unsigned char *p;
32725 + tgt->vd_deblk_sz = deblk_sz;
32726 + p = au_krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS,
32727 + /*may_shrink*/1);
32728 + if (unlikely(!p))
32730 + tgt->vd_deblk[0] = p;
32732 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
32733 + tgt->vd_version = src->vd_version;
32734 + tgt->vd_jiffy = src->vd_jiffy;
32736 + n = src->vd_nblk;
32737 + for (ul = 1; ul < n; ul++) {
32738 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
32740 + if (unlikely(!tgt->vd_deblk[ul]))
32744 + tgt->vd_nblk = n;
32745 + tgt->vd_last.ul = tgt->vd_last.ul;
32746 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
32747 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
32748 + - src->vd_deblk[src->vd_last.ul];
32750 + return 0; /* success */
32753 + rerr = reinit_vdir(tgt);
32758 +int au_vdir_init(struct file *file)
32761 + struct inode *inode;
32762 + struct au_vdir *vdir_cache, *allocated;
32764 + /* test file->f_pos here instead of ctx->pos */
32765 + err = read_vdir(file, !file->f_pos);
32766 + if (unlikely(err))
32769 + allocated = NULL;
32770 + vdir_cache = au_fvdir_cache(file);
32771 + if (!vdir_cache) {
32772 + vdir_cache = alloc_vdir(file);
32773 + err = PTR_ERR(vdir_cache);
32774 + if (IS_ERR(vdir_cache))
32776 + allocated = vdir_cache;
32777 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
32778 + /* test file->f_pos here instead of ctx->pos */
32779 + err = reinit_vdir(vdir_cache);
32780 + if (unlikely(err))
32783 + return 0; /* success */
32785 + inode = file_inode(file);
32786 + err = copy_vdir(vdir_cache, au_ivdir(inode));
32788 + file->f_version = inode_query_iversion(inode);
32790 + au_set_fvdir_cache(file, allocated);
32791 + } else if (allocated)
32792 + au_vdir_free(allocated);
32798 +static loff_t calc_offset(struct au_vdir *vdir)
32801 + union au_vdir_deblk_p p;
32803 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
32804 + offset = vdir->vd_last.p.deblk - p.deblk;
32805 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
32809 +/* returns true or false */
32810 +static int seek_vdir(struct file *file, struct dir_context *ctx)
32813 + unsigned int deblk_sz;
32814 + unsigned long ul, n;
32816 + union au_vdir_deblk_p p, deblk_end;
32817 + struct au_vdir *vdir_cache;
32820 + vdir_cache = au_fvdir_cache(file);
32821 + offset = calc_offset(vdir_cache);
32822 + AuDbg("offset %lld\n", offset);
32823 + if (ctx->pos == offset)
32826 + vdir_cache->vd_last.ul = 0;
32827 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
32832 + deblk_sz = vdir_cache->vd_deblk_sz;
32833 + ul = div64_u64(ctx->pos, deblk_sz);
32834 + AuDbg("ul %lu\n", ul);
32835 + if (ul >= vdir_cache->vd_nblk)
32838 + n = vdir_cache->vd_nblk;
32839 + for (; ul < n; ul++) {
32840 + p.deblk = vdir_cache->vd_deblk[ul];
32841 + deblk_end.deblk = p.deblk + deblk_sz;
32843 + offset *= deblk_sz;
32844 + while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
32847 + l = calc_size(p.de->de_str.len);
32851 + if (!is_deblk_end(&p, &deblk_end)) {
32853 + vdir_cache->vd_last.ul = ul;
32854 + vdir_cache->vd_last.p = p;
32862 + AuDbg("valid %d\n", !valid);
32866 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
32868 + unsigned int l, deblk_sz;
32869 + union au_vdir_deblk_p deblk_end;
32870 + struct au_vdir *vdir_cache;
32871 + struct au_vdir_de *de;
32873 + if (!seek_vdir(file, ctx))
32876 + vdir_cache = au_fvdir_cache(file);
32877 + deblk_sz = vdir_cache->vd_deblk_sz;
32879 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32880 + deblk_end.deblk += deblk_sz;
32881 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
32882 + de = vdir_cache->vd_last.p.de;
32883 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
32884 + de->de_str.len, de->de_str.name, ctx->pos,
32885 + (unsigned long)de->de_ino, de->de_type);
32886 + if (unlikely(!dir_emit(ctx, de->de_str.name,
32887 + de->de_str.len, de->de_ino,
32889 + /* todo: ignore the error caused by udba? */
32890 + /* return err; */
32894 + l = calc_size(de->de_str.len);
32895 + vdir_cache->vd_last.p.deblk += l;
32898 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
32899 + vdir_cache->vd_last.ul++;
32900 + vdir_cache->vd_last.p.deblk
32901 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32902 + ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
32911 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
32912 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
32913 +++ linux/fs/aufs/vfsub.c 2022-03-21 14:49:05.733299677 +0100
32915 +// SPDX-License-Identifier: GPL-2.0
32917 + * Copyright (C) 2005-2021 Junjiro R. Okajima
32919 + * This program, aufs is free software; you can redistribute it and/or modify
32920 + * it under the terms of the GNU General Public License as published by
32921 + * the Free Software Foundation; either version 2 of the License, or
32922 + * (at your option) any later version.
32924 + * This program is distributed in the hope that it will be useful,
32925 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32926 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32927 + * GNU General Public License for more details.
32929 + * You should have received a copy of the GNU General Public License
32930 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32934 + * sub-routines for VFS
32937 +#include <linux/mnt_namespace.h>
32938 +#include <linux/nsproxy.h>
32939 +#include <linux/security.h>
32940 +#include <linux/splice.h>
32943 +#ifdef CONFIG_AUFS_BR_FUSE
32944 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb)
32946 + if (!au_test_fuse(h_sb) || !au_userns)
32949 + return is_current_mnt_ns(mnt) ? 0 : -EACCES;
32953 +int vfsub_sync_filesystem(struct super_block *h_sb)
32958 + down_read(&h_sb->s_umount);
32959 + err = sync_filesystem(h_sb);
32960 + up_read(&h_sb->s_umount);
32966 +/* ---------------------------------------------------------------------- */
32968 +int vfsub_update_h_iattr(struct path *h_path, int *did)
32972 + struct super_block *h_sb;
32975 + * Always needs h_path->mnt for LSM or FUSE branch.
32977 + AuDebugOn(!h_path->mnt);
32979 + /* for remote fs, leave work for its getattr or d_revalidate */
32980 + /* for bad i_attr fs, handle them in aufs_getattr() */
32981 + /* still some fs may acquire i_mutex. we need to skip them */
32985 + h_sb = h_path->dentry->d_sb;
32986 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
32988 + err = vfsub_getattr(h_path, &st);
32993 +/* ---------------------------------------------------------------------- */
32995 +struct file *vfsub_dentry_open(struct path *path, int flags)
32997 + return dentry_open(path, flags /* | __FMODE_NONOTIFY */,
33001 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
33003 + struct file *file;
33006 + file = filp_open(path,
33007 + oflags /* | __FMODE_NONOTIFY */,
33010 + if (IS_ERR(file))
33012 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33019 + * Ideally this function should call VFS:do_last() in order to keep all its
33020 + * checkings. But it is very hard for aufs to regenerate several VFS internal
33021 + * structure such as nameidata. This is a second (or third) best approach.
33022 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
33024 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33025 + struct vfsub_aopen_args *args)
33028 + struct au_branch *br = args->br;
33029 + struct file *file = args->file;
33030 + /* copied from linux/fs/namei.c:atomic_open() */
33031 + struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
33034 + AuDebugOn(!dir->i_op->atomic_open);
33036 + err = au_br_test_oflag(args->open_flag, br);
33037 + if (unlikely(err))
33040 + au_lcnt_inc(&br->br_nfiles);
33041 + file->f_path.dentry = DENTRY_NOT_SET;
33042 + file->f_path.mnt = au_br_mnt(br);
33043 + AuDbg("%ps\n", dir->i_op->atomic_open);
33044 + err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
33045 + args->create_mode);
33046 + if (unlikely(err < 0)) {
33047 + au_lcnt_dec(&br->br_nfiles);
33051 + /* temporary workaround for nfsv4 branch */
33052 + if (au_test_nfs(dir->i_sb))
33053 + nfs_mark_for_revalidate(dir);
33055 + if (file->f_mode & FMODE_CREATED)
33056 + fsnotify_create(dir, dentry);
33057 + if (!(file->f_mode & FMODE_OPENED)) {
33058 + au_lcnt_dec(&br->br_nfiles);
33062 + /* todo: call VFS:may_open() here */
33063 + /* todo: ima_file_check() too? */
33064 + if (!err && (args->open_flag & __FMODE_EXEC))
33065 + err = deny_write_access(file);
33067 + fsnotify_open(file);
33069 + au_lcnt_dec(&br->br_nfiles);
33070 + /* note that the file is created and still opened */
33076 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
33080 + err = kern_path(name, flags, path);
33081 + if (!err && d_is_positive(path->dentry))
33082 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33086 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33087 + struct path *ppath, int len)
33089 + struct path path;
33091 + path.dentry = lookup_one_len_unlocked(name, ppath->dentry, len);
33092 + if (IS_ERR(path.dentry))
33094 + if (d_is_positive(path.dentry)) {
33095 + path.mnt = ppath->mnt;
33096 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33100 + AuTraceErrPtr(path.dentry);
33101 + return path.dentry;
33104 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33107 + struct path path;
33109 + /* VFS checks it too, but by WARN_ON_ONCE() */
33110 + IMustLock(d_inode(ppath->dentry));
33112 + path.dentry = lookup_one_len(name, ppath->dentry, len);
33113 + if (IS_ERR(path.dentry))
33115 + if (d_is_positive(path.dentry)) {
33116 + path.mnt = ppath->mnt;
33117 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33121 + AuTraceErrPtr(path.dentry);
33122 + return path.dentry;
33125 +void vfsub_call_lkup_one(void *args)
33127 + struct vfsub_lkup_one_args *a = args;
33128 + *a->errp = vfsub_lkup_one(a->name, a->ppath);
33131 +/* ---------------------------------------------------------------------- */
33133 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33134 + struct dentry *d2, struct au_hinode *hdir2)
33136 + struct dentry *d;
33139 + d = lock_rename(d1, d2);
33141 + au_hn_suspend(hdir1);
33142 + if (hdir1 != hdir2)
33143 + au_hn_suspend(hdir2);
33148 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33149 + struct dentry *d2, struct au_hinode *hdir2)
33151 + au_hn_resume(hdir1);
33152 + if (hdir1 != hdir2)
33153 + au_hn_resume(hdir2);
33155 + unlock_rename(d1, d2);
33159 +/* ---------------------------------------------------------------------- */
33161 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
33164 + struct dentry *d;
33165 + struct user_namespace *userns;
33169 + d = path->dentry;
33170 + path->dentry = d->d_parent;
33171 + err = security_path_mknod(path, d, mode, 0);
33172 + path->dentry = d;
33173 + if (unlikely(err))
33175 + userns = mnt_user_ns(path->mnt);
33178 + err = vfs_create(userns, dir, path->dentry, mode, want_excl);
33181 + struct path tmp = *path;
33184 + vfsub_update_h_iattr(&tmp, &did);
33186 + tmp.dentry = path->dentry->d_parent;
33187 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33196 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
33199 + struct dentry *d;
33200 + struct user_namespace *userns;
33204 + d = path->dentry;
33205 + path->dentry = d->d_parent;
33206 + err = security_path_symlink(path, d, symname);
33207 + path->dentry = d;
33208 + if (unlikely(err))
33210 + userns = mnt_user_ns(path->mnt);
33213 + err = vfs_symlink(userns, dir, path->dentry, symname);
33216 + struct path tmp = *path;
33219 + vfsub_update_h_iattr(&tmp, &did);
33221 + tmp.dentry = path->dentry->d_parent;
33222 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33231 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
33234 + struct dentry *d;
33235 + struct user_namespace *userns;
33239 + d = path->dentry;
33240 + path->dentry = d->d_parent;
33241 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
33242 + path->dentry = d;
33243 + if (unlikely(err))
33245 + userns = mnt_user_ns(path->mnt);
33248 + err = vfs_mknod(userns, dir, path->dentry, mode, dev);
33251 + struct path tmp = *path;
33254 + vfsub_update_h_iattr(&tmp, &did);
33256 + tmp.dentry = path->dentry->d_parent;
33257 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33266 +static int au_test_nlink(struct inode *inode)
33268 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
33270 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
33271 + || inode->i_nlink < link_max)
33276 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
33277 + struct inode **delegated_inode)
33280 + struct dentry *d;
33281 + struct user_namespace *userns;
33285 + err = au_test_nlink(d_inode(src_dentry));
33286 + if (unlikely(err))
33289 + /* we don't call may_linkat() */
33290 + d = path->dentry;
33291 + path->dentry = d->d_parent;
33292 + err = security_path_link(src_dentry, path, d);
33293 + path->dentry = d;
33294 + if (unlikely(err))
33296 + userns = mnt_user_ns(path->mnt);
33299 + err = vfs_link(src_dentry, userns, dir, path->dentry, delegated_inode);
33302 + struct path tmp = *path;
33305 + /* fuse has different memory inode for the same inumber */
33306 + vfsub_update_h_iattr(&tmp, &did);
33308 + tmp.dentry = path->dentry->d_parent;
33309 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33310 + tmp.dentry = src_dentry;
33311 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33320 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
33321 + struct inode *dir, struct path *path,
33322 + struct inode **delegated_inode, unsigned int flags)
33325 + struct renamedata rd;
33326 + struct path tmp = {
33329 + struct dentry *d;
33332 + IMustLock(src_dir);
33334 + d = path->dentry;
33335 + path->dentry = d->d_parent;
33336 + tmp.dentry = src_dentry->d_parent;
33337 + err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
33338 + path->dentry = d;
33339 + if (unlikely(err))
33342 + rd.old_mnt_userns = mnt_user_ns(path->mnt);
33343 + rd.old_dir = src_dir;
33344 + rd.old_dentry = src_dentry;
33345 + rd.new_mnt_userns = rd.old_mnt_userns;
33346 + rd.new_dir = dir;
33347 + rd.new_dentry = path->dentry;
33348 + rd.delegated_inode = delegated_inode;
33349 + rd.flags = flags;
33351 + err = vfs_rename(&rd);
33356 + tmp.dentry = d->d_parent;
33357 + vfsub_update_h_iattr(&tmp, &did);
33359 + tmp.dentry = src_dentry;
33360 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33361 + tmp.dentry = src_dentry->d_parent;
33362 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33371 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
33374 + struct dentry *d;
33375 + struct user_namespace *userns;
33379 + d = path->dentry;
33380 + path->dentry = d->d_parent;
33381 + err = security_path_mkdir(path, d, mode);
33382 + path->dentry = d;
33383 + if (unlikely(err))
33385 + userns = mnt_user_ns(path->mnt);
33388 + err = vfs_mkdir(userns, dir, path->dentry, mode);
33391 + struct path tmp = *path;
33394 + vfsub_update_h_iattr(&tmp, &did);
33396 + tmp.dentry = path->dentry->d_parent;
33397 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33406 +int vfsub_rmdir(struct inode *dir, struct path *path)
33409 + struct dentry *d;
33410 + struct user_namespace *userns;
33414 + d = path->dentry;
33415 + path->dentry = d->d_parent;
33416 + err = security_path_rmdir(path, d);
33417 + path->dentry = d;
33418 + if (unlikely(err))
33420 + userns = mnt_user_ns(path->mnt);
33423 + err = vfs_rmdir(userns, dir, path->dentry);
33426 + struct path tmp = {
33427 + .dentry = path->dentry->d_parent,
33431 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33438 +/* ---------------------------------------------------------------------- */
33440 +/* todo: support mmap_sem? */
33441 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
33447 + err = vfs_read(file, ubuf, count, ppos);
33450 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33454 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
33460 + err = kernel_read(file, kbuf, count, ppos);
33464 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33468 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
33474 + err = vfs_write(file, ubuf, count, ppos);
33477 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33481 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
33486 + err = kernel_write(file, kbuf, count, ppos);
33489 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33493 +int vfsub_flush(struct file *file, fl_owner_t id)
33498 + if (file->f_op->flush) {
33499 + if (!au_test_nfs(file->f_path.dentry->d_sb))
33500 + err = file->f_op->flush(file, id);
33503 + err = file->f_op->flush(file, id);
33507 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
33513 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
33517 + AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
33520 + err = iterate_dir(file, ctx);
33523 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33528 +long vfsub_splice_to(struct file *in, loff_t *ppos,
33529 + struct pipe_inode_info *pipe, size_t len,
33530 + unsigned int flags)
33535 + err = do_splice_to(in, ppos, pipe, len, flags);
33537 + file_accessed(in);
33539 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
33543 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
33544 + loff_t *ppos, size_t len, unsigned int flags)
33549 + err = do_splice_from(pipe, out, ppos, len, flags);
33552 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
33556 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
33560 + /* file can be NULL */
33562 + err = vfs_fsync(file, datasync);
33566 + AuDebugOn(!file);
33567 + path = &file->f_path;
33569 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33574 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
33575 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33576 + struct file *h_file)
33579 + struct inode *h_inode;
33580 + struct super_block *h_sb;
33581 + struct user_namespace *h_userns;
33584 + err = vfsub_truncate(h_path, length);
33588 + h_inode = d_inode(h_path->dentry);
33589 + h_sb = h_inode->i_sb;
33591 + sb_start_write(h_sb);
33593 + err = security_path_truncate(h_path);
33595 + h_userns = mnt_user_ns(h_path->mnt);
33597 + err = do_truncate(h_userns, h_path->dentry, length, attr,
33602 + sb_end_write(h_sb);
33609 +/* ---------------------------------------------------------------------- */
33611 +struct au_vfsub_mkdir_args {
33613 + struct inode *dir;
33614 + struct path *path;
33618 +static void au_call_vfsub_mkdir(void *args)
33620 + struct au_vfsub_mkdir_args *a = args;
33621 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
33624 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
33626 + int err, do_sio, wkq_err;
33627 + struct user_namespace *userns;
33629 + userns = mnt_user_ns(path->mnt);
33630 + do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33633 + err = vfsub_mkdir(dir, path, mode);
33636 + struct au_vfsub_mkdir_args args = {
33642 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
33643 + if (unlikely(wkq_err))
33650 +struct au_vfsub_rmdir_args {
33652 + struct inode *dir;
33653 + struct path *path;
33656 +static void au_call_vfsub_rmdir(void *args)
33658 + struct au_vfsub_rmdir_args *a = args;
33659 + *a->errp = vfsub_rmdir(a->dir, a->path);
33662 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
33664 + int err, do_sio, wkq_err;
33665 + struct user_namespace *userns;
33667 + userns = mnt_user_ns(path->mnt);
33668 + do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33671 + err = vfsub_rmdir(dir, path);
33674 + struct au_vfsub_rmdir_args args = {
33679 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
33680 + if (unlikely(wkq_err))
33687 +/* ---------------------------------------------------------------------- */
33689 +struct notify_change_args {
33691 + struct path *path;
33692 + struct iattr *ia;
33693 + struct inode **delegated_inode;
33696 +static void call_notify_change(void *args)
33698 + struct notify_change_args *a = args;
33699 + struct inode *h_inode;
33700 + struct user_namespace *userns;
33702 + h_inode = d_inode(a->path->dentry);
33703 + IMustLock(h_inode);
33705 + *a->errp = -EPERM;
33706 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
33707 + userns = mnt_user_ns(a->path->mnt);
33709 + *a->errp = notify_change(userns, a->path->dentry, a->ia,
33710 + a->delegated_inode);
33713 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
33715 + AuTraceErr(*a->errp);
33718 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33719 + struct inode **delegated_inode)
33722 + struct notify_change_args args = {
33726 + .delegated_inode = delegated_inode
33729 + call_notify_change(&args);
33734 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33735 + struct inode **delegated_inode)
33737 + int err, wkq_err;
33738 + struct notify_change_args args = {
33742 + .delegated_inode = delegated_inode
33745 + wkq_err = au_wkq_wait(call_notify_change, &args);
33746 + if (unlikely(wkq_err))
33752 +/* ---------------------------------------------------------------------- */
33754 +struct unlink_args {
33756 + struct inode *dir;
33757 + struct path *path;
33758 + struct inode **delegated_inode;
33761 +static void call_unlink(void *args)
33763 + struct unlink_args *a = args;
33764 + struct dentry *d = a->path->dentry;
33765 + struct inode *h_inode;
33766 + struct user_namespace *userns;
33767 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
33768 + && au_dcount(d) == 1);
33770 + IMustLock(a->dir);
33772 + a->path->dentry = d->d_parent;
33773 + *a->errp = security_path_unlink(a->path, d);
33774 + a->path->dentry = d;
33775 + if (unlikely(*a->errp))
33778 + if (!stop_sillyrename)
33781 + if (d_is_positive(d)) {
33782 + h_inode = d_inode(d);
33786 + userns = mnt_user_ns(a->path->mnt);
33788 + *a->errp = vfs_unlink(userns, a->dir, d, a->delegated_inode);
33791 + struct path tmp = {
33792 + .dentry = d->d_parent,
33793 + .mnt = a->path->mnt
33795 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33798 + if (!stop_sillyrename)
33803 + AuTraceErr(*a->errp);
33807 + * @dir: must be locked.
33808 + * @dentry: target dentry.
33810 +int vfsub_unlink(struct inode *dir, struct path *path,
33811 + struct inode **delegated_inode, int force)
33814 + struct unlink_args args = {
33818 + .delegated_inode = delegated_inode
33822 + call_unlink(&args);
33826 + wkq_err = au_wkq_wait(call_unlink, &args);
33827 + if (unlikely(wkq_err))
33833 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
33834 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
33835 +++ linux/fs/aufs/vfsub.h 2022-10-03 07:51:14.929542593 +0200
33837 +/* SPDX-License-Identifier: GPL-2.0 */
33839 + * Copyright (C) 2005-2021 Junjiro R. Okajima
33841 + * This program, aufs is free software; you can redistribute it and/or modify
33842 + * it under the terms of the GNU General Public License as published by
33843 + * the Free Software Foundation; either version 2 of the License, or
33844 + * (at your option) any later version.
33846 + * This program is distributed in the hope that it will be useful,
33847 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33848 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33849 + * GNU General Public License for more details.
33851 + * You should have received a copy of the GNU General Public License
33852 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33856 + * sub-routines for VFS
33859 +#ifndef __AUFS_VFSUB_H__
33860 +#define __AUFS_VFSUB_H__
33864 +#include <linux/fs.h>
33865 +#include <linux/mount.h>
33866 +#include <linux/posix_acl.h>
33867 +#include <linux/xattr.h>
33868 +#include "debug.h"
33870 +/* copied from linux/fs/internal.h */
33871 +/* todo: BAD approach!! */
33872 +extern void __mnt_drop_write(struct vfsmount *);
33873 +extern struct file *alloc_empty_file(int, const struct cred *);
33875 +/* ---------------------------------------------------------------------- */
33877 +/* lock subclass for lower inode */
33878 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
33879 +/* reduce? gave up. */
33881 + AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
33882 + AuLsc_I_PARENT, /* lower inode, parent first */
33883 + AuLsc_I_PARENT2, /* copyup dirs */
33884 + AuLsc_I_PARENT3, /* copyup wh */
33890 +/* to debug easier, do not make them inlined functions */
33891 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
33892 +#define IMustLock(i) AuDebugOn(!inode_is_locked(i))
33894 +/* ---------------------------------------------------------------------- */
33896 +static inline void vfsub_drop_nlink(struct inode *inode)
33898 + AuDebugOn(!inode->i_nlink);
33899 + drop_nlink(inode);
33902 +static inline void vfsub_dead_dir(struct inode *inode)
33904 + AuDebugOn(!S_ISDIR(inode->i_mode));
33905 + inode->i_flags |= S_DEAD;
33906 + clear_nlink(inode);
33909 +static inline int vfsub_native_ro(struct inode *inode)
33911 + return sb_rdonly(inode->i_sb)
33912 + || IS_RDONLY(inode)
33913 + /* || IS_APPEND(inode) */
33914 + || IS_IMMUTABLE(inode);
33917 +#ifdef CONFIG_AUFS_BR_FUSE
33918 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb);
33920 +AuStubInt0(vfsub_test_mntns, struct vfsmount *mnt, struct super_block *h_sb);
33923 +int vfsub_sync_filesystem(struct super_block *h_sb);
33925 +/* ---------------------------------------------------------------------- */
33927 +int vfsub_update_h_iattr(struct path *h_path, int *did);
33928 +struct file *vfsub_dentry_open(struct path *path, int flags);
33929 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
33931 +struct vfsub_aopen_args {
33932 + struct file *file;
33933 + unsigned int open_flag;
33934 + umode_t create_mode;
33935 + struct au_branch *br;
33937 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33938 + struct vfsub_aopen_args *args);
33939 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
33941 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33942 + struct path *ppath, int len);
33943 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33946 +struct vfsub_lkup_one_args {
33947 + struct dentry **errp;
33948 + struct qstr *name;
33949 + struct path *ppath;
33952 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
33953 + struct path *ppath)
33955 + return vfsub_lookup_one_len(name->name, ppath, name->len);
33958 +void vfsub_call_lkup_one(void *args);
33960 +/* ---------------------------------------------------------------------- */
33962 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
33967 + err = mnt_want_write(mnt);
33972 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
33975 + mnt_drop_write(mnt);
33979 +#if 0 /* reserved */
33980 +static inline void vfsub_mnt_drop_write_file(struct file *file)
33983 + mnt_drop_write_file(file);
33988 +/* ---------------------------------------------------------------------- */
33991 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33992 + struct dentry *d2, struct au_hinode *hdir2);
33993 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33994 + struct dentry *d2, struct au_hinode *hdir2);
33996 +int vfsub_create(struct inode *dir, struct path *path, int mode,
33998 +int vfsub_symlink(struct inode *dir, struct path *path,
33999 + const char *symname);
34000 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
34001 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
34002 + struct path *path, struct inode **delegated_inode);
34003 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
34004 + struct inode *hdir, struct path *path,
34005 + struct inode **delegated_inode, unsigned int flags);
34006 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
34007 +int vfsub_rmdir(struct inode *dir, struct path *path);
34009 +/* ---------------------------------------------------------------------- */
34011 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
34013 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
34015 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
34017 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
34019 +int vfsub_flush(struct file *file, fl_owner_t id);
34020 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
34022 +static inline loff_t vfsub_f_size_read(struct file *file)
34024 + return i_size_read(file_inode(file));
34027 +static inline unsigned int vfsub_file_flags(struct file *file)
34029 + unsigned int flags;
34031 + spin_lock(&file->f_lock);
34032 + flags = file->f_flags;
34033 + spin_unlock(&file->f_lock);
34038 +static inline int vfsub_file_execed(struct file *file)
34040 + /* todo: direct access f_flags */
34041 + return !!(vfsub_file_flags(file) & __FMODE_EXEC);
34044 +#if 0 /* reserved */
34045 +static inline void vfsub_file_accessed(struct file *h_file)
34047 + file_accessed(h_file);
34048 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
34052 +#if 0 /* reserved */
34053 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
34054 + struct dentry *h_dentry)
34056 + struct path h_path = {
34057 + .dentry = h_dentry,
34060 + touch_atime(&h_path);
34061 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
34065 +static inline int vfsub_update_time(struct inode *h_inode,
34066 + struct timespec64 *ts, int flags)
34068 + return inode_update_time(h_inode, ts, flags);
34069 + /* no vfsub_update_h_iattr() since we don't have struct path */
34072 +#ifdef CONFIG_FS_POSIX_ACL
34073 +static inline int vfsub_acl_chmod(struct user_namespace *h_userns,
34074 + struct inode *h_inode, umode_t h_mode)
34078 + err = posix_acl_chmod(h_userns, h_inode, h_mode);
34079 + if (err == -EOPNOTSUPP)
34084 +AuStubInt0(vfsub_acl_chmod, struct user_namespace *h_userns,
34085 + struct inode *h_inode, umode_t h_mode);
34088 +long vfsub_splice_to(struct file *in, loff_t *ppos,
34089 + struct pipe_inode_info *pipe, size_t len,
34090 + unsigned int flags);
34091 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
34092 + loff_t *ppos, size_t len, unsigned int flags);
34094 +static inline long vfsub_truncate(struct path *path, loff_t length)
34099 + err = vfs_truncate(path, length);
34104 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
34105 + struct file *h_file);
34106 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
34109 + * re-use branch fs's ioctl(FICLONE) while aufs itself doesn't support such
34112 +static inline loff_t vfsub_clone_file_range(struct file *src, struct file *dst,
34118 + err = vfs_clone_file_range(src, 0, dst, 0, len, /*remap_flags*/0);
34124 +/* copy_file_range(2) is a systemcall */
34125 +static inline ssize_t vfsub_copy_file_range(struct file *src, loff_t src_pos,
34126 + struct file *dst, loff_t dst_pos,
34127 + size_t len, unsigned int flags)
34132 + ssz = vfs_copy_file_range(src, src_pos, dst, dst_pos, len, flags);
34138 +/* ---------------------------------------------------------------------- */
34140 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
34145 + err = vfs_llseek(file, offset, origin);
34150 +/* ---------------------------------------------------------------------- */
34152 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
34153 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
34154 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
34155 + struct inode **delegated_inode);
34156 +int vfsub_notify_change(struct path *path, struct iattr *ia,
34157 + struct inode **delegated_inode);
34158 +int vfsub_unlink(struct inode *dir, struct path *path,
34159 + struct inode **delegated_inode, int force);
34161 +static inline int vfsub_getattr(const struct path *path, struct kstat *st)
34163 + return vfs_getattr(path, st, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
34166 +/* ---------------------------------------------------------------------- */
34169 + * The parameter 'value' is not const since the id in ACL_USER/ACL_GROUP may be
34170 + * changed by idmapping.
34172 +static inline int vfsub_setxattr(struct user_namespace *userns,
34173 + struct dentry *dentry, const char *name,
34174 + void *value, size_t size, int flags)
34179 + err = vfs_setxattr(userns, dentry, name, value, size, flags);
34185 +static inline int vfsub_removexattr(struct user_namespace *userns,
34186 + struct dentry *dentry, const char *name)
34191 + err = vfs_removexattr(userns, dentry, name);
34197 +#endif /* __KERNEL__ */
34198 +#endif /* __AUFS_VFSUB_H__ */
34199 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
34200 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
34201 +++ linux/fs/aufs/wbr_policy.c 2022-03-21 14:49:05.733299677 +0100
34203 +// SPDX-License-Identifier: GPL-2.0
34205 + * Copyright (C) 2005-2021 Junjiro R. Okajima
34207 + * This program, aufs is free software; you can redistribute it and/or modify
34208 + * it under the terms of the GNU General Public License as published by
34209 + * the Free Software Foundation; either version 2 of the License, or
34210 + * (at your option) any later version.
34212 + * This program is distributed in the hope that it will be useful,
34213 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
34214 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
34215 + * GNU General Public License for more details.
34217 + * You should have received a copy of the GNU General Public License
34218 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
34222 + * policies for selecting one among multiple writable branches
34225 +#include <linux/statfs.h>
34228 +/* subset of cpup_attr() */
34229 +static noinline_for_stack
34230 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
34234 + struct inode *h_isrc;
34236 + h_isrc = d_inode(h_src);
34237 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
34238 + ia.ia_mode = h_isrc->i_mode;
34239 + ia.ia_uid = h_isrc->i_uid;
34240 + ia.ia_gid = h_isrc->i_gid;
34241 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
34242 + au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
34243 + /* no delegation since it is just created */
34244 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34246 + /* is this nfs only? */
34247 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
34248 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
34249 + ia.ia_mode = h_isrc->i_mode;
34250 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34256 +#define AuCpdown_PARENT_OPQ 1
34257 +#define AuCpdown_WHED (1 << 1)
34258 +#define AuCpdown_MADE_DIR (1 << 2)
34259 +#define AuCpdown_DIROPQ (1 << 3)
34260 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
34261 +#define au_fset_cpdown(flags, name) \
34262 + do { (flags) |= AuCpdown_##name; } while (0)
34263 +#define au_fclr_cpdown(flags, name) \
34264 + do { (flags) &= ~AuCpdown_##name; } while (0)
34266 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
34267 + unsigned int *flags)
34270 + struct dentry *opq_dentry;
34272 + opq_dentry = au_diropq_create(dentry, bdst);
34273 + err = PTR_ERR(opq_dentry);
34274 + if (IS_ERR(opq_dentry))
34276 + dput(opq_dentry);
34277 + au_fset_cpdown(*flags, DIROPQ);
34283 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
34284 + struct inode *dir, aufs_bindex_t bdst)
34287 + struct path h_path;
34288 + struct au_branch *br;
34290 + br = au_sbr(dentry->d_sb, bdst);
34291 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
34292 + err = PTR_ERR(h_path.dentry);
34293 + if (IS_ERR(h_path.dentry))
34297 + if (d_is_positive(h_path.dentry)) {
34298 + h_path.mnt = au_br_mnt(br);
34299 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
34302 + dput(h_path.dentry);
34308 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
34309 + struct au_pin *pin,
34310 + struct dentry *h_parent, void *arg)
34313 + aufs_bindex_t bopq, btop;
34314 + struct path h_path;
34315 + struct dentry *parent;
34316 + struct inode *h_dir, *h_inode, *inode, *dir;
34317 + unsigned int *flags = arg;
34319 + btop = au_dbtop(dentry);
34320 + /* dentry is di-locked */
34321 + parent = dget_parent(dentry);
34322 + dir = d_inode(parent);
34323 + h_dir = d_inode(h_parent);
34324 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
34325 + IMustLock(h_dir);
34327 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
34328 + if (unlikely(err < 0))
34330 + h_path.dentry = au_h_dptr(dentry, bdst);
34331 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
34332 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path, 0755);
34333 + if (unlikely(err))
34335 + au_fset_cpdown(*flags, MADE_DIR);
34337 + bopq = au_dbdiropq(dentry);
34338 + au_fclr_cpdown(*flags, WHED);
34339 + au_fclr_cpdown(*flags, DIROPQ);
34340 + if (au_dbwh(dentry) == bdst)
34341 + au_fset_cpdown(*flags, WHED);
34342 + if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
34343 + au_fset_cpdown(*flags, PARENT_OPQ);
34344 + h_inode = d_inode(h_path.dentry);
34345 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
34346 + if (au_ftest_cpdown(*flags, WHED)) {
34347 + err = au_cpdown_dir_opq(dentry, bdst, flags);
34348 + if (unlikely(err)) {
34349 + inode_unlock(h_inode);
34354 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, btop));
34355 + inode_unlock(h_inode);
34356 + if (unlikely(err))
34359 + if (au_ftest_cpdown(*flags, WHED)) {
34360 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
34361 + if (unlikely(err))
34365 + inode = d_inode(dentry);
34366 + if (au_ibbot(inode) < bdst)
34367 + au_set_ibbot(inode, bdst);
34368 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
34369 + au_hi_flags(inode, /*isdir*/1));
34370 + au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
34371 + goto out; /* success */
34375 + if (au_ftest_cpdown(*flags, DIROPQ)) {
34376 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
34377 + rerr = au_diropq_remove(dentry, bdst);
34378 + inode_unlock(h_inode);
34379 + if (unlikely(rerr)) {
34380 + AuIOErr("failed removing diropq for %pd b%d (%d)\n",
34381 + dentry, bdst, rerr);
34387 + if (au_ftest_cpdown(*flags, MADE_DIR)) {
34388 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
34389 + if (unlikely(rerr)) {
34390 + AuIOErr("failed removing %pd b%d (%d)\n",
34391 + dentry, bdst, rerr);
34396 + au_set_h_dptr(dentry, bdst, NULL);
34397 + if (au_dbbot(dentry) == bdst)
34398 + au_update_dbbot(dentry);
34404 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
34407 + unsigned int flags;
34410 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
34415 +/* ---------------------------------------------------------------------- */
34417 +/* policies for create */
34419 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
34421 + int err, i, j, ndentry;
34422 + aufs_bindex_t bopq;
34423 + struct au_dcsub_pages dpages;
34424 + struct au_dpage *dpage;
34425 + struct dentry **dentries, *parent, *d;
34427 + err = au_dpages_init(&dpages, GFP_NOFS);
34428 + if (unlikely(err))
34430 + parent = dget_parent(dentry);
34431 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
34432 + if (unlikely(err))
34436 + for (i = 0; i < dpages.ndpage; i++) {
34437 + dpage = dpages.dpages + i;
34438 + dentries = dpage->dentries;
34439 + ndentry = dpage->ndentry;
34440 + for (j = 0; j < ndentry; j++) {
34442 + di_read_lock_parent2(d, !AuLock_IR);
34443 + bopq = au_dbdiropq(d);
34444 + di_read_unlock(d, !AuLock_IR);
34445 + if (bopq >= 0 && bopq < err)
34452 + au_dpages_free(&dpages);
34457 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
34459 + for (; bindex >= 0; bindex--)
34460 + if (!au_br_rdonly(au_sbr(sb, bindex)))
34465 +/* top down parent */
34466 +static int au_wbr_create_tdp(struct dentry *dentry,
34467 + unsigned int flags __maybe_unused)
34470 + aufs_bindex_t btop, bindex;
34471 + struct super_block *sb;
34472 + struct dentry *parent, *h_parent;
34474 + sb = dentry->d_sb;
34475 + btop = au_dbtop(dentry);
34477 + if (!au_br_rdonly(au_sbr(sb, btop)))
34481 + parent = dget_parent(dentry);
34482 + for (bindex = au_dbtop(parent); bindex < btop; bindex++) {
34483 + h_parent = au_h_dptr(parent, bindex);
34484 + if (!h_parent || d_is_negative(h_parent))
34487 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
34494 + /* bottom up here */
34495 + if (unlikely(err < 0)) {
34496 + err = au_wbr_bu(sb, btop - 1);
34498 + err = au_wbr_nonopq(dentry, err);
34502 + AuDbg("b%d\n", err);
34506 +/* ---------------------------------------------------------------------- */
34508 +/* an exception for the policy other than tdp */
34509 +static int au_wbr_create_exp(struct dentry *dentry)
34512 + aufs_bindex_t bwh, bdiropq;
34513 + struct dentry *parent;
34516 + bwh = au_dbwh(dentry);
34517 + parent = dget_parent(dentry);
34518 + bdiropq = au_dbdiropq(parent);
34520 + if (bdiropq >= 0)
34521 + err = min(bdiropq, bwh);
34524 + AuDbg("%d\n", err);
34525 + } else if (bdiropq >= 0) {
34527 + AuDbg("%d\n", err);
34532 + err = au_wbr_nonopq(dentry, err);
34534 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
34537 + AuDbg("%d\n", err);
34541 +/* ---------------------------------------------------------------------- */
34544 +static int au_wbr_create_init_rr(struct super_block *sb)
34548 + err = au_wbr_bu(sb, au_sbbot(sb));
34549 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
34552 + AuDbg("b%d\n", err);
34556 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
34560 + aufs_bindex_t bindex, bbot;
34561 + struct super_block *sb;
34564 + err = au_wbr_create_exp(dentry);
34568 + sb = dentry->d_sb;
34569 + next = &au_sbi(sb)->si_wbr_rr_next;
34570 + bbot = au_sbbot(sb);
34572 + for (bindex = 0; bindex <= bbot; bindex++) {
34573 + if (!au_ftest_wbr(flags, DIR)) {
34574 + err = atomic_dec_return(next) + 1;
34575 + /* modulo for 0 is meaningless */
34576 + if (unlikely(!err))
34577 + err = atomic_dec_return(next) + 1;
34579 + err = atomic_read(next);
34580 + AuDbg("%d\n", err);
34583 + AuDbg("%d\n", err);
34584 + if (!au_br_rdonly(au_sbr(sb, err)))
34590 + err = au_wbr_nonopq(dentry, err);
34593 + AuDbg("%d\n", err);
34597 +/* ---------------------------------------------------------------------- */
34599 +/* most free space */
34600 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
34602 + struct super_block *sb;
34603 + struct au_branch *br;
34604 + struct au_wbr_mfs *mfs;
34605 + struct dentry *h_parent;
34606 + aufs_bindex_t bindex, bbot;
34608 + unsigned long long b, bavail;
34609 + struct path h_path;
34610 + /* reduce the stack usage */
34611 + struct kstatfs *st;
34613 + st = kmalloc(sizeof(*st), GFP_NOFS);
34614 + if (unlikely(!st)) {
34615 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
34620 + sb = dentry->d_sb;
34621 + mfs = &au_sbi(sb)->si_wbr_mfs;
34622 + MtxMustLock(&mfs->mfs_lock);
34623 + mfs->mfs_bindex = -EROFS;
34624 + mfs->mfsrr_bytes = 0;
34627 + bbot = au_sbbot(sb);
34629 + bindex = au_dbtop(parent);
34630 + bbot = au_dbtaildir(parent);
34633 + for (; bindex <= bbot; bindex++) {
34635 + h_parent = au_h_dptr(parent, bindex);
34636 + if (!h_parent || d_is_negative(h_parent))
34639 + br = au_sbr(sb, bindex);
34640 + if (au_br_rdonly(br))
34643 + /* sb->s_root for NFS is unreliable */
34644 + h_path.mnt = au_br_mnt(br);
34645 + h_path.dentry = h_path.mnt->mnt_root;
34646 + err = vfs_statfs(&h_path, st);
34647 + if (unlikely(err)) {
34648 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
34652 + /* when the available size is equal, select the lower one */
34653 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
34654 + || sizeof(b) < sizeof(st->f_bsize));
34655 + b = st->f_bavail * st->f_bsize;
34656 + br->br_wbr->wbr_bytes = b;
34657 + if (b >= bavail) {
34659 + mfs->mfs_bindex = bindex;
34660 + mfs->mfs_jiffy = jiffies;
34664 + mfs->mfsrr_bytes = bavail;
34665 + AuDbg("b%d\n", mfs->mfs_bindex);
34666 + au_kfree_rcu(st);
34669 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
34672 + struct dentry *parent;
34673 + struct super_block *sb;
34674 + struct au_wbr_mfs *mfs;
34676 + err = au_wbr_create_exp(dentry);
34680 + sb = dentry->d_sb;
34682 + if (au_ftest_wbr(flags, PARENT))
34683 + parent = dget_parent(dentry);
34684 + mfs = &au_sbi(sb)->si_wbr_mfs;
34685 + mutex_lock(&mfs->mfs_lock);
34686 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34687 + || mfs->mfs_bindex < 0
34688 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
34689 + au_mfs(dentry, parent);
34690 + mutex_unlock(&mfs->mfs_lock);
34691 + err = mfs->mfs_bindex;
34695 + err = au_wbr_nonopq(dentry, err);
34698 + AuDbg("b%d\n", err);
34702 +static int au_wbr_create_init_mfs(struct super_block *sb)
34704 + struct au_wbr_mfs *mfs;
34706 + mfs = &au_sbi(sb)->si_wbr_mfs;
34707 + mutex_init(&mfs->mfs_lock);
34708 + mfs->mfs_jiffy = 0;
34709 + mfs->mfs_bindex = -EROFS;
34714 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
34716 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
34720 +/* ---------------------------------------------------------------------- */
34722 +/* top down regardless parent, and then mfs */
34723 +static int au_wbr_create_tdmfs(struct dentry *dentry,
34724 + unsigned int flags __maybe_unused)
34727 + aufs_bindex_t bwh, btail, bindex, bfound, bmfs;
34728 + unsigned long long watermark;
34729 + struct super_block *sb;
34730 + struct au_wbr_mfs *mfs;
34731 + struct au_branch *br;
34732 + struct dentry *parent;
34734 + sb = dentry->d_sb;
34735 + mfs = &au_sbi(sb)->si_wbr_mfs;
34736 + mutex_lock(&mfs->mfs_lock);
34737 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34738 + || mfs->mfs_bindex < 0)
34739 + au_mfs(dentry, /*parent*/NULL);
34740 + watermark = mfs->mfsrr_watermark;
34741 + bmfs = mfs->mfs_bindex;
34742 + mutex_unlock(&mfs->mfs_lock);
34744 + /* another style of au_wbr_create_exp() */
34745 + bwh = au_dbwh(dentry);
34746 + parent = dget_parent(dentry);
34747 + btail = au_dbtaildir(parent);
34748 + if (bwh >= 0 && bwh < btail)
34751 + err = au_wbr_nonopq(dentry, btail);
34752 + if (unlikely(err < 0))
34756 + for (bindex = 0; bindex <= btail; bindex++) {
34757 + br = au_sbr(sb, bindex);
34758 + if (au_br_rdonly(br))
34760 + if (br->br_wbr->wbr_bytes > watermark) {
34771 + AuDbg("b%d\n", err);
34775 +/* ---------------------------------------------------------------------- */
34777 +/* most free space and then round robin */
34778 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
34781 + struct au_wbr_mfs *mfs;
34783 + err = au_wbr_create_mfs(dentry, flags);
34785 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
34786 + mutex_lock(&mfs->mfs_lock);
34787 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
34788 + err = au_wbr_create_rr(dentry, flags);
34789 + mutex_unlock(&mfs->mfs_lock);
34792 + AuDbg("b%d\n", err);
34796 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
34800 + au_wbr_create_init_mfs(sb); /* ignore */
34801 + err = au_wbr_create_init_rr(sb);
34806 +/* ---------------------------------------------------------------------- */
34808 +/* top down parent and most free space */
34809 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
34812 + unsigned long long b;
34813 + aufs_bindex_t bindex, btop, bbot;
34814 + struct super_block *sb;
34815 + struct dentry *parent, *h_parent;
34816 + struct au_branch *br;
34818 + err = au_wbr_create_tdp(dentry, flags);
34819 + if (unlikely(err < 0))
34821 + parent = dget_parent(dentry);
34822 + btop = au_dbtop(parent);
34823 + bbot = au_dbtaildir(parent);
34824 + if (btop == bbot)
34825 + goto out_parent; /* success */
34827 + e2 = au_wbr_create_mfs(dentry, flags);
34829 + goto out_parent; /* success */
34831 + /* when the available size is equal, select upper one */
34832 + sb = dentry->d_sb;
34833 + br = au_sbr(sb, err);
34834 + b = br->br_wbr->wbr_bytes;
34835 + AuDbg("b%d, %llu\n", err, b);
34837 + for (bindex = btop; bindex <= bbot; bindex++) {
34838 + h_parent = au_h_dptr(parent, bindex);
34839 + if (!h_parent || d_is_negative(h_parent))
34842 + br = au_sbr(sb, bindex);
34843 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
34844 + b = br->br_wbr->wbr_bytes;
34846 + AuDbg("b%d, %llu\n", err, b);
34851 + err = au_wbr_nonopq(dentry, err);
34856 + AuDbg("b%d\n", err);
34860 +/* ---------------------------------------------------------------------- */
34863 + * - top down parent
34864 + * - most free space with parent
34865 + * - most free space round-robin regardless parent
34867 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
34870 + unsigned long long watermark;
34871 + struct super_block *sb;
34872 + struct au_branch *br;
34873 + struct au_wbr_mfs *mfs;
34875 + err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
34876 + if (unlikely(err < 0))
34879 + sb = dentry->d_sb;
34880 + br = au_sbr(sb, err);
34881 + mfs = &au_sbi(sb)->si_wbr_mfs;
34882 + mutex_lock(&mfs->mfs_lock);
34883 + watermark = mfs->mfsrr_watermark;
34884 + mutex_unlock(&mfs->mfs_lock);
34885 + if (br->br_wbr->wbr_bytes < watermark)
34886 + /* regardless the parent dir */
34887 + err = au_wbr_create_mfsrr(dentry, flags);
34890 + AuDbg("b%d\n", err);
34894 +/* ---------------------------------------------------------------------- */
34896 +/* policies for copyup */
34898 +/* top down parent */
34899 +static int au_wbr_copyup_tdp(struct dentry *dentry)
34901 + return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
34904 +/* bottom up parent */
34905 +static int au_wbr_copyup_bup(struct dentry *dentry)
34908 + aufs_bindex_t bindex, btop;
34909 + struct dentry *parent, *h_parent;
34910 + struct super_block *sb;
34913 + sb = dentry->d_sb;
34914 + parent = dget_parent(dentry);
34915 + btop = au_dbtop(parent);
34916 + for (bindex = au_dbtop(dentry); bindex >= btop; bindex--) {
34917 + h_parent = au_h_dptr(parent, bindex);
34918 + if (!h_parent || d_is_negative(h_parent))
34921 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
34928 + /* bottom up here */
34929 + if (unlikely(err < 0))
34930 + err = au_wbr_bu(sb, btop - 1);
34932 + AuDbg("b%d\n", err);
34937 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop)
34941 + err = au_wbr_bu(dentry->d_sb, btop);
34942 + AuDbg("b%d\n", err);
34944 + err = au_wbr_nonopq(dentry, err);
34946 + AuDbg("b%d\n", err);
34950 +static int au_wbr_copyup_bu(struct dentry *dentry)
34953 + aufs_bindex_t btop;
34955 + btop = au_dbtop(dentry);
34956 + err = au_wbr_do_copyup_bu(dentry, btop);
34960 +/* ---------------------------------------------------------------------- */
34962 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
34963 + [AuWbrCopyup_TDP] = {
34964 + .copyup = au_wbr_copyup_tdp
34966 + [AuWbrCopyup_BUP] = {
34967 + .copyup = au_wbr_copyup_bup
34969 + [AuWbrCopyup_BU] = {
34970 + .copyup = au_wbr_copyup_bu
34974 +struct au_wbr_create_operations au_wbr_create_ops[] = {
34975 + [AuWbrCreate_TDP] = {
34976 + .create = au_wbr_create_tdp
34978 + [AuWbrCreate_RR] = {
34979 + .create = au_wbr_create_rr,
34980 + .init = au_wbr_create_init_rr
34982 + [AuWbrCreate_MFS] = {
34983 + .create = au_wbr_create_mfs,
34984 + .init = au_wbr_create_init_mfs,
34985 + .fin = au_wbr_create_fin_mfs
34987 + [AuWbrCreate_MFSV] = {
34988 + .create = au_wbr_create_mfs,
34989 + .init = au_wbr_create_init_mfs,
34990 + .fin = au_wbr_create_fin_mfs
34992 + [AuWbrCreate_MFSRR] = {
34993 + .create = au_wbr_create_mfsrr,
34994 + .init = au_wbr_create_init_mfsrr,
34995 + .fin = au_wbr_create_fin_mfs
34997 + [AuWbrCreate_MFSRRV] = {
34998 + .create = au_wbr_create_mfsrr,
34999 + .init = au_wbr_create_init_mfsrr,
35000 + .fin = au_wbr_create_fin_mfs
35002 + [AuWbrCreate_TDMFS] = {
35003 + .create = au_wbr_create_tdmfs,
35004 + .init = au_wbr_create_init_mfs,
35005 + .fin = au_wbr_create_fin_mfs
35007 + [AuWbrCreate_TDMFSV] = {
35008 + .create = au_wbr_create_tdmfs,
35009 + .init = au_wbr_create_init_mfs,
35010 + .fin = au_wbr_create_fin_mfs
35012 + [AuWbrCreate_PMFS] = {
35013 + .create = au_wbr_create_pmfs,
35014 + .init = au_wbr_create_init_mfs,
35015 + .fin = au_wbr_create_fin_mfs
35017 + [AuWbrCreate_PMFSV] = {
35018 + .create = au_wbr_create_pmfs,
35019 + .init = au_wbr_create_init_mfs,
35020 + .fin = au_wbr_create_fin_mfs
35022 + [AuWbrCreate_PMFSRR] = {
35023 + .create = au_wbr_create_pmfsrr,
35024 + .init = au_wbr_create_init_mfsrr,
35025 + .fin = au_wbr_create_fin_mfs
35027 + [AuWbrCreate_PMFSRRV] = {
35028 + .create = au_wbr_create_pmfsrr,
35029 + .init = au_wbr_create_init_mfsrr,
35030 + .fin = au_wbr_create_fin_mfs
35033 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
35034 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
35035 +++ linux/fs/aufs/whout.c 2022-03-21 14:49:05.733299677 +0100
35037 +// SPDX-License-Identifier: GPL-2.0
35039 + * Copyright (C) 2005-2021 Junjiro R. Okajima
35041 + * This program, aufs is free software; you can redistribute it and/or modify
35042 + * it under the terms of the GNU General Public License as published by
35043 + * the Free Software Foundation; either version 2 of the License, or
35044 + * (at your option) any later version.
35046 + * This program is distributed in the hope that it will be useful,
35047 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35048 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35049 + * GNU General Public License for more details.
35051 + * You should have received a copy of the GNU General Public License
35052 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35056 + * whiteout for logical deletion and opaque directory
35061 +#define WH_MASK 0444
35064 + * If a directory contains this file, then it is opaque. We start with the
35065 + * .wh. flag so that it is blocked by lookup.
35067 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
35068 + sizeof(AUFS_WH_DIROPQ) - 1);
35071 + * generate whiteout name, which is NOT terminated by NULL.
35072 + * @name: original d_name.name
35073 + * @len: original d_name.len
35074 + * @wh: whiteout qstr
35075 + * returns zero when succeeds, otherwise error.
35076 + * succeeded value as wh->name should be freed by kfree().
35078 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
35082 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
35083 + return -ENAMETOOLONG;
35085 + wh->len = name->len + AUFS_WH_PFX_LEN;
35086 + p = kmalloc(wh->len, GFP_NOFS);
35089 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35090 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
35097 +/* ---------------------------------------------------------------------- */
35100 + * test if the @wh_name exists under @h_ppath.
35101 + * @try_sio specifies the necessary of super-io.
35103 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
35104 + struct qstr *wh_name, int try_sio)
35107 + struct dentry *wh_dentry;
35110 + wh_dentry = vfsub_lkup_one(wh_name, h_ppath);
35112 + wh_dentry = au_sio_lkup_one(h_userns, wh_name, h_ppath);
35113 + err = PTR_ERR(wh_dentry);
35114 + if (IS_ERR(wh_dentry)) {
35115 + if (err == -ENAMETOOLONG)
35121 + if (d_is_negative(wh_dentry))
35122 + goto out_wh; /* success */
35125 + if (d_is_reg(wh_dentry))
35126 + goto out_wh; /* success */
35129 + AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
35130 + wh_dentry, d_inode(wh_dentry)->i_mode);
35139 + * test if the @h_path->dentry sets opaque or not.
35141 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path)
35144 + struct inode *h_dir;
35146 + h_dir = d_inode(h_path->dentry);
35147 + err = au_wh_test(h_userns, h_path, &diropq_name,
35148 + au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC));
35153 + * returns a negative dentry whose name is unique and temporary.
35155 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
35156 + struct qstr *prefix)
35158 + struct dentry *dentry;
35160 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
35162 + /* strict atomic_t is unnecessary here */
35163 + static unsigned short cnt;
35165 + struct path h_ppath;
35166 + struct user_namespace *h_userns;
35168 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
35171 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
35172 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
35173 + dentry = ERR_PTR(-ENAMETOOLONG);
35174 + if (unlikely(qs.len > NAME_MAX))
35176 + dentry = ERR_PTR(-ENOMEM);
35177 + name = kmalloc(qs.len + 1, GFP_NOFS);
35178 + if (unlikely(!name))
35182 + /* doubly whiteout-ed */
35183 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
35184 + p = name + AUFS_WH_PFX_LEN * 2;
35185 + memcpy(p, prefix->name, prefix->len);
35186 + p += prefix->len;
35188 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
35190 + h_ppath.dentry = h_parent;
35191 + h_ppath.mnt = au_br_mnt(br);
35192 + h_userns = au_br_userns(br);
35194 + for (i = 0; i < 3; i++) {
35195 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
35196 + dentry = au_sio_lkup_one(h_userns, &qs, &h_ppath);
35197 + if (IS_ERR(dentry) || d_is_negative(dentry))
35201 + /* pr_warn("could not get random name\n"); */
35202 + dentry = ERR_PTR(-EEXIST);
35203 + AuDbg("%.*s\n", AuLNPair(&qs));
35207 + if (name != defname)
35208 + au_kfree_try_rcu(name);
35210 + AuTraceErrPtr(dentry);
35215 + * rename the @h_dentry on @br to the whiteouted temporary name.
35217 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
35220 + struct path h_path = {
35221 + .mnt = au_br_mnt(br)
35223 + struct inode *h_dir, *delegated;
35224 + struct dentry *h_parent;
35226 + h_parent = h_dentry->d_parent; /* dir inode is locked */
35227 + h_dir = d_inode(h_parent);
35228 + IMustLock(h_dir);
35230 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
35231 + err = PTR_ERR(h_path.dentry);
35232 + if (IS_ERR(h_path.dentry))
35235 + /* under the same dir, no need to lock_rename() */
35236 + delegated = NULL;
35237 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated,
35240 + if (unlikely(err == -EWOULDBLOCK)) {
35241 + pr_warn("cannot retry for NFSv4 delegation"
35242 + " for an internal rename\n");
35245 + dput(h_path.dentry);
35252 +/* ---------------------------------------------------------------------- */
35254 + * functions for removing a whiteout
35257 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
35260 + struct inode *delegated;
35263 + * forces superio when the dir has a sticky bit.
35264 + * this may be a violation of unix fs semantics.
35266 + force = (h_dir->i_mode & S_ISVTX)
35267 + && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
35268 + delegated = NULL;
35269 + err = vfsub_unlink(h_dir, h_path, &delegated, force);
35270 + if (unlikely(err == -EWOULDBLOCK)) {
35271 + pr_warn("cannot retry for NFSv4 delegation"
35272 + " for an internal unlink\n");
35278 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
35279 + struct dentry *dentry)
35283 + err = do_unlink_wh(h_dir, h_path);
35284 + if (!err && dentry)
35285 + au_set_dbwh(dentry, -1);
35290 +static int unlink_wh_name(struct path *h_ppath, struct qstr *wh)
35293 + struct path h_path;
35296 + h_path.dentry = vfsub_lkup_one(wh, h_ppath);
35297 + if (IS_ERR(h_path.dentry))
35298 + err = PTR_ERR(h_path.dentry);
35300 + if (d_is_reg(h_path.dentry)) {
35301 + h_path.mnt = h_ppath->mnt;
35302 + err = do_unlink_wh(d_inode(h_ppath->dentry), &h_path);
35304 + dput(h_path.dentry);
35310 +/* ---------------------------------------------------------------------- */
35312 + * initialize/clean whiteout for a branch
35315 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
35319 + struct inode *delegated;
35321 + if (d_is_negative(whpath->dentry))
35325 + err = vfsub_rmdir(h_dir, whpath);
35327 + delegated = NULL;
35328 + err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
35329 + if (unlikely(err == -EWOULDBLOCK)) {
35330 + pr_warn("cannot retry for NFSv4 delegation"
35331 + " for an internal unlink\n");
35335 + if (unlikely(err))
35336 + pr_warn("failed removing %pd (%d), ignored.\n",
35337 + whpath->dentry, err);
35340 +static int test_linkable(struct dentry *h_root)
35342 + struct inode *h_dir = d_inode(h_root);
35344 + if (h_dir->i_op->link)
35347 + pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
35348 + h_root, au_sbtype(h_root->d_sb));
35349 + return -ENOSYS; /* the branch doesn't have its ->link() */
35352 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
35353 +static int au_whdir(struct inode *h_dir, struct path *path)
35358 + if (d_is_negative(path->dentry)) {
35361 + if (au_test_nfs(path->dentry->d_sb))
35363 + err = vfsub_mkdir(h_dir, path, mode);
35364 + } else if (d_is_dir(path->dentry))
35367 + pr_err("unknown %pd exists\n", path->dentry);
35372 +struct au_wh_base {
35373 + const struct qstr *name;
35374 + struct dentry *dentry;
35377 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
35378 + struct path *h_path)
35380 + h_path->dentry = base[AuBrWh_BASE].dentry;
35381 + au_wh_clean(h_dir, h_path, /*isdir*/0);
35382 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35383 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35384 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35385 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35389 + * returns tri-state,
35390 + * minus: error, caller should print the message
35392 + * plus: error, caller should NOT print the message
35394 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
35395 + int do_plink, struct au_wh_base base[],
35396 + struct path *h_path)
35399 + struct inode *h_dir;
35401 + h_dir = d_inode(h_root);
35402 + h_path->dentry = base[AuBrWh_BASE].dentry;
35403 + au_wh_clean(h_dir, h_path, /*isdir*/0);
35404 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35406 + err = test_linkable(h_root);
35407 + if (unlikely(err)) {
35412 + err = au_whdir(h_dir, h_path);
35413 + if (unlikely(err))
35415 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35417 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35418 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35419 + err = au_whdir(h_dir, h_path);
35420 + if (unlikely(err))
35422 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35429 + * for the moment, aufs supports the branch filesystem which does not support
35430 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
35431 + * copyup failed. finally, such filesystem will not be used as the writable
35434 + * returns tri-state, see above.
35436 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
35437 + int do_plink, struct au_wh_base base[],
35438 + struct path *h_path)
35441 + struct inode *h_dir;
35443 + WbrWhMustWriteLock(wbr);
35445 + err = test_linkable(h_root);
35446 + if (unlikely(err)) {
35452 + * todo: should this create be done in /sbin/mount.aufs helper?
35455 + h_dir = d_inode(h_root);
35456 + if (d_is_negative(base[AuBrWh_BASE].dentry)) {
35457 + h_path->dentry = base[AuBrWh_BASE].dentry;
35458 + err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
35459 + } else if (d_is_reg(base[AuBrWh_BASE].dentry))
35462 + pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
35463 + if (unlikely(err))
35466 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35468 + err = au_whdir(h_dir, h_path);
35469 + if (unlikely(err))
35471 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35473 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35474 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
35476 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35477 + err = au_whdir(h_dir, h_path);
35478 + if (unlikely(err))
35480 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35487 + * initialize the whiteout base file/dir for @br.
35489 +int au_wh_init(struct au_branch *br, struct super_block *sb)
35492 + const unsigned char do_plink
35493 + = !!au_opt_test(au_mntflags(sb), PLINK);
35494 + struct inode *h_dir;
35495 + struct path path = br->br_path;
35496 + struct dentry *h_root = path.dentry;
35497 + struct au_wbr *wbr = br->br_wbr;
35498 + static const struct qstr base_name[] = {
35499 + [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
35500 + sizeof(AUFS_BASE_NAME) - 1),
35501 + [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
35502 + sizeof(AUFS_PLINKDIR_NAME) - 1),
35503 + [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
35504 + sizeof(AUFS_ORPHDIR_NAME) - 1)
35506 + struct au_wh_base base[] = {
35507 + [AuBrWh_BASE] = {
35508 + .name = base_name + AuBrWh_BASE,
35511 + [AuBrWh_PLINK] = {
35512 + .name = base_name + AuBrWh_PLINK,
35515 + [AuBrWh_ORPH] = {
35516 + .name = base_name + AuBrWh_ORPH,
35522 + WbrWhMustWriteLock(wbr);
35524 + for (i = 0; i < AuBrWh_Last; i++) {
35525 + /* doubly whiteouted */
35526 + struct dentry *d;
35528 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
35529 + err = PTR_ERR(d);
35533 + base[i].dentry = d;
35535 + && wbr->wbr_wh[i]
35536 + && wbr->wbr_wh[i] != base[i].dentry);
35540 + for (i = 0; i < AuBrWh_Last; i++) {
35541 + dput(wbr->wbr_wh[i]);
35542 + wbr->wbr_wh[i] = NULL;
35546 + if (!au_br_writable(br->br_perm)) {
35547 + h_dir = d_inode(h_root);
35548 + au_wh_init_ro(h_dir, base, &path);
35549 + } else if (!au_br_wh_linkable(br->br_perm)) {
35550 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
35556 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
35562 + goto out; /* success */
35565 + pr_err("an error(%d) on the writable branch %pd(%s)\n",
35566 + err, h_root, au_sbtype(h_root->d_sb));
35568 + for (i = 0; i < AuBrWh_Last; i++)
35569 + dput(base[i].dentry);
35573 +/* ---------------------------------------------------------------------- */
35575 + * whiteouts are all hard-linked usually.
35576 + * when its link count reaches a ceiling, we create a new whiteout base
35577 + * asynchronously.
35580 +struct reinit_br_wh {
35581 + struct super_block *sb;
35582 + struct au_branch *br;
35585 +static void reinit_br_wh(void *arg)
35588 + aufs_bindex_t bindex;
35589 + struct path h_path;
35590 + struct reinit_br_wh *a = arg;
35591 + struct au_wbr *wbr;
35592 + struct inode *dir, *delegated;
35593 + struct dentry *h_root;
35594 + struct au_hinode *hdir;
35597 + wbr = a->br->br_wbr;
35598 + /* big aufs lock */
35599 + si_noflush_write_lock(a->sb);
35600 + if (!au_br_writable(a->br->br_perm))
35602 + bindex = au_br_index(a->sb, a->br->br_id);
35603 + if (unlikely(bindex < 0))
35606 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
35607 + dir = d_inode(a->sb->s_root);
35608 + hdir = au_hi(dir, bindex);
35609 + h_root = au_h_dptr(a->sb->s_root, bindex);
35610 + AuDebugOn(h_root != au_br_dentry(a->br));
35612 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35613 + wbr_wh_write_lock(wbr);
35614 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
35617 + h_path.dentry = wbr->wbr_whbase;
35618 + h_path.mnt = au_br_mnt(a->br);
35619 + delegated = NULL;
35620 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
35622 + if (unlikely(err == -EWOULDBLOCK)) {
35623 + pr_warn("cannot retry for NFSv4 delegation"
35624 + " for an internal unlink\n");
35628 + pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
35631 + dput(wbr->wbr_whbase);
35632 + wbr->wbr_whbase = NULL;
35634 + err = au_wh_init(a->br, a->sb);
35635 + wbr_wh_write_unlock(wbr);
35636 + au_hn_inode_unlock(hdir);
35637 + di_read_unlock(a->sb->s_root, AuLock_IR);
35639 + au_fhsm_wrote(a->sb, bindex, /*force*/0);
35643 + atomic_dec(&wbr->wbr_wh_running);
35644 + au_lcnt_dec(&a->br->br_count);
35645 + si_write_unlock(a->sb);
35646 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
35648 + if (unlikely(err))
35649 + AuIOErr("err %d\n", err);
35652 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
35654 + int do_dec, wkq_err;
35655 + struct reinit_br_wh *arg;
35658 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
35661 + /* ignore ENOMEM */
35662 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
35665 + * dec(wh_running), kfree(arg) and dec(br_count)
35666 + * in reinit function
35670 + au_lcnt_inc(&br->br_count);
35671 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
35672 + if (unlikely(wkq_err)) {
35673 + atomic_dec(&br->br_wbr->wbr_wh_running);
35674 + au_lcnt_dec(&br->br_count);
35675 + au_kfree_rcu(arg);
35682 + atomic_dec(&br->br_wbr->wbr_wh_running);
35685 +/* ---------------------------------------------------------------------- */
35688 + * create the whiteout @wh.
35690 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
35691 + struct dentry *wh)
35694 + struct path h_path = {
35697 + struct au_branch *br;
35698 + struct au_wbr *wbr;
35699 + struct dentry *h_parent;
35700 + struct inode *h_dir, *delegated;
35702 + h_parent = wh->d_parent; /* dir inode is locked */
35703 + h_dir = d_inode(h_parent);
35704 + IMustLock(h_dir);
35706 + br = au_sbr(sb, bindex);
35707 + h_path.mnt = au_br_mnt(br);
35708 + wbr = br->br_wbr;
35709 + wbr_wh_read_lock(wbr);
35710 + if (wbr->wbr_whbase) {
35711 + delegated = NULL;
35712 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
35713 + if (unlikely(err == -EWOULDBLOCK)) {
35714 + pr_warn("cannot retry for NFSv4 delegation"
35715 + " for an internal link\n");
35718 + if (!err || err != -EMLINK)
35721 + /* link count full. re-initialize br_whbase. */
35722 + kick_reinit_br_wh(sb, br);
35725 + /* return this error in this context */
35726 + err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
35728 + au_fhsm_wrote(sb, bindex, /*force*/0);
35731 + wbr_wh_read_unlock(wbr);
35735 +/* ---------------------------------------------------------------------- */
35738 + * create or remove the diropq.
35740 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
35741 + unsigned int flags)
35743 + struct dentry *opq_dentry;
35744 + struct super_block *sb;
35745 + struct au_branch *br;
35746 + struct path h_path;
35749 + sb = dentry->d_sb;
35750 + br = au_sbr(sb, bindex);
35751 + h_path.dentry = au_h_dptr(dentry, bindex);
35752 + h_path.mnt = au_br_mnt(br);
35753 + opq_dentry = vfsub_lkup_one(&diropq_name, &h_path);
35754 + if (IS_ERR(opq_dentry))
35757 + if (au_ftest_diropq(flags, CREATE)) {
35758 + err = link_or_create_wh(sb, bindex, opq_dentry);
35760 + au_set_dbdiropq(dentry, bindex);
35761 + goto out; /* success */
35764 + h_path.dentry = opq_dentry;
35765 + err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &h_path);
35767 + au_set_dbdiropq(dentry, -1);
35769 + dput(opq_dentry);
35770 + opq_dentry = ERR_PTR(err);
35773 + return opq_dentry;
35776 +struct do_diropq_args {
35777 + struct dentry **errp;
35778 + struct dentry *dentry;
35779 + aufs_bindex_t bindex;
35780 + unsigned int flags;
35783 +static void call_do_diropq(void *args)
35785 + struct do_diropq_args *a = args;
35786 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
35789 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35790 + unsigned int flags)
35792 + struct dentry *diropq, *h_dentry;
35793 + struct user_namespace *h_userns;
35795 + h_userns = au_sbr_userns(dentry->d_sb, bindex);
35796 + h_dentry = au_h_dptr(dentry, bindex);
35797 + if (!au_test_h_perm_sio(h_userns, d_inode(h_dentry),
35798 + MAY_EXEC | MAY_WRITE))
35799 + diropq = do_diropq(dentry, bindex, flags);
35802 + struct do_diropq_args args = {
35804 + .dentry = dentry,
35805 + .bindex = bindex,
35809 + wkq_err = au_wkq_wait(call_do_diropq, &args);
35810 + if (unlikely(wkq_err))
35811 + diropq = ERR_PTR(wkq_err);
35817 +/* ---------------------------------------------------------------------- */
35820 + * lookup whiteout dentry.
35821 + * @h_parent: lower parent dentry which must exist and be locked
35822 + * @base_name: name of dentry which will be whiteouted
35823 + * returns dentry for whiteout.
35825 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35826 + struct au_branch *br)
35829 + struct qstr wh_name;
35830 + struct dentry *wh_dentry;
35831 + struct path h_path;
35833 + err = au_wh_name_alloc(&wh_name, base_name);
35834 + wh_dentry = ERR_PTR(err);
35836 + h_path.dentry = h_parent;
35837 + h_path.mnt = au_br_mnt(br);
35838 + wh_dentry = vfsub_lkup_one(&wh_name, &h_path);
35839 + au_kfree_try_rcu(wh_name.name);
35841 + return wh_dentry;
35845 + * link/create a whiteout for @dentry on @bindex.
35847 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35848 + struct dentry *h_parent)
35850 + struct dentry *wh_dentry;
35851 + struct super_block *sb;
35854 + sb = dentry->d_sb;
35855 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
35856 + if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
35857 + err = link_or_create_wh(sb, bindex, wh_dentry);
35859 + au_set_dbwh(dentry, bindex);
35860 + au_fhsm_wrote(sb, bindex, /*force*/0);
35863 + wh_dentry = ERR_PTR(err);
35867 + return wh_dentry;
35870 +/* ---------------------------------------------------------------------- */
35872 +/* Delete all whiteouts in this directory on branch bindex. */
35873 +static int del_wh_children(struct path *h_path, struct au_nhash *whlist,
35874 + aufs_bindex_t bindex)
35877 + unsigned long ul, n;
35878 + struct qstr wh_name;
35880 + struct hlist_head *head;
35881 + struct au_vdir_wh *pos;
35882 + struct au_vdir_destr *str;
35885 + p = (void *)__get_free_page(GFP_NOFS);
35886 + wh_name.name = p;
35887 + if (unlikely(!wh_name.name))
35891 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35892 + p += AUFS_WH_PFX_LEN;
35893 + n = whlist->nh_num;
35894 + head = whlist->nh_head;
35895 + for (ul = 0; !err && ul < n; ul++, head++) {
35896 + hlist_for_each_entry(pos, head, wh_hash) {
35897 + if (pos->wh_bindex != bindex)
35900 + str = &pos->wh_str;
35901 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
35902 + memcpy(p, str->name, str->len);
35903 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
35904 + err = unlink_wh_name(h_path, &wh_name);
35909 + AuIOErr("whiteout name too long %.*s\n",
35910 + str->len, str->name);
35915 + free_page((unsigned long)wh_name.name);
35921 +struct del_wh_children_args {
35923 + struct path *h_path;
35924 + struct au_nhash *whlist;
35925 + aufs_bindex_t bindex;
35928 +static void call_del_wh_children(void *args)
35930 + struct del_wh_children_args *a = args;
35931 + *a->errp = del_wh_children(a->h_path, a->whlist, a->bindex);
35934 +/* ---------------------------------------------------------------------- */
35936 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
35938 + struct au_whtmp_rmdir *whtmp;
35940 + unsigned int rdhash;
35942 + SiMustAnyLock(sb);
35944 + whtmp = kzalloc(sizeof(*whtmp), gfp);
35945 + if (unlikely(!whtmp)) {
35946 + whtmp = ERR_PTR(-ENOMEM);
35950 + /* no estimation for dir size */
35951 + rdhash = au_sbi(sb)->si_rdhash;
35953 + rdhash = AUFS_RDHASH_DEF;
35954 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
35955 + if (unlikely(err)) {
35956 + au_kfree_rcu(whtmp);
35957 + whtmp = ERR_PTR(err);
35964 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
35967 + au_lcnt_dec(&whtmp->br->br_count);
35968 + dput(whtmp->wh_dentry);
35969 + iput(whtmp->dir);
35970 + au_nhash_wh_free(&whtmp->whlist);
35971 + au_kfree_rcu(whtmp);
35975 + * rmdir the whiteouted temporary named dir @h_dentry.
35976 + * @whlist: whiteouted children.
35978 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
35979 + struct dentry *wh_dentry, struct au_nhash *whlist)
35982 + unsigned int h_nlink;
35983 + struct path wh_path;
35984 + struct inode *wh_inode, *h_dir;
35985 + struct au_branch *br;
35986 + struct user_namespace *h_userns;
35988 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
35989 + IMustLock(h_dir);
35991 + br = au_sbr(dir->i_sb, bindex);
35992 + wh_path.dentry = wh_dentry;
35993 + wh_path.mnt = au_br_mnt(br);
35994 + h_userns = au_br_userns(br);
35995 + wh_inode = d_inode(wh_dentry);
35996 + inode_lock_nested(wh_inode, AuLsc_I_CHILD);
35999 + * someone else might change some whiteouts while we were sleeping.
36000 + * it means this whlist may have an obsoleted entry.
36002 + if (!au_test_h_perm_sio(h_userns, wh_inode, MAY_EXEC | MAY_WRITE))
36003 + err = del_wh_children(&wh_path, whlist, bindex);
36006 + struct del_wh_children_args args = {
36008 + .h_path = &wh_path,
36009 + .whlist = whlist,
36013 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
36014 + if (unlikely(wkq_err))
36017 + inode_unlock(wh_inode);
36020 + h_nlink = h_dir->i_nlink;
36021 + err = vfsub_rmdir(h_dir, &wh_path);
36022 + /* some fs doesn't change the parent nlink in some cases */
36023 + h_nlink -= h_dir->i_nlink;
36027 + if (au_ibtop(dir) == bindex) {
36028 + /* todo: dir->i_mutex is necessary */
36029 + au_cpup_attr_timesizes(dir);
36031 + vfsub_drop_nlink(dir);
36033 + return 0; /* success */
36036 + pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
36040 +static void call_rmdir_whtmp(void *args)
36043 + aufs_bindex_t bindex;
36044 + struct au_whtmp_rmdir *a = args;
36045 + struct super_block *sb;
36046 + struct dentry *h_parent;
36047 + struct inode *h_dir;
36048 + struct au_hinode *hdir;
36050 + /* rmdir by nfsd may cause deadlock with this i_mutex */
36051 + /* inode_lock(a->dir); */
36053 + sb = a->dir->i_sb;
36054 + si_read_lock(sb, !AuLock_FLUSH);
36055 + if (!au_br_writable(a->br->br_perm))
36057 + bindex = au_br_index(sb, a->br->br_id);
36058 + if (unlikely(bindex < 0))
36062 + ii_write_lock_parent(a->dir);
36063 + h_parent = dget_parent(a->wh_dentry);
36064 + h_dir = d_inode(h_parent);
36065 + hdir = au_hi(a->dir, bindex);
36066 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
36067 + if (unlikely(err))
36069 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
36070 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
36073 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
36074 + au_hn_inode_unlock(hdir);
36075 + vfsub_mnt_drop_write(au_br_mnt(a->br));
36079 + ii_write_unlock(a->dir);
36081 + /* inode_unlock(a->dir); */
36082 + au_whtmp_rmdir_free(a);
36083 + si_read_unlock(sb);
36084 + au_nwt_done(&au_sbi(sb)->si_nowait);
36085 + if (unlikely(err))
36086 + AuIOErr("err %d\n", err);
36089 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36090 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
36093 + struct super_block *sb;
36097 + /* all post-process will be done in do_rmdir_whtmp(). */
36099 + args->dir = au_igrab(dir);
36100 + args->br = au_sbr(sb, bindex);
36101 + au_lcnt_inc(&args->br->br_count);
36102 + args->wh_dentry = dget(wh_dentry);
36103 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
36104 + if (unlikely(wkq_err)) {
36105 + pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
36106 + au_whtmp_rmdir_free(args);
36109 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
36110 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
36111 +++ linux/fs/aufs/whout.h 2022-03-21 14:49:05.733299677 +0100
36113 +/* SPDX-License-Identifier: GPL-2.0 */
36115 + * Copyright (C) 2005-2021 Junjiro R. Okajima
36117 + * This program, aufs is free software; you can redistribute it and/or modify
36118 + * it under the terms of the GNU General Public License as published by
36119 + * the Free Software Foundation; either version 2 of the License, or
36120 + * (at your option) any later version.
36122 + * This program is distributed in the hope that it will be useful,
36123 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36124 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36125 + * GNU General Public License for more details.
36127 + * You should have received a copy of the GNU General Public License
36128 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36132 + * whiteout for logical deletion and opaque directory
36135 +#ifndef __AUFS_WHOUT_H__
36136 +#define __AUFS_WHOUT_H__
36143 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
36144 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
36145 + struct qstr *wh_name, int try_sio);
36146 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path);
36148 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
36149 + struct qstr *prefix);
36150 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
36151 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
36152 + struct dentry *dentry);
36153 +int au_wh_init(struct au_branch *br, struct super_block *sb);
36155 +/* diropq flags */
36156 +#define AuDiropq_CREATE 1
36157 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
36158 +#define au_fset_diropq(flags, name) \
36159 + do { (flags) |= AuDiropq_##name; } while (0)
36160 +#define au_fclr_diropq(flags, name) \
36161 + do { (flags) &= ~AuDiropq_##name; } while (0)
36163 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
36164 + unsigned int flags);
36165 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
36166 + struct au_branch *br);
36167 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
36168 + struct dentry *h_parent);
36170 +/* real rmdir for the whiteout-ed dir */
36171 +struct au_whtmp_rmdir {
36172 + struct inode *dir;
36173 + struct au_branch *br;
36174 + struct dentry *wh_dentry;
36175 + struct au_nhash whlist;
36178 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
36179 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
36180 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36181 + struct dentry *wh_dentry, struct au_nhash *whlist);
36182 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36183 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
36185 +/* ---------------------------------------------------------------------- */
36187 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
36188 + aufs_bindex_t bindex)
36190 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
36193 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
36195 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
36198 +#endif /* __KERNEL__ */
36199 +#endif /* __AUFS_WHOUT_H__ */
36200 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
36201 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
36202 +++ linux/fs/aufs/wkq.c 2022-03-21 14:49:05.733299677 +0100
36204 +// SPDX-License-Identifier: GPL-2.0
36206 + * Copyright (C) 2005-2021 Junjiro R. Okajima
36208 + * This program, aufs is free software; you can redistribute it and/or modify
36209 + * it under the terms of the GNU General Public License as published by
36210 + * the Free Software Foundation; either version 2 of the License, or
36211 + * (at your option) any later version.
36213 + * This program is distributed in the hope that it will be useful,
36214 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36215 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36216 + * GNU General Public License for more details.
36218 + * You should have received a copy of the GNU General Public License
36219 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36223 + * workqueue for asynchronous/super-io operations
36224 + * todo: try new credential scheme
36227 +#include <linux/module.h>
36230 +/* internal workqueue named AUFS_WKQ_NAME */
36232 +static struct workqueue_struct *au_wkq;
36234 +struct au_wkinfo {
36235 + struct work_struct wk;
36236 + struct kobject *kobj;
36238 + unsigned int flags; /* see wkq.h */
36240 + au_wkq_func_t func;
36243 +#ifdef CONFIG_LOCKDEP
36245 + struct held_lock **hlock;
36248 + struct completion *comp;
36251 +/* ---------------------------------------------------------------------- */
36253 + * Aufs passes some operations to the workqueue such as the internal copyup.
36254 + * This scheme looks rather unnatural for LOCKDEP debugging feature, since the
36255 + * job run by workqueue depends upon the locks acquired in the other task.
36256 + * Delegating a small operation to the workqueue, aufs passes its lockdep
36257 + * information too. And the job in the workqueue restores the info in order to
36258 + * pretend as if it acquired those locks. This is just to make LOCKDEP work
36259 + * correctly and expectedly.
36262 +#ifndef CONFIG_LOCKDEP
36263 +AuStubInt0(au_wkq_lockdep_alloc, struct au_wkinfo *wkinfo);
36264 +AuStubVoid(au_wkq_lockdep_free, struct au_wkinfo *wkinfo);
36265 +AuStubVoid(au_wkq_lockdep_pre, struct au_wkinfo *wkinfo);
36266 +AuStubVoid(au_wkq_lockdep_post, struct au_wkinfo *wkinfo);
36267 +AuStubVoid(au_wkq_lockdep_init, struct au_wkinfo *wkinfo);
36269 +static void au_wkq_lockdep_init(struct au_wkinfo *wkinfo)
36271 + wkinfo->hlock = NULL;
36272 + wkinfo->dont_check = 0;
36279 +static int au_wkq_lockdep_test(struct lock_class_key *key, const char *name)
36281 + static DEFINE_SPINLOCK(spin);
36284 + struct lock_class_key *key;
36286 + { .name = "&sbinfo->si_rwsem" },
36287 + { .name = "&finfo->fi_rwsem" },
36288 + { .name = "&dinfo->di_rwsem" },
36289 + { .name = "&iinfo->ii_rwsem" }
36294 + /* lockless read from 'set.' see below */
36295 + if (set == ARRAY_SIZE(a)) {
36296 + for (i = 0; i < ARRAY_SIZE(a); i++)
36297 + if (a[i].key == key)
36302 + spin_lock(&spin);
36304 + for (i = 0; i < ARRAY_SIZE(a); i++)
36305 + if (a[i].key == key) {
36306 + spin_unlock(&spin);
36309 + for (i = 0; i < ARRAY_SIZE(a); i++) {
36311 + if (unlikely(a[i].key == key)) { /* rare but possible */
36312 + spin_unlock(&spin);
36317 + if (strstr(a[i].name, name)) {
36319 + * the order of these three lines is important for the
36320 + * lockless read above.
36323 + spin_unlock(&spin);
36325 + /* AuDbg("%d, %s\n", set, name); */
36329 + spin_unlock(&spin);
36338 +static int au_wkq_lockdep_alloc(struct au_wkinfo *wkinfo)
36341 + struct task_struct *curr;
36342 + struct held_lock **hl, *held_locks, *p;
36346 + wkinfo->dont_check = lockdep_recursing(curr);
36347 + if (wkinfo->dont_check)
36349 + n = curr->lockdep_depth;
36354 + wkinfo->hlock = kmalloc_array(n + 1, sizeof(*wkinfo->hlock), GFP_NOFS);
36355 + if (unlikely(!wkinfo->hlock))
36359 +#if 0 /* left for debugging */
36360 + if (0 && au_debug_test())
36361 + lockdep_print_held_locks(curr);
36363 + held_locks = curr->held_locks;
36364 + hl = wkinfo->hlock;
36366 + p = held_locks++;
36367 + if (au_wkq_lockdep_test(p->instance->key, p->instance->name))
36376 +static void au_wkq_lockdep_free(struct au_wkinfo *wkinfo)
36378 + au_kfree_try_rcu(wkinfo->hlock);
36381 +static void au_wkq_lockdep_pre(struct au_wkinfo *wkinfo)
36383 + struct held_lock *p, **hl = wkinfo->hlock;
36386 + if (wkinfo->dont_check)
36390 + while ((p = *hl++)) { /* assignment */
36391 + subclass = lockdep_hlock_class(p)->subclass;
36392 + /* AuDbg("%s, %d\n", p->instance->name, subclass); */
36394 + rwsem_acquire_read(p->instance, subclass, 0,
36395 + /*p->acquire_ip*/_RET_IP_);
36397 + rwsem_acquire(p->instance, subclass, 0,
36398 + /*p->acquire_ip*/_RET_IP_);
36402 +static void au_wkq_lockdep_post(struct au_wkinfo *wkinfo)
36404 + struct held_lock *p, **hl = wkinfo->hlock;
36406 + if (wkinfo->dont_check)
36410 + while ((p = *hl++)) /* assignment */
36411 + rwsem_release(p->instance, /*p->acquire_ip*/_RET_IP_);
36415 +static void wkq_func(struct work_struct *wk)
36417 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
36419 + AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
36420 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
36422 + au_wkq_lockdep_pre(wkinfo);
36423 + wkinfo->func(wkinfo->args);
36424 + au_wkq_lockdep_post(wkinfo);
36425 + if (au_ftest_wkq(wkinfo->flags, WAIT))
36426 + complete(wkinfo->comp);
36428 + kobject_put(wkinfo->kobj);
36429 + module_put(THIS_MODULE); /* todo: ?? */
36430 + au_kfree_rcu(wkinfo);
36435 + * Since struct completion is large, try allocating it dynamically.
36437 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
36439 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
36441 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
36443 + init_completion(*comp);
36444 + wkinfo->comp = *comp;
36450 +static void au_wkq_comp_free(struct completion *comp)
36452 + au_kfree_rcu(comp);
36455 +static void au_wkq_run(struct au_wkinfo *wkinfo)
36457 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
36458 + if (au_wkq_test()) {
36459 + AuWarn1("wkq from wkq, unless silly-rename on NFS,"
36460 + " due to a dead dir by UDBA,"
36461 + " or async xino write?\n");
36462 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
36465 + au_dbg_verify_kthread();
36467 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
36468 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
36469 + queue_work(au_wkq, &wkinfo->wk);
36471 + INIT_WORK(&wkinfo->wk, wkq_func);
36472 + schedule_work(&wkinfo->wk);
36477 + * Be careful. It is easy to make deadlock happen.
36478 + * processA: lock, wkq and wait
36479 + * processB: wkq and wait, lock in wkq
36482 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
36485 + AuWkqCompDeclare(comp);
36486 + struct au_wkinfo wkinfo = {
36492 + err = au_wkq_comp_alloc(&wkinfo, &comp);
36493 + if (unlikely(err))
36495 + err = au_wkq_lockdep_alloc(&wkinfo);
36496 + if (unlikely(err))
36499 + au_wkq_run(&wkinfo);
36500 + /* no timeout, no interrupt */
36501 + wait_for_completion(wkinfo.comp);
36503 + au_wkq_lockdep_free(&wkinfo);
36506 + au_wkq_comp_free(comp);
36508 + destroy_work_on_stack(&wkinfo.wk);
36513 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
36514 + * problem in a concurrent umounting.
36516 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36517 + unsigned int flags)
36520 + struct au_wkinfo *wkinfo;
36522 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
36525 + * wkq_func() must free this wkinfo.
36526 + * it highly depends upon the implementation of workqueue.
36529 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
36531 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
36532 + wkinfo->flags = flags & ~AuWkq_WAIT;
36533 + wkinfo->func = func;
36534 + wkinfo->args = args;
36535 + wkinfo->comp = NULL;
36536 + au_wkq_lockdep_init(wkinfo);
36537 + kobject_get(wkinfo->kobj);
36538 + __module_get(THIS_MODULE); /* todo: ?? */
36540 + au_wkq_run(wkinfo);
36543 + au_nwt_done(&au_sbi(sb)->si_nowait);
36549 +/* ---------------------------------------------------------------------- */
36551 +void au_nwt_init(struct au_nowait_tasks *nwt)
36553 + atomic_set(&nwt->nw_len, 0);
36554 + /* smp_mb(); */ /* atomic_set */
36555 + init_waitqueue_head(&nwt->nw_wq);
36558 +void au_wkq_fin(void)
36560 + destroy_workqueue(au_wkq);
36563 +int __init au_wkq_init(void)
36568 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
36569 + if (IS_ERR(au_wkq))
36570 + err = PTR_ERR(au_wkq);
36571 + else if (!au_wkq)
36576 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
36577 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
36578 +++ linux/fs/aufs/wkq.h 2022-03-21 14:49:05.733299677 +0100
36580 +/* SPDX-License-Identifier: GPL-2.0 */
36582 + * Copyright (C) 2005-2021 Junjiro R. Okajima
36584 + * This program, aufs is free software; you can redistribute it and/or modify
36585 + * it under the terms of the GNU General Public License as published by
36586 + * the Free Software Foundation; either version 2 of the License, or
36587 + * (at your option) any later version.
36589 + * This program is distributed in the hope that it will be useful,
36590 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36591 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36592 + * GNU General Public License for more details.
36594 + * You should have received a copy of the GNU General Public License
36595 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36599 + * workqueue for asynchronous/super-io operations
36600 + * todo: try new credentials management scheme
36603 +#ifndef __AUFS_WKQ_H__
36604 +#define __AUFS_WKQ_H__
36608 +#include <linux/wait.h>
36610 +struct super_block;
36612 +/* ---------------------------------------------------------------------- */
36615 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
36617 +struct au_nowait_tasks {
36619 + wait_queue_head_t nw_wq;
36622 +/* ---------------------------------------------------------------------- */
36624 +typedef void (*au_wkq_func_t)(void *args);
36627 +#define AuWkq_WAIT 1
36628 +#define AuWkq_NEST (1 << 1)
36629 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
36630 +#define au_fset_wkq(flags, name) \
36631 + do { (flags) |= AuWkq_##name; } while (0)
36632 +#define au_fclr_wkq(flags, name) \
36633 + do { (flags) &= ~AuWkq_##name; } while (0)
36636 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
36637 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36638 + unsigned int flags);
36639 +void au_nwt_init(struct au_nowait_tasks *nwt);
36640 +int __init au_wkq_init(void);
36641 +void au_wkq_fin(void);
36643 +/* ---------------------------------------------------------------------- */
36645 +static inline int au_wkq_test(void)
36647 + return current->flags & PF_WQ_WORKER;
36650 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
36652 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
36655 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
36657 + if (atomic_dec_and_test(&nwt->nw_len))
36658 + wake_up_all(&nwt->nw_wq);
36661 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
36663 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
36667 +#endif /* __KERNEL__ */
36668 +#endif /* __AUFS_WKQ_H__ */
36669 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
36670 --- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
36671 +++ linux/fs/aufs/xattr.c 2022-10-03 07:51:14.929542593 +0200
36673 +// SPDX-License-Identifier: GPL-2.0
36675 + * Copyright (C) 2014-2021 Junjiro R. Okajima
36677 + * This program, aufs is free software; you can redistribute it and/or modify
36678 + * it under the terms of the GNU General Public License as published by
36679 + * the Free Software Foundation; either version 2 of the License, or
36680 + * (at your option) any later version.
36682 + * This program is distributed in the hope that it will be useful,
36683 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36684 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36685 + * GNU General Public License for more details.
36687 + * You should have received a copy of the GNU General Public License
36688 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36692 + * handling xattr functions
36695 +#include <linux/fs.h>
36696 +#include <linux/posix_acl_xattr.h>
36697 +#include <linux/xattr.h>
36700 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
36702 + if (!ignore_flags)
36710 + if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
36715 +#define cmp(brattr, prefix) do { \
36716 + if (!strncmp(name, XATTR_##prefix##_PREFIX, \
36717 + XATTR_##prefix##_PREFIX_LEN)) { \
36718 + if (ignore_flags & AuBrAttr_ICEX_##brattr) \
36724 + cmp(SEC, SECURITY);
36725 + cmp(SYS, SYSTEM);
36726 + cmp(TR, TRUSTED);
36730 + if (ignore_flags & AuBrAttr_ICEX_OTH)
36737 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
36739 +static int au_do_cpup_xattr(struct path *h_dst, struct path *h_src,
36740 + char *name, char **buf, unsigned int ignore_flags,
36741 + unsigned int verbose)
36745 + struct inode *h_idst;
36746 + struct dentry *h_dst_dentry, *h_src_dentry;
36747 + struct user_namespace *h_dst_userns, *h_src_userns;
36749 + h_src_userns = mnt_user_ns(h_src->mnt);
36750 + h_src_dentry = h_src->dentry;
36751 + ssz = vfs_getxattr_alloc(h_src_userns, h_src_dentry, name, buf, 0,
36754 + if (unlikely(err <= 0)) {
36755 + if (err == -ENODATA
36756 + || (err == -EOPNOTSUPP
36757 + && ((ignore_flags & au_xattr_out_of_list)
36758 + || (au_test_nfs_noacl(d_inode(h_src_dentry))
36759 + && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
36761 + XATTR_NAME_POSIX_ACL_DEFAULT))))
36764 + if (err && (verbose || au_debug_test()))
36765 + pr_err("%s, err %d\n", name, err);
36769 + /* unlock it temporary */
36770 + h_dst_userns = mnt_user_ns(h_dst->mnt);
36771 + h_dst_dentry = h_dst->dentry;
36772 + h_idst = d_inode(h_dst_dentry);
36773 + inode_unlock(h_idst);
36774 + err = vfsub_setxattr(h_dst_userns, h_dst_dentry, name, *buf, ssz,
36776 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36777 + if (unlikely(err)) {
36778 + if (verbose || au_debug_test())
36779 + pr_err("%s, err %d\n", name, err);
36780 + err = au_xattr_ignore(err, name, ignore_flags);
36787 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
36788 + unsigned int verbose)
36790 + int err, unlocked, acl_access, acl_default;
36792 + struct dentry *h_dst_dentry, *h_src_dentry;
36793 + struct inode *h_isrc, *h_idst;
36794 + char *value, *p, *o, *e;
36796 + /* try stopping to update the source inode while we are referencing */
36797 + /* there should not be the parent-child relationship between them */
36798 + h_dst_dentry = h_dst->dentry;
36799 + h_idst = d_inode(h_dst_dentry);
36800 + h_src_dentry = h_src->dentry;
36801 + h_isrc = d_inode(h_src_dentry);
36802 + inode_unlock(h_idst);
36803 + inode_lock_shared_nested(h_isrc, AuLsc_I_CHILD);
36804 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36807 + /* some filesystems don't list POSIX ACL, for example tmpfs */
36808 + ssz = vfs_listxattr(h_src_dentry, NULL, 0);
36810 + if (unlikely(err < 0)) {
36812 + if (err == -ENODATA
36813 + || err == -EOPNOTSUPP)
36814 + err = 0; /* ignore */
36823 + p = kmalloc(ssz, GFP_NOFS);
36825 + if (unlikely(!p))
36827 + err = vfs_listxattr(h_src_dentry, p, ssz);
36829 + inode_unlock_shared(h_isrc);
36831 + AuDbg("err %d, ssz %zd\n", err, ssz);
36832 + if (unlikely(err < 0))
36840 + while (!err && p < e) {
36841 + acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
36842 + sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
36843 + acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
36844 + sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
36846 + err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
36848 + p += strlen(p) + 1;
36851 + ignore_flags |= au_xattr_out_of_list;
36852 + if (!err && !acl_access) {
36853 + err = au_do_cpup_xattr(h_dst, h_src,
36854 + XATTR_NAME_POSIX_ACL_ACCESS, &value,
36855 + ignore_flags, verbose);
36858 + if (!err && !acl_default) {
36859 + err = au_do_cpup_xattr(h_dst, h_src,
36860 + XATTR_NAME_POSIX_ACL_DEFAULT, &value,
36861 + ignore_flags, verbose);
36865 + au_kfree_try_rcu(value);
36868 + au_kfree_try_rcu(o);
36871 + inode_unlock_shared(h_isrc);
36876 +/* ---------------------------------------------------------------------- */
36878 +static int au_smack_reentering(struct super_block *sb)
36880 +#if IS_ENABLED(CONFIG_SECURITY_SMACK) || IS_ENABLED(CONFIG_SECURITY_SELINUX)
36882 + * as a part of lookup, smack_d_instantiate() is called, and it calls
36883 + * i_op->getxattr(). ouch.
36885 + return si_pid_test(sb);
36896 +struct au_lgxattr {
36904 + const char *name;
36911 +static ssize_t au_lgxattr(struct dentry *dentry, struct inode *inode,
36912 + struct au_lgxattr *arg)
36916 + struct path h_path;
36917 + struct super_block *sb;
36919 + sb = dentry->d_sb;
36920 + reenter = au_smack_reentering(sb);
36922 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
36923 + if (unlikely(err))
36926 + err = au_h_path_getattr(dentry, inode, /*force*/1, &h_path, reenter);
36927 + if (unlikely(err))
36929 + if (unlikely(!h_path.dentry))
36930 + /* illegally overlapped or something */
36931 + goto out_di; /* pretending success */
36933 + /* always topmost entry only */
36934 + switch (arg->type) {
36935 + case AU_XATTR_LIST:
36936 + err = vfs_listxattr(h_path.dentry,
36937 + arg->u.list.list, arg->u.list.size);
36939 + case AU_XATTR_GET:
36940 + AuDebugOn(d_is_negative(h_path.dentry));
36941 + err = vfs_getxattr(mnt_user_ns(h_path.mnt), h_path.dentry,
36942 + arg->u.get.name, arg->u.get.value,
36943 + arg->u.get.size);
36949 + di_read_unlock(dentry, AuLock_IR);
36952 + si_read_unlock(sb);
36958 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
36960 + struct au_lgxattr arg = {
36961 + .type = AU_XATTR_LIST,
36968 + return au_lgxattr(dentry, /*inode*/NULL, &arg);
36971 +static ssize_t au_getxattr(struct dentry *dentry, struct inode *inode,
36972 + const char *name, void *value, size_t size)
36974 + struct au_lgxattr arg = {
36975 + .type = AU_XATTR_GET,
36983 + return au_lgxattr(dentry, inode, &arg);
36986 +static int au_setxattr(struct dentry *dentry, struct inode *inode,
36987 + const char *name, void *value, size_t size,
36990 + struct au_sxattr arg = {
36991 + .type = AU_XATTR_SET,
37000 + return au_sxattr(dentry, inode, &arg);
37003 +/* ---------------------------------------------------------------------- */
37005 +static int au_xattr_get(const struct xattr_handler *handler,
37006 + struct dentry *dentry, struct inode *inode,
37007 + const char *name, void *buffer, size_t size)
37009 + return au_getxattr(dentry, inode, name, buffer, size);
37013 + * The prameter 'void *value' of vfs_setxattr() is NOT const, but
37014 + * xattr_handler->set() requires const. Is it OK? I don't know.
37016 +static int au_xattr_set(const struct xattr_handler *handler,
37017 + struct user_namespace *userns,
37018 + struct dentry *dentry, struct inode *inode,
37019 + const char *name, const void *value, size_t size,
37022 + return au_setxattr(dentry, inode, name, (void *)value, size, flags);
37025 +static const struct xattr_handler au_xattr_handler = {
37028 + .get = au_xattr_get,
37029 + .set = au_xattr_set
37032 +static const struct xattr_handler *au_xattr_handlers[] = {
37033 +#ifdef CONFIG_FS_POSIX_ACL
37034 + &posix_acl_access_xattr_handler,
37035 + &posix_acl_default_xattr_handler,
37037 + &au_xattr_handler, /* must be last */
37041 +void au_xattr_init(struct super_block *sb)
37043 + sb->s_xattr = au_xattr_handlers;
37045 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
37046 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
37047 +++ linux/fs/aufs/xino.c 2022-03-21 14:49:05.733299677 +0100
37049 +// SPDX-License-Identifier: GPL-2.0
37051 + * Copyright (C) 2005-2021 Junjiro R. Okajima
37053 + * This program, aufs is free software; you can redistribute it and/or modify
37054 + * it under the terms of the GNU General Public License as published by
37055 + * the Free Software Foundation; either version 2 of the License, or
37056 + * (at your option) any later version.
37058 + * This program is distributed in the hope that it will be useful,
37059 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
37060 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
37061 + * GNU General Public License for more details.
37063 + * You should have received a copy of the GNU General Public License
37064 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
37068 + * external inode number translation table and bitmap
37070 + * things to consider
37072 + * + au_xino object
37073 + * + XINO files (xino, xib, xigen)
37074 + * + dynamic debugfs entries (xiN)
37075 + * + static debugfs entries (xib, xigen)
37076 + * + static sysfs entry (xi_path)
37077 + * - several entry points to handle them.
37078 + * + mount(2) without xino option (default)
37079 + * + mount(2) with xino option
37080 + * + mount(2) with noxino option
37082 + * + remount with add/del branches
37083 + * + remount with xino/noxino options
37086 +#include <linux/seq_file.h>
37087 +#include <linux/statfs.h>
37090 +static aufs_bindex_t sbr_find_shared(struct super_block *sb, aufs_bindex_t btop,
37091 + aufs_bindex_t bbot,
37092 + struct super_block *h_sb)
37094 + /* todo: try binary-search if the branches are many */
37095 + for (; btop <= bbot; btop++)
37096 + if (h_sb == au_sbr_sb(sb, btop))
37102 + * find another branch who is on the same filesystem of the specified
37103 + * branch{@btgt}. search until @bbot.
37105 +static aufs_bindex_t is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
37106 + aufs_bindex_t bbot)
37108 + aufs_bindex_t bindex;
37109 + struct super_block *tgt_sb;
37111 + tgt_sb = au_sbr_sb(sb, btgt);
37112 + bindex = sbr_find_shared(sb, /*btop*/0, btgt - 1, tgt_sb);
37114 + bindex = sbr_find_shared(sb, btgt + 1, bbot, tgt_sb);
37119 +/* ---------------------------------------------------------------------- */
37122 + * stop unnecessary notify events at creating xino files
37125 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry)
37127 + aufs_bindex_t bfound, bindex, bbot;
37128 + struct dentry *parent;
37129 + struct au_branch *br;
37132 + parent = dentry->d_parent; /* safe d_parent access */
37133 + bbot = au_sbbot(sb);
37134 + for (bindex = 0; bindex <= bbot; bindex++) {
37135 + br = au_sbr(sb, bindex);
37136 + if (au_br_dentry(br) == parent) {
37142 + AuDbg("bfound b%d\n", bfound);
37146 +struct au_xino_lock_dir {
37147 + struct au_hinode *hdir;
37148 + struct dentry *parent;
37149 + struct inode *dir;
37152 +static struct dentry *au_dget_parent_lock(struct dentry *dentry,
37153 + unsigned int lsc)
37155 + struct dentry *parent;
37156 + struct inode *dir;
37158 + parent = dget_parent(dentry);
37159 + dir = d_inode(parent);
37160 + inode_lock_nested(dir, lsc);
37161 +#if 0 /* it should not happen */
37162 + spin_lock(&dentry->d_lock);
37163 + if (unlikely(dentry->d_parent != parent)) {
37164 + spin_unlock(&dentry->d_lock);
37165 + inode_unlock(dir);
37170 + spin_unlock(&dentry->d_lock);
37177 +static void au_xino_lock_dir(struct super_block *sb, struct path *xipath,
37178 + struct au_xino_lock_dir *ldir)
37180 + aufs_bindex_t bindex;
37182 + ldir->hdir = NULL;
37183 + bindex = au_xi_root(sb, xipath->dentry);
37184 + if (bindex >= 0) {
37185 + /* rw branch root */
37186 + ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
37187 + au_hn_inode_lock_nested(ldir->hdir, AuLsc_I_PARENT);
37190 + ldir->parent = au_dget_parent_lock(xipath->dentry,
37192 + ldir->dir = d_inode(ldir->parent);
37196 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
37199 + au_hn_inode_unlock(ldir->hdir);
37201 + inode_unlock(ldir->dir);
37202 + dput(ldir->parent);
37206 +/* ---------------------------------------------------------------------- */
37209 + * create and set a new xino file
37211 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
37214 + struct file *file;
37215 + struct dentry *h_parent, *d;
37216 + struct inode *h_dir, *inode;
37218 + static DEFINE_MUTEX(mtx);
37221 + * at mount-time, and the xino file is the default path,
37222 + * hnotify is disabled so we have no notify events to ignore.
37223 + * when a user specified the xino, we cannot get au_hdir to be ignored.
37226 + mutex_lock(&mtx);
37227 + file = vfsub_filp_open(fpath, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37228 + /* | __FMODE_NONOTIFY */,
37230 + if (IS_ERR(file)) {
37232 + mutex_unlock(&mtx);
37234 + pr_err("open %s(%ld)\n", fpath, PTR_ERR(file));
37238 + /* keep file count */
37240 + d = file->f_path.dentry;
37241 + h_parent = au_dget_parent_lock(d, AuLsc_I_PARENT);
37243 + mutex_unlock(&mtx);
37244 + /* mnt_want_write() is unnecessary here */
37245 + h_dir = d_inode(h_parent);
37246 + inode = file_inode(file);
37247 + /* no delegation since it is just created */
37248 + if (inode->i_nlink)
37249 + err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
37251 + inode_unlock(h_dir);
37253 + if (unlikely(err)) {
37255 + pr_err("unlink %s(%d)\n", fpath, err);
37260 + if (unlikely(sb && sb == d->d_sb)) {
37262 + pr_err("%s must be outside\n", fpath);
37265 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
37267 + pr_err("xino doesn't support %s(%s)\n",
37268 + fpath, au_sbtype(d->d_sb));
37271 + return file; /* success */
37275 + file = ERR_PTR(err);
37280 + * create a new xinofile at the same place/path as @base.
37282 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
37283 + struct file *copy_src)
37285 + struct file *file;
37286 + struct dentry *dentry;
37287 + struct inode *dir, *delegated;
37288 + struct qstr *name;
37289 + struct path ppath, path;
37290 + int err, do_unlock;
37291 + struct au_xino_lock_dir ldir;
37294 + au_xino_lock_dir(sb, base, &ldir);
37295 + dentry = base->dentry;
37296 + ppath.dentry = dentry->d_parent; /* dir inode is locked */
37297 + ppath.mnt = base->mnt;
37298 + dir = d_inode(ppath.dentry);
37301 + name = &dentry->d_name;
37302 + path.dentry = vfsub_lookup_one_len(name->name, &ppath, name->len);
37303 + if (IS_ERR(path.dentry)) {
37304 + file = (void *)path.dentry;
37305 + pr_err("%pd lookup err %ld\n", dentry, PTR_ERR(path.dentry));
37309 + /* no need to mnt_want_write() since we call dentry_open() later */
37310 + err = vfs_create(mnt_user_ns(base->mnt), dir, path.dentry, 0666, NULL);
37311 + if (unlikely(err)) {
37312 + file = ERR_PTR(err);
37313 + pr_err("%pd create err %d\n", dentry, err);
37317 + path.mnt = base->mnt;
37318 + file = vfsub_dentry_open(&path,
37319 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37320 + /* | __FMODE_NONOTIFY */);
37321 + if (IS_ERR(file)) {
37322 + pr_err("%pd open err %ld\n", dentry, PTR_ERR(file));
37326 + delegated = NULL;
37327 + err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
37328 + au_xino_unlock_dir(&ldir);
37330 + if (unlikely(err == -EWOULDBLOCK)) {
37331 + pr_warn("cannot retry for NFSv4 delegation"
37332 + " for an internal unlink\n");
37335 + if (unlikely(err)) {
37336 + pr_err("%pd unlink err %d\n", dentry, err);
37341 + /* no one can touch copy_src xino */
37342 + err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
37343 + if (unlikely(err)) {
37344 + pr_err("%pd copy err %d\n", dentry, err);
37348 + goto out_dput; /* success */
37352 + file = ERR_PTR(err);
37354 + dput(path.dentry);
37357 + au_xino_unlock_dir(&ldir);
37361 +struct file *au_xino_file1(struct au_xino *xi)
37363 + struct file *file;
37364 + unsigned int u, nfile;
37367 + nfile = xi->xi_nfile;
37368 + for (u = 0; u < nfile; u++) {
37369 + file = xi->xi_file[u];
37377 +static int au_xino_file_set(struct au_xino *xi, int idx, struct file *file)
37388 + if (idx < xi->xi_nfile) {
37389 + f = xi->xi_file[idx];
37393 + p = au_kzrealloc(xi->xi_file,
37394 + sizeof(*xi->xi_file) * xi->xi_nfile,
37395 + sizeof(*xi->xi_file) * (idx + 1),
37396 + GFP_NOFS, /*may_shrink*/0);
37398 + MtxMustLock(&xi->xi_mtx);
37400 + xi->xi_nfile = idx + 1;
37408 + xi->xi_file[idx] = file;
37415 + * if @xinew->xi is not set, then create new xigen file.
37417 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew)
37419 + struct file *file;
37422 + SiMustAnyLock(sb);
37424 + file = au_xino_create2(sb, xinew->base, xinew->copy_src);
37425 + if (IS_ERR(file)) {
37426 + err = PTR_ERR(file);
37427 + pr_err("%s[%d], err %d\n",
37428 + xinew->xi ? "xino" : "xigen",
37429 + xinew->idx, err);
37434 + err = au_xino_file_set(xinew->xi, xinew->idx, file);
37437 + /* todo: make xigen file an array */
37438 + /* err = au_xigen_file_set(sb, xinew->idx, file); */
37441 + if (unlikely(err))
37442 + file = ERR_PTR(err);
37448 +/* ---------------------------------------------------------------------- */
37451 + * truncate xino files
37453 +static int au_xino_do_trunc(struct super_block *sb, aufs_bindex_t bindex,
37454 + int idx, struct kstatfs *st)
37458 + struct file *file, *new_xino;
37459 + struct au_xi_new xinew = {
37464 + xinew.xi = au_sbr(sb, bindex)->br_xino;
37465 + file = au_xino_file(xinew.xi, idx);
37469 + xinew.base = &file->f_path;
37470 + err = vfs_statfs(xinew.base, st);
37471 + if (unlikely(err)) {
37472 + AuErr1("statfs err %d, ignored\n", err);
37477 + blocks = file_inode(file)->i_blocks;
37478 + pr_info("begin truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37479 + bindex, idx, (u64)blocks, st->f_bfree, st->f_blocks);
37481 + xinew.copy_src = file;
37482 + new_xino = au_xi_new(sb, &xinew);
37483 + if (IS_ERR(new_xino)) {
37484 + err = PTR_ERR(new_xino);
37485 + pr_err("xino(b%d-%d), err %d, ignored\n", bindex, idx, err);
37489 + err = vfs_statfs(&new_xino->f_path, st);
37491 + pr_info("end truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37492 + bindex, idx, (u64)file_inode(new_xino)->i_blocks,
37493 + st->f_bfree, st->f_blocks);
37495 + AuErr1("statfs err %d, ignored\n", err);
37503 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin)
37506 + unsigned long jiffy;
37507 + aufs_bindex_t bbot;
37508 + struct kstatfs *st;
37509 + struct au_branch *br;
37510 + struct au_xino *xi;
37513 + st = kmalloc(sizeof(*st), GFP_NOFS);
37514 + if (unlikely(!st))
37518 + bbot = au_sbbot(sb);
37519 + if (unlikely(bindex < 0 || bbot < bindex))
37524 + br = au_sbr(sb, bindex);
37525 + xi = br->br_xino;
37526 + for (i = idx_begin; !err && i < xi->xi_nfile; i++)
37527 + err = au_xino_do_trunc(sb, bindex, i, st);
37529 + au_sbi(sb)->si_xino_jiffy = jiffy;
37532 + au_kfree_rcu(st);
37537 +struct xino_do_trunc_args {
37538 + struct super_block *sb;
37539 + struct au_branch *br;
37543 +static void xino_do_trunc(void *_args)
37545 + struct xino_do_trunc_args *args = _args;
37546 + struct super_block *sb;
37547 + struct au_branch *br;
37548 + struct inode *dir;
37550 + aufs_bindex_t bindex;
37554 + dir = d_inode(sb->s_root);
37558 + si_noflush_write_lock(sb);
37559 + ii_read_lock_parent(dir);
37560 + bindex = au_br_index(sb, br->br_id);
37561 + err = au_xino_trunc(sb, bindex, idx);
37562 + ii_read_unlock(dir);
37563 + if (unlikely(err))
37564 + pr_warn("err b%d, (%d)\n", bindex, err);
37565 + atomic_dec(&br->br_xino->xi_truncating);
37566 + au_lcnt_dec(&br->br_count);
37567 + si_write_unlock(sb);
37568 + au_nwt_done(&au_sbi(sb)->si_nowait);
37569 + au_kfree_rcu(args);
37573 + * returns the index in the xi_file array whose corresponding file is necessary
37574 + * to truncate, or -1 which means no need to truncate.
37576 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
37580 + struct kstatfs st;
37581 + struct au_sbinfo *sbinfo;
37582 + struct au_xino *xi;
37583 + struct file *file;
37585 + /* todo: si_xino_expire and the ratio should be customizable */
37586 + sbinfo = au_sbi(sb);
37587 + if (time_before(jiffies,
37588 + sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
37591 + /* truncation border */
37592 + xi = br->br_xino;
37593 + for (u = 0; u < xi->xi_nfile; u++) {
37594 + file = au_xino_file(xi, u);
37598 + err = vfs_statfs(&file->f_path, &st);
37599 + if (unlikely(err)) {
37600 + AuErr1("statfs err %d, ignored\n", err);
37603 + if (div64_u64(st.f_bfree * 100, st.f_blocks)
37604 + >= AUFS_XINO_DEF_TRUNC)
37611 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
37614 + struct xino_do_trunc_args *args;
37617 + idx = xino_trunc_test(sb, br);
37621 + if (atomic_inc_return(&br->br_xino->xi_truncating) > 1)
37624 + /* lock and kfree() will be called in trunc_xino() */
37625 + args = kmalloc(sizeof(*args), GFP_NOFS);
37626 + if (unlikely(!args)) {
37627 + AuErr1("no memory\n");
37631 + au_lcnt_inc(&br->br_count);
37635 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
37637 + return; /* success */
37639 + pr_err("wkq %d\n", wkq_err);
37640 + au_lcnt_dec(&br->br_count);
37641 + au_kfree_rcu(args);
37644 + atomic_dec(&br->br_xino->xi_truncating);
37647 +/* ---------------------------------------------------------------------- */
37649 +struct au_xi_calc {
37654 +static void au_xi_calc(struct super_block *sb, ino_t h_ino,
37655 + struct au_xi_calc *calc)
37659 + maxent = au_xi_maxent(sb);
37660 + calc->idx = div64_u64_rem(h_ino, maxent, &calc->pos);
37661 + calc->pos *= sizeof(ino_t);
37664 +static int au_xino_do_new_async(struct super_block *sb, struct au_branch *br,
37665 + struct au_xi_calc *calc)
37668 + struct file *file;
37669 + struct au_xino *xi = br->br_xino;
37670 + struct au_xi_new xinew = {
37674 + SiMustAnyLock(sb);
37680 + mutex_lock(&xi->xi_mtx);
37681 + file = au_xino_file(xi, calc->idx);
37685 + file = au_xino_file(xi, /*idx*/-1);
37686 + AuDebugOn(!file);
37687 + xinew.idx = calc->idx;
37688 + xinew.base = &file->f_path;
37689 + /* xinew.copy_src = NULL; */
37690 + file = au_xi_new(sb, &xinew);
37691 + if (IS_ERR(file))
37692 + err = PTR_ERR(file);
37695 + mutex_unlock(&xi->xi_mtx);
37700 +struct au_xino_do_new_async_args {
37701 + struct super_block *sb;
37702 + struct au_branch *br;
37703 + struct au_xi_calc calc;
37707 +struct au_xi_writing {
37708 + struct hlist_bl_node node;
37709 + ino_t h_ino, ino;
37712 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37715 +static void au_xino_call_do_new_async(void *args)
37717 + struct au_xino_do_new_async_args *a = args;
37718 + struct au_branch *br;
37719 + struct super_block *sb;
37720 + struct au_sbinfo *sbi;
37721 + struct inode *root;
37722 + struct file *file;
37723 + struct au_xi_writing *del, *p;
37724 + struct hlist_bl_head *hbl;
37725 + struct hlist_bl_node *pos;
37730 + sbi = au_sbi(sb);
37731 + si_noflush_read_lock(sb);
37732 + root = d_inode(sb->s_root);
37733 + ii_read_lock_child(root);
37734 + err = au_xino_do_new_async(sb, br, &a->calc);
37735 + if (unlikely(err)) {
37736 + AuIOErr("err %d\n", err);
37740 + file = au_xino_file(br->br_xino, a->calc.idx);
37741 + AuDebugOn(!file);
37742 + err = au_xino_do_write(file, &a->calc, a->ino);
37743 + if (unlikely(err)) {
37744 + AuIOErr("err %d\n", err);
37749 + hbl = &br->br_xino->xi_writing;
37750 + hlist_bl_lock(hbl);
37751 + au_hbl_for_each(pos, hbl) {
37752 + p = container_of(pos, struct au_xi_writing, node);
37753 + if (p->ino == a->ino) {
37755 + hlist_bl_del(&p->node);
37759 + hlist_bl_unlock(hbl);
37760 + au_kfree_rcu(del);
37763 + au_lcnt_dec(&br->br_count);
37764 + ii_read_unlock(root);
37765 + si_read_unlock(sb);
37766 + au_nwt_done(&sbi->si_nowait);
37771 + * create a new xino file asynchronously
37773 +static int au_xino_new_async(struct super_block *sb, struct au_branch *br,
37774 + struct au_xi_calc *calc, ino_t ino)
37777 + struct au_xino_do_new_async_args *arg;
37780 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
37781 + if (unlikely(!arg))
37786 + arg->calc = *calc;
37788 + au_lcnt_inc(&br->br_count);
37789 + err = au_wkq_nowait(au_xino_call_do_new_async, arg, sb, AuWkq_NEST);
37790 + if (unlikely(err)) {
37791 + pr_err("wkq %d\n", err);
37792 + au_lcnt_dec(&br->br_count);
37793 + au_kfree_rcu(arg);
37801 + * read @ino from xinofile for the specified branch{@sb, @bindex}
37802 + * at the position of @h_ino.
37804 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37809 + struct au_xi_calc calc;
37810 + struct au_sbinfo *sbinfo;
37811 + struct file *file;
37812 + struct au_xino *xi;
37813 + struct hlist_bl_head *hbl;
37814 + struct hlist_bl_node *pos;
37815 + struct au_xi_writing *p;
37818 + if (!au_opt_test(au_mntflags(sb), XINO))
37819 + return 0; /* no xino */
37822 + au_xi_calc(sb, h_ino, &calc);
37823 + xi = au_sbr(sb, bindex)->br_xino;
37824 + file = au_xino_file(xi, calc.idx);
37826 + hbl = &xi->xi_writing;
37827 + hlist_bl_lock(hbl);
37828 + au_hbl_for_each(pos, hbl) {
37829 + p = container_of(pos, struct au_xi_writing, node);
37830 + if (p->h_ino == h_ino) {
37831 + AuDbg("hi%llu, i%llu, found\n",
37832 + (u64)p->h_ino, (u64)p->ino);
37837 + hlist_bl_unlock(hbl);
37839 + } else if (vfsub_f_size_read(file) < calc.pos + sizeof(*ino))
37840 + return 0; /* no xino */
37842 + sbinfo = au_sbi(sb);
37843 + sz = xino_fread(file, ino, sizeof(*ino), &calc.pos);
37844 + if (sz == sizeof(*ino))
37845 + return 0; /* success */
37848 + if (unlikely(sz >= 0)) {
37850 + AuIOErr("xino read error (%zd)\n", sz);
37855 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37860 + sz = xino_fwrite(file, &ino, sizeof(ino), &calc->pos);
37861 + if (sz == sizeof(ino))
37862 + return 0; /* success */
37864 + AuIOErr("write failed (%zd)\n", sz);
37869 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
37870 + * at the position of @h_ino.
37871 + * even if @ino is zero, it is written to the xinofile and means no entry.
37872 + * if the size of the xino file on a specific filesystem exceeds the watermark,
37873 + * try truncating it.
37875 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37879 + unsigned int mnt_flags;
37880 + struct au_xi_calc calc;
37881 + struct file *file;
37882 + struct au_branch *br;
37883 + struct au_xino *xi;
37884 + struct au_xi_writing *p;
37886 + SiMustAnyLock(sb);
37888 + mnt_flags = au_mntflags(sb);
37889 + if (!au_opt_test(mnt_flags, XINO))
37892 + au_xi_calc(sb, h_ino, &calc);
37893 + br = au_sbr(sb, bindex);
37894 + xi = br->br_xino;
37895 + file = au_xino_file(xi, calc.idx);
37897 + /* store the inum pair into the list */
37898 + p = kmalloc(sizeof(*p), GFP_NOFS | __GFP_NOFAIL);
37899 + p->h_ino = h_ino;
37901 + au_hbl_add(&p->node, &xi->xi_writing);
37903 + /* create and write a new xino file asynchronously */
37904 + err = au_xino_new_async(sb, br, &calc, ino);
37906 + return 0; /* success */
37910 + err = au_xino_do_write(file, &calc, ino);
37912 + br = au_sbr(sb, bindex);
37913 + if (au_opt_test(mnt_flags, TRUNC_XINO)
37914 + && au_test_fs_trunc_xino(au_br_sb(br)))
37915 + xino_try_trunc(sb, br);
37916 + return 0; /* success */
37920 + AuIOErr("write failed (%d)\n", err);
37924 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37927 +/* todo: unnecessary to support mmap_sem since kernel-space? */
37928 +ssize_t xino_fread(struct file *file, void *kbuf, size_t size, loff_t *pos)
37932 + const int prevent_endless = 10;
37936 + err = vfsub_read_k(file, kbuf, size, pos);
37937 + if (err == -EINTR
37938 + && !au_wkq_test()
37939 + && fatal_signal_pending(current)) {
37940 + err = xino_fread_wkq(file, kbuf, size, pos);
37941 + BUG_ON(err == -EINTR);
37943 + } while (i++ < prevent_endless
37944 + && (err == -EAGAIN || err == -EINTR));
37946 +#if 0 /* reserved for future use */
37948 + fsnotify_access(file->f_path.dentry);
37954 +struct xino_fread_args {
37956 + struct file *file;
37962 +static void call_xino_fread(void *args)
37964 + struct xino_fread_args *a = args;
37965 + *a->errp = xino_fread(a->file, a->buf, a->size, a->pos);
37968 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37973 + struct xino_fread_args args = {
37981 + wkq_err = au_wkq_wait(call_xino_fread, &args);
37982 + if (unlikely(wkq_err))
37988 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
37991 +static ssize_t do_xino_fwrite(struct file *file, void *kbuf, size_t size,
37996 + const int prevent_endless = 10;
38000 + err = vfsub_write_k(file, kbuf, size, pos);
38001 + if (err == -EINTR
38002 + && !au_wkq_test()
38003 + && fatal_signal_pending(current)) {
38004 + err = xino_fwrite_wkq(file, kbuf, size, pos);
38005 + BUG_ON(err == -EINTR);
38007 + } while (i++ < prevent_endless
38008 + && (err == -EAGAIN || err == -EINTR));
38010 +#if 0 /* reserved for future use */
38012 + fsnotify_modify(file->f_path.dentry);
38018 +struct do_xino_fwrite_args {
38020 + struct file *file;
38026 +static void call_do_xino_fwrite(void *args)
38028 + struct do_xino_fwrite_args *a = args;
38029 + *a->errp = do_xino_fwrite(a->file, a->buf, a->size, a->pos);
38032 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38037 + struct do_xino_fwrite_args args = {
38046 + * it breaks RLIMIT_FSIZE and normal user's limit,
38047 + * users should care about quota and real 'filesystem full.'
38049 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
38050 + if (unlikely(wkq_err))
38056 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos)
38060 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
38062 + err = do_xino_fwrite(file, buf, size, pos);
38066 + err = xino_fwrite_wkq(file, buf, size, pos);
38073 +/* ---------------------------------------------------------------------- */
38076 + * inode number bitmap
38078 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
38079 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
38083 + AuDebugOn(bit < 0 || page_bits <= bit);
38084 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
38088 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
38090 + AuDebugOn(ino < AUFS_FIRST_INO);
38091 + ino -= AUFS_FIRST_INO;
38092 + *pindex = ino / page_bits;
38093 + *bit = ino % page_bits;
38096 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
38101 + struct au_sbinfo *sbinfo;
38102 + struct file *xib;
38103 + unsigned long *p;
38105 + sbinfo = au_sbi(sb);
38106 + MtxMustLock(&sbinfo->si_xib_mtx);
38107 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
38108 + || !au_opt_test(sbinfo->si_mntflags, XINO));
38110 + if (pindex == sbinfo->si_xib_last_pindex)
38113 + xib = sbinfo->si_xib;
38114 + p = sbinfo->si_xib_buf;
38115 + pos = sbinfo->si_xib_last_pindex;
38116 + pos *= PAGE_SIZE;
38117 + sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38118 + if (unlikely(sz != PAGE_SIZE))
38122 + pos *= PAGE_SIZE;
38123 + if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
38124 + sz = xino_fread(xib, p, PAGE_SIZE, &pos);
38126 + memset(p, 0, PAGE_SIZE);
38127 + sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38129 + if (sz == PAGE_SIZE) {
38130 + sbinfo->si_xib_last_pindex = pindex;
38131 + return 0; /* success */
38135 + AuIOErr1("write failed (%zd)\n", sz);
38142 +static void au_xib_clear_bit(struct inode *inode)
38145 + unsigned long pindex;
38146 + struct super_block *sb;
38147 + struct au_sbinfo *sbinfo;
38149 + AuDebugOn(inode->i_nlink);
38151 + sb = inode->i_sb;
38152 + xib_calc_bit(inode->i_ino, &pindex, &bit);
38153 + AuDebugOn(page_bits <= bit);
38154 + sbinfo = au_sbi(sb);
38155 + mutex_lock(&sbinfo->si_xib_mtx);
38156 + err = xib_pindex(sb, pindex);
38158 + clear_bit(bit, sbinfo->si_xib_buf);
38159 + sbinfo->si_xib_next_bit = bit;
38161 + mutex_unlock(&sbinfo->si_xib_mtx);
38164 +/* ---------------------------------------------------------------------- */
38167 + * truncate a xino bitmap file
38171 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
38175 + unsigned long pindex;
38176 + loff_t pos, pend;
38177 + struct au_sbinfo *sbinfo;
38179 + unsigned long *p;
38182 + sbinfo = au_sbi(sb);
38183 + MtxMustLock(&sbinfo->si_xib_mtx);
38184 + p = sbinfo->si_xib_buf;
38185 + pend = vfsub_f_size_read(file);
38187 + while (pos < pend) {
38188 + sz = xino_fread(file, page, PAGE_SIZE, &pos);
38190 + if (unlikely(sz <= 0))
38194 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
38195 + if (unlikely(*ino < AUFS_FIRST_INO))
38198 + xib_calc_bit(*ino, &pindex, &bit);
38199 + AuDebugOn(page_bits <= bit);
38200 + err = xib_pindex(sb, pindex);
38212 +static int xib_restore(struct super_block *sb)
38215 + unsigned int nfile;
38216 + aufs_bindex_t bindex, bbot;
38218 + struct au_branch *br;
38219 + struct au_xino *xi;
38220 + struct file *file;
38223 + page = (void *)__get_free_page(GFP_NOFS);
38224 + if (unlikely(!page))
38228 + bbot = au_sbbot(sb);
38229 + for (bindex = 0; !err && bindex <= bbot; bindex++)
38230 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0) {
38231 + br = au_sbr(sb, bindex);
38232 + xi = br->br_xino;
38233 + nfile = xi->xi_nfile;
38234 + for (i = 0; i < nfile; i++) {
38235 + file = au_xino_file(xi, i);
38237 + err = do_xib_restore(sb, file, page);
38240 + AuDbg("skip shared b%d\n", bindex);
38241 + free_page((unsigned long)page);
38247 +int au_xib_trunc(struct super_block *sb)
38252 + struct au_sbinfo *sbinfo;
38253 + unsigned long *p;
38254 + struct file *file;
38256 + SiMustWriteLock(sb);
38259 + sbinfo = au_sbi(sb);
38260 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
38263 + file = sbinfo->si_xib;
38264 + if (vfsub_f_size_read(file) <= PAGE_SIZE)
38267 + file = au_xino_create2(sb, &sbinfo->si_xib->f_path, NULL);
38268 + err = PTR_ERR(file);
38269 + if (IS_ERR(file))
38271 + fput(sbinfo->si_xib);
38272 + sbinfo->si_xib = file;
38274 + p = sbinfo->si_xib_buf;
38275 + memset(p, 0, PAGE_SIZE);
38277 + sz = xino_fwrite(sbinfo->si_xib, p, PAGE_SIZE, &pos);
38278 + if (unlikely(sz != PAGE_SIZE)) {
38280 + AuIOErr("err %d\n", err);
38286 + mutex_lock(&sbinfo->si_xib_mtx);
38287 + /* mnt_want_write() is unnecessary here */
38288 + err = xib_restore(sb);
38289 + mutex_unlock(&sbinfo->si_xib_mtx);
38295 +/* ---------------------------------------------------------------------- */
38297 +struct au_xino *au_xino_alloc(unsigned int nfile)
38299 + struct au_xino *xi;
38301 + xi = kzalloc(sizeof(*xi), GFP_NOFS);
38302 + if (unlikely(!xi))
38304 + xi->xi_nfile = nfile;
38305 + xi->xi_file = kcalloc(nfile, sizeof(*xi->xi_file), GFP_NOFS);
38306 + if (unlikely(!xi->xi_file))
38309 + xi->xi_nondir.total = 8; /* initial size */
38310 + xi->xi_nondir.array = kcalloc(xi->xi_nondir.total, sizeof(ino_t),
38312 + if (unlikely(!xi->xi_nondir.array))
38315 + spin_lock_init(&xi->xi_nondir.spin);
38316 + init_waitqueue_head(&xi->xi_nondir.wqh);
38317 + mutex_init(&xi->xi_mtx);
38318 + INIT_HLIST_BL_HEAD(&xi->xi_writing);
38319 + atomic_set(&xi->xi_truncating, 0);
38320 + kref_init(&xi->xi_kref);
38321 + goto out; /* success */
38324 + au_kfree_try_rcu(xi->xi_file);
38326 + au_kfree_rcu(xi);
38332 +static int au_xino_init(struct au_branch *br, int idx, struct file *file)
38335 + struct au_xino *xi;
38338 + xi = au_xino_alloc(idx + 1);
38339 + if (unlikely(!xi)) {
38346 + xi->xi_file[idx] = file;
38347 + AuDebugOn(br->br_xino);
38348 + br->br_xino = xi;
38354 +static void au_xino_release(struct kref *kref)
38356 + struct au_xino *xi;
38358 + unsigned long ul;
38359 + struct hlist_bl_head *hbl;
38360 + struct hlist_bl_node *pos, *n;
38361 + struct au_xi_writing *p;
38363 + xi = container_of(kref, struct au_xino, xi_kref);
38364 + for (i = 0; i < xi->xi_nfile; i++)
38365 + if (xi->xi_file[i])
38366 + fput(xi->xi_file[i]);
38367 + for (i = xi->xi_nondir.total - 1; i >= 0; i--)
38368 + AuDebugOn(xi->xi_nondir.array[i]);
38369 + mutex_destroy(&xi->xi_mtx);
38370 + hbl = &xi->xi_writing;
38371 + ul = au_hbl_count(hbl);
38372 + if (unlikely(ul)) {
38373 + pr_warn("xi_writing %lu\n", ul);
38374 + hlist_bl_lock(hbl);
38375 + hlist_bl_for_each_entry_safe(p, pos, n, hbl, node) {
38376 + hlist_bl_del(&p->node);
38377 + /* kmemleak reported au_kfree_rcu() doesn't free it */
38380 + hlist_bl_unlock(hbl);
38382 + au_kfree_try_rcu(xi->xi_file);
38383 + au_kfree_try_rcu(xi->xi_nondir.array);
38384 + au_kfree_rcu(xi);
38387 +int au_xino_put(struct au_branch *br)
38390 + struct au_xino *xi;
38393 + xi = br->br_xino;
38395 + br->br_xino = NULL;
38396 + ret = kref_put(&xi->xi_kref, au_xino_release);
38402 +/* ---------------------------------------------------------------------- */
38405 + * xino mount option handlers
38409 +static void xino_clear_xib(struct super_block *sb)
38411 + struct au_sbinfo *sbinfo;
38413 + SiMustWriteLock(sb);
38415 + sbinfo = au_sbi(sb);
38416 + if (sbinfo->si_xib)
38417 + fput(sbinfo->si_xib);
38418 + sbinfo->si_xib = NULL;
38419 + if (sbinfo->si_xib_buf)
38420 + free_page((unsigned long)sbinfo->si_xib_buf);
38421 + sbinfo->si_xib_buf = NULL;
38424 +static int au_xino_set_xib(struct super_block *sb, struct path *path)
38428 + struct au_sbinfo *sbinfo;
38429 + struct file *file;
38430 + struct super_block *xi_sb;
38432 + SiMustWriteLock(sb);
38434 + sbinfo = au_sbi(sb);
38435 + file = au_xino_create2(sb, path, sbinfo->si_xib);
38436 + err = PTR_ERR(file);
38437 + if (IS_ERR(file))
38439 + if (sbinfo->si_xib)
38440 + fput(sbinfo->si_xib);
38441 + sbinfo->si_xib = file;
38442 + xi_sb = file_inode(file)->i_sb;
38443 + sbinfo->si_ximaxent = xi_sb->s_maxbytes;
38444 + if (unlikely(sbinfo->si_ximaxent < PAGE_SIZE)) {
38446 + pr_err("s_maxbytes(%llu) on %s is too small\n",
38447 + (u64)sbinfo->si_ximaxent, au_sbtype(xi_sb));
38450 + sbinfo->si_ximaxent /= sizeof(ino_t);
38453 + if (!sbinfo->si_xib_buf)
38454 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
38455 + if (unlikely(!sbinfo->si_xib_buf))
38458 + sbinfo->si_xib_last_pindex = 0;
38459 + sbinfo->si_xib_next_bit = 0;
38460 + if (vfsub_f_size_read(file) < PAGE_SIZE) {
38462 + err = xino_fwrite(file, sbinfo->si_xib_buf, PAGE_SIZE, &pos);
38463 + if (unlikely(err != PAGE_SIZE))
38467 + goto out; /* success */
38470 + if (sbinfo->si_xib_buf)
38471 + free_page((unsigned long)sbinfo->si_xib_buf);
38472 + sbinfo->si_xib_buf = NULL;
38476 + fput(sbinfo->si_xib);
38477 + sbinfo->si_xib = NULL;
38483 +/* xino for each branch */
38484 +static void xino_clear_br(struct super_block *sb)
38486 + aufs_bindex_t bindex, bbot;
38487 + struct au_branch *br;
38489 + bbot = au_sbbot(sb);
38490 + for (bindex = 0; bindex <= bbot; bindex++) {
38491 + br = au_sbr(sb, bindex);
38497 +static void au_xino_set_br_shared(struct super_block *sb, struct au_branch *br,
38498 + aufs_bindex_t bshared)
38500 + struct au_branch *brshared;
38502 + brshared = au_sbr(sb, bshared);
38503 + AuDebugOn(!brshared->br_xino);
38504 + AuDebugOn(!brshared->br_xino->xi_file);
38505 + if (br->br_xino != brshared->br_xino) {
38506 + au_xino_get(brshared);
38508 + br->br_xino = brshared->br_xino;
38512 +struct au_xino_do_set_br {
38513 + struct au_branch *br;
38515 + aufs_bindex_t bshared;
38518 +static int au_xino_do_set_br(struct super_block *sb, struct path *path,
38519 + struct au_xino_do_set_br *args)
38522 + struct au_xi_calc calc;
38523 + struct file *file;
38524 + struct au_branch *br;
38525 + struct au_xi_new xinew = {
38530 + xinew.xi = br->br_xino;
38531 + au_xi_calc(sb, args->h_ino, &calc);
38532 + xinew.copy_src = au_xino_file(xinew.xi, calc.idx);
38533 + if (args->bshared >= 0)
38534 + /* shared xino */
38535 + au_xino_set_br_shared(sb, br, args->bshared);
38536 + else if (!xinew.xi) {
38538 + err = au_xino_init(br, calc.idx, xinew.copy_src);
38539 + if (unlikely(err))
38543 + /* force re-creating */
38544 + xinew.xi = br->br_xino;
38545 + xinew.idx = calc.idx;
38546 + mutex_lock(&xinew.xi->xi_mtx);
38547 + file = au_xi_new(sb, &xinew);
38548 + mutex_unlock(&xinew.xi->xi_mtx);
38549 + err = PTR_ERR(file);
38550 + if (IS_ERR(file))
38552 + AuDebugOn(!file);
38554 + err = au_xino_do_write(file, &calc, AUFS_ROOT_INO);
38555 + if (unlikely(err))
38563 +static int au_xino_set_br(struct super_block *sb, struct path *path)
38566 + aufs_bindex_t bindex, bbot;
38567 + struct au_xino_do_set_br args;
38568 + struct inode *inode;
38570 + SiMustWriteLock(sb);
38572 + bbot = au_sbbot(sb);
38573 + inode = d_inode(sb->s_root);
38574 + for (bindex = 0; bindex <= bbot; bindex++) {
38575 + args.h_ino = au_h_iptr(inode, bindex)->i_ino;
38576 + args.br = au_sbr(sb, bindex);
38577 + args.bshared = is_sb_shared(sb, bindex, bindex - 1);
38578 + err = au_xino_do_set_br(sb, path, &args);
38579 + if (unlikely(err))
38587 +void au_xino_clr(struct super_block *sb)
38589 + struct au_sbinfo *sbinfo;
38591 + au_xigen_clr(sb);
38592 + xino_clear_xib(sb);
38593 + xino_clear_br(sb);
38594 + dbgaufs_brs_del(sb, 0);
38595 + sbinfo = au_sbi(sb);
38596 + /* lvalue, do not call au_mntflags() */
38597 + au_opt_clr(sbinfo->si_mntflags, XINO);
38600 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount)
38603 + struct dentry *dentry, *parent, *cur_dentry, *cur_parent;
38604 + struct qstr *dname, *cur_name;
38605 + struct file *cur_xino;
38606 + struct au_sbinfo *sbinfo;
38607 + struct path *path, *cur_path;
38609 + SiMustWriteLock(sb);
38612 + sbinfo = au_sbi(sb);
38613 + path = &xiopt->file->f_path;
38614 + dentry = path->dentry;
38615 + parent = dget_parent(dentry);
38618 + cur_xino = sbinfo->si_xib;
38620 + cur_path = &cur_xino->f_path;
38621 + cur_dentry = cur_path->dentry;
38622 + cur_parent = dget_parent(cur_dentry);
38623 + cur_name = &cur_dentry->d_name;
38624 + dname = &dentry->d_name;
38625 + skip = (cur_parent == parent
38626 + && au_qstreq(dname, cur_name));
38627 + dput(cur_parent);
38633 + au_opt_set(sbinfo->si_mntflags, XINO);
38634 + err = au_xino_set_xib(sb, path);
38635 + /* si_x{read,write} are set */
38637 + err = au_xigen_set(sb, path);
38639 + err = au_xino_set_br(sb, path);
38641 + dbgaufs_brs_add(sb, 0, /*topdown*/1);
38642 + goto out; /* success */
38646 + AuIOErr("failed setting xino(%d).\n", err);
38655 + * create a xinofile at the default place/path.
38657 +struct file *au_xino_def(struct super_block *sb)
38659 + struct file *file;
38661 + struct au_branch *br;
38662 + struct super_block *h_sb;
38663 + struct path path;
38664 + aufs_bindex_t bbot, bindex, bwr;
38667 + bbot = au_sbbot(sb);
38669 + for (bindex = 0; bindex <= bbot; bindex++) {
38670 + br = au_sbr(sb, bindex);
38671 + if (au_br_writable(br->br_perm)
38672 + && !au_test_fs_bad_xino(au_br_sb(br))) {
38679 + file = ERR_PTR(-ENOMEM);
38680 + page = (void *)__get_free_page(GFP_NOFS);
38681 + if (unlikely(!page))
38683 + path.mnt = au_br_mnt(br);
38684 + path.dentry = au_h_dptr(sb->s_root, bwr);
38685 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
38686 + file = (void *)p;
38687 + if (!IS_ERR(p)) {
38688 + strcat(p, "/" AUFS_XINO_FNAME);
38689 + AuDbg("%s\n", p);
38690 + file = au_xino_create(sb, p, /*silent*/0, /*wbrtop*/1);
38692 + free_page((unsigned long)page);
38694 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0,
38696 + if (IS_ERR(file))
38698 + h_sb = file->f_path.dentry->d_sb;
38699 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
38700 + pr_err("xino doesn't support %s(%s)\n",
38701 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
38703 + file = ERR_PTR(-EINVAL);
38711 +/* ---------------------------------------------------------------------- */
38714 + * initialize the xinofile for the specified branch @br
38715 + * at the place/path where @base_file indicates.
38716 + * test whether another branch is on the same filesystem or not,
38717 + * if found then share the xinofile with another branch.
38719 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
38720 + struct path *base)
38723 + struct au_xino_do_set_br args = {
38728 + args.bshared = sbr_find_shared(sb, /*btop*/0, au_sbbot(sb),
38730 + err = au_xino_do_set_br(sb, base, &args);
38731 + if (unlikely(err))
38737 +/* ---------------------------------------------------------------------- */
38740 + * get an unused inode number from bitmap
38742 +ino_t au_xino_new_ino(struct super_block *sb)
38745 + unsigned long *p, pindex, ul, pend;
38746 + struct au_sbinfo *sbinfo;
38747 + struct file *file;
38748 + int free_bit, err;
38750 + if (!au_opt_test(au_mntflags(sb), XINO))
38751 + return iunique(sb, AUFS_FIRST_INO);
38753 + sbinfo = au_sbi(sb);
38754 + mutex_lock(&sbinfo->si_xib_mtx);
38755 + p = sbinfo->si_xib_buf;
38756 + free_bit = sbinfo->si_xib_next_bit;
38757 + if (free_bit < page_bits && !test_bit(free_bit, p))
38758 + goto out; /* success */
38759 + free_bit = find_first_zero_bit(p, page_bits);
38760 + if (free_bit < page_bits)
38761 + goto out; /* success */
38763 + pindex = sbinfo->si_xib_last_pindex;
38764 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
38765 + err = xib_pindex(sb, ul);
38766 + if (unlikely(err))
38768 + free_bit = find_first_zero_bit(p, page_bits);
38769 + if (free_bit < page_bits)
38770 + goto out; /* success */
38773 + file = sbinfo->si_xib;
38774 + pend = vfsub_f_size_read(file) / PAGE_SIZE;
38775 + for (ul = pindex + 1; ul <= pend; ul++) {
38776 + err = xib_pindex(sb, ul);
38777 + if (unlikely(err))
38779 + free_bit = find_first_zero_bit(p, page_bits);
38780 + if (free_bit < page_bits)
38781 + goto out; /* success */
38786 + set_bit(free_bit, p);
38787 + sbinfo->si_xib_next_bit = free_bit + 1;
38788 + pindex = sbinfo->si_xib_last_pindex;
38789 + mutex_unlock(&sbinfo->si_xib_mtx);
38790 + ino = xib_calc_ino(pindex, free_bit);
38791 + AuDbg("i%lu\n", (unsigned long)ino);
38794 + mutex_unlock(&sbinfo->si_xib_mtx);
38799 +/* for s_op->delete_inode() */
38800 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
38803 + unsigned int mnt_flags;
38804 + aufs_bindex_t bindex, bbot, bi;
38805 + unsigned char try_trunc;
38806 + struct au_iinfo *iinfo;
38807 + struct super_block *sb;
38808 + struct au_hinode *hi;
38809 + struct inode *h_inode;
38810 + struct au_branch *br;
38811 + struct au_xi_calc calc;
38812 + struct file *file;
38814 + AuDebugOn(au_is_bad_inode(inode));
38816 + sb = inode->i_sb;
38817 + mnt_flags = au_mntflags(sb);
38818 + if (!au_opt_test(mnt_flags, XINO)
38819 + || inode->i_ino == AUFS_ROOT_INO)
38823 + au_xigen_inc(inode);
38824 + au_xib_clear_bit(inode);
38827 + iinfo = au_ii(inode);
38828 + bindex = iinfo->ii_btop;
38832 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
38833 + hi = au_hinode(iinfo, bindex);
38834 + bbot = iinfo->ii_bbot;
38835 + for (; bindex <= bbot; bindex++, hi++) {
38836 + h_inode = hi->hi_inode;
38838 + || (!unlinked && h_inode->i_nlink))
38841 + /* inode may not be revalidated */
38842 + bi = au_br_index(sb, hi->hi_id);
38846 + br = au_sbr(sb, bi);
38847 + au_xi_calc(sb, h_inode->i_ino, &calc);
38848 + file = au_xino_file(br->br_xino, calc.idx);
38849 + if (IS_ERR_OR_NULL(file))
38852 + err = au_xino_do_write(file, &calc, /*ino*/0);
38853 + if (!err && try_trunc
38854 + && au_test_fs_trunc_xino(au_br_sb(br)))
38855 + xino_try_trunc(sb, br);
38859 +/* ---------------------------------------------------------------------- */
38861 +static int au_xinondir_find(struct au_xino *xi, ino_t h_ino)
38863 + int found, total, i;
38866 + total = xi->xi_nondir.total;
38867 + for (i = 0; i < total; i++) {
38868 + if (xi->xi_nondir.array[i] != h_ino)
38877 +static int au_xinondir_expand(struct au_xino *xi)
38882 + BUILD_BUG_ON(KMALLOC_MAX_SIZE > INT_MAX);
38885 + sz = xi->xi_nondir.total * sizeof(ino_t);
38886 + if (unlikely(sz > KMALLOC_MAX_SIZE / 2))
38888 + p = au_kzrealloc(xi->xi_nondir.array, sz, sz << 1, GFP_ATOMIC,
38889 + /*may_shrink*/0);
38891 + xi->xi_nondir.array = p;
38892 + xi->xi_nondir.total <<= 1;
38893 + AuDbg("xi_nondir.total %d\n", xi->xi_nondir.total);
38901 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
38902 + ino_t h_ino, int idx)
38904 + struct au_xino *xi;
38906 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
38907 + xi = au_sbr(sb, bindex)->br_xino;
38908 + AuDebugOn(idx < 0 || xi->xi_nondir.total <= idx);
38910 + spin_lock(&xi->xi_nondir.spin);
38911 + AuDebugOn(xi->xi_nondir.array[idx] != h_ino);
38912 + xi->xi_nondir.array[idx] = 0;
38913 + spin_unlock(&xi->xi_nondir.spin);
38914 + wake_up_all(&xi->xi_nondir.wqh);
38917 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
38920 + int err, found, empty;
38921 + struct au_xino *xi;
38925 + if (!au_opt_test(au_mntflags(sb), XINO))
38926 + goto out; /* no xino */
38928 + xi = au_sbr(sb, bindex)->br_xino;
38931 + spin_lock(&xi->xi_nondir.spin);
38932 + found = au_xinondir_find(xi, h_ino);
38933 + if (found == -1) {
38934 + empty = au_xinondir_find(xi, /*h_ino*/0);
38935 + if (empty == -1) {
38936 + empty = xi->xi_nondir.total;
38937 + err = au_xinondir_expand(xi);
38938 + if (unlikely(err))
38941 + xi->xi_nondir.array[empty] = h_ino;
38944 + spin_unlock(&xi->xi_nondir.spin);
38945 + wait_event(xi->xi_nondir.wqh,
38946 + xi->xi_nondir.array[found] != h_ino);
38951 + spin_unlock(&xi->xi_nondir.spin);
38956 +/* ---------------------------------------------------------------------- */
38958 +int au_xino_path(struct seq_file *seq, struct file *file)
38962 + err = au_seq_path(seq, &file->f_path);
38963 + if (unlikely(err))
38966 +#define Deleted "\\040(deleted)"
38967 + seq->count -= sizeof(Deleted) - 1;
38968 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
38969 + sizeof(Deleted) - 1));
38975 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
38976 --- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
38977 +++ linux/include/uapi/linux/aufs_type.h 2022-10-03 07:51:14.929542593 +0200
38979 +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
38981 + * Copyright (C) 2005-2021 Junjiro R. Okajima
38983 + * This program, aufs is free software; you can redistribute it and/or modify
38984 + * it under the terms of the GNU General Public License as published by
38985 + * the Free Software Foundation; either version 2 of the License, or
38986 + * (at your option) any later version.
38988 + * This program is distributed in the hope that it will be useful,
38989 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
38990 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
38991 + * GNU General Public License for more details.
38993 + * You should have received a copy of the GNU General Public License
38994 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
38997 +#ifndef __AUFS_TYPE_H__
38998 +#define __AUFS_TYPE_H__
39000 +#define AUFS_NAME "aufs"
39004 + * define it before including all other headers.
39005 + * sched.h may use pr_* macros before defining "current", so define the
39006 + * no-current version first, and re-define later.
39008 +#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
39009 +#include <linux/sched.h>
39011 +#define pr_fmt(fmt) \
39012 + AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
39013 + (int)sizeof(current->comm), current->comm, current->pid
39014 +#include <linux/limits.h>
39016 +#include <stdint.h>
39017 +#include <sys/types.h>
39018 +#include <limits.h>
39019 +#endif /* __KERNEL__ */
39021 +#define AUFS_VERSION "5.x-rcN-20220912"
39023 +/* todo? move this to linux-2.6.19/include/magic.h */
39024 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
39026 +/* ---------------------------------------------------------------------- */
39029 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
39030 +typedef int8_t aufs_bindex_t;
39031 +#define AUFS_BRANCH_MAX 127
39033 +typedef int16_t aufs_bindex_t;
39034 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
39035 +#define AUFS_BRANCH_MAX 511
39036 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
39037 +#define AUFS_BRANCH_MAX 1023
39038 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
39039 +#define AUFS_BRANCH_MAX 32767
39043 +#ifndef AUFS_BRANCH_MAX
39044 +#error unknown CONFIG_AUFS_BRANCH_MAX value
39046 +#endif /* __KERNEL__ */
39048 +/* ---------------------------------------------------------------------- */
39050 +#define AUFS_FSTYPE AUFS_NAME
39052 +#define AUFS_ROOT_INO 2
39053 +#define AUFS_FIRST_INO 11
39055 +#define AUFS_WH_PFX ".wh."
39056 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
39057 +#define AUFS_WH_TMP_LEN 4
39058 +/* a limit for rmdir/rename a dir and copyup */
39059 +#define AUFS_MAX_NAMELEN (NAME_MAX \
39060 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
39062 + - AUFS_WH_TMP_LEN) /* hex */
39063 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
39064 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
39065 +#define AUFS_XINO_DEF_SEC 30 /* seconds */
39066 +#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
39067 +#define AUFS_DIRWH_DEF 3
39068 +#define AUFS_RDCACHE_DEF 10 /* seconds */
39069 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
39070 +#define AUFS_RDBLK_DEF 512 /* bytes */
39071 +#define AUFS_RDHASH_DEF 32
39072 +#define AUFS_WKQ_NAME AUFS_NAME "d"
39073 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
39074 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
39075 +#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
39076 +#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
39078 +/* pseudo-link maintenace under /proc */
39079 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
39080 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
39081 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
39083 +/* dirren, renamed dir */
39084 +#define AUFS_DR_INFO_PFX AUFS_WH_PFX ".dr."
39085 +#define AUFS_DR_BRHINO_NAME AUFS_WH_PFX "hino"
39086 +/* whiteouted doubly */
39087 +#define AUFS_WH_DR_INFO_PFX AUFS_WH_PFX AUFS_DR_INFO_PFX
39088 +#define AUFS_WH_DR_BRHINO AUFS_WH_PFX AUFS_DR_BRHINO_NAME
39090 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
39091 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
39093 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
39094 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
39095 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
39097 +/* doubly whiteouted */
39098 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
39099 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
39100 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
39102 +/* branch permissions and attributes */
39103 +#define AUFS_BRPERM_RW "rw"
39104 +#define AUFS_BRPERM_RO "ro"
39105 +#define AUFS_BRPERM_RR "rr"
39106 +#define AUFS_BRATTR_COO_REG "coo_reg"
39107 +#define AUFS_BRATTR_COO_ALL "coo_all"
39108 +#define AUFS_BRATTR_FHSM "fhsm"
39109 +#define AUFS_BRATTR_UNPIN "unpin"
39110 +#define AUFS_BRATTR_ICEX "icex"
39111 +#define AUFS_BRATTR_ICEX_SEC "icexsec"
39112 +#define AUFS_BRATTR_ICEX_SYS "icexsys"
39113 +#define AUFS_BRATTR_ICEX_TR "icextr"
39114 +#define AUFS_BRATTR_ICEX_USR "icexusr"
39115 +#define AUFS_BRATTR_ICEX_OTH "icexoth"
39116 +#define AUFS_BRRATTR_WH "wh"
39117 +#define AUFS_BRWATTR_NLWH "nolwh"
39118 +#define AUFS_BRWATTR_MOO "moo"
39120 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
39121 +#define AuBrPerm_RO (1 << 1) /* readonly */
39122 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
39123 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
39125 +#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
39126 +#define AuBrAttr_COO_ALL (1 << 4)
39127 +#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
39129 +#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
39130 +#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
39131 + branch. meaningless since
39132 + linux-3.18-rc1 */
39134 +/* ignore error in copying XATTR */
39135 +#define AuBrAttr_ICEX_SEC (1 << 7)
39136 +#define AuBrAttr_ICEX_SYS (1 << 8)
39137 +#define AuBrAttr_ICEX_TR (1 << 9)
39138 +#define AuBrAttr_ICEX_USR (1 << 10)
39139 +#define AuBrAttr_ICEX_OTH (1 << 11)
39140 +#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
39141 + | AuBrAttr_ICEX_SYS \
39142 + | AuBrAttr_ICEX_TR \
39143 + | AuBrAttr_ICEX_USR \
39144 + | AuBrAttr_ICEX_OTH)
39146 +#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
39147 +#define AuBrRAttr_Mask AuBrRAttr_WH
39149 +#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
39150 +#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
39151 +#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
39153 +#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
39155 +/* #warning test userspace */
39157 +#ifndef CONFIG_AUFS_FHSM
39158 +#undef AuBrAttr_FHSM
39159 +#define AuBrAttr_FHSM 0
39161 +#ifndef CONFIG_AUFS_XATTR
39162 +#undef AuBrAttr_ICEX
39163 +#define AuBrAttr_ICEX 0
39164 +#undef AuBrAttr_ICEX_SEC
39165 +#define AuBrAttr_ICEX_SEC 0
39166 +#undef AuBrAttr_ICEX_SYS
39167 +#define AuBrAttr_ICEX_SYS 0
39168 +#undef AuBrAttr_ICEX_TR
39169 +#define AuBrAttr_ICEX_TR 0
39170 +#undef AuBrAttr_ICEX_USR
39171 +#define AuBrAttr_ICEX_USR 0
39172 +#undef AuBrAttr_ICEX_OTH
39173 +#define AuBrAttr_ICEX_OTH 0
39177 +/* the longest combination */
39178 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
39179 +#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
39180 + "+" AUFS_BRATTR_COO_REG \
39181 + "+" AUFS_BRATTR_FHSM \
39182 + "+" AUFS_BRATTR_UNPIN \
39183 + "+" AUFS_BRATTR_ICEX_SEC \
39184 + "+" AUFS_BRATTR_ICEX_SYS \
39185 + "+" AUFS_BRATTR_ICEX_USR \
39186 + "+" AUFS_BRATTR_ICEX_OTH \
39187 + "+" AUFS_BRWATTR_NLWH)
39190 + char a[AuBrPermStrSz];
39191 +} au_br_perm_str_t;
39193 +static inline int au_br_writable(int brperm)
39195 + return brperm & AuBrPerm_RW;
39198 +static inline int au_br_whable(int brperm)
39200 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
39203 +static inline int au_br_wh_linkable(int brperm)
39205 + return !(brperm & AuBrWAttr_NoLinkWH);
39208 +static inline int au_br_cmoo(int brperm)
39210 + return brperm & AuBrAttr_CMOO_Mask;
39213 +static inline int au_br_fhsm(int brperm)
39215 + return brperm & AuBrAttr_FHSM;
39218 +/* ---------------------------------------------------------------------- */
39222 + /* readdir in userspace */
39226 + AuCtl_WBR_FD, /* pathconf wrapper */
39227 + AuCtl_IBUSY, /* busy inode */
39228 + AuCtl_MVDOWN, /* move-down */
39229 + AuCtl_BR, /* info about branches */
39230 + AuCtl_FHSM_FD /* connection for fhsm */
39233 +/* borrowed from linux/include/linux/kernel.h */
39235 +#ifdef _GNU_SOURCE
39236 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
39238 +#define ALIGN(x, a) (((x) + (a) - 1) & ~((a) - 1))
39240 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
39243 +/* borrowed from linux/include/linux/compiler-gcc3.h */
39245 +#define __aligned(x) __attribute__((aligned(x)))
39250 +#define __packed __attribute__((packed))
39254 +struct au_rdu_cookie {
39259 + uint32_t generation;
39262 +struct au_rdu_ent {
39271 +static inline int au_rdu_len(int nlen)
39273 + /* include the terminating NULL */
39274 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
39275 + sizeof(uint64_t));
39278 +union au_rdu_ent_ul {
39279 + struct au_rdu_ent __user *e;
39291 + uint64_t sz; /* AuCtl_RDU */
39292 + uint64_t nent; /* AuCtl_RDU_INO */
39294 + union au_rdu_ent_ul ent;
39295 + uint16_t verify[AufsCtlRduV_End];
39297 + /* input/output */
39301 + union au_rdu_ent_ul tail;
39302 + /* number of entries which were added in a single call */
39307 + struct au_rdu_cookie cookie;
39310 +/* ---------------------------------------------------------------------- */
39312 +/* dirren. the branch is identified by the filename who contains this */
39313 +struct au_drinfo {
39316 + uint8_t oldnamelen;
39317 + uint64_t _padding;
39319 + uint8_t oldname[];
39322 +struct au_drinfo_fdata {
39324 + struct au_drinfo drinfo;
39327 +#define AUFS_DRINFO_MAGIC_V1 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x01)
39329 +#define AUFS_DRINFO_MAGIC_V2 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x02)
39331 +/* ---------------------------------------------------------------------- */
39333 +struct aufs_wbr_fd {
39338 +/* ---------------------------------------------------------------------- */
39340 +struct aufs_ibusy {
39341 + uint64_t ino, h_ino;
39345 +/* ---------------------------------------------------------------------- */
39347 +/* error code for move-down */
39348 +/* the actual message strings are implemented in aufs-util.git */
39350 + EAU_MVDOWN_OPAQUE = 1,
39351 + EAU_MVDOWN_WHITEOUT,
39352 + EAU_MVDOWN_UPPER,
39353 + EAU_MVDOWN_BOTTOM,
39354 + EAU_MVDOWN_NOUPPER,
39355 + EAU_MVDOWN_NOLOWERBR,
39359 +/* flags for move-down */
39360 +#define AUFS_MVDOWN_DMSG 1
39361 +#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
39362 +#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
39363 +#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
39364 +#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
39365 +#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
39366 +#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
39367 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
39368 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
39369 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
39370 +#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
39371 +#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
39372 +#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
39374 +/* index for move-down */
39376 + AUFS_MVDOWN_UPPER,
39377 + AUFS_MVDOWN_LOWER,
39378 + AUFS_MVDOWN_NARRAY
39382 + * additional info of move-down
39383 + * number of free blocks and inodes.
39384 + * subset of struct kstatfs, but smaller and always 64bit.
39386 +struct aufs_stfs {
39387 + uint64_t f_blocks;
39388 + uint64_t f_bavail;
39389 + uint64_t f_files;
39390 + uint64_t f_ffree;
39393 +struct aufs_stbr {
39394 + int16_t brid; /* optional input */
39395 + int16_t bindex; /* output */
39396 + struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
39399 +struct aufs_mvdown {
39400 + uint32_t flags; /* input/output */
39401 + struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
39402 + int8_t au_errno; /* output */
39405 +/* ---------------------------------------------------------------------- */
39407 +union aufs_brinfo {
39408 + /* PATH_MAX may differ between kernel-space and user-space */
39409 + char _spacer[4096];
39417 +/* ---------------------------------------------------------------------- */
39419 +#define AuCtlType 'A'
39420 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
39421 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
39422 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
39423 + struct aufs_wbr_fd)
39424 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
39425 +#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
39426 + struct aufs_mvdown)
39427 +#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
39428 +#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
39430 +#endif /* __AUFS_TYPE_H__ */
39431 SPDX-License-Identifier: GPL-2.0
39432 aufs5.x-rcN loopback patch
39434 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
39435 index 8ca56da3ca69..9ac6f67edf09 100644
39436 --- a/drivers/block/loop.c
39437 +++ b/drivers/block/loop.c
39438 @@ -54,7 +54,7 @@ struct loop_device {
39440 char lo_file_name[LO_NAME_SIZE];
39442 - struct file * lo_backing_file;
39443 + struct file *lo_backing_file, *lo_backing_virt_file;
39444 struct block_device *lo_device;
39446 gfp_t old_gfp_mask;
39447 @@ -510,6 +510,15 @@ static inline void loop_update_dio(struct loop_device *lo)
39451 +static struct file *loop_real_file(struct file *file)
39453 + struct file *f = NULL;
39455 + if (file->f_path.dentry->d_sb->s_op->real_loop)
39456 + f = file->f_path.dentry->d_sb->s_op->real_loop(file);
39460 static void loop_reread_partitions(struct loop_device *lo)
39463 @@ -567,6 +576,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39465 struct file *file = fget(arg);
39466 struct file *old_file;
39467 + struct file *f, *virt_file = NULL, *old_virt_file;
39471 @@ -590,11 +600,19 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39472 if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
39475 + f = loop_real_file(file);
39477 + virt_file = file;
39482 error = loop_validate_file(file, bdev);
39486 old_file = lo->lo_backing_file;
39487 + old_virt_file = lo->lo_backing_virt_file;
39491 @@ -607,6 +625,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39492 blk_mq_freeze_queue(lo->lo_queue);
39493 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
39494 lo->lo_backing_file = file;
39495 + lo->lo_backing_virt_file = virt_file;
39496 lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
39497 mapping_set_gfp_mask(file->f_mapping,
39498 lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39499 @@ -629,6 +648,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39503 + if (old_virt_file)
39504 + fput(old_virt_file);
39506 loop_reread_partitions(lo);
39508 @@ -642,6 +663,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39509 loop_global_unlock(lo, is_loop);
39517 @@ -1013,6 +1036,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39518 const struct loop_config *config)
39520 struct file *file = fget(config->fd);
39521 + struct file *f, *virt_file = NULL;
39522 struct inode *inode;
39523 struct address_space *mapping;
39525 @@ -1031,6 +1055,13 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39526 /* suppress uevents while reconfiguring the device */
39527 dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 1);
39529 + f = loop_real_file(file);
39531 + virt_file = file;
39537 * If we don't hold exclusive handle for the device, upgrade to it
39538 * here to avoid changing device under exclusive owner.
39539 @@ -1091,6 +1122,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39540 lo->use_dio = lo->lo_flags & LO_FLAGS_DIRECT_IO;
39541 lo->lo_device = bdev;
39542 lo->lo_backing_file = file;
39543 + lo->lo_backing_virt_file = virt_file;
39544 lo->old_gfp_mask = mapping_gfp_mask(mapping);
39545 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39547 @@ -1146,6 +1178,8 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39548 bd_abort_claiming(bdev, loop_configure);
39553 /* This is safe: open() is still holding a reference. */
39554 module_put(THIS_MODULE);
39556 @@ -1154,6 +1188,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39557 static void __loop_clr_fd(struct loop_device *lo, bool release)
39560 + struct file *virt_filp = lo->lo_backing_virt_file;
39561 gfp_t gfp = lo->old_gfp_mask;
39563 if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags))
39564 @@ -1170,6 +1205,7 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39565 spin_lock_irq(&lo->lo_lock);
39566 filp = lo->lo_backing_file;
39567 lo->lo_backing_file = NULL;
39568 + lo->lo_backing_virt_file = NULL;
39569 spin_unlock_irq(&lo->lo_lock);
39571 lo->lo_device = NULL;
39572 @@ -1232,6 +1268,8 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39573 * fput can take open_mutex which is usually taken before lo_mutex.
39580 static int loop_clr_fd(struct loop_device *lo)
39581 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
39582 index 99ceca144044..e49dfe855032 100644
39583 --- a/fs/aufs/f_op.c
39584 +++ b/fs/aufs/f_op.c
39585 @@ -304,7 +304,7 @@ static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
39586 if (IS_ERR(h_file))
39589 - if (au_test_loopback_kthread()) {
39590 + if (0 && au_test_loopback_kthread()) {
39591 au_warn_loopback(h_file->f_path.dentry->d_sb);
39592 if (file->f_mapping != h_file->f_mapping) {
39593 file->f_mapping = h_file->f_mapping;
39594 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
39595 index 74347bd75b38..5ef888a1d53f 100644
39596 --- a/fs/aufs/loop.c
39597 +++ b/fs/aufs/loop.c
39598 @@ -133,3 +133,19 @@ void au_loopback_fin(void)
39599 symbol_put(loop_backing_file);
39600 au_kfree_try_rcu(au_warn_loopback_array);
39603 +/* ---------------------------------------------------------------------- */
39605 +/* support the loopback block device insude aufs */
39607 +struct file *aufs_real_loop(struct file *file)
39611 + BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
39612 + fi_read_lock(file);
39613 + f = au_hf_top(file);
39614 + fi_read_unlock(file);
39618 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
39619 index 7293bee427f9..3345c098d0d4 100644
39620 --- a/fs/aufs/loop.h
39621 +++ b/fs/aufs/loop.h
39622 @@ -26,6 +26,8 @@ void au_warn_loopback(struct super_block *h_sb);
39624 int au_loopback_init(void);
39625 void au_loopback_fin(void);
39627 +struct file *aufs_real_loop(struct file *file);
39629 AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
39631 @@ -36,6 +38,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
39633 AuStubInt0(au_loopback_init, void)
39634 AuStubVoid(au_loopback_fin, void)
39636 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
39637 #endif /* BLK_DEV_LOOP */
39639 #endif /* __KERNEL__ */
39640 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
39641 index 7d9bcc074c87..0eefe350c01c 100644
39642 --- a/fs/aufs/super.c
39643 +++ b/fs/aufs/super.c
39644 @@ -758,7 +758,10 @@ const struct super_operations aufs_sop = {
39645 .show_options = aufs_show_options,
39646 .statfs = aufs_statfs,
39647 .put_super = aufs_put_super,
39648 - .sync_fs = aufs_sync_fs
39649 + .sync_fs = aufs_sync_fs,
39650 +#ifdef CONFIG_AUFS_BDEV_LOOP
39651 + .real_loop = aufs_real_loop
39655 /* ---------------------------------------------------------------------- */
39656 diff --git a/include/linux/fs.h b/include/linux/fs.h
39657 index 773ebd5696e4..4baca7d85d52 100644
39658 --- a/include/linux/fs.h
39659 +++ b/include/linux/fs.h
39660 @@ -2253,6 +2253,10 @@ struct super_operations {
39661 struct shrink_control *);
39662 long (*free_cached_objects)(struct super_block *,
39663 struct shrink_control *);
39664 +#if IS_ENABLED(CONFIG_BLK_DEV_LOOP) || IS_ENABLED(CONFIG_BLK_DEV_LOOP_MODULE)
39666 + struct file *(*real_loop)(struct file *);