1 SPDX-License-Identifier: GPL-2.0
2 aufs5.x-rcN kbuild patch
4 diff --git a/fs/Kconfig b/fs/Kconfig
5 index 30b751c7f11a..fb8214374858 100644
8 @@ -312,6 +312,7 @@ 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 208a74e0b00e..57d47a0762e0 100644
20 @@ -140,3 +140,4 @@ 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 61d9f114c37f..95cb37e9c555 100644
32 @@ -3323,6 +3323,19 @@ 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 a58595f5ee2c..ba35d0d6847a 100644
54 --- a/drivers/block/loop.c
55 +++ b/drivers/block/loop.c
56 @@ -635,6 +635,24 @@ 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 93f4f5ee07bf..cddcf2e44930 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 f15d885b9796..d91d1bf393cc 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 @@ -64,6 +64,8 @@ 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 a0a36bfa3aa0..7a54c6b514a2 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 047b79db8eb5..a21caa5c54c8 100644
137 @@ -759,8 +759,8 @@ 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 @@ -770,9 +770,9 @@ 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 bbde95387a23..06c9242932a2 100644
163 --- a/include/linux/fs.h
164 +++ b/include/linux/fs.h
165 @@ -1347,6 +1347,7 @@ 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 @@ -1979,6 +1980,7 @@ 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 @@ -2486,6 +2488,7 @@ 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 @@ -2664,6 +2667,7 @@ 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 467b94257105..dab20282c82e 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 c06cab6546ed..9d7c317899e1 100644
247 --- a/kernel/locking/lockdep.c
248 +++ b/kernel/locking/lockdep.c
249 @@ -187,7 +187,7 @@ 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 @@ -208,6 +208,7 @@ 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 c1031843cc6a..6597f53969bf 100644
273 @@ -2190,7 +2190,7 @@ 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 13452b32e2bd..38acccfef9d4 100644
284 --- a/fs/proc/nommu.c
285 +++ b/fs/proc/nommu.c
286 @@ -40,7 +40,10 @@ 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 f46060eb91b5..2a7c8709acad 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 @@ -1903,7 +1906,7 @@ 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 e34edb775334..08ac00a002a7 100644
341 --- a/include/linux/mm.h
342 +++ b/include/linux/mm.h
343 @@ -1902,6 +1902,43 @@ 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 8834e38c06a4..628b7d1eedbd 100644
389 --- a/include/linux/mm_types.h
390 +++ b/include/linux/mm_types.h
391 @@ -362,6 +362,9 @@ 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 @@ -458,6 +461,9 @@ 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 9796897560ab..67c5046f47b8 100644
415 @@ -664,7 +664,7 @@ 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 4cc13f3179a5..a1bbb3af3fc2 100644
428 @@ -133,3 +133,4 @@ obj-$(CONFIG_PAGE_REPORTING) += page_reporting.o
429 obj-$(CONFIG_IO_MAPPING) += io-mapping.o
430 obj-$(CONFIG_HAVE_BOOTMEM_INFO_NODE) += bootmem_info.o
431 obj-$(CONFIG_GENERIC_IOREMAP) += ioremap.o
432 +obj-$(CONFIG_AUFS_FS:m=y) += prfile.o
433 diff --git a/mm/filemap.c b/mm/filemap.c
434 index 3a5ffb5587cd..6d12a6337a60 100644
437 @@ -3422,7 +3422,7 @@ 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 3aa839f81e63..9dfe979e8c2b 100644
450 @@ -184,7 +184,7 @@ 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 @@ -953,7 +953,7 @@ 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 @@ -1877,7 +1877,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
472 - fput(vma->vm_file);
476 /* Undo any partial mapping done by a device driver. */
477 @@ -2737,7 +2737,7 @@ 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 @@ -2756,7 +2756,7 @@ 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 @@ -2929,6 +2929,9 @@ 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/vm/remap_file_pages.rst.\n",
504 current->comm, current->pid);
505 @@ -2984,10 +2987,34 @@ 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 @@ -3261,7 +3288,7 @@ 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 55a9e48a7a02..8dc77ce96335 100644
553 @@ -522,7 +522,7 @@ 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 @@ -654,7 +654,7 @@ 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 @@ -1174,7 +1174,7 @@ 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 @@ -1251,10 +1251,10 @@ 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 cddcf2e44930..8e24a158ef4f 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 @@ -2996,6 +2997,7 @@ 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 e3e55d5e0be1..145d95db8bc9 100644
712 @@ -111,6 +111,7 @@ 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 d91d1bf393cc..3417421651d5 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 ada8fe814db9..4ebcd56ea4f0 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 7a54c6b514a2..ddae5f68ac19 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 b7d4d64f87c2..2d2d2c6d33e4 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 1315253e0247..46d9235f3a85 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 e643aec2b0ef..87acd61a0180 100644
798 --- a/fs/read_write.c
799 +++ b/fs/read_write.c
800 @@ -489,6 +489,7 @@ 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 @@ -599,6 +600,7 @@ 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 a21caa5c54c8..ccf9e1681eb7 100644
820 @@ -766,6 +766,7 @@ 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 @@ -795,6 +796,7 @@ 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 5c8c5175b385..ff7e9ff774b7 100644
840 @@ -384,6 +384,7 @@ 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 9d7c317899e1..ddfdeeec39d6 100644
850 --- a/kernel/locking/lockdep.c
851 +++ b/kernel/locking/lockdep.c
852 @@ -208,6 +208,7 @@ 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 c59e1a49bc40..8af926a7276d 100644
862 --- a/kernel/task_work.c
863 +++ b/kernel/task_work.c
864 @@ -167,3 +167,4 @@ 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 b7cf5cbfdc67..37f2d9d3d08c 100644
871 --- a/security/security.c
872 +++ b/security/security.c
873 @@ -1164,6 +1164,7 @@ 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 @@ -1180,6 +1181,7 @@ 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 @@ -1188,6 +1190,7 @@ 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 @@ -1215,6 +1218,7 @@ 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 @@ -1222,6 +1226,7 @@ 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 @@ -1229,6 +1234,7 @@ 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 @@ -1329,6 +1335,7 @@ 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 dentry *dentry, struct iattr *attr)
929 @@ -1526,6 +1533,7 @@ 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 2022-03-21 14:49:05.713299677 +0100
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 refered 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 2022-03-21 14:49:05.713299677 +0100
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).
2814 +Thank you very much.
2815 +Donations are always, including future donations, very important and
2816 +helpful for me to keep on developing aufs.
2820 +----------------------------------------
2821 +If you are an experienced user, no explanation is needed. Aufs is
2822 +just a linux filesystem.
2827 +# Local variables: ;
2830 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2831 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
2832 +++ linux/fs/aufs/aufs.h 2022-03-21 14:49:05.716633010 +0100
2834 +/* SPDX-License-Identifier: GPL-2.0 */
2836 + * Copyright (C) 2005-2021 Junjiro R. Okajima
2838 + * This program, aufs is free software; you can redistribute it and/or modify
2839 + * it under the terms of the GNU General Public License as published by
2840 + * the Free Software Foundation; either version 2 of the License, or
2841 + * (at your option) any later version.
2843 + * This program is distributed in the hope that it will be useful,
2844 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2845 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2846 + * GNU General Public License for more details.
2848 + * You should have received a copy of the GNU General Public License
2849 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2853 + * all header files
2861 +#define AuStub(type, name, body, ...) \
2862 + static inline type name(__VA_ARGS__) { body; }
2864 +#define AuStubVoid(name, ...) \
2865 + AuStub(void, name, , __VA_ARGS__)
2866 +#define AuStubInt0(name, ...) \
2867 + AuStub(int, name, return 0, __VA_ARGS__)
2871 +#include "branch.h"
2874 +#include "dbgaufs.h"
2875 +#include "dentry.h"
2877 +#include "dirren.h"
2880 +#include "fstype.h"
2885 +#include "module.h"
2889 +#include "sysaufs.h"
2894 +#endif /* __KERNEL__ */
2895 +#endif /* __AUFS_H__ */
2896 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2897 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
2898 +++ linux/fs/aufs/branch.c 2022-03-21 14:49:05.719966344 +0100
2900 +// SPDX-License-Identifier: GPL-2.0
2902 + * Copyright (C) 2005-2021 Junjiro R. Okajima
2904 + * This program, aufs is free software; you can redistribute it and/or modify
2905 + * it under the terms of the GNU General Public License as published by
2906 + * the Free Software Foundation; either version 2 of the License, or
2907 + * (at your option) any later version.
2909 + * This program is distributed in the hope that it will be useful,
2910 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2911 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2912 + * GNU General Public License for more details.
2914 + * You should have received a copy of the GNU General Public License
2915 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2919 + * branch management
2922 +#include <linux/compat.h>
2923 +#include <linux/statfs.h>
2927 + * free a single branch
2929 +static void au_br_do_free(struct au_branch *br)
2932 + struct au_wbr *wbr;
2933 + struct au_dykey **key;
2935 + au_hnotify_fin_br(br);
2936 + /* always, regardless the mount option */
2937 + au_dr_hino_free(&br->br_dirren);
2940 + AuLCntZero(au_lcnt_read(&br->br_nfiles, /*do_rev*/0));
2941 + au_lcnt_fin(&br->br_nfiles, /*do_sync*/0);
2942 + AuLCntZero(au_lcnt_read(&br->br_count, /*do_rev*/0));
2943 + au_lcnt_fin(&br->br_count, /*do_sync*/0);
2947 + for (i = 0; i < AuBrWh_Last; i++)
2948 + dput(wbr->wbr_wh[i]);
2949 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2950 + AuRwDestroy(&wbr->wbr_wh_rwsem);
2953 + if (br->br_fhsm) {
2954 + au_br_fhsm_fin(br->br_fhsm);
2955 + au_kfree_try_rcu(br->br_fhsm);
2958 + key = br->br_dykey;
2959 + for (i = 0; i < AuBrDynOp; i++, key++)
2965 + /* recursive lock, s_umount of branch's */
2966 + /* synchronize_rcu(); */ /* why? */
2968 + path_put(&br->br_path);
2970 + au_kfree_rcu(wbr);
2971 + au_lcnt_wait_for_fin(&br->br_nfiles);
2972 + au_lcnt_wait_for_fin(&br->br_count);
2973 + /* I don't know why, but percpu_refcount requires this */
2974 + /* synchronize_rcu(); */
2979 + * frees all branches
2981 +void au_br_free(struct au_sbinfo *sbinfo)
2983 + aufs_bindex_t bmax;
2984 + struct au_branch **br;
2986 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2988 + bmax = sbinfo->si_bbot + 1;
2989 + br = sbinfo->si_branch;
2991 + au_br_do_free(*br++);
2995 + * find the index of a branch which is specified by @br_id.
2997 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2999 + aufs_bindex_t bindex, bbot;
3001 + bbot = au_sbbot(sb);
3002 + for (bindex = 0; bindex <= bbot; bindex++)
3003 + if (au_sbr_id(sb, bindex) == br_id)
3008 +/* ---------------------------------------------------------------------- */
3014 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
3015 + struct dentry *h_root)
3017 + if (unlikely(h_adding == h_root
3018 + || au_test_loopback_overlap(sb, h_adding)))
3020 + if (h_adding->d_sb != h_root->d_sb)
3022 + return au_test_subdir(h_adding, h_root)
3023 + || au_test_subdir(h_root, h_adding);
3027 + * returns a newly allocated branch. @new_nbranch is a number of branches
3028 + * after adding a branch.
3030 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
3033 + struct au_branch *add_branch;
3034 + struct dentry *root;
3035 + struct inode *inode;
3039 + add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
3040 + if (unlikely(!add_branch))
3042 + add_branch->br_xino = au_xino_alloc(/*nfile*/1);
3043 + if (unlikely(!add_branch->br_xino))
3045 + err = au_hnotify_init_br(add_branch, perm);
3046 + if (unlikely(err))
3049 + if (au_br_writable(perm)) {
3050 + /* may be freed separately at changing the branch permission */
3051 + add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
3053 + if (unlikely(!add_branch->br_wbr))
3057 + if (au_br_fhsm(perm)) {
3058 + err = au_fhsm_br_alloc(add_branch);
3059 + if (unlikely(err))
3063 + root = sb->s_root;
3064 + err = au_sbr_realloc(au_sbi(sb), new_nbranch, /*may_shrink*/0);
3066 + err = au_di_realloc(au_di(root), new_nbranch, /*may_shrink*/0);
3068 + inode = d_inode(root);
3069 + err = au_hinode_realloc(au_ii(inode), new_nbranch,
3073 + return add_branch; /* success */
3076 + au_kfree_rcu(add_branch->br_wbr);
3078 + au_hnotify_fin_br(add_branch);
3080 + au_xino_put(add_branch);
3082 + au_kfree_rcu(add_branch);
3084 + return ERR_PTR(err);
3088 + * test if the branch permission is legal or not.
3090 +static int test_br(struct inode *inode, int brperm, char *path)
3094 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
3099 + pr_err("write permission for readonly mount or inode, %s\n", path);
3107 + * 0: success, the caller will add it
3108 + * plus: success, it is already unified, the caller should ignore it
3111 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
3114 + aufs_bindex_t bbot, bindex;
3115 + struct dentry *root, *h_dentry;
3116 + struct inode *inode, *h_inode;
3118 + root = sb->s_root;
3119 + bbot = au_sbbot(sb);
3120 + if (unlikely(bbot >= 0
3121 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
3125 + pr_err("%s duplicated\n", add->pathname);
3130 + err = -ENOSPC; /* -E2BIG; */
3131 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
3132 + || AUFS_BRANCH_MAX - 1 <= bbot)) {
3133 + pr_err("number of branches exceeded %s\n", add->pathname);
3138 + if (unlikely(add->bindex < 0 || bbot + 1 < add->bindex)) {
3139 + pr_err("bad index %d\n", add->bindex);
3143 + inode = d_inode(add->path.dentry);
3145 + if (unlikely(!inode->i_nlink)) {
3146 + pr_err("no existence %s\n", add->pathname);
3151 + if (unlikely(inode->i_sb == sb)) {
3152 + pr_err("%s must be outside\n", add->pathname);
3156 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
3157 + pr_err("unsupported filesystem, %s (%s)\n",
3158 + add->pathname, au_sbtype(inode->i_sb));
3162 + if (unlikely(inode->i_sb->s_stack_depth)) {
3163 + pr_err("already stacked, %s (%s)\n",
3164 + add->pathname, au_sbtype(inode->i_sb));
3168 + err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3169 + if (unlikely(err))
3173 + return 0; /* success */
3176 + for (bindex = 0; bindex <= bbot; bindex++)
3177 + if (unlikely(test_overlap(sb, add->path.dentry,
3178 + au_h_dptr(root, bindex)))) {
3179 + pr_err("%s is overlapped\n", add->pathname);
3184 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3185 + h_dentry = au_h_dptr(root, 0);
3186 + h_inode = d_inode(h_dentry);
3187 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3188 + || !uid_eq(h_inode->i_uid, inode->i_uid)
3189 + || !gid_eq(h_inode->i_gid, inode->i_gid))
3190 + pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3192 + i_uid_read(inode), i_gid_read(inode),
3193 + (inode->i_mode & S_IALLUGO),
3194 + i_uid_read(h_inode), i_gid_read(h_inode),
3195 + (h_inode->i_mode & S_IALLUGO));
3203 + * initialize or clean the whiteouts for an adding branch
3205 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3208 + int err, old_perm;
3209 + aufs_bindex_t bindex;
3210 + struct inode *h_inode;
3211 + struct au_wbr *wbr;
3212 + struct au_hinode *hdir;
3213 + struct dentry *h_dentry;
3215 + err = vfsub_mnt_want_write(au_br_mnt(br));
3216 + if (unlikely(err))
3220 + old_perm = br->br_perm;
3221 + br->br_perm = new_perm;
3224 + bindex = au_br_index(sb, br->br_id);
3225 + if (0 <= bindex) {
3226 + hdir = au_hi(d_inode(sb->s_root), bindex);
3227 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
3229 + h_dentry = au_br_dentry(br);
3230 + h_inode = d_inode(h_dentry);
3231 + inode_lock_nested(h_inode, AuLsc_I_PARENT);
3234 + err = au_wh_init(br, sb);
3236 + wbr_wh_write_lock(wbr);
3237 + err = au_wh_init(br, sb);
3238 + wbr_wh_write_unlock(wbr);
3241 + au_hn_inode_unlock(hdir);
3243 + inode_unlock(h_inode);
3244 + vfsub_mnt_drop_write(au_br_mnt(br));
3245 + br->br_perm = old_perm;
3247 + if (!err && wbr && !au_br_writable(new_perm)) {
3248 + au_kfree_rcu(wbr);
3249 + br->br_wbr = NULL;
3256 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3260 + struct kstatfs kst;
3261 + struct au_wbr *wbr;
3264 + au_rw_init(&wbr->wbr_wh_rwsem);
3265 + atomic_set(&wbr->wbr_wh_running, 0);
3268 + * a limit for rmdir/rename a dir
3269 + * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3271 + err = vfs_statfs(&br->br_path, &kst);
3272 + if (unlikely(err))
3275 + if (kst.f_namelen >= NAME_MAX)
3276 + err = au_br_init_wh(sb, br, perm);
3278 + pr_err("%pd(%s), unsupported namelen %ld\n",
3280 + au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3286 +/* initialize a new branch */
3287 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3288 + struct au_opt_add *add)
3291 + struct au_branch *brbase;
3293 + struct inode *h_inode;
3296 + br->br_perm = add->perm;
3297 + br->br_path = add->path; /* set first, path_get() later */
3298 + spin_lock_init(&br->br_dykey_lock);
3299 + au_lcnt_init(&br->br_nfiles, /*release*/NULL);
3300 + au_lcnt_init(&br->br_count, /*release*/NULL);
3301 + br->br_id = au_new_br_id(sb);
3302 + AuDebugOn(br->br_id < 0);
3304 + /* always, regardless the given option */
3305 + err = au_dr_br_init(sb, br, &add->path);
3306 + if (unlikely(err))
3309 + if (au_br_writable(add->perm)) {
3310 + err = au_wbr_init(br, sb, add->perm);
3311 + if (unlikely(err))
3315 + if (au_opt_test(au_mntflags(sb), XINO)) {
3316 + brbase = au_sbr(sb, 0);
3317 + xf = au_xino_file(brbase->br_xino, /*idx*/-1);
3319 + h_inode = d_inode(add->path.dentry);
3320 + err = au_xino_init_br(sb, br, h_inode->i_ino, &xf->f_path);
3321 + if (unlikely(err)) {
3322 + AuDebugOn(au_xino_file(br->br_xino, /*idx*/-1));
3327 + sysaufs_br_init(br);
3328 + path_get(&br->br_path);
3329 + goto out; /* success */
3332 + memset(&br->br_path, 0, sizeof(br->br_path));
3337 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3338 + struct au_branch *br, aufs_bindex_t bbot,
3339 + aufs_bindex_t amount)
3341 + struct au_branch **brp;
3343 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3345 + brp = sbinfo->si_branch + bindex;
3346 + memmove(brp + 1, brp, sizeof(*brp) * amount);
3348 + sbinfo->si_bbot++;
3349 + if (unlikely(bbot < 0))
3350 + sbinfo->si_bbot = 0;
3353 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3354 + aufs_bindex_t bbot, aufs_bindex_t amount)
3356 + struct au_hdentry *hdp;
3358 + AuRwMustWriteLock(&dinfo->di_rwsem);
3360 + hdp = au_hdentry(dinfo, bindex);
3361 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3362 + au_h_dentry_init(hdp);
3364 + if (unlikely(bbot < 0))
3365 + dinfo->di_btop = 0;
3368 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3369 + aufs_bindex_t bbot, aufs_bindex_t amount)
3371 + struct au_hinode *hip;
3373 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3375 + hip = au_hinode(iinfo, bindex);
3376 + memmove(hip + 1, hip, sizeof(*hip) * amount);
3377 + au_hinode_init(hip);
3379 + if (unlikely(bbot < 0))
3380 + iinfo->ii_btop = 0;
3383 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3384 + aufs_bindex_t bindex)
3386 + struct dentry *root, *h_dentry;
3387 + struct inode *root_inode, *h_inode;
3388 + aufs_bindex_t bbot, amount;
3390 + root = sb->s_root;
3391 + root_inode = d_inode(root);
3392 + bbot = au_sbbot(sb);
3393 + amount = bbot + 1 - bindex;
3394 + h_dentry = au_br_dentry(br);
3395 + au_sbilist_lock();
3396 + au_br_do_add_brp(au_sbi(sb), bindex, br, bbot, amount);
3397 + au_br_do_add_hdp(au_di(root), bindex, bbot, amount);
3398 + au_br_do_add_hip(au_ii(root_inode), bindex, bbot, amount);
3399 + au_set_h_dptr(root, bindex, dget(h_dentry));
3400 + h_inode = d_inode(h_dentry);
3401 + au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3402 + au_sbilist_unlock();
3405 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3408 + aufs_bindex_t bbot, add_bindex;
3409 + struct dentry *root, *h_dentry;
3410 + struct inode *root_inode;
3411 + struct au_branch *add_branch;
3413 + root = sb->s_root;
3414 + root_inode = d_inode(root);
3415 + IMustLock(root_inode);
3416 + IiMustWriteLock(root_inode);
3417 + err = test_add(sb, add, remount);
3418 + if (unlikely(err < 0))
3422 + goto out; /* success */
3425 + bbot = au_sbbot(sb);
3426 + add_branch = au_br_alloc(sb, bbot + 2, add->perm);
3427 + err = PTR_ERR(add_branch);
3428 + if (IS_ERR(add_branch))
3431 + err = au_br_init(add_branch, sb, add);
3432 + if (unlikely(err)) {
3433 + au_br_do_free(add_branch);
3437 + add_bindex = add->bindex;
3438 + sysaufs_brs_del(sb, add_bindex); /* remove successors */
3439 + au_br_do_add(sb, add_branch, add_bindex);
3440 + sysaufs_brs_add(sb, add_bindex); /* append successors */
3441 + dbgaufs_brs_add(sb, add_bindex, /*topdown*/0); /* rename successors */
3443 + h_dentry = add->path.dentry;
3444 + if (!add_bindex) {
3445 + au_cpup_attr_all(root_inode, /*force*/1);
3446 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3448 + au_add_nlink(root_inode, d_inode(h_dentry));
3454 +/* ---------------------------------------------------------------------- */
3456 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3457 + unsigned long long max __maybe_unused,
3460 + unsigned long long n;
3461 + struct file **p, *f;
3462 + struct hlist_bl_head *files;
3463 + struct hlist_bl_node *pos;
3464 + struct au_finfo *finfo;
3468 + files = &au_sbi(sb)->si_files;
3469 + hlist_bl_lock(files);
3470 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
3471 + f = finfo->fi_file;
3473 + && !special_file(file_inode(f)->i_mode)) {
3477 + AuDebugOn(n > max);
3480 + hlist_bl_unlock(files);
3485 +static struct file **au_farray_alloc(struct super_block *sb,
3486 + unsigned long long *max)
3488 + struct au_sbinfo *sbi;
3491 + *max = au_lcnt_read(&sbi->si_nfiles, /*do_rev*/1);
3492 + return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3495 +static void au_farray_free(struct file **a, unsigned long long max)
3497 + unsigned long long ull;
3499 + for (ull = 0; ull < max; ull++)
3505 +/* ---------------------------------------------------------------------- */
3511 +/* to show the line number, do not make it inlined function */
3512 +#define AuVerbose(do_info, fmt, ...) do { \
3514 + pr_info(fmt, ##__VA_ARGS__); \
3517 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t btop,
3518 + aufs_bindex_t bbot)
3520 + return (inode && !S_ISDIR(inode->i_mode)) || btop == bbot;
3523 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t btop,
3524 + aufs_bindex_t bbot)
3526 + return au_test_ibusy(d_inode(dentry), btop, bbot);
3530 + * test if the branch is deletable or not.
3532 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3533 + unsigned int sigen, const unsigned int verbose)
3535 + int err, i, j, ndentry;
3536 + aufs_bindex_t btop, bbot;
3537 + struct au_dcsub_pages dpages;
3538 + struct au_dpage *dpage;
3541 + err = au_dpages_init(&dpages, GFP_NOFS);
3542 + if (unlikely(err))
3544 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
3545 + if (unlikely(err))
3548 + for (i = 0; !err && i < dpages.ndpage; i++) {
3549 + dpage = dpages.dpages + i;
3550 + ndentry = dpage->ndentry;
3551 + for (j = 0; !err && j < ndentry; j++) {
3552 + d = dpage->dentries[j];
3553 + AuDebugOn(au_dcount(d) <= 0);
3554 + if (!au_digen_test(d, sigen)) {
3555 + di_read_lock_child(d, AuLock_IR);
3556 + if (unlikely(au_dbrange_test(d))) {
3557 + di_read_unlock(d, AuLock_IR);
3561 + di_write_lock_child(d);
3562 + if (unlikely(au_dbrange_test(d))) {
3563 + di_write_unlock(d);
3566 + err = au_reval_dpath(d, sigen);
3568 + di_downgrade_lock(d, AuLock_IR);
3570 + di_write_unlock(d);
3575 + /* AuDbgDentry(d); */
3576 + btop = au_dbtop(d);
3577 + bbot = au_dbbot(d);
3578 + if (btop <= bindex
3580 + && au_h_dptr(d, bindex)
3581 + && au_test_dbusy(d, btop, bbot)) {
3583 + AuVerbose(verbose, "busy %pd\n", d);
3586 + di_read_unlock(d, AuLock_IR);
3591 + au_dpages_free(&dpages);
3596 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3597 + unsigned int sigen, const unsigned int verbose)
3600 + unsigned long long max, ull;
3601 + struct inode *i, **array;
3602 + aufs_bindex_t btop, bbot;
3604 + array = au_iarray_alloc(sb, &max);
3605 + err = PTR_ERR(array);
3606 + if (IS_ERR(array))
3610 + AuDbg("b%d\n", bindex);
3611 + for (ull = 0; !err && ull < max; ull++) {
3615 + if (i->i_ino == AUFS_ROOT_INO)
3618 + /* AuDbgInode(i); */
3619 + if (au_iigen(i, NULL) == sigen)
3620 + ii_read_lock_child(i);
3622 + ii_write_lock_child(i);
3623 + err = au_refresh_hinode_self(i);
3626 + ii_downgrade_lock(i);
3628 + ii_write_unlock(i);
3633 + btop = au_ibtop(i);
3634 + bbot = au_ibbot(i);
3635 + if (btop <= bindex
3637 + && au_h_iptr(i, bindex)
3638 + && au_test_ibusy(i, btop, bbot)) {
3640 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3643 + ii_read_unlock(i);
3645 + au_iarray_free(array, max);
3651 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3652 + const unsigned int verbose)
3655 + unsigned int sigen;
3657 + sigen = au_sigen(root->d_sb);
3658 + DiMustNoWaiters(root);
3659 + IiMustNoWaiters(d_inode(root));
3660 + di_write_unlock(root);
3661 + err = test_dentry_busy(root, bindex, sigen, verbose);
3663 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3664 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3669 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3670 + struct file **to_free, int *idx)
3673 + unsigned char matched, root;
3674 + aufs_bindex_t bindex, bbot;
3675 + struct au_fidir *fidir;
3676 + struct au_hfile *hfile;
3679 + root = IS_ROOT(file->f_path.dentry);
3682 + to_free[*idx] = file;
3688 + fidir = au_fi(file)->fi_hdir;
3689 + AuDebugOn(!fidir);
3690 + bbot = au_fbbot_dir(file);
3691 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++) {
3692 + hfile = fidir->fd_hfile + bindex;
3693 + if (!hfile->hf_file)
3696 + if (hfile->hf_br->br_id == br_id) {
3708 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3709 + struct file **to_free, int opened)
3712 + unsigned long long ull, max;
3713 + aufs_bindex_t btop;
3714 + struct file *file, **array;
3715 + struct dentry *root;
3716 + struct au_hfile *hfile;
3718 + array = au_farray_alloc(sb, &max);
3719 + err = PTR_ERR(array);
3720 + if (IS_ERR(array))
3725 + root = sb->s_root;
3726 + di_write_unlock(root);
3727 + for (ull = 0; ull < max; ull++) {
3728 + file = array[ull];
3729 + if (unlikely(!file))
3732 + /* AuDbg("%pD\n", file); */
3733 + fi_read_lock(file);
3734 + btop = au_fbtop(file);
3735 + if (!d_is_dir(file->f_path.dentry)) {
3736 + hfile = &au_fi(file)->fi_htop;
3737 + if (hfile->hf_br->br_id == br_id)
3740 + err = test_dir_busy(file, br_id, to_free, &idx);
3741 + fi_read_unlock(file);
3742 + if (unlikely(err))
3745 + di_write_lock_child(root);
3746 + au_farray_free(array, max);
3747 + AuDebugOn(idx > opened);
3753 +static void br_del_file(struct file **to_free, unsigned long long opened,
3754 + aufs_bindex_t br_id)
3756 + unsigned long long ull;
3757 + aufs_bindex_t bindex, btop, bbot, bfound;
3758 + struct file *file;
3759 + struct au_fidir *fidir;
3760 + struct au_hfile *hfile;
3762 + for (ull = 0; ull < opened; ull++) {
3763 + file = to_free[ull];
3764 + if (unlikely(!file))
3767 + /* AuDbg("%pD\n", file); */
3768 + AuDebugOn(!d_is_dir(file->f_path.dentry));
3770 + fidir = au_fi(file)->fi_hdir;
3771 + AuDebugOn(!fidir);
3772 + fi_write_lock(file);
3773 + btop = au_fbtop(file);
3774 + bbot = au_fbbot_dir(file);
3775 + for (bindex = btop; bindex <= bbot; bindex++) {
3776 + hfile = fidir->fd_hfile + bindex;
3777 + if (!hfile->hf_file)
3780 + if (hfile->hf_br->br_id == br_id) {
3785 + AuDebugOn(bfound < 0);
3786 + au_set_h_fptr(file, bfound, NULL);
3787 + if (bfound == btop) {
3788 + for (btop++; btop <= bbot; btop++)
3789 + if (au_hf_dir(file, btop)) {
3790 + au_set_fbtop(file, btop);
3794 + fi_write_unlock(file);
3798 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3799 + const aufs_bindex_t bindex,
3800 + const aufs_bindex_t bbot)
3802 + struct au_branch **brp, **p;
3804 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3806 + brp = sbinfo->si_branch + bindex;
3807 + if (bindex < bbot)
3808 + memmove(brp, brp + 1, sizeof(*brp) * (bbot - bindex));
3809 + sbinfo->si_branch[0 + bbot] = NULL;
3810 + sbinfo->si_bbot--;
3812 + p = au_krealloc(sbinfo->si_branch, sizeof(*p) * bbot, AuGFP_SBILIST,
3815 + sbinfo->si_branch = p;
3816 + /* harmless error */
3819 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3820 + const aufs_bindex_t bbot)
3822 + struct au_hdentry *hdp, *p;
3824 + AuRwMustWriteLock(&dinfo->di_rwsem);
3826 + hdp = au_hdentry(dinfo, bindex);
3827 + if (bindex < bbot)
3828 + memmove(hdp, hdp + 1, sizeof(*hdp) * (bbot - bindex));
3829 + /* au_h_dentry_init(au_hdentry(dinfo, bbot); */
3832 + p = au_krealloc(dinfo->di_hdentry, sizeof(*p) * bbot, AuGFP_SBILIST,
3835 + dinfo->di_hdentry = p;
3836 + /* harmless error */
3839 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3840 + const aufs_bindex_t bbot)
3842 + struct au_hinode *hip, *p;
3844 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3846 + hip = au_hinode(iinfo, bindex);
3847 + if (bindex < bbot)
3848 + memmove(hip, hip + 1, sizeof(*hip) * (bbot - bindex));
3849 + /* au_hinode_init(au_hinode(iinfo, bbot)); */
3852 + p = au_krealloc(iinfo->ii_hinode, sizeof(*p) * bbot, AuGFP_SBILIST,
3855 + iinfo->ii_hinode = p;
3856 + /* harmless error */
3859 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3860 + struct au_branch *br)
3862 + aufs_bindex_t bbot;
3863 + struct au_sbinfo *sbinfo;
3864 + struct dentry *root, *h_root;
3865 + struct inode *inode, *h_inode;
3866 + struct au_hinode *hinode;
3868 + SiMustWriteLock(sb);
3870 + root = sb->s_root;
3871 + inode = d_inode(root);
3872 + sbinfo = au_sbi(sb);
3873 + bbot = sbinfo->si_bbot;
3875 + h_root = au_h_dptr(root, bindex);
3876 + hinode = au_hi(inode, bindex);
3877 + h_inode = au_igrab(hinode->hi_inode);
3880 + au_sbilist_lock();
3881 + au_br_do_del_brp(sbinfo, bindex, bbot);
3882 + au_br_do_del_hdp(au_di(root), bindex, bbot);
3883 + au_br_do_del_hip(au_ii(inode), bindex, bbot);
3884 + au_sbilist_unlock();
3886 + /* ignore an error */
3887 + au_dr_br_fin(sb, br); /* always, regardless the mount option */
3891 + au_br_do_free(br);
3894 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3895 + unsigned long long max, void *arg)
3900 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3903 + unsigned long long opened;
3904 + unsigned int mnt_flags;
3905 + aufs_bindex_t bindex, bbot, br_id;
3906 + unsigned char do_wh, verbose;
3907 + struct au_branch *br;
3908 + struct au_wbr *wbr;
3909 + struct dentry *root;
3910 + struct file **to_free;
3915 + root = sb->s_root;
3916 + bindex = au_find_dbindex(root, del->h_path.dentry);
3919 + goto out; /* success */
3921 + pr_err("%s no such branch\n", del->pathname);
3924 + AuDbg("bindex b%d\n", bindex);
3927 + mnt_flags = au_mntflags(sb);
3928 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3929 + bbot = au_sbbot(sb);
3930 + if (unlikely(!bbot)) {
3931 + AuVerbose(verbose, "no more branches left\n");
3935 + br = au_sbr(sb, bindex);
3936 + AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3937 + if (unlikely(au_lcnt_read(&br->br_count, /*do_rev*/1))) {
3938 + AuVerbose(verbose, "br %pd2 is busy now\n", del->h_path.dentry);
3942 + br_id = br->br_id;
3943 + opened = au_lcnt_read(&br->br_nfiles, /*do_rev*/1);
3944 + if (unlikely(opened)) {
3945 + to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3946 + err = PTR_ERR(to_free);
3947 + if (IS_ERR(to_free))
3950 + err = test_file_busy(sb, br_id, to_free, opened);
3951 + if (unlikely(err)) {
3952 + AuVerbose(verbose, "%llu file(s) opened\n", opened);
3958 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3960 + /* instead of WbrWhMustWriteLock(wbr) */
3961 + SiMustWriteLock(sb);
3962 + for (i = 0; i < AuBrWh_Last; i++) {
3963 + dput(wbr->wbr_wh[i]);
3964 + wbr->wbr_wh[i] = NULL;
3968 + err = test_children_busy(root, bindex, verbose);
3969 + if (unlikely(err)) {
3978 + * now we confirmed the branch is deletable.
3979 + * let's free the remaining opened dirs on the branch.
3981 + di_write_unlock(root);
3982 + br_del_file(to_free, opened, br_id);
3983 + di_write_lock_child(root);
3986 + sysaufs_brs_del(sb, bindex); /* remove successors */
3987 + dbgaufs_xino_del(br); /* remove one */
3988 + au_br_do_del(sb, bindex, br);
3989 + sysaufs_brs_add(sb, bindex); /* append successors */
3990 + dbgaufs_brs_add(sb, bindex, /*topdown*/1); /* rename successors */
3993 + au_cpup_attr_all(d_inode(root), /*force*/1);
3994 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3996 + au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
3997 + if (au_opt_test(mnt_flags, PLINK))
3998 + au_plink_half_refresh(sb, br_id);
4000 + goto out; /* success */
4004 + rerr = au_br_init_wh(sb, br, br->br_perm);
4006 + pr_warn("failed re-creating base whiteout, %s. (%d)\n",
4007 + del->pathname, rerr);
4010 + au_farray_free(to_free, opened);
4014 +/* ---------------------------------------------------------------------- */
4016 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
4019 + aufs_bindex_t btop, bbot;
4020 + struct aufs_ibusy ibusy;
4021 + struct inode *inode, *h_inode;
4024 + if (unlikely(!capable(CAP_SYS_ADMIN)))
4027 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
4029 + /* VERIFY_WRITE */
4030 + err = !access_ok(&arg->h_ino, sizeof(arg->h_ino));
4031 + if (unlikely(err)) {
4038 + si_read_lock(sb, AuLock_FLUSH);
4039 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbbot(sb)))
4043 + ibusy.h_ino = 0; /* invalid */
4044 + inode = ilookup(sb, ibusy.ino);
4046 + || inode->i_ino == AUFS_ROOT_INO
4047 + || au_is_bad_inode(inode))
4050 + ii_read_lock_child(inode);
4051 + btop = au_ibtop(inode);
4052 + bbot = au_ibbot(inode);
4053 + if (btop <= ibusy.bindex && ibusy.bindex <= bbot) {
4054 + h_inode = au_h_iptr(inode, ibusy.bindex);
4055 + if (h_inode && au_test_ibusy(inode, btop, bbot))
4056 + ibusy.h_ino = h_inode->i_ino;
4058 + ii_read_unlock(inode);
4062 + si_read_unlock(sb);
4064 + err = __put_user(ibusy.h_ino, &arg->h_ino);
4065 + if (unlikely(err)) {
4074 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
4076 + return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
4079 +#ifdef CONFIG_COMPAT
4080 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
4082 + return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
4086 +/* ---------------------------------------------------------------------- */
4089 + * change a branch permission
4092 +static void au_warn_ima(void)
4095 + /* since it doesn't support mark_files_ro() */
4096 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
4100 +static int do_need_sigen_inc(int a, int b)
4102 + return au_br_whable(a) && !au_br_whable(b);
4105 +static int need_sigen_inc(int old, int new)
4107 + return do_need_sigen_inc(old, new)
4108 + || do_need_sigen_inc(new, old);
4111 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
4114 + unsigned int mnt_flags;
4115 + unsigned long long ull, max;
4116 + aufs_bindex_t br_id;
4117 + unsigned char verbose, writer;
4118 + struct file *file, *hf, **array;
4119 + struct au_hfile *hfile;
4120 + struct inode *h_inode;
4122 + mnt_flags = au_mntflags(sb);
4123 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4125 + array = au_farray_alloc(sb, &max);
4126 + err = PTR_ERR(array);
4127 + if (IS_ERR(array))
4131 + br_id = au_sbr_id(sb, bindex);
4132 + for (ull = 0; ull < max; ull++) {
4133 + file = array[ull];
4134 + if (unlikely(!file))
4137 + /* AuDbg("%pD\n", file); */
4138 + fi_read_lock(file);
4139 + if (unlikely(au_test_mmapped(file))) {
4141 + AuVerbose(verbose, "mmapped %pD\n", file);
4143 + FiMustNoWaiters(file);
4144 + fi_read_unlock(file);
4148 + hfile = &au_fi(file)->fi_htop;
4149 + hf = hfile->hf_file;
4150 + if (!d_is_reg(file->f_path.dentry)
4151 + || !(file->f_mode & FMODE_WRITE)
4152 + || hfile->hf_br->br_id != br_id
4153 + || !(hf->f_mode & FMODE_WRITE))
4154 + array[ull] = NULL;
4160 + FiMustNoWaiters(file);
4161 + fi_read_unlock(file);
4169 + for (ull = 0; ull < max; ull++) {
4170 + file = array[ull];
4174 + /* todo: already flushed? */
4176 + * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4177 + * approach which resets f_mode and calls mnt_drop_write() and
4178 + * file_release_write() for each file, because the branch
4179 + * attribute in aufs world is totally different from the native
4182 + /* fi_read_lock(file); */
4183 + hfile = &au_fi(file)->fi_htop;
4184 + hf = hfile->hf_file;
4185 + /* fi_read_unlock(file); */
4186 + spin_lock(&hf->f_lock);
4187 + writer = !!(hf->f_mode & FMODE_WRITER);
4188 + hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4189 + spin_unlock(&hf->f_lock);
4191 + h_inode = file_inode(hf);
4192 + if (hf->f_mode & FMODE_READ)
4193 + i_readcount_inc(h_inode);
4194 + put_write_access(h_inode);
4195 + __mnt_drop_write(hf->f_path.mnt);
4200 + au_farray_free(array, max);
4206 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4210 + aufs_bindex_t bindex;
4211 + struct dentry *root;
4212 + struct au_branch *br;
4213 + struct au_br_fhsm *bf;
4215 + root = sb->s_root;
4216 + bindex = au_find_dbindex(root, mod->h_root);
4219 + return 0; /* success */
4221 + pr_err("%s no such branch\n", mod->path);
4224 + AuDbg("bindex b%d\n", bindex);
4226 + err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4227 + if (unlikely(err))
4230 + br = au_sbr(sb, bindex);
4231 + AuDebugOn(mod->h_root != au_br_dentry(br));
4232 + if (br->br_perm == mod->perm)
4233 + return 0; /* success */
4235 + /* pre-allocate for non-fhsm --> fhsm */
4237 + if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4238 + err = au_fhsm_br_alloc(br);
4239 + if (unlikely(err))
4242 + br->br_fhsm = NULL;
4245 + if (au_br_writable(br->br_perm)) {
4246 + /* remove whiteout base */
4247 + err = au_br_init_wh(sb, br, mod->perm);
4248 + if (unlikely(err))
4251 + if (!au_br_writable(mod->perm)) {
4252 + /* rw --> ro, file might be mmapped */
4253 + DiMustNoWaiters(root);
4254 + IiMustNoWaiters(d_inode(root));
4255 + di_write_unlock(root);
4256 + err = au_br_mod_files_ro(sb, bindex);
4257 + /* aufs_write_lock() calls ..._child() */
4258 + di_write_lock_child(root);
4260 + if (unlikely(err)) {
4262 + br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4265 + rerr = au_wbr_init(br, sb, br->br_perm);
4266 + if (unlikely(rerr)) {
4267 + AuIOErr("nested error %d (%d)\n",
4269 + br->br_perm = mod->perm;
4273 + } else if (au_br_writable(mod->perm)) {
4276 + br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4278 + err = au_wbr_init(br, sb, mod->perm);
4279 + if (unlikely(err)) {
4280 + au_kfree_rcu(br->br_wbr);
4281 + br->br_wbr = NULL;
4285 + if (unlikely(err))
4288 + if (au_br_fhsm(br->br_perm)) {
4289 + if (!au_br_fhsm(mod->perm)) {
4290 + /* fhsm --> non-fhsm */
4291 + au_br_fhsm_fin(br->br_fhsm);
4292 + au_kfree_rcu(br->br_fhsm);
4293 + br->br_fhsm = NULL;
4295 + } else if (au_br_fhsm(mod->perm))
4296 + /* non-fhsm --> fhsm */
4299 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4300 + br->br_perm = mod->perm;
4301 + goto out; /* success */
4304 + au_kfree_try_rcu(bf);
4310 +/* ---------------------------------------------------------------------- */
4312 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4315 + struct kstatfs kstfs;
4317 + err = vfs_statfs(&br->br_path, &kstfs);
4319 + stfs->f_blocks = kstfs.f_blocks;
4320 + stfs->f_bavail = kstfs.f_bavail;
4321 + stfs->f_files = kstfs.f_files;
4322 + stfs->f_ffree = kstfs.f_ffree;
4327 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4328 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
4329 +++ linux/fs/aufs/branch.h 2022-03-21 14:49:05.719966344 +0100
4331 +/* SPDX-License-Identifier: GPL-2.0 */
4333 + * Copyright (C) 2005-2021 Junjiro R. Okajima
4335 + * This program, aufs is free software; you can redistribute it and/or modify
4336 + * it under the terms of the GNU General Public License as published by
4337 + * the Free Software Foundation; either version 2 of the License, or
4338 + * (at your option) any later version.
4340 + * This program is distributed in the hope that it will be useful,
4341 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4342 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4343 + * GNU General Public License for more details.
4345 + * You should have received a copy of the GNU General Public License
4346 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4350 + * branch filesystems and xino for them
4353 +#ifndef __AUFS_BRANCH_H__
4354 +#define __AUFS_BRANCH_H__
4358 +#include <linux/mount.h>
4359 +#include "dirren.h"
4365 +/* ---------------------------------------------------------------------- */
4369 + struct file **xi_file;
4370 + unsigned int xi_nfile;
4376 + /* reserved for future use */
4377 + /* unsigned long *bitmap; */
4378 + wait_queue_head_t wqh;
4381 + struct mutex xi_mtx; /* protects xi_file array */
4382 + struct hlist_bl_head xi_writing;
4384 + atomic_t xi_truncating;
4386 + struct kref xi_kref;
4389 +/* File-based Hierarchical Storage Management */
4390 +struct au_br_fhsm {
4391 +#ifdef CONFIG_AUFS_FHSM
4392 + struct mutex bf_lock;
4393 + unsigned long bf_jiffy;
4394 + struct aufs_stfs bf_stfs;
4399 +/* members for writable branch only */
4400 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4402 + struct au_rwsem wbr_wh_rwsem;
4403 + struct dentry *wbr_wh[AuBrWh_Last];
4404 + atomic_t wbr_wh_running;
4405 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4406 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4407 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4410 + unsigned long long wbr_bytes;
4413 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4414 +#define AuBrDynOp (AuDyLast * 4)
4416 +#ifdef CONFIG_AUFS_HFSNOTIFY
4417 +/* support for asynchronous destruction */
4418 +struct au_br_hfsnotify {
4419 + struct fsnotify_group *hfsn_group;
4423 +/* sysfs entries */
4424 +struct au_brsysfs {
4426 + struct attribute attr;
4435 +/* protected by superblock rwsem */
4437 + struct au_xino *br_xino;
4439 + aufs_bindex_t br_id;
4442 + struct path br_path;
4443 + spinlock_t br_dykey_lock;
4444 + struct au_dykey *br_dykey[AuBrDynOp];
4445 + au_lcnt_t br_nfiles; /* opened files */
4446 + au_lcnt_t br_count; /* in-use for other */
4448 + struct au_wbr *br_wbr;
4449 + struct au_br_fhsm *br_fhsm;
4451 +#ifdef CONFIG_AUFS_HFSNOTIFY
4452 + struct au_br_hfsnotify *br_hfsn;
4455 +#ifdef CONFIG_SYSFS
4456 + /* entries under sysfs per mount-point */
4457 + struct au_brsysfs br_sysfs[AuBrSysfs_Last];
4460 +#ifdef CONFIG_DEBUG_FS
4461 + struct dentry *br_dbgaufs; /* xino */
4464 + struct au_dr_br br_dirren;
4467 +/* ---------------------------------------------------------------------- */
4469 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4471 + return br->br_path.mnt;
4474 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4476 + return br->br_path.dentry;
4479 +static inline struct user_namespace *au_br_userns(struct au_branch *br)
4481 + return mnt_user_ns(br->br_path.mnt);
4484 +static inline struct super_block *au_br_sb(struct au_branch *br)
4486 + return au_br_mnt(br)->mnt_sb;
4489 +static inline int au_br_rdonly(struct au_branch *br)
4491 + return (sb_rdonly(au_br_sb(br))
4492 + || !au_br_writable(br->br_perm))
4496 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4498 +#ifdef CONFIG_AUFS_HNOTIFY
4499 + return !(brperm & AuBrPerm_RR);
4505 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4507 + int err, exec_flag;
4510 + exec_flag = oflag & __FMODE_EXEC;
4511 + if (unlikely(exec_flag && path_noexec(&br->br_path)))
4517 +static inline void au_xino_get(struct au_branch *br)
4519 + struct au_xino *xi;
4523 + kref_get(&xi->xi_kref);
4526 +static inline int au_xino_count(struct au_branch *br)
4529 + struct au_xino *xi;
4534 + v = kref_read(&xi->xi_kref);
4539 +/* ---------------------------------------------------------------------- */
4543 +void au_br_free(struct au_sbinfo *sinfo);
4544 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4546 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4548 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4549 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4550 +#ifdef CONFIG_COMPAT
4551 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4554 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4557 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4560 +static const loff_t au_loff_max = LLONG_MAX;
4562 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry);
4563 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
4565 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
4566 + struct file *copy_src);
4568 + struct au_xino *xi; /* switch between xino and xigen */
4570 + struct path *base;
4571 + struct file *copy_src;
4573 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew);
4575 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4577 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4579 +ssize_t xino_fread(struct file *file, void *buf, size_t size, loff_t *pos);
4580 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos);
4582 +int au_xib_trunc(struct super_block *sb);
4583 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin);
4585 +struct au_xino *au_xino_alloc(unsigned int nfile);
4586 +int au_xino_put(struct au_branch *br);
4587 +struct file *au_xino_file1(struct au_xino *xi);
4589 +struct au_opt_xino;
4590 +void au_xino_clr(struct super_block *sb);
4591 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount);
4592 +struct file *au_xino_def(struct super_block *sb);
4593 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4594 + struct path *base);
4596 +ino_t au_xino_new_ino(struct super_block *sb);
4597 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4599 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
4600 + ino_t h_ino, int idx);
4601 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4604 +int au_xino_path(struct seq_file *seq, struct file *file);
4606 +/* ---------------------------------------------------------------------- */
4608 +/* @idx is signed to accept -1 meaning the first file */
4609 +static inline struct file *au_xino_file(struct au_xino *xi, int idx)
4611 + struct file *file;
4618 + if (idx < xi->xi_nfile)
4619 + file = xi->xi_file[idx];
4621 + file = au_xino_file1(xi);
4627 +/* ---------------------------------------------------------------------- */
4629 +/* Superblock to branch */
4631 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4633 + return au_sbr(sb, bindex)->br_id;
4637 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4639 + return au_br_mnt(au_sbr(sb, bindex));
4643 +struct user_namespace *au_sbr_userns(struct super_block *sb, aufs_bindex_t bindex)
4645 + return au_br_userns(au_sbr(sb, bindex));
4649 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4651 + return au_br_sb(au_sbr(sb, bindex));
4654 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4656 + return au_sbr(sb, bindex)->br_perm;
4659 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4661 + return au_br_whable(au_sbr_perm(sb, bindex));
4664 +/* ---------------------------------------------------------------------- */
4666 +#define wbr_wh_read_lock(wbr) au_rw_read_lock(&(wbr)->wbr_wh_rwsem)
4667 +#define wbr_wh_write_lock(wbr) au_rw_write_lock(&(wbr)->wbr_wh_rwsem)
4668 +#define wbr_wh_read_trylock(wbr) au_rw_read_trylock(&(wbr)->wbr_wh_rwsem)
4669 +#define wbr_wh_write_trylock(wbr) au_rw_write_trylock(&(wbr)->wbr_wh_rwsem)
4671 +#define wbr_wh_read_trylock_nested(wbr) \
4672 + au_rw_read_trylock_nested(&(wbr)->wbr_wh_rwsem)
4673 +#define wbr_wh_write_trylock_nested(wbr) \
4674 + au_rw_write_trylock_nested(&(wbr)->wbr_wh_rwsem)
4677 +#define wbr_wh_read_unlock(wbr) au_rw_read_unlock(&(wbr)->wbr_wh_rwsem)
4678 +#define wbr_wh_write_unlock(wbr) au_rw_write_unlock(&(wbr)->wbr_wh_rwsem)
4679 +#define wbr_wh_downgrade_lock(wbr) au_rw_dgrade_lock(&(wbr)->wbr_wh_rwsem)
4681 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&(wbr)->wbr_wh_rwsem)
4682 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&(wbr)->wbr_wh_rwsem)
4683 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&(wbr)->wbr_wh_rwsem)
4685 +/* ---------------------------------------------------------------------- */
4687 +#ifdef CONFIG_AUFS_FHSM
4688 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4690 + mutex_init(&brfhsm->bf_lock);
4691 + brfhsm->bf_jiffy = 0;
4692 + brfhsm->bf_readable = 0;
4695 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4697 + mutex_destroy(&brfhsm->bf_lock);
4700 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4701 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4704 +#endif /* __KERNEL__ */
4705 +#endif /* __AUFS_BRANCH_H__ */
4706 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4707 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
4708 +++ linux/fs/aufs/conf.mk 2022-03-21 14:49:05.719966344 +0100
4710 +# SPDX-License-Identifier: GPL-2.0
4712 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4716 +AuConfStr += ${1}=${${1}}
4720 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4722 + HNOTIFY HFSNOTIFY \
4734 +$(foreach i, ${AuConfAll}, \
4735 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
4737 +AuConfName = ${obj}/conf.str
4738 +${AuConfName}.tmp: FORCE
4739 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4740 +${AuConfName}: ${AuConfName}.tmp
4741 + @diff -q $< $@ > /dev/null 2>&1 || { \
4742 + echo ' GEN ' $@; \
4746 +clean-files += ${AuConfName} ${AuConfName}.tmp
4747 +${obj}/sysfs.o: ${AuConfName}
4749 +-include ${srctree}/${src}/conf_priv.mk
4750 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4751 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
4752 +++ linux/fs/aufs/cpup.c 2022-03-21 14:49:05.719966344 +0100
4754 +// SPDX-License-Identifier: GPL-2.0
4756 + * Copyright (C) 2005-2021 Junjiro R. Okajima
4758 + * This program, aufs is free software; you can redistribute it and/or modify
4759 + * it under the terms of the GNU General Public License as published by
4760 + * the Free Software Foundation; either version 2 of the License, or
4761 + * (at your option) any later version.
4763 + * This program is distributed in the hope that it will be useful,
4764 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4765 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4766 + * GNU General Public License for more details.
4768 + * You should have received a copy of the GNU General Public License
4769 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4773 + * copy-up functions, see wbr_policy.c for copy-down
4776 +#include <linux/fs_stack.h>
4777 +#include <linux/mm.h>
4778 +#include <linux/task_work.h>
4781 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4783 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4784 + | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4786 + BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4788 + dst->i_flags |= iflags & ~mask;
4789 + if (au_test_fs_notime(dst->i_sb))
4790 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
4793 +void au_cpup_attr_timesizes(struct inode *inode)
4795 + struct inode *h_inode;
4797 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4798 + fsstack_copy_attr_times(inode, h_inode);
4799 + fsstack_copy_inode_size(inode, h_inode);
4802 +void au_cpup_attr_nlink(struct inode *inode, int force)
4804 + struct inode *h_inode;
4805 + struct super_block *sb;
4806 + aufs_bindex_t bindex, bbot;
4809 + bindex = au_ibtop(inode);
4810 + h_inode = au_h_iptr(inode, bindex);
4812 + && !S_ISDIR(h_inode->i_mode)
4813 + && au_opt_test(au_mntflags(sb), PLINK)
4814 + && au_plink_test(inode))
4818 + * 0 can happen in revalidating.
4819 + * h_inode->i_mutex may not be held here, but it is harmless since once
4820 + * i_nlink reaches 0, it will never become positive except O_TMPFILE
4822 + * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4823 + * the incorrect link count.
4825 + set_nlink(inode, h_inode->i_nlink);
4828 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4829 + * it may includes whplink directory.
4831 + if (S_ISDIR(h_inode->i_mode)) {
4832 + bbot = au_ibbot(inode);
4833 + for (bindex++; bindex <= bbot; bindex++) {
4834 + h_inode = au_h_iptr(inode, bindex);
4836 + au_add_nlink(inode, h_inode);
4841 +void au_cpup_attr_changeable(struct inode *inode)
4843 + struct inode *h_inode;
4845 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4846 + inode->i_mode = h_inode->i_mode;
4847 + inode->i_uid = h_inode->i_uid;
4848 + inode->i_gid = h_inode->i_gid;
4849 + au_cpup_attr_timesizes(inode);
4850 + au_cpup_attr_flags(inode, h_inode->i_flags);
4853 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4855 + struct au_iinfo *iinfo = au_ii(inode);
4857 + IiMustWriteLock(inode);
4859 + iinfo->ii_higen = h_inode->i_generation;
4860 + iinfo->ii_hsb1 = h_inode->i_sb;
4863 +void au_cpup_attr_all(struct inode *inode, int force)
4865 + struct inode *h_inode;
4867 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4868 + au_cpup_attr_changeable(inode);
4869 + if (inode->i_nlink > 0)
4870 + au_cpup_attr_nlink(inode, force);
4871 + inode->i_rdev = h_inode->i_rdev;
4872 + inode->i_blkbits = h_inode->i_blkbits;
4873 + au_cpup_igen(inode, h_inode);
4876 +/* ---------------------------------------------------------------------- */
4878 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4880 +/* keep the timestamps of the parent dir when cpup */
4881 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4882 + struct path *h_path)
4884 + struct inode *h_inode;
4886 + dt->dt_dentry = dentry;
4887 + dt->dt_h_path = *h_path;
4888 + h_inode = d_inode(h_path->dentry);
4889 + dt->dt_atime = h_inode->i_atime;
4890 + dt->dt_mtime = h_inode->i_mtime;
4894 +void au_dtime_revert(struct au_dtime *dt)
4896 + struct iattr attr;
4899 + attr.ia_atime = dt->dt_atime;
4900 + attr.ia_mtime = dt->dt_mtime;
4901 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4902 + | ATTR_ATIME | ATTR_ATIME_SET;
4904 + /* no delegation since this is a directory */
4905 + err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4906 + if (unlikely(err))
4907 + pr_warn("restoring timestamps failed(%d). ignored\n", err);
4910 +/* ---------------------------------------------------------------------- */
4912 +/* internal use only */
4913 +struct au_cpup_reg_attr {
4916 + unsigned int iflags; /* inode->i_flags */
4919 +static noinline_for_stack
4920 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct path *h_src,
4921 + struct au_cpup_reg_attr *h_src_attr)
4923 + int err, sbits, icex;
4924 + unsigned int mnt_flags;
4925 + unsigned char verbose;
4927 + struct path h_path;
4928 + struct inode *h_isrc, *h_idst;
4929 + struct kstat *h_st;
4930 + struct au_branch *br;
4932 + br = au_sbr(dst->d_sb, bindex);
4933 + h_path.mnt = au_br_mnt(br);
4934 + h_path.dentry = au_h_dptr(dst, bindex);
4935 + h_idst = d_inode(h_path.dentry);
4936 + h_isrc = d_inode(h_src->dentry);
4937 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4938 + | ATTR_ATIME | ATTR_MTIME
4939 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
4940 + if (h_src_attr && h_src_attr->valid) {
4941 + h_st = &h_src_attr->st;
4942 + ia.ia_uid = h_st->uid;
4943 + ia.ia_gid = h_st->gid;
4944 + ia.ia_atime = h_st->atime;
4945 + ia.ia_mtime = h_st->mtime;
4946 + if (h_idst->i_mode != h_st->mode
4947 + && !S_ISLNK(h_idst->i_mode)) {
4948 + ia.ia_valid |= ATTR_MODE;
4949 + ia.ia_mode = h_st->mode;
4951 + sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4952 + au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4954 + ia.ia_uid = h_isrc->i_uid;
4955 + ia.ia_gid = h_isrc->i_gid;
4956 + ia.ia_atime = h_isrc->i_atime;
4957 + ia.ia_mtime = h_isrc->i_mtime;
4958 + if (h_idst->i_mode != h_isrc->i_mode
4959 + && !S_ISLNK(h_idst->i_mode)) {
4960 + ia.ia_valid |= ATTR_MODE;
4961 + ia.ia_mode = h_isrc->i_mode;
4963 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4964 + au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4966 + /* no delegation since it is just created */
4967 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4969 + /* is this nfs only? */
4970 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4971 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4972 + ia.ia_mode = h_isrc->i_mode;
4973 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4976 + icex = br->br_perm & AuBrAttr_ICEX;
4978 + mnt_flags = au_mntflags(dst->d_sb);
4979 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4980 + err = au_cpup_xattr(&h_path, h_src, icex, verbose);
4986 +/* ---------------------------------------------------------------------- */
4988 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4989 + char *buf, unsigned long blksize)
4992 + size_t sz, rbytes, wbytes;
4993 + unsigned char all_zero;
4995 + struct inode *h_inode;
4996 + /* reduce stack usage */
4999 + zp = page_address(ZERO_PAGE(0));
5000 + if (unlikely(!zp))
5001 + return -ENOMEM; /* possible? */
5006 + AuDbg("len %lld\n", len);
5008 + if (len < blksize)
5012 + /* todo: signal_pending? */
5013 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
5014 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
5017 + if (unlikely(err < 0))
5021 + if (len >= rbytes && rbytes == blksize)
5022 + all_zero = !memcmp(buf, zp, rbytes);
5029 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
5031 + /* todo: signal_pending? */
5032 + if (unlikely(err == -EAGAIN || err == -EINTR))
5034 + if (unlikely(err < 0))
5039 + if (unlikely(err < 0))
5045 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
5047 + if (unlikely(res < 0))
5054 + /* the last block may be a hole */
5055 + if (!err && all_zero) {
5056 + AuLabel(last hole);
5059 + if (au_test_nfs(dst->f_path.dentry->d_sb)) {
5060 + /* nfs requires this step to make last hole */
5061 + /* is this only nfs? */
5063 + /* todo: signal_pending? */
5064 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
5065 + } while (err == -EAGAIN || err == -EINTR);
5072 + ia->ia_size = dst->f_pos;
5073 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
5074 + ia->ia_file = dst;
5075 + h_inode = file_inode(dst);
5076 + inode_lock_nested(h_inode, AuLsc_I_CHILD2);
5077 + /* no delegation since it is just created */
5078 + err = vfsub_notify_change(&dst->f_path, ia,
5079 + /*delegated*/NULL);
5080 + inode_unlock(h_inode);
5087 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
5090 + unsigned long blksize;
5091 + unsigned char do_kfree;
5093 + struct super_block *h_sb;
5096 + h_sb = file_inode(dst)->i_sb;
5097 + blksize = h_sb->s_blocksize;
5098 + if (!blksize || PAGE_SIZE < blksize)
5099 + blksize = PAGE_SIZE;
5100 + AuDbg("blksize %lu\n", blksize);
5101 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
5103 + buf = kmalloc(blksize, GFP_NOFS);
5105 + buf = (void *)__get_free_page(GFP_NOFS);
5106 + if (unlikely(!buf))
5109 + if (len > (1 << 22))
5110 + AuDbg("copying a large file %lld\n", (long long)len);
5114 + err = au_do_copy_file(dst, src, len, buf, blksize);
5116 + AuDebugOn(!au_kfree_do_sz_test(blksize));
5117 + au_kfree_do_rcu(buf);
5119 + free_page((unsigned long)buf);
5125 +static int au_do_copy(struct file *dst, struct file *src, loff_t len)
5128 + struct super_block *h_src_sb;
5129 + struct inode *h_src_inode;
5131 + h_src_inode = file_inode(src);
5132 + h_src_sb = h_src_inode->i_sb;
5134 + /* XFS acquires inode_lock */
5135 + if (!au_test_xfs(h_src_sb))
5136 + err = au_copy_file(dst, src, len);
5138 + inode_unlock_shared(h_src_inode);
5139 + err = au_copy_file(dst, src, len);
5140 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5146 +static int au_clone_or_copy(struct file *dst, struct file *src, loff_t len)
5150 + struct super_block *h_src_sb;
5151 + struct inode *h_src_inode;
5153 + h_src_inode = file_inode(src);
5154 + h_src_sb = h_src_inode->i_sb;
5155 + if (h_src_sb != file_inode(dst)->i_sb
5156 + || !dst->f_op->remap_file_range) {
5157 + err = au_do_copy(dst, src, len);
5161 + if (!au_test_nfs(h_src_sb)) {
5162 + inode_unlock_shared(h_src_inode);
5163 + lo = vfsub_clone_file_range(src, dst, len);
5164 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5166 + lo = vfsub_clone_file_range(src, dst, len);
5169 + goto out; /* success */
5170 + } else if (lo >= 0)
5171 + /* todo: possible? */
5172 + /* paritially succeeded */
5173 + AuDbg("lo %lld, len %lld. Retrying.\n", lo, len);
5174 + else if (lo != -EOPNOTSUPP) {
5175 + /* older XFS has a condition in cloning */
5180 + /* the backend fs on NFS may not support cloning */
5181 + err = au_do_copy(dst, src, len);
5189 + * to support a sparse file which is opened with O_APPEND,
5190 + * we need to close the file.
5192 +static int au_cp_regular(struct au_cp_generic *cpg)
5195 + enum { SRC, DST };
5197 + aufs_bindex_t bindex;
5198 + unsigned int flags;
5199 + struct dentry *dentry;
5201 + struct file *file;
5204 + .bindex = cpg->bsrc,
5205 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
5208 + .bindex = cpg->bdst,
5209 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
5210 + .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
5213 + struct au_branch *br;
5214 + struct super_block *sb, *h_src_sb;
5215 + struct inode *h_src_inode;
5216 + struct task_struct *tsk = current;
5218 + /* bsrc branch can be ro/rw. */
5219 + sb = cpg->dentry->d_sb;
5221 + for (i = 0; i < 2; i++, f++) {
5222 + f->dentry = au_h_dptr(cpg->dentry, f->bindex);
5223 + f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
5224 + /*file*/NULL, f->force_wr);
5225 + if (IS_ERR(f->file)) {
5226 + err = PTR_ERR(f->file);
5234 + /* try stopping to update while we copyup */
5235 + h_src_inode = d_inode(file[SRC].dentry);
5236 + h_src_sb = h_src_inode->i_sb;
5237 + if (!au_test_nfs(h_src_sb))
5238 + IMustLock(h_src_inode);
5239 + err = au_clone_or_copy(file[DST].file, file[SRC].file, cpg->len);
5241 + /* i wonder if we had O_NO_DELAY_FPUT flag */
5242 + if (tsk->flags & PF_KTHREAD)
5243 + __fput_sync(file[DST].file);
5245 + /* it happened actually */
5246 + fput(file[DST].file);
5249 + * we have to call both since we don't know which place the file
5253 + flush_delayed_fput();
5255 + br = au_sbr(sb, file[DST].bindex);
5256 + au_lcnt_dec(&br->br_nfiles);
5259 + fput(file[SRC].file);
5260 + br = au_sbr(sb, file[SRC].bindex);
5261 + au_lcnt_dec(&br->br_nfiles);
5266 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
5267 + struct au_cpup_reg_attr *h_src_attr)
5271 + struct path h_path;
5272 + struct inode *h_src_inode, *h_dst_inode;
5275 + h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
5276 + l = i_size_read(h_src_inode);
5277 + if (cpg->len == -1 || l < cpg->len)
5280 + /* try stopping to update while we are referencing */
5281 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5282 + au_pin_hdir_unlock(cpg->pin);
5284 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5285 + h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
5286 + h_src_attr->iflags = h_src_inode->i_flags;
5287 + if (!au_test_nfs(h_src_inode->i_sb))
5288 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5290 + inode_unlock_shared(h_src_inode);
5291 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5292 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5294 + if (unlikely(err)) {
5295 + inode_unlock_shared(h_src_inode);
5298 + h_src_attr->valid = 1;
5299 + if (!au_test_nfs(h_src_inode->i_sb)) {
5300 + err = au_cp_regular(cpg);
5301 + inode_unlock_shared(h_src_inode);
5303 + inode_unlock_shared(h_src_inode);
5304 + err = au_cp_regular(cpg);
5306 + rerr = au_pin_hdir_relock(cpg->pin);
5310 + if (!err && (h_src_inode->i_state & I_LINKABLE)) {
5311 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5312 + h_dst_inode = d_inode(h_path.dentry);
5313 + spin_lock(&h_dst_inode->i_lock);
5314 + h_dst_inode->i_state |= I_LINKABLE;
5315 + spin_unlock(&h_dst_inode->i_lock);
5322 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
5323 + struct inode *h_dir)
5326 + DEFINE_DELAYED_CALL(done);
5329 + sym = vfs_get_link(h_src, &done);
5330 + err = PTR_ERR(sym);
5334 + err = vfsub_symlink(h_dir, h_path, sym);
5337 + do_delayed_call(&done);
5342 + * regardless 'acl' option, reset all ACL.
5343 + * All ACL will be copied up later from the original entry on the lower branch.
5345 +static int au_reset_acl(struct inode *h_dir, struct path *h_path, umode_t mode)
5348 + struct dentry *h_dentry;
5349 + struct inode *h_inode;
5350 + struct user_namespace *h_userns;
5352 + h_userns = mnt_user_ns(h_path->mnt);
5353 + h_dentry = h_path->dentry;
5354 + h_inode = d_inode(h_dentry);
5355 + /* forget_all_cached_acls(h_inode)); */
5356 + err = vfsub_removexattr(h_userns, h_dentry, XATTR_NAME_POSIX_ACL_ACCESS);
5358 + if (err == -EOPNOTSUPP)
5361 + err = vfsub_acl_chmod(h_userns, h_inode, mode);
5367 +static int au_do_cpup_dir(struct au_cp_generic *cpg, struct dentry *dst_parent,
5368 + struct inode *h_dir, struct path *h_path)
5371 + struct inode *dir, *inode;
5372 + struct user_namespace *h_userns;
5374 + h_userns = mnt_user_ns(h_path->mnt);
5375 + err = vfsub_removexattr(h_userns, h_path->dentry,
5376 + XATTR_NAME_POSIX_ACL_DEFAULT);
5378 + if (err == -EOPNOTSUPP)
5380 + if (unlikely(err))
5384 + * strange behaviour from the users view,
5385 + * particularly setattr case
5387 + dir = d_inode(dst_parent);
5388 + if (au_ibtop(dir) == cpg->bdst)
5389 + au_cpup_attr_nlink(dir, /*force*/1);
5390 + inode = d_inode(cpg->dentry);
5391 + au_cpup_attr_nlink(inode, /*force*/1);
5397 +static noinline_for_stack
5398 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5399 + struct au_cpup_reg_attr *h_src_attr)
5403 + unsigned int mnt_flags;
5404 + unsigned char isdir, isreg, force;
5405 + const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5406 + struct au_dtime dt;
5407 + struct path h_path;
5408 + struct dentry *h_src, *h_dst, *h_parent;
5409 + struct inode *h_inode, *h_dir;
5410 + struct super_block *sb;
5412 + /* bsrc branch can be ro/rw. */
5413 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5414 + h_inode = d_inode(h_src);
5415 + AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5417 + /* try stopping to be referenced while we are creating */
5418 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5419 + if (au_ftest_cpup(cpg->flags, RENAME))
5420 + AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5421 + AUFS_WH_PFX_LEN));
5422 + h_parent = h_dst->d_parent; /* dir inode is locked */
5423 + h_dir = d_inode(h_parent);
5425 + AuDebugOn(h_parent != h_dst->d_parent);
5427 + sb = cpg->dentry->d_sb;
5428 + h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5430 + h_path.dentry = h_parent;
5431 + au_dtime_store(&dt, dst_parent, &h_path);
5433 + h_path.dentry = h_dst;
5437 + mode = h_inode->i_mode;
5438 + switch (mode & S_IFMT) {
5441 + err = vfsub_create(h_dir, &h_path, 0600, /*want_excl*/true);
5443 + err = au_do_cpup_regular(cpg, h_src_attr);
5447 + err = vfsub_mkdir(h_dir, &h_path, mode);
5449 + err = au_do_cpup_dir(cpg, dst_parent, h_dir, &h_path);
5452 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5456 + AuDebugOn(!capable(CAP_MKNOD));
5460 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5463 + AuIOErr("Unknown inode type 0%o\n", mode);
5467 + err = au_reset_acl(h_dir, &h_path, mode);
5469 + mnt_flags = au_mntflags(sb);
5470 + if (!au_opt_test(mnt_flags, UDBA_NONE)
5472 + && au_opt_test(mnt_flags, XINO)
5473 + && (h_inode->i_nlink == 1
5474 + || (h_inode->i_state & I_LINKABLE))
5475 + /* todo: unnecessary? */
5476 + /* && d_inode(cpg->dentry)->i_nlink == 1 */
5477 + && cpg->bdst < cpg->bsrc
5478 + && !au_ftest_cpup(cpg->flags, KEEPLINO))
5479 + au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5480 + /* ignore this error */
5485 + force = !!cpg->len;
5486 + if (cpg->len == -1)
5487 + force = !!i_size_read(h_inode);
5489 + au_fhsm_wrote(sb, cpg->bdst, force);
5493 + au_dtime_revert(&dt);
5497 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5500 + struct dentry *dentry, *h_dentry, *h_parent, *parent;
5501 + struct path h_ppath;
5502 + struct inode *h_dir;
5503 + aufs_bindex_t bdst;
5505 + dentry = cpg->dentry;
5507 + h_ppath.mnt = au_sbr_mnt(dentry->d_sb, bdst);
5508 + h_dentry = au_h_dptr(dentry, bdst);
5509 + if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5511 + au_set_h_dptr(dentry, bdst, NULL);
5512 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
5514 + h_path->dentry = dget(au_h_dptr(dentry, bdst));
5515 + au_set_h_dptr(dentry, bdst, h_dentry);
5518 + parent = dget_parent(dentry);
5519 + h_ppath.dentry = au_h_dptr(parent, bdst);
5521 + h_path->dentry = vfsub_lkup_one(&dentry->d_name, &h_ppath);
5522 + if (IS_ERR(h_path->dentry))
5523 + err = PTR_ERR(h_path->dentry);
5525 + if (unlikely(err))
5528 + h_parent = h_dentry->d_parent; /* dir inode is locked */
5529 + h_dir = d_inode(h_parent);
5531 + AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5532 + /* no delegation since it is just created */
5533 + err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL,
5535 + dput(h_path->dentry);
5542 + * copyup the @dentry from @bsrc to @bdst.
5543 + * the caller must set the both of lower dentries.
5544 + * @len is for truncating when it is -1 copyup the entire file.
5545 + * in link/rename cases, @dst_parent may be different from the real one.
5546 + * basic->bsrc can be larger than basic->bdst.
5547 + * aufs doesn't touch the credential so
5548 + * security_inode_copy_up{,_xattr}() are unnecessary.
5550 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5553 + aufs_bindex_t old_ibtop;
5554 + unsigned char isdir, plink;
5555 + struct dentry *h_src, *h_dst, *h_parent;
5556 + struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5557 + struct super_block *sb;
5558 + struct au_branch *br;
5559 + struct path h_src_path;
5560 + /* to reduce stack size */
5562 + struct au_dtime dt;
5563 + struct path h_path;
5564 + struct au_cpup_reg_attr h_src_attr;
5568 + a = kmalloc(sizeof(*a), GFP_NOFS);
5571 + a->h_src_attr.valid = 0;
5573 + sb = cpg->dentry->d_sb;
5574 + br = au_sbr(sb, cpg->bdst);
5575 + a->h_path.mnt = au_br_mnt(br);
5576 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5577 + h_parent = h_dst->d_parent; /* dir inode is locked */
5578 + h_dir = d_inode(h_parent);
5581 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5582 + inode = d_inode(cpg->dentry);
5585 + dst_parent = dget_parent(cpg->dentry);
5589 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
5590 + dst_inode = au_h_iptr(inode, cpg->bdst);
5592 + if (unlikely(!plink)) {
5594 + AuIOErr("hi%lu(i%lu) exists on b%d "
5595 + "but plink is disabled\n",
5596 + dst_inode->i_ino, inode->i_ino, cpg->bdst);
5600 + if (dst_inode->i_nlink) {
5601 + const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5603 + h_src = au_plink_lkup(inode, cpg->bdst);
5604 + err = PTR_ERR(h_src);
5605 + if (IS_ERR(h_src))
5607 + if (unlikely(d_is_negative(h_src))) {
5609 + AuIOErr("i%lu exists on b%d "
5610 + "but not pseudo-linked\n",
5611 + inode->i_ino, cpg->bdst);
5617 + a->h_path.dentry = h_parent;
5618 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5621 + a->h_path.dentry = h_dst;
5623 + err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5624 + if (!err && au_ftest_cpup(cpg->flags, RENAME))
5625 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5627 + au_dtime_revert(&a->dt);
5628 + if (unlikely(err == -EWOULDBLOCK)) {
5629 + pr_warn("cannot retry for NFSv4 delegation"
5630 + " for an internal link\n");
5636 + /* todo: cpup_wh_file? */
5638 + au_update_ibrange(inode, /*do_put_zero*/1);
5641 + isdir = S_ISDIR(inode->i_mode);
5642 + old_ibtop = au_ibtop(inode);
5643 + err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5644 + if (unlikely(err))
5646 + dst_inode = d_inode(h_dst);
5647 + inode_lock_nested(dst_inode, AuLsc_I_CHILD2);
5648 + /* todo: necessary? */
5649 + /* au_pin_hdir_unlock(cpg->pin); */
5651 + h_src_path.dentry = h_src;
5652 + h_src_path.mnt = au_sbr_mnt(sb, cpg->bsrc);
5653 + err = cpup_iattr(cpg->dentry, cpg->bdst, &h_src_path, &a->h_src_attr);
5654 + if (unlikely(err)) {
5655 + /* todo: necessary? */
5656 + /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5657 + inode_unlock(dst_inode);
5661 + if (cpg->bdst < old_ibtop) {
5662 + if (S_ISREG(inode->i_mode)) {
5663 + err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5664 + if (unlikely(err)) {
5665 + /* ignore an error */
5666 + /* au_pin_hdir_relock(cpg->pin); */
5667 + inode_unlock(dst_inode);
5671 + au_set_ibtop(inode, cpg->bdst);
5673 + au_set_ibbot(inode, cpg->bdst);
5674 + au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5675 + au_hi_flags(inode, isdir));
5677 + /* todo: necessary? */
5678 + /* err = au_pin_hdir_relock(cpg->pin); */
5679 + inode_unlock(dst_inode);
5680 + if (unlikely(err))
5683 + src_inode = d_inode(h_src);
5685 + && (src_inode->i_nlink > 1
5686 + || src_inode->i_state & I_LINKABLE)
5688 + au_plink_append(inode, cpg->bdst, h_dst);
5690 + if (au_ftest_cpup(cpg->flags, RENAME)) {
5691 + a->h_path.dentry = h_dst;
5692 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5695 + goto out_parent; /* success */
5699 + a->h_path.dentry = h_parent;
5700 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5701 + a->h_path.dentry = h_dst;
5703 + if (d_is_positive(h_dst)) {
5705 + /* no delegation since it is just created */
5706 + rerr = vfsub_unlink(h_dir, &a->h_path,
5707 + /*delegated*/NULL, /*force*/0);
5709 + rerr = vfsub_rmdir(h_dir, &a->h_path);
5711 + au_dtime_revert(&a->dt);
5713 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5723 +#if 0 /* reserved */
5724 +struct au_cpup_single_args {
5726 + struct au_cp_generic *cpg;
5727 + struct dentry *dst_parent;
5730 +static void au_call_cpup_single(void *args)
5732 + struct au_cpup_single_args *a = args;
5734 + au_pin_hdir_acquire_nest(a->cpg->pin);
5735 + *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5736 + au_pin_hdir_release(a->cpg->pin);
5741 + * prevent SIGXFSZ in copy-up.
5742 + * testing CAP_MKNOD is for generic fs,
5743 + * but CAP_FSETID is for xfs only, currently.
5745 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5748 + struct super_block *sb;
5749 + struct inode *h_dir;
5752 + sb = au_pinned_parent(pin)->d_sb;
5753 + if (!au_wkq_test()
5754 + && (!au_sbi(sb)->si_plink_maint_pid
5755 + || au_plink_maint(sb, AuLock_NOPLM))) {
5756 + switch (mode & S_IFMT) {
5758 + /* no condition about RLIMIT_FSIZE and the file size */
5763 + do_sio = !capable(CAP_MKNOD);
5767 + do_sio = ((mode & (S_ISUID | S_ISGID))
5768 + && !capable(CAP_FSETID));
5769 + /* this workaround may be removed in the future */
5771 + h_dir = au_pinned_h_dir(pin);
5772 + do_sio = h_dir->i_mode & S_ISVTX;
5779 +#if 0 /* reserved */
5780 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5783 + struct dentry *h_dentry;
5785 + h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5786 + if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5787 + err = au_cpup_single(cpg, dst_parent);
5789 + struct au_cpup_single_args args = {
5792 + .dst_parent = dst_parent
5794 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5795 + if (unlikely(wkq_err))
5804 + * copyup the @dentry from the first active lower branch to @bdst,
5805 + * using au_cpup_single().
5807 +static int au_cpup_simple(struct au_cp_generic *cpg)
5810 + unsigned int flags_orig;
5811 + struct dentry *dentry;
5813 + AuDebugOn(cpg->bsrc < 0);
5815 + dentry = cpg->dentry;
5816 + DiMustWriteLock(dentry);
5818 + err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5820 + flags_orig = cpg->flags;
5821 + au_fset_cpup(cpg->flags, RENAME);
5822 + err = au_cpup_single(cpg, NULL);
5823 + cpg->flags = flags_orig;
5825 + return 0; /* success */
5828 + au_set_h_dptr(dentry, cpg->bdst, NULL);
5829 + au_set_dbtop(dentry, cpg->bsrc);
5835 +struct au_cpup_simple_args {
5837 + struct au_cp_generic *cpg;
5840 +static void au_call_cpup_simple(void *args)
5842 + struct au_cpup_simple_args *a = args;
5844 + au_pin_hdir_acquire_nest(a->cpg->pin);
5845 + *a->errp = au_cpup_simple(a->cpg);
5846 + au_pin_hdir_release(a->cpg->pin);
5849 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5852 + struct dentry *dentry, *parent;
5853 + struct file *h_file;
5854 + struct inode *h_dir;
5855 + struct user_namespace *h_userns;
5857 + dentry = cpg->dentry;
5859 + if (au_ftest_cpup(cpg->flags, HOPEN)) {
5860 + AuDebugOn(cpg->bsrc < 0);
5861 + h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5862 + err = PTR_ERR(h_file);
5863 + if (IS_ERR(h_file))
5867 + parent = dget_parent(dentry);
5868 + h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5869 + h_userns = au_sbr_userns(dentry->d_sb, cpg->bdst);
5870 + if (!au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC | MAY_WRITE)
5871 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5872 + err = au_cpup_simple(cpg);
5874 + struct au_cpup_simple_args args = {
5878 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5879 + if (unlikely(wkq_err))
5885 + au_h_open_post(dentry, cpg->bsrc, h_file);
5891 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5893 + aufs_bindex_t bsrc, bbot;
5894 + struct dentry *dentry, *h_dentry;
5896 + if (cpg->bsrc < 0) {
5897 + dentry = cpg->dentry;
5898 + bbot = au_dbbot(dentry);
5899 + for (bsrc = cpg->bdst + 1; bsrc <= bbot; bsrc++) {
5900 + h_dentry = au_h_dptr(dentry, bsrc);
5902 + AuDebugOn(d_is_negative(h_dentry));
5906 + AuDebugOn(bsrc > bbot);
5909 + AuDebugOn(cpg->bsrc <= cpg->bdst);
5910 + return au_do_sio_cpup_simple(cpg);
5913 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5915 + AuDebugOn(cpg->bdst <= cpg->bsrc);
5916 + return au_do_sio_cpup_simple(cpg);
5919 +/* ---------------------------------------------------------------------- */
5922 + * copyup the deleted file for writing.
5924 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5925 + struct file *file)
5928 + unsigned int flags_orig;
5929 + aufs_bindex_t bsrc_orig;
5930 + struct au_dinfo *dinfo;
5932 + struct au_hdentry *hd;
5933 + struct dentry *h_dentry;
5936 + dinfo = au_di(cpg->dentry);
5937 + AuRwMustWriteLock(&dinfo->di_rwsem);
5939 + bsrc_orig = cpg->bsrc;
5940 + cpg->bsrc = dinfo->di_btop;
5941 + hdst.hd = au_hdentry(dinfo, cpg->bdst);
5942 + hdst.h_dentry = hdst.hd->hd_dentry;
5943 + hdst.hd->hd_dentry = wh_dentry;
5944 + dinfo->di_btop = cpg->bdst;
5946 + hsrc.h_dentry = NULL;
5948 + hsrc.hd = au_hdentry(dinfo, cpg->bsrc);
5949 + hsrc.h_dentry = hsrc.hd->hd_dentry;
5950 + hsrc.hd->hd_dentry = au_hf_top(file)->f_path.dentry;
5952 + flags_orig = cpg->flags;
5953 + cpg->flags = !AuCpup_DTIME;
5954 + err = au_cpup_single(cpg, /*h_parent*/NULL);
5955 + cpg->flags = flags_orig;
5958 + err = au_reopen_nondir(file);
5959 + hsrc.hd->hd_dentry = hsrc.h_dentry;
5961 + hdst.hd->hd_dentry = hdst.h_dentry;
5962 + dinfo->di_btop = cpg->bsrc;
5963 + cpg->bsrc = bsrc_orig;
5968 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5971 + aufs_bindex_t bdst;
5972 + struct au_dtime dt;
5973 + struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5974 + struct au_branch *br;
5975 + struct path h_path;
5977 + dentry = cpg->dentry;
5979 + br = au_sbr(dentry->d_sb, bdst);
5980 + parent = dget_parent(dentry);
5981 + h_parent = au_h_dptr(parent, bdst);
5982 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5983 + err = PTR_ERR(wh_dentry);
5984 + if (IS_ERR(wh_dentry))
5987 + h_path.dentry = h_parent;
5988 + h_path.mnt = au_br_mnt(br);
5989 + au_dtime_store(&dt, parent, &h_path);
5990 + err = au_do_cpup_wh(cpg, wh_dentry, file);
5991 + if (unlikely(err))
5995 + h_path.dentry = wh_dentry;
5996 + if (!d_is_dir(wh_dentry)) {
5997 + /* no delegation since it is just created */
5998 + err = vfsub_unlink(d_inode(h_parent), &h_path,
5999 + /*delegated*/NULL, /*force*/0);
6001 + err = vfsub_rmdir(d_inode(h_parent), &h_path);
6002 + if (unlikely(err)) {
6003 + AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
6007 + au_dtime_revert(&dt);
6008 + au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
6017 +struct au_cpup_wh_args {
6019 + struct au_cp_generic *cpg;
6020 + struct file *file;
6023 +static void au_call_cpup_wh(void *args)
6025 + struct au_cpup_wh_args *a = args;
6027 + au_pin_hdir_acquire_nest(a->cpg->pin);
6028 + *a->errp = au_cpup_wh(a->cpg, a->file);
6029 + au_pin_hdir_release(a->cpg->pin);
6032 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6035 + aufs_bindex_t bdst;
6036 + struct dentry *dentry, *parent, *h_orph, *h_parent;
6037 + struct inode *dir, *h_dir, *h_tmpdir;
6038 + struct au_wbr *wbr;
6039 + struct au_pin wh_pin, *pin_orig;
6040 + struct user_namespace *h_userns;
6042 + dentry = cpg->dentry;
6044 + parent = dget_parent(dentry);
6045 + dir = d_inode(parent);
6048 + h_dir = au_igrab(au_h_iptr(dir, bdst));
6051 + if (!h_dir->i_nlink) {
6052 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
6053 + h_orph = wbr->wbr_orph;
6055 + h_parent = dget(au_h_dptr(parent, bdst));
6056 + au_set_h_dptr(parent, bdst, dget(h_orph));
6057 + h_tmpdir = d_inode(h_orph);
6058 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
6060 + inode_lock_nested(h_tmpdir, AuLsc_I_PARENT3);
6061 + /* todo: au_h_open_pre()? */
6063 + pin_orig = cpg->pin;
6064 + au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
6065 + AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
6066 + cpg->pin = &wh_pin;
6069 + h_userns = au_sbr_userns(dentry->d_sb, bdst);
6070 + if (!au_test_h_perm_sio(h_userns, h_tmpdir, MAY_EXEC | MAY_WRITE)
6071 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6072 + err = au_cpup_wh(cpg, file);
6074 + struct au_cpup_wh_args args = {
6079 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
6080 + if (unlikely(wkq_err))
6085 + inode_unlock(h_tmpdir);
6086 + /* todo: au_h_open_post()? */
6087 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
6088 + au_set_h_dptr(parent, bdst, h_parent);
6089 + AuDebugOn(!pin_orig);
6090 + cpg->pin = pin_orig;
6098 +/* ---------------------------------------------------------------------- */
6101 + * generic routine for both of copy-up and copy-down.
6103 +/* cf. revalidate function in file.c */
6104 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6105 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6106 + struct au_pin *pin,
6107 + struct dentry *h_parent, void *arg),
6111 + struct au_pin pin;
6112 + struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
6115 + parent = dget_parent(dentry);
6116 + if (IS_ROOT(parent))
6119 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
6120 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
6122 + /* do not use au_dpage */
6123 + real_parent = parent;
6126 + parent = dget_parent(dentry);
6127 + h_parent = au_h_dptr(parent, bdst);
6129 + goto out; /* success */
6131 + /* find top dir which is necessary to cpup */
6135 + parent = dget_parent(d);
6136 + di_read_lock_parent3(parent, !AuLock_IR);
6137 + h_parent = au_h_dptr(parent, bdst);
6138 + di_read_unlock(parent, !AuLock_IR);
6139 + } while (!h_parent);
6141 + if (d != real_parent)
6142 + di_write_lock_child3(d);
6144 + /* somebody else might create while we were sleeping */
6145 + h_dentry = au_h_dptr(d, bdst);
6146 + if (!h_dentry || d_is_negative(h_dentry)) {
6148 + au_update_dbtop(d);
6150 + au_pin_set_dentry(&pin, d);
6151 + err = au_do_pin(&pin);
6153 + err = cp(d, bdst, &pin, h_parent, arg);
6158 + if (d != real_parent)
6159 + di_write_unlock(d);
6160 + if (unlikely(err))
6169 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
6170 + struct au_pin *pin,
6171 + struct dentry *h_parent __maybe_unused,
6172 + void *arg __maybe_unused)
6174 + struct au_cp_generic cpg = {
6180 + .flags = AuCpup_DTIME
6182 + return au_sio_cpup_simple(&cpg);
6185 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6187 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
6190 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6193 + struct dentry *parent;
6194 + struct inode *dir;
6196 + parent = dget_parent(dentry);
6197 + dir = d_inode(parent);
6199 + if (au_h_iptr(dir, bdst))
6202 + di_read_unlock(parent, AuLock_IR);
6203 + di_write_lock_parent(parent);
6204 + /* someone else might change our inode while we were sleeping */
6205 + if (!au_h_iptr(dir, bdst))
6206 + err = au_cpup_dirs(dentry, bdst);
6207 + di_downgrade_lock(parent, AuLock_IR);
6213 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
6214 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
6215 +++ linux/fs/aufs/cpup.h 2022-03-21 14:49:05.719966344 +0100
6217 +/* SPDX-License-Identifier: GPL-2.0 */
6219 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6221 + * This program, aufs is free software; you can redistribute it and/or modify
6222 + * it under the terms of the GNU General Public License as published by
6223 + * the Free Software Foundation; either version 2 of the License, or
6224 + * (at your option) any later version.
6226 + * This program is distributed in the hope that it will be useful,
6227 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6228 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6229 + * GNU General Public License for more details.
6231 + * You should have received a copy of the GNU General Public License
6232 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6236 + * copy-up/down functions
6239 +#ifndef __AUFS_CPUP_H__
6240 +#define __AUFS_CPUP_H__
6244 +#include <linux/path.h>
6250 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
6251 +void au_cpup_attr_timesizes(struct inode *inode);
6252 +void au_cpup_attr_nlink(struct inode *inode, int force);
6253 +void au_cpup_attr_changeable(struct inode *inode);
6254 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
6255 +void au_cpup_attr_all(struct inode *inode, int force);
6257 +/* ---------------------------------------------------------------------- */
6259 +struct au_cp_generic {
6260 + struct dentry *dentry;
6261 + aufs_bindex_t bdst, bsrc;
6263 + struct au_pin *pin;
6264 + unsigned int flags;
6268 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
6269 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
6271 +#define AuCpup_RENAME (1 << 2) /* rename after cpup */
6272 +#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
6274 +#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
6276 +#define AuCpup_RWDST (1 << 5) /* force write target even if
6277 + the branch is marked as RO */
6279 +#ifndef CONFIG_AUFS_BR_HFSPLUS
6280 +#undef AuCpup_HOPEN
6281 +#define AuCpup_HOPEN 0
6284 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
6285 +#define au_fset_cpup(flags, name) \
6286 + do { (flags) |= AuCpup_##name; } while (0)
6287 +#define au_fclr_cpup(flags, name) \
6288 + do { (flags) &= ~AuCpup_##name; } while (0)
6290 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
6291 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
6292 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
6293 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
6295 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6296 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6297 + struct au_pin *pin,
6298 + struct dentry *h_parent, void *arg),
6300 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6301 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6303 +/* ---------------------------------------------------------------------- */
6305 +/* keep timestamps when copyup */
6307 + struct dentry *dt_dentry;
6308 + struct path dt_h_path;
6309 + struct timespec64 dt_atime, dt_mtime;
6311 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
6312 + struct path *h_path);
6313 +void au_dtime_revert(struct au_dtime *dt);
6315 +#endif /* __KERNEL__ */
6316 +#endif /* __AUFS_CPUP_H__ */
6317 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
6318 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
6319 +++ linux/fs/aufs/dbgaufs.c 2022-03-21 14:49:05.719966344 +0100
6321 +// SPDX-License-Identifier: GPL-2.0
6323 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6325 + * This program, aufs is free software; you can redistribute it and/or modify
6326 + * it under the terms of the GNU General Public License as published by
6327 + * the Free Software Foundation; either version 2 of the License, or
6328 + * (at your option) any later version.
6330 + * This program is distributed in the hope that it will be useful,
6331 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6332 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6333 + * GNU General Public License for more details.
6335 + * You should have received a copy of the GNU General Public License
6336 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6340 + * debugfs interface
6343 +#include <linux/debugfs.h>
6346 +#ifndef CONFIG_SYSFS
6347 +#error DEBUG_FS depends upon SYSFS
6350 +static struct dentry *dbgaufs;
6351 +static const mode_t dbgaufs_mode = 0444;
6353 +/* 20 is max digits length of ulong 64 */
6354 +struct dbgaufs_arg {
6360 + * common function for all XINO files
6362 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
6363 + struct file *file)
6367 + p = file->private_data;
6369 + /* this is struct dbgaufs_arg */
6370 + AuDebugOn(!au_kfree_sz_test(p));
6371 + au_kfree_do_rcu(p);
6376 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt,
6381 + struct dbgaufs_arg *p;
6384 + p = kmalloc(sizeof(*p), GFP_NOFS);
6390 + file->private_data = p;
6394 + err = vfsub_getattr(&xf->f_path, &st);
6398 + (p->a, sizeof(p->a), "%d, %llux%u %lld\n",
6399 + cnt, st.blocks, st.blksize,
6400 + (long long)st.size);
6402 + p->n = snprintf(p->a, sizeof(p->a), "%llux%u %lld\n",
6403 + st.blocks, st.blksize,
6404 + (long long)st.size);
6405 + AuDebugOn(p->n >= sizeof(p->a));
6407 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
6415 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
6416 + size_t count, loff_t *ppos)
6418 + struct dbgaufs_arg *p;
6420 + p = file->private_data;
6421 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6424 +/* ---------------------------------------------------------------------- */
6426 +struct dbgaufs_plink_arg {
6431 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6432 + struct file *file)
6434 + free_page((unsigned long)file->private_data);
6438 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6440 + int err, i, limit;
6441 + unsigned long n, sum;
6442 + struct dbgaufs_plink_arg *p;
6443 + struct au_sbinfo *sbinfo;
6444 + struct super_block *sb;
6445 + struct hlist_bl_head *hbl;
6448 + p = (void *)get_zeroed_page(GFP_NOFS);
6453 + sbinfo = inode->i_private;
6454 + sb = sbinfo->si_sb;
6455 + si_noflush_read_lock(sb);
6456 + if (au_opt_test(au_mntflags(sb), PLINK)) {
6457 + limit = PAGE_SIZE - sizeof(p->n);
6459 + /* the number of buckets */
6460 + n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6465 + for (i = 0, hbl = sbinfo->si_plink; i < AuPlink_NHASH;
6467 + n = au_hbl_count(hbl);
6470 + n = snprintf(p->a + p->n, limit, "%lu ", n);
6473 + if (unlikely(limit <= 0))
6476 + p->a[p->n - 1] = '\n';
6478 + /* the sum of plinks */
6479 + n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6482 + if (unlikely(limit <= 0))
6485 +#define str "1\n0\n0\n"
6486 + p->n = sizeof(str) - 1;
6487 + strcpy(p->a, str);
6490 + si_read_unlock(sb);
6493 + file->private_data = p;
6494 + goto out; /* success */
6497 + free_page((unsigned long)p);
6502 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6503 + size_t count, loff_t *ppos)
6505 + struct dbgaufs_plink_arg *p;
6507 + p = file->private_data;
6508 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6511 +static const struct file_operations dbgaufs_plink_fop = {
6512 + .owner = THIS_MODULE,
6513 + .open = dbgaufs_plink_open,
6514 + .release = dbgaufs_plink_release,
6515 + .read = dbgaufs_plink_read
6518 +/* ---------------------------------------------------------------------- */
6520 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6523 + struct au_sbinfo *sbinfo;
6524 + struct super_block *sb;
6526 + sbinfo = inode->i_private;
6527 + sb = sbinfo->si_sb;
6528 + si_noflush_read_lock(sb);
6529 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0, /*cnt*/0);
6530 + si_read_unlock(sb);
6534 +static const struct file_operations dbgaufs_xib_fop = {
6535 + .owner = THIS_MODULE,
6536 + .open = dbgaufs_xib_open,
6537 + .release = dbgaufs_xi_release,
6538 + .read = dbgaufs_xi_read
6541 +/* ---------------------------------------------------------------------- */
6543 +#define DbgaufsXi_PREFIX "xi"
6545 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6549 + aufs_bindex_t bindex;
6550 + char *p, a[sizeof(DbgaufsXi_PREFIX) + 8];
6551 + struct au_sbinfo *sbinfo;
6552 + struct super_block *sb;
6553 + struct au_xino *xi;
6555 + struct qstr *name;
6556 + struct au_branch *br;
6559 + name = &file->f_path.dentry->d_name;
6560 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6561 + || memcmp(name->name, DbgaufsXi_PREFIX,
6562 + sizeof(DbgaufsXi_PREFIX) - 1)))
6565 + AuDebugOn(name->len >= sizeof(a));
6566 + memcpy(a, name->name, name->len);
6567 + a[name->len] = '\0';
6568 + p = strchr(a, '-');
6571 + err = kstrtol(a + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6572 + if (unlikely(err))
6577 + err = kstrtol(p + 1, 10, &l);
6578 + if (unlikely(err))
6584 + sbinfo = inode->i_private;
6585 + sb = sbinfo->si_sb;
6586 + si_noflush_read_lock(sb);
6587 + if (unlikely(bindex < 0 || bindex > au_sbbot(sb)))
6589 + br = au_sbr(sb, bindex);
6591 + if (unlikely(idx >= xi->xi_nfile))
6593 + xf = au_xino_file(xi, idx);
6595 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1,
6596 + au_xino_count(br));
6599 + si_read_unlock(sb);
6605 +static const struct file_operations dbgaufs_xino_fop = {
6606 + .owner = THIS_MODULE,
6607 + .open = dbgaufs_xino_open,
6608 + .release = dbgaufs_xi_release,
6609 + .read = dbgaufs_xi_read
6612 +void dbgaufs_xino_del(struct au_branch *br)
6614 + struct dentry *dbgaufs;
6616 + dbgaufs = br->br_dbgaufs;
6620 + br->br_dbgaufs = NULL;
6621 + /* debugfs acquires the parent i_mutex */
6623 + debugfs_remove(dbgaufs);
6627 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6629 + aufs_bindex_t bbot;
6630 + struct au_branch *br;
6632 + if (!au_sbi(sb)->si_dbgaufs)
6635 + bbot = au_sbbot(sb);
6636 + for (; bindex <= bbot; bindex++) {
6637 + br = au_sbr(sb, bindex);
6638 + dbgaufs_xino_del(br);
6642 +static void dbgaufs_br_do_add(struct super_block *sb, aufs_bindex_t bindex,
6643 + unsigned int idx, struct dentry *parent,
6644 + struct au_sbinfo *sbinfo)
6646 + struct au_branch *br;
6648 + /* "xi" bindex(5) "-" idx(2) NULL */
6649 + char name[sizeof(DbgaufsXi_PREFIX) + 8];
6652 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6654 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d-%u",
6656 + br = au_sbr(sb, bindex);
6657 + if (br->br_dbgaufs) {
6658 + struct qstr qstr = QSTR_INIT(name, strlen(name));
6660 + if (!au_qstreq(&br->br_dbgaufs->d_name, &qstr)) {
6661 + /* debugfs acquires the parent i_mutex */
6663 + d = debugfs_rename(parent, br->br_dbgaufs, parent,
6667 + pr_warn("failed renaming %pd/%s, ignored.\n",
6672 + br->br_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6673 + sbinfo, &dbgaufs_xino_fop);
6675 + if (unlikely(!br->br_dbgaufs))
6676 + pr_warn("failed creating %pd/%s, ignored.\n",
6681 +static void dbgaufs_br_add(struct super_block *sb, aufs_bindex_t bindex,
6682 + struct dentry *parent, struct au_sbinfo *sbinfo)
6684 + struct au_branch *br;
6685 + struct au_xino *xi;
6688 + br = au_sbr(sb, bindex);
6690 + for (u = 0; u < xi->xi_nfile; u++)
6691 + dbgaufs_br_do_add(sb, bindex, u, parent, sbinfo);
6694 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown)
6696 + struct au_sbinfo *sbinfo;
6697 + struct dentry *parent;
6698 + aufs_bindex_t bbot;
6700 + if (!au_opt_test(au_mntflags(sb), XINO))
6703 + sbinfo = au_sbi(sb);
6704 + parent = sbinfo->si_dbgaufs;
6708 + bbot = au_sbbot(sb);
6710 + for (; bindex <= bbot; bindex++)
6711 + dbgaufs_br_add(sb, bindex, parent, sbinfo);
6713 + for (; bbot >= bindex; bbot--)
6714 + dbgaufs_br_add(sb, bbot, parent, sbinfo);
6717 +/* ---------------------------------------------------------------------- */
6719 +#ifdef CONFIG_AUFS_EXPORT
6720 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6723 + struct au_sbinfo *sbinfo;
6724 + struct super_block *sb;
6726 + sbinfo = inode->i_private;
6727 + sb = sbinfo->si_sb;
6728 + si_noflush_read_lock(sb);
6729 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0, /*cnt*/0);
6730 + si_read_unlock(sb);
6734 +static const struct file_operations dbgaufs_xigen_fop = {
6735 + .owner = THIS_MODULE,
6736 + .open = dbgaufs_xigen_open,
6737 + .release = dbgaufs_xi_release,
6738 + .read = dbgaufs_xi_read
6741 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6746 + * This function is a dynamic '__init' function actually,
6747 + * so the tiny check for si_rwsem is unnecessary.
6749 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6752 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
6753 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6754 + &dbgaufs_xigen_fop);
6755 + if (sbinfo->si_dbgaufs_xigen)
6761 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6765 +#endif /* CONFIG_AUFS_EXPORT */
6767 +/* ---------------------------------------------------------------------- */
6769 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6772 + * This function is a dynamic '__fin' function actually,
6773 + * so the tiny check for si_rwsem is unnecessary.
6775 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6777 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
6778 + sbinfo->si_dbgaufs = NULL;
6781 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6784 + char name[SysaufsSiNameLen];
6787 + * This function is a dynamic '__init' function actually,
6788 + * so the tiny check for si_rwsem is unnecessary.
6790 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6794 + AuErr1("/debug/aufs is uninitialized\n");
6799 + sysaufs_name(sbinfo, name);
6800 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6801 + if (unlikely(!sbinfo->si_dbgaufs))
6804 + /* regardless plink/noplink option */
6805 + sbinfo->si_dbgaufs_plink = debugfs_create_file
6806 + ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6807 + &dbgaufs_plink_fop);
6808 + if (unlikely(!sbinfo->si_dbgaufs_plink))
6811 + /* regardless xino/noxino option */
6812 + sbinfo->si_dbgaufs_xib = debugfs_create_file
6813 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6814 + &dbgaufs_xib_fop);
6815 + if (unlikely(!sbinfo->si_dbgaufs_xib))
6818 + err = dbgaufs_xigen_init(sbinfo);
6820 + goto out; /* success */
6823 + dbgaufs_si_fin(sbinfo);
6825 + if (unlikely(err))
6826 + pr_err("debugfs/aufs failed\n");
6830 +/* ---------------------------------------------------------------------- */
6832 +void dbgaufs_fin(void)
6834 + debugfs_remove(dbgaufs);
6837 +int __init dbgaufs_init(void)
6842 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6847 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6848 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
6849 +++ linux/fs/aufs/dbgaufs.h 2022-03-21 14:49:05.719966344 +0100
6851 +/* SPDX-License-Identifier: GPL-2.0 */
6853 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6855 + * This program, aufs is free software; you can redistribute it and/or modify
6856 + * it under the terms of the GNU General Public License as published by
6857 + * the Free Software Foundation; either version 2 of the License, or
6858 + * (at your option) any later version.
6860 + * This program is distributed in the hope that it will be useful,
6861 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6862 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6863 + * GNU General Public License for more details.
6865 + * You should have received a copy of the GNU General Public License
6866 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6870 + * debugfs interface
6873 +#ifndef __DBGAUFS_H__
6874 +#define __DBGAUFS_H__
6878 +struct super_block;
6882 +#ifdef CONFIG_DEBUG_FS
6884 +void dbgaufs_xino_del(struct au_branch *br);
6885 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6886 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown);
6887 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6888 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6889 +void dbgaufs_fin(void);
6890 +int __init dbgaufs_init(void);
6892 +AuStubVoid(dbgaufs_xino_del, struct au_branch *br)
6893 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6894 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex,
6896 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6897 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6898 +AuStubVoid(dbgaufs_fin, void)
6899 +AuStubInt0(__init dbgaufs_init, void)
6900 +#endif /* CONFIG_DEBUG_FS */
6902 +#endif /* __KERNEL__ */
6903 +#endif /* __DBGAUFS_H__ */
6904 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6905 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
6906 +++ linux/fs/aufs/dcsub.c 2022-03-21 14:49:05.719966344 +0100
6908 +// SPDX-License-Identifier: GPL-2.0
6910 + * Copyright (C) 2005-2021 Junjiro R. Okajima
6912 + * This program, aufs is free software; you can redistribute it and/or modify
6913 + * it under the terms of the GNU General Public License as published by
6914 + * the Free Software Foundation; either version 2 of the License, or
6915 + * (at your option) any later version.
6917 + * This program is distributed in the hope that it will be useful,
6918 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6919 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6920 + * GNU General Public License for more details.
6922 + * You should have received a copy of the GNU General Public License
6923 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6927 + * sub-routines for dentry cache
6932 +static void au_dpage_free(struct au_dpage *dpage)
6935 + struct dentry **p;
6937 + p = dpage->dentries;
6938 + for (i = 0; i < dpage->ndentry; i++)
6940 + free_page((unsigned long)dpage->dentries);
6943 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6949 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6950 + if (unlikely(!dpages->dpages))
6953 + p = (void *)__get_free_page(gfp);
6957 + dpages->dpages[0].ndentry = 0;
6958 + dpages->dpages[0].dentries = p;
6959 + dpages->ndpage = 1;
6960 + return 0; /* success */
6963 + au_kfree_try_rcu(dpages->dpages);
6968 +void au_dpages_free(struct au_dcsub_pages *dpages)
6971 + struct au_dpage *p;
6973 + p = dpages->dpages;
6974 + for (i = 0; i < dpages->ndpage; i++)
6975 + au_dpage_free(p++);
6976 + au_kfree_try_rcu(dpages->dpages);
6979 +static int au_dpages_append(struct au_dcsub_pages *dpages,
6980 + struct dentry *dentry, gfp_t gfp)
6983 + struct au_dpage *dpage;
6986 + dpage = dpages->dpages + dpages->ndpage - 1;
6987 + sz = PAGE_SIZE / sizeof(dentry);
6988 + if (unlikely(dpage->ndentry >= sz)) {
6989 + AuLabel(new dpage);
6991 + sz = dpages->ndpage * sizeof(*dpages->dpages);
6992 + p = au_kzrealloc(dpages->dpages, sz,
6993 + sz + sizeof(*dpages->dpages), gfp,
6998 + dpages->dpages = p;
6999 + dpage = dpages->dpages + dpages->ndpage;
7000 + p = (void *)__get_free_page(gfp);
7004 + dpage->ndentry = 0;
7005 + dpage->dentries = p;
7009 + AuDebugOn(au_dcount(dentry) <= 0);
7010 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
7011 + return 0; /* success */
7017 +/* todo: BAD approach */
7018 +/* copied from linux/fs/dcache.c */
7026 +extern void d_walk(struct dentry *parent, void *data,
7027 + enum d_walk_ret (*enter)(void *, struct dentry *));
7029 +struct ac_dpages_arg {
7031 + struct au_dcsub_pages *dpages;
7032 + struct super_block *sb;
7033 + au_dpages_test test;
7037 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
7039 + enum d_walk_ret ret;
7040 + struct ac_dpages_arg *arg = _arg;
7042 + ret = D_WALK_CONTINUE;
7043 + if (dentry->d_sb == arg->sb
7044 + && !IS_ROOT(dentry)
7045 + && au_dcount(dentry) > 0
7047 + && (!arg->test || arg->test(dentry, arg->arg))) {
7048 + arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
7049 + if (unlikely(arg->err))
7050 + ret = D_WALK_QUIT;
7056 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7057 + au_dpages_test test, void *arg)
7059 + struct ac_dpages_arg args = {
7067 + d_walk(root, &args, au_call_dpages_append);
7072 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7073 + int do_include, au_dpages_test test, void *arg)
7078 + write_seqlock(&rename_lock);
7079 + spin_lock(&dentry->d_lock);
7081 + && au_dcount(dentry) > 0
7082 + && (!test || test(dentry, arg)))
7083 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7084 + spin_unlock(&dentry->d_lock);
7085 + if (unlikely(err))
7089 + * RCU for vfsmount is unnecessary since this is a traverse in a single
7092 + while (!IS_ROOT(dentry)) {
7093 + dentry = dentry->d_parent; /* rename_lock is locked */
7094 + spin_lock(&dentry->d_lock);
7095 + if (au_dcount(dentry) > 0
7096 + && (!test || test(dentry, arg)))
7097 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7098 + spin_unlock(&dentry->d_lock);
7099 + if (unlikely(err))
7104 + write_sequnlock(&rename_lock);
7108 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
7110 + return au_di(dentry) && dentry->d_sb == arg;
7113 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7114 + struct dentry *dentry, int do_include)
7116 + return au_dcsub_pages_rev(dpages, dentry, do_include,
7117 + au_dcsub_dpages_aufs, dentry->d_sb);
7120 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
7122 + struct path path[2] = {
7131 + return path_is_under(path + 0, path + 1);
7133 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
7134 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
7135 +++ linux/fs/aufs/dcsub.h 2022-03-21 14:49:05.719966344 +0100
7137 +/* SPDX-License-Identifier: GPL-2.0 */
7139 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7141 + * This program, aufs is free software; you can redistribute it and/or modify
7142 + * it under the terms of the GNU General Public License as published by
7143 + * the Free Software Foundation; either version 2 of the License, or
7144 + * (at your option) any later version.
7146 + * This program is distributed in the hope that it will be useful,
7147 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7148 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7149 + * GNU General Public License for more details.
7151 + * You should have received a copy of the GNU General Public License
7152 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7156 + * sub-routines for dentry cache
7159 +#ifndef __AUFS_DCSUB_H__
7160 +#define __AUFS_DCSUB_H__
7164 +#include <linux/dcache.h>
7165 +#include <linux/fs.h>
7169 + struct dentry **dentries;
7172 +struct au_dcsub_pages {
7174 + struct au_dpage *dpages;
7177 +/* ---------------------------------------------------------------------- */
7180 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
7181 +void au_dpages_free(struct au_dcsub_pages *dpages);
7182 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
7183 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7184 + au_dpages_test test, void *arg);
7185 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7186 + int do_include, au_dpages_test test, void *arg);
7187 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7188 + struct dentry *dentry, int do_include);
7189 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
7191 +/* ---------------------------------------------------------------------- */
7194 + * todo: in linux-3.13, several similar (but faster) helpers are added to
7195 + * include/linux/dcache.h. Try them (in the future).
7198 +static inline int au_d_hashed_positive(struct dentry *d)
7201 + struct inode *inode = d_inode(d);
7204 + if (unlikely(d_unhashed(d)
7205 + || d_is_negative(d)
7206 + || !inode->i_nlink))
7211 +static inline int au_d_linkable(struct dentry *d)
7214 + struct inode *inode = d_inode(d);
7216 + err = au_d_hashed_positive(d);
7218 + && d_is_positive(d)
7219 + && (inode->i_state & I_LINKABLE))
7224 +static inline int au_d_alive(struct dentry *d)
7227 + struct inode *inode;
7231 + err = au_d_hashed_positive(d);
7233 + inode = d_inode(d);
7234 + if (unlikely(d_unlinked(d)
7235 + || d_is_negative(d)
7236 + || !inode->i_nlink))
7242 +static inline int au_alive_dir(struct dentry *d)
7246 + err = au_d_alive(d);
7247 + if (unlikely(err || IS_DEADDIR(d_inode(d))))
7252 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
7254 + return a->len == b->len
7255 + && !memcmp(a->name, b->name, a->len);
7260 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
7262 + * the type of d_lockref.count became int, but the inlined function d_count()
7263 + * still returns unsigned int.
7264 + * I don't know why. Maybe it is for every d_count() users?
7265 + * Anyway au_dcount() lives on.
7267 +static inline int au_dcount(struct dentry *d)
7269 + return (int)d_count(d);
7272 +#endif /* __KERNEL__ */
7273 +#endif /* __AUFS_DCSUB_H__ */
7274 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
7275 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
7276 +++ linux/fs/aufs/debug.c 2022-03-21 14:49:05.719966344 +0100
7278 +// SPDX-License-Identifier: GPL-2.0
7280 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7282 + * This program, aufs is free software; you can redistribute it and/or modify
7283 + * it under the terms of the GNU General Public License as published by
7284 + * the Free Software Foundation; either version 2 of the License, or
7285 + * (at your option) any later version.
7287 + * This program is distributed in the hope that it will be useful,
7288 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7289 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7290 + * GNU General Public License for more details.
7292 + * You should have received a copy of the GNU General Public License
7293 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7297 + * debug print functions
7300 +#include <linux/iversion.h>
7303 +/* Returns 0, or -errno. arg is in kp->arg. */
7304 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
7308 + err = kstrtoint(val, 0, &n);
7318 +/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
7319 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
7324 + return sprintf(buffer, "%d", atomic_read(a));
7327 +static struct kernel_param_ops param_ops_atomic_t = {
7328 + .set = param_atomic_t_set,
7329 + .get = param_atomic_t_get
7330 + /* void (*free)(void *arg) */
7333 +atomic_t aufs_debug = ATOMIC_INIT(0);
7334 +MODULE_PARM_DESC(debug, "debug print");
7335 +module_param_named(debug, aufs_debug, atomic_t, 0664);
7337 +DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
7338 +char *au_plevel = KERN_DEBUG;
7339 +#define dpri(fmt, ...) do { \
7341 + && strcmp(au_plevel, KERN_DEBUG)) \
7342 + || au_debug_test()) \
7343 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
7346 +/* ---------------------------------------------------------------------- */
7348 +void au_dpri_whlist(struct au_nhash *whlist)
7350 + unsigned long ul, n;
7351 + struct hlist_head *head;
7352 + struct au_vdir_wh *pos;
7354 + n = whlist->nh_num;
7355 + head = whlist->nh_head;
7356 + for (ul = 0; ul < n; ul++) {
7357 + hlist_for_each_entry(pos, head, wh_hash)
7358 + dpri("b%d, %.*s, %d\n",
7360 + pos->wh_str.len, pos->wh_str.name,
7366 +void au_dpri_vdir(struct au_vdir *vdir)
7369 + union au_vdir_deblk_p p;
7372 + if (!vdir || IS_ERR(vdir)) {
7373 + dpri("err %ld\n", PTR_ERR(vdir));
7377 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %llu\n",
7378 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
7379 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
7380 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
7381 + p.deblk = vdir->vd_deblk[ul];
7383 + dpri("[%lu]: %p\n", ul, o);
7387 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
7388 + struct dentry *wh)
7393 + if (!inode || IS_ERR(inode)) {
7394 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
7398 + /* the type of i_blocks depends upon CONFIG_LBDAF */
7399 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
7400 + && sizeof(inode->i_blocks) != sizeof(u64));
7402 + n = (void *)wh->d_name.name;
7403 + l = wh->d_name.len;
7406 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
7407 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
7409 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
7410 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
7411 + i_size_read(inode), (unsigned long long)inode->i_blocks,
7412 + hn, (long long)timespec64_to_ns(&inode->i_ctime) & 0x0ffff,
7413 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
7414 + inode->i_state, inode->i_flags, inode_peek_iversion(inode),
7415 + inode->i_generation,
7416 + l ? ", wh " : "", l, n);
7420 +void au_dpri_inode(struct inode *inode)
7422 + struct au_iinfo *iinfo;
7423 + struct au_hinode *hi;
7424 + aufs_bindex_t bindex;
7427 + err = do_pri_inode(-1, inode, -1, NULL);
7428 + if (err || !au_test_aufs(inode->i_sb) || au_is_bad_inode(inode))
7431 + iinfo = au_ii(inode);
7432 + dpri("i-1: btop %d, bbot %d, gen %d\n",
7433 + iinfo->ii_btop, iinfo->ii_bbot, au_iigen(inode, NULL));
7434 + if (iinfo->ii_btop < 0)
7437 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot; bindex++) {
7438 + hi = au_hinode(iinfo, bindex);
7440 + do_pri_inode(bindex, hi->hi_inode, hn, hi->hi_whdentry);
7444 +void au_dpri_dalias(struct inode *inode)
7448 + spin_lock(&inode->i_lock);
7449 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
7450 + au_dpri_dentry(d);
7451 + spin_unlock(&inode->i_lock);
7454 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
7456 + struct dentry *wh = NULL;
7458 + struct inode *inode;
7459 + struct au_iinfo *iinfo;
7460 + struct au_hinode *hi;
7462 + if (!dentry || IS_ERR(dentry)) {
7463 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
7466 + /* do not call dget_parent() here */
7467 + /* note: access d_xxx without d_lock */
7468 + dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
7469 + bindex, dentry, dentry,
7470 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
7471 + au_dcount(dentry), dentry->d_flags,
7472 + d_unhashed(dentry) ? "un" : "");
7475 + if (d_is_positive(dentry))
7476 + inode = d_inode(dentry);
7478 + && au_test_aufs(dentry->d_sb)
7480 + && !au_is_bad_inode(inode)) {
7481 + iinfo = au_ii(inode);
7482 + hi = au_hinode(iinfo, bindex);
7484 + wh = hi->hi_whdentry;
7486 + do_pri_inode(bindex, inode, hn, wh);
7490 +void au_dpri_dentry(struct dentry *dentry)
7492 + struct au_dinfo *dinfo;
7493 + aufs_bindex_t bindex;
7496 + err = do_pri_dentry(-1, dentry);
7497 + if (err || !au_test_aufs(dentry->d_sb))
7500 + dinfo = au_di(dentry);
7503 + dpri("d-1: btop %d, bbot %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
7504 + dinfo->di_btop, dinfo->di_bbot,
7505 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
7506 + dinfo->di_tmpfile);
7507 + if (dinfo->di_btop < 0)
7509 + for (bindex = dinfo->di_btop; bindex <= dinfo->di_bbot; bindex++)
7510 + do_pri_dentry(bindex, au_hdentry(dinfo, bindex)->hd_dentry);
7513 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7517 + if (!file || IS_ERR(file)) {
7518 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7523 + && !IS_ERR_OR_NULL(file->f_path.dentry)
7524 + && au_test_aufs(file->f_path.dentry->d_sb)
7526 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7527 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7528 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7529 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
7530 + file->f_version, file->f_pos, a);
7531 + if (!IS_ERR_OR_NULL(file->f_path.dentry))
7532 + do_pri_dentry(bindex, file->f_path.dentry);
7536 +void au_dpri_file(struct file *file)
7538 + struct au_finfo *finfo;
7539 + struct au_fidir *fidir;
7540 + struct au_hfile *hfile;
7541 + aufs_bindex_t bindex;
7544 + err = do_pri_file(-1, file);
7546 + || IS_ERR_OR_NULL(file->f_path.dentry)
7547 + || !au_test_aufs(file->f_path.dentry->d_sb))
7550 + finfo = au_fi(file);
7553 + if (finfo->fi_btop < 0)
7555 + fidir = finfo->fi_hdir;
7557 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7559 + for (bindex = finfo->fi_btop;
7560 + bindex >= 0 && bindex <= fidir->fd_bbot;
7562 + hfile = fidir->fd_hfile + bindex;
7563 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7567 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7569 + struct vfsmount *mnt;
7570 + struct super_block *sb;
7572 + if (!br || IS_ERR(br))
7574 + mnt = au_br_mnt(br);
7575 + if (!mnt || IS_ERR(mnt))
7578 + if (!sb || IS_ERR(sb))
7581 + dpri("s%d: {perm 0x%x, id %d, wbr %p}, "
7582 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7584 + bindex, br->br_perm, br->br_id, br->br_wbr,
7585 + au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7586 + sb->s_flags, sb->s_count,
7587 + atomic_read(&sb->s_active),
7588 + !!au_xino_file(br->br_xino, /*idx*/-1));
7592 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7596 +void au_dpri_sb(struct super_block *sb)
7598 + struct au_sbinfo *sbinfo;
7599 + aufs_bindex_t bindex;
7601 + /* to reduce stack size */
7603 + struct vfsmount mnt;
7604 + struct au_branch fake;
7607 + /* this function can be called from magic sysrq */
7608 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
7609 + if (unlikely(!a)) {
7610 + dpri("no memory\n");
7614 + a->mnt.mnt_sb = sb;
7615 + a->fake.br_path.mnt = &a->mnt;
7616 + err = do_pri_br(-1, &a->fake);
7618 + dpri("dev 0x%x\n", sb->s_dev);
7619 + if (err || !au_test_aufs(sb))
7622 + sbinfo = au_sbi(sb);
7625 + dpri("nw %d, gen %u, kobj %d\n",
7626 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7627 + kref_read(&sbinfo->si_kobj.kref));
7628 + for (bindex = 0; bindex <= sbinfo->si_bbot; bindex++)
7629 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7632 +/* ---------------------------------------------------------------------- */
7634 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7636 + struct inode *h_inode, *inode = d_inode(dentry);
7637 + struct dentry *h_dentry;
7638 + aufs_bindex_t bindex, bbot, bi;
7640 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7643 + bbot = au_dbbot(dentry);
7644 + bi = au_ibbot(inode);
7647 + bindex = au_dbtop(dentry);
7648 + bi = au_ibtop(inode);
7652 + for (; bindex <= bbot; bindex++) {
7653 + h_dentry = au_h_dptr(dentry, bindex);
7656 + h_inode = au_h_iptr(inode, bindex);
7657 + if (unlikely(h_inode != d_inode(h_dentry))) {
7659 + AuDbg("b%d, %s:%d\n", bindex, func, line);
7660 + AuDbgDentry(dentry);
7661 + AuDbgInode(inode);
7663 + if (au_test_fuse(h_inode->i_sb))
7671 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7674 + struct au_dcsub_pages dpages;
7675 + struct au_dpage *dpage;
7676 + struct dentry **dentries;
7678 + err = au_dpages_init(&dpages, GFP_NOFS);
7680 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7682 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7683 + dpage = dpages.dpages + i;
7684 + dentries = dpage->dentries;
7685 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7686 + AuDebugOn(au_digen_test(dentries[j], sigen));
7688 + au_dpages_free(&dpages);
7691 +void au_dbg_verify_kthread(void)
7693 + if (au_wkq_test()) {
7696 + * It may be recursive, but udba=notify between two aufs mounts,
7697 + * where a single ro branch is shared, is not a problem.
7703 +/* ---------------------------------------------------------------------- */
7705 +int __init au_debug_init(void)
7707 + aufs_bindex_t bindex;
7708 + struct au_vdir_destr destr;
7711 + AuDebugOn(bindex >= 0);
7714 + AuDebugOn(destr.len < NAME_MAX);
7716 +#ifdef CONFIG_4KSTACKS
7717 + pr_warn("CONFIG_4KSTACKS is defined.\n");
7722 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7723 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
7724 +++ linux/fs/aufs/debug.h 2022-03-21 14:49:05.719966344 +0100
7726 +/* SPDX-License-Identifier: GPL-2.0 */
7728 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7730 + * This program, aufs is free software; you can redistribute it and/or modify
7731 + * it under the terms of the GNU General Public License as published by
7732 + * the Free Software Foundation; either version 2 of the License, or
7733 + * (at your option) any later version.
7735 + * This program is distributed in the hope that it will be useful,
7736 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7737 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7738 + * GNU General Public License for more details.
7740 + * You should have received a copy of the GNU General Public License
7741 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7745 + * debug print functions
7748 +#ifndef __AUFS_DEBUG_H__
7749 +#define __AUFS_DEBUG_H__
7753 +#include <linux/atomic.h>
7754 +#include <linux/module.h>
7755 +#include <linux/kallsyms.h>
7756 +#include <linux/sysrq.h>
7758 +#ifdef CONFIG_AUFS_DEBUG
7759 +#define AuDebugOn(a) BUG_ON(a)
7761 +/* module parameter */
7762 +extern atomic_t aufs_debug;
7763 +static inline void au_debug_on(void)
7765 + atomic_inc(&aufs_debug);
7767 +static inline void au_debug_off(void)
7769 + atomic_dec_if_positive(&aufs_debug);
7772 +static inline int au_debug_test(void)
7774 + return atomic_read(&aufs_debug) > 0;
7777 +#define AuDebugOn(a) do {} while (0)
7778 +AuStubVoid(au_debug_on, void)
7779 +AuStubVoid(au_debug_off, void)
7780 +AuStubInt0(au_debug_test, void)
7781 +#endif /* CONFIG_AUFS_DEBUG */
7783 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7785 +/* ---------------------------------------------------------------------- */
7789 +#define AuDbg(fmt, ...) do { \
7790 + if (au_debug_test()) \
7791 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7793 +#define AuLabel(l) AuDbg(#l "\n")
7794 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7795 +#define AuWarn1(fmt, ...) do { \
7796 + static unsigned char _c; \
7798 + pr_warn(fmt, ##__VA_ARGS__); \
7801 +#define AuErr1(fmt, ...) do { \
7802 + static unsigned char _c; \
7804 + pr_err(fmt, ##__VA_ARGS__); \
7807 +#define AuIOErr1(fmt, ...) do { \
7808 + static unsigned char _c; \
7810 + AuIOErr(fmt, ##__VA_ARGS__); \
7813 +#define AuUnsupportMsg "This operation is not supported." \
7814 + " Please report this application to aufs-users ML."
7815 +#define AuUnsupport(fmt, ...) do { \
7816 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7820 +#define AuTraceErr(e) do { \
7821 + if (unlikely((e) < 0)) \
7822 + AuDbg("err %d\n", (int)(e)); \
7825 +#define AuTraceErrPtr(p) do { \
7827 + AuDbg("err %ld\n", PTR_ERR(p)); \
7830 +/* dirty macros for debug print, use with "%.*s" and caution */
7831 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
7833 +/* ---------------------------------------------------------------------- */
7836 +#ifdef CONFIG_AUFS_DEBUG
7837 +extern struct mutex au_dbg_mtx;
7838 +extern char *au_plevel;
7840 +void au_dpri_whlist(struct au_nhash *whlist);
7842 +void au_dpri_vdir(struct au_vdir *vdir);
7844 +void au_dpri_inode(struct inode *inode);
7845 +void au_dpri_dalias(struct inode *inode);
7846 +void au_dpri_dentry(struct dentry *dentry);
7848 +void au_dpri_file(struct file *filp);
7849 +struct super_block;
7850 +void au_dpri_sb(struct super_block *sb);
7852 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7853 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7854 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7855 +void au_dbg_verify_kthread(void);
7857 +int __init au_debug_init(void);
7859 +#define AuDbgWhlist(w) do { \
7860 + mutex_lock(&au_dbg_mtx); \
7862 + au_dpri_whlist(w); \
7863 + mutex_unlock(&au_dbg_mtx); \
7866 +#define AuDbgVdir(v) do { \
7867 + mutex_lock(&au_dbg_mtx); \
7869 + au_dpri_vdir(v); \
7870 + mutex_unlock(&au_dbg_mtx); \
7873 +#define AuDbgInode(i) do { \
7874 + mutex_lock(&au_dbg_mtx); \
7876 + au_dpri_inode(i); \
7877 + mutex_unlock(&au_dbg_mtx); \
7880 +#define AuDbgDAlias(i) do { \
7881 + mutex_lock(&au_dbg_mtx); \
7883 + au_dpri_dalias(i); \
7884 + mutex_unlock(&au_dbg_mtx); \
7887 +#define AuDbgDentry(d) do { \
7888 + mutex_lock(&au_dbg_mtx); \
7890 + au_dpri_dentry(d); \
7891 + mutex_unlock(&au_dbg_mtx); \
7894 +#define AuDbgFile(f) do { \
7895 + mutex_lock(&au_dbg_mtx); \
7897 + au_dpri_file(f); \
7898 + mutex_unlock(&au_dbg_mtx); \
7901 +#define AuDbgSb(sb) do { \
7902 + mutex_lock(&au_dbg_mtx); \
7903 + AuDbg(#sb "\n"); \
7905 + mutex_unlock(&au_dbg_mtx); \
7908 +#define AuDbgSym(addr) do { \
7909 + char sym[KSYM_SYMBOL_LEN]; \
7910 + sprint_symbol(sym, (unsigned long)addr); \
7911 + AuDbg("%s\n", sym); \
7914 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7915 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7916 +AuStubVoid(au_dbg_verify_kthread, void)
7917 +AuStubInt0(__init au_debug_init, void)
7919 +#define AuDbgWhlist(w) do {} while (0)
7920 +#define AuDbgVdir(v) do {} while (0)
7921 +#define AuDbgInode(i) do {} while (0)
7922 +#define AuDbgDAlias(i) do {} while (0)
7923 +#define AuDbgDentry(d) do {} while (0)
7924 +#define AuDbgFile(f) do {} while (0)
7925 +#define AuDbgSb(sb) do {} while (0)
7926 +#define AuDbgSym(addr) do {} while (0)
7927 +#endif /* CONFIG_AUFS_DEBUG */
7929 +/* ---------------------------------------------------------------------- */
7931 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7932 +int __init au_sysrq_init(void);
7933 +void au_sysrq_fin(void);
7935 +#ifdef CONFIG_HW_CONSOLE
7936 +#define au_dbg_blocked() do { \
7938 + handle_sysrq('w'); \
7941 +AuStubVoid(au_dbg_blocked, void)
7945 +AuStubInt0(__init au_sysrq_init, void)
7946 +AuStubVoid(au_sysrq_fin, void)
7947 +AuStubVoid(au_dbg_blocked, void)
7948 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7950 +#endif /* __KERNEL__ */
7951 +#endif /* __AUFS_DEBUG_H__ */
7952 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7953 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
7954 +++ linux/fs/aufs/dentry.c 2022-03-21 14:49:05.719966344 +0100
7956 +// SPDX-License-Identifier: GPL-2.0
7958 + * Copyright (C) 2005-2021 Junjiro R. Okajima
7960 + * This program, aufs is free software; you can redistribute it and/or modify
7961 + * it under the terms of the GNU General Public License as published by
7962 + * the Free Software Foundation; either version 2 of the License, or
7963 + * (at your option) any later version.
7965 + * This program is distributed in the hope that it will be useful,
7966 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7967 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7968 + * GNU General Public License for more details.
7970 + * You should have received a copy of the GNU General Public License
7971 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7975 + * lookup and dentry operations
7978 +#include <linux/iversion.h>
7982 + * returns positive/negative dentry, NULL or an error.
7983 + * NULL means whiteout-ed or not-found.
7985 +static struct dentry*
7986 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7987 + aufs_bindex_t bindex, struct au_do_lookup_args *args)
7989 + struct dentry *h_dentry;
7990 + struct inode *h_inode;
7991 + struct au_branch *br;
7992 + struct user_namespace *h_userns;
7993 + struct path h_path;
7994 + int wh_found, opq;
7995 + unsigned char wh_able;
7996 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
7997 + const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
8001 + br = au_sbr(dentry->d_sb, bindex);
8002 + h_path.dentry = h_parent;
8003 + h_path.mnt = au_br_mnt(br);
8004 + h_userns = au_br_userns(br);
8005 + wh_able = !!au_br_whable(br->br_perm);
8007 + wh_found = au_wh_test(h_userns, &h_path, &args->whname,
8009 + h_dentry = ERR_PTR(wh_found);
8012 + if (unlikely(wh_found < 0))
8015 + /* We found a whiteout */
8016 + /* au_set_dbbot(dentry, bindex); */
8017 + au_set_dbwh(dentry, bindex);
8019 + return NULL; /* success */
8023 + h_dentry = vfsub_lkup_one(args->name, &h_path);
8025 + h_dentry = au_sio_lkup_one(h_userns, args->name, &h_path);
8026 + if (IS_ERR(h_dentry)) {
8027 + if (PTR_ERR(h_dentry) == -ENAMETOOLONG
8033 + h_inode = d_inode(h_dentry);
8034 + if (d_is_negative(h_dentry)) {
8037 + } else if (wh_found
8038 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
8040 + else if (au_ftest_lkup(args->flags, DIRREN)
8042 + && !au_dr_lkup_h_ino(args, bindex, h_inode->i_ino)) {
8043 + AuDbg("b%d %pd ignored hi%llu\n", bindex, h_dentry,
8044 + (unsigned long long)h_inode->i_ino);
8048 + if (au_dbbot(dentry) <= bindex)
8049 + au_set_dbbot(dentry, bindex);
8050 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
8051 + au_set_dbtop(dentry, bindex);
8052 + au_set_h_dptr(dentry, bindex, h_dentry);
8054 + if (!d_is_dir(h_dentry)
8056 + || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
8057 + goto out; /* success */
8059 + h_path.dentry = h_dentry;
8060 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
8061 + opq = au_diropq_test(h_userns, &h_path);
8062 + inode_unlock_shared(h_inode);
8064 + au_set_dbdiropq(dentry, bindex);
8065 + else if (unlikely(opq < 0)) {
8066 + au_set_h_dptr(dentry, bindex, NULL);
8067 + h_dentry = ERR_PTR(opq);
8078 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
8080 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
8081 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
8087 + * returns the number of lower positive dentries,
8088 + * otherwise an error.
8089 + * can be called at unlinking with @type is zero.
8091 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
8092 + unsigned int flags)
8094 + int npositive, err;
8095 + aufs_bindex_t bindex, btail, bdiropq;
8096 + unsigned char isdir, dirperm1, dirren;
8097 + struct au_do_lookup_args args = {
8099 + .name = &dentry->d_name
8101 + struct dentry *parent;
8102 + struct super_block *sb;
8104 + sb = dentry->d_sb;
8105 + err = au_test_shwh(sb, args.name);
8106 + if (unlikely(err))
8109 + err = au_wh_name_alloc(&args.whname, args.name);
8110 + if (unlikely(err))
8113 + isdir = !!d_is_dir(dentry);
8114 + dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
8115 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
8117 + au_fset_lkup(args.flags, DIRREN);
8120 + parent = dget_parent(dentry);
8121 + btail = au_dbtaildir(parent);
8122 + for (bindex = btop; bindex <= btail; bindex++) {
8123 + struct dentry *h_parent, *h_dentry;
8124 + struct inode *h_inode, *h_dir;
8125 + struct au_branch *br;
8127 + h_dentry = au_h_dptr(dentry, bindex);
8129 + if (d_is_positive(h_dentry))
8133 + h_parent = au_h_dptr(parent, bindex);
8134 + if (!h_parent || !d_is_dir(h_parent))
8138 + /* if the inum matches, then use the prepared name */
8139 + err = au_dr_lkup_name(&args, bindex);
8140 + if (unlikely(err))
8144 + h_dir = d_inode(h_parent);
8145 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
8146 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &args);
8147 + inode_unlock_shared(h_dir);
8148 + err = PTR_ERR(h_dentry);
8149 + if (IS_ERR(h_dentry))
8152 + au_fclr_lkup(args.flags, ALLOW_NEG);
8154 + au_fset_lkup(args.flags, IGNORE_PERM);
8156 + if (au_dbwh(dentry) == bindex)
8160 + if (d_is_negative(h_dentry))
8162 + h_inode = d_inode(h_dentry);
8165 + args.type = h_inode->i_mode & S_IFMT;
8166 + if (args.type != S_IFDIR)
8169 + /* the type of lower may be different */
8170 + bdiropq = au_dbdiropq(dentry);
8171 + if (bdiropq >= 0 && bdiropq <= bindex)
8174 + br = au_sbr(sb, bindex);
8176 + && au_dr_hino_test_add(&br->br_dirren, h_inode->i_ino,
8177 + /*add_ent*/NULL)) {
8178 + /* prepare next name to lookup */
8179 + err = au_dr_lkup(&args, dentry, bindex);
8180 + if (unlikely(err))
8186 + AuLabel(positive);
8187 + au_update_dbtop(dentry);
8190 + if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
8191 + && au_dbtop(dentry) < 0)) {
8193 + AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
8199 + au_kfree_try_rcu(args.whname.name);
8201 + au_dr_lkup_fin(&args);
8206 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
8207 + struct path *ppath)
8209 + struct dentry *dentry;
8212 + if (!au_test_h_perm_sio(userns, d_inode(ppath->dentry), MAY_EXEC))
8213 + dentry = vfsub_lkup_one(name, ppath);
8215 + struct vfsub_lkup_one_args args = {
8221 + wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
8222 + if (unlikely(wkq_err))
8223 + dentry = ERR_PTR(wkq_err);
8230 + * lookup @dentry on @bindex which should be negative.
8232 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
8235 + struct dentry *parent, *h_dentry;
8236 + struct au_branch *br;
8237 + struct user_namespace *h_userns;
8238 + struct path h_ppath;
8240 + parent = dget_parent(dentry);
8241 + br = au_sbr(dentry->d_sb, bindex);
8242 + h_ppath.dentry = au_h_dptr(parent, bindex);
8243 + h_ppath.mnt = au_br_mnt(br);
8244 + h_userns = au_br_userns(br);
8246 + h_dentry = au_whtmp_lkup(h_ppath.dentry, br, &dentry->d_name);
8248 + h_dentry = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
8249 + err = PTR_ERR(h_dentry);
8250 + if (IS_ERR(h_dentry))
8252 + if (unlikely(d_is_positive(h_dentry))) {
8254 + AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
8260 + if (bindex < au_dbtop(dentry))
8261 + au_set_dbtop(dentry, bindex);
8262 + if (au_dbbot(dentry) < bindex)
8263 + au_set_dbbot(dentry, bindex);
8264 + au_set_h_dptr(dentry, bindex, h_dentry);
8271 +/* ---------------------------------------------------------------------- */
8273 +/* subset of struct inode */
8275 + unsigned long i_ino;
8276 + /* unsigned int i_nlink; */
8282 + blkcnt_t i_blocks;
8287 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
8289 + ia->i_ino = h_inode->i_ino;
8290 + /* ia->i_nlink = h_inode->i_nlink; */
8291 + ia->i_uid = h_inode->i_uid;
8292 + ia->i_gid = h_inode->i_gid;
8293 + ia->i_version = inode_query_iversion(h_inode);
8295 + ia->i_size = h_inode->i_size;
8296 + ia->i_blocks = h_inode->i_blocks;
8298 + ia->i_mode = (h_inode->i_mode & S_IFMT);
8301 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
8303 + return ia->i_ino != h_inode->i_ino
8304 + /* || ia->i_nlink != h_inode->i_nlink */
8305 + || !uid_eq(ia->i_uid, h_inode->i_uid)
8306 + || !gid_eq(ia->i_gid, h_inode->i_gid)
8307 + || !inode_eq_iversion(h_inode, ia->i_version)
8309 + || ia->i_size != h_inode->i_size
8310 + || ia->i_blocks != h_inode->i_blocks
8312 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
8315 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
8316 + struct au_branch *br)
8319 + struct au_iattr ia;
8320 + struct inode *h_inode;
8321 + struct dentry *h_d;
8322 + struct super_block *h_sb;
8323 + struct path h_ppath;
8326 + memset(&ia, -1, sizeof(ia));
8327 + h_sb = h_dentry->d_sb;
8329 + if (d_is_positive(h_dentry)) {
8330 + h_inode = d_inode(h_dentry);
8331 + au_iattr_save(&ia, h_inode);
8332 + } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
8333 + /* nfs d_revalidate may return 0 for negative dentry */
8334 + /* fuse d_revalidate always return 0 for negative dentry */
8337 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
8338 + h_ppath.dentry = h_parent;
8339 + h_ppath.mnt = au_br_mnt(br);
8340 + h_d = vfsub_lkup_one(&h_dentry->d_name, &h_ppath);
8341 + err = PTR_ERR(h_d);
8346 + if (unlikely(h_d != h_dentry
8347 + || d_inode(h_d) != h_inode
8348 + || (h_inode && au_iattr_test(&ia, h_inode))))
8349 + err = au_busy_or_stale();
8357 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8358 + struct dentry *h_parent, struct au_branch *br)
8363 + if (udba == AuOpt_UDBA_REVAL
8364 + && !au_test_fs_remote(h_dentry->d_sb)) {
8366 + err = (d_inode(h_dentry->d_parent) != h_dir);
8367 + } else if (udba != AuOpt_UDBA_NONE)
8368 + err = au_h_verify_dentry(h_dentry, h_parent, br);
8373 +/* ---------------------------------------------------------------------- */
8375 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
8378 + aufs_bindex_t new_bindex, bindex, bbot, bwh, bdiropq;
8379 + struct au_hdentry tmp, *p, *q;
8380 + struct au_dinfo *dinfo;
8381 + struct super_block *sb;
8383 + DiMustWriteLock(dentry);
8385 + sb = dentry->d_sb;
8386 + dinfo = au_di(dentry);
8387 + bbot = dinfo->di_bbot;
8388 + bwh = dinfo->di_bwh;
8389 + bdiropq = dinfo->di_bdiropq;
8390 + bindex = dinfo->di_btop;
8391 + p = au_hdentry(dinfo, bindex);
8392 + for (; bindex <= bbot; bindex++, p++) {
8393 + if (!p->hd_dentry)
8396 + new_bindex = au_br_index(sb, p->hd_id);
8397 + if (new_bindex == bindex)
8400 + if (dinfo->di_bwh == bindex)
8402 + if (dinfo->di_bdiropq == bindex)
8403 + bdiropq = new_bindex;
8404 + if (new_bindex < 0) {
8406 + p->hd_dentry = NULL;
8410 + /* swap two lower dentries, and loop again */
8411 + q = au_hdentry(dinfo, new_bindex);
8415 + if (tmp.hd_dentry) {
8421 + dinfo->di_bwh = -1;
8422 + if (bwh >= 0 && bwh <= au_sbbot(sb) && au_sbr_whable(sb, bwh))
8423 + dinfo->di_bwh = bwh;
8425 + dinfo->di_bdiropq = -1;
8427 + && bdiropq <= au_sbbot(sb)
8428 + && au_sbr_whable(sb, bdiropq))
8429 + dinfo->di_bdiropq = bdiropq;
8432 + dinfo->di_btop = -1;
8433 + dinfo->di_bbot = -1;
8434 + bbot = au_dbbot(parent);
8436 + p = au_hdentry(dinfo, bindex);
8437 + for (; bindex <= bbot; bindex++, p++)
8438 + if (p->hd_dentry) {
8439 + dinfo->di_btop = bindex;
8443 + if (dinfo->di_btop >= 0) {
8445 + p = au_hdentry(dinfo, bindex);
8446 + for (; bindex >= 0; bindex--, p--)
8447 + if (p->hd_dentry) {
8448 + dinfo->di_bbot = bindex;
8457 +static void au_do_hide(struct dentry *dentry)
8459 + struct inode *inode;
8461 + if (d_really_is_positive(dentry)) {
8462 + inode = d_inode(dentry);
8463 + if (!d_is_dir(dentry)) {
8464 + if (inode->i_nlink && !d_unhashed(dentry))
8465 + drop_nlink(inode);
8467 + clear_nlink(inode);
8468 + /* stop next lookup */
8469 + inode->i_flags |= S_DEAD;
8471 + smp_mb(); /* necessary? */
8476 +static int au_hide_children(struct dentry *parent)
8478 + int err, i, j, ndentry;
8479 + struct au_dcsub_pages dpages;
8480 + struct au_dpage *dpage;
8481 + struct dentry *dentry;
8483 + err = au_dpages_init(&dpages, GFP_NOFS);
8484 + if (unlikely(err))
8486 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
8487 + if (unlikely(err))
8490 + /* in reverse order */
8491 + for (i = dpages.ndpage - 1; i >= 0; i--) {
8492 + dpage = dpages.dpages + i;
8493 + ndentry = dpage->ndentry;
8494 + for (j = ndentry - 1; j >= 0; j--) {
8495 + dentry = dpage->dentries[j];
8496 + if (dentry != parent)
8497 + au_do_hide(dentry);
8502 + au_dpages_free(&dpages);
8507 +static void au_hide(struct dentry *dentry)
8511 + AuDbgDentry(dentry);
8512 + if (d_is_dir(dentry)) {
8513 + /* shrink_dcache_parent(dentry); */
8514 + err = au_hide_children(dentry);
8515 + if (unlikely(err))
8516 + AuIOErr("%pd, failed hiding children, ignored %d\n",
8519 + au_do_hide(dentry);
8523 + * By adding a dirty branch, a cached dentry may be affected in various ways.
8525 + * a dirty branch is added
8526 + * - on the top of layers
8527 + * - in the middle of layers
8528 + * - to the bottom of layers
8530 + * on the added branch there exists
8533 + * - a same named entry
8535 + * * negative --> positive
8536 + * * positive --> positive
8537 + * - type is unchanged
8538 + * - type is changed
8540 + * * negative --> negative
8541 + * * positive --> negative (rejected by au_br_del() for non-dir case)
8544 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8545 + struct au_dinfo *tmp)
8548 + aufs_bindex_t bindex, bbot;
8550 + struct dentry *dentry;
8551 + struct inode *inode;
8553 + } orig_h, tmp_h = {
8556 + struct au_hdentry *hd;
8557 + struct inode *inode, *h_inode;
8558 + struct dentry *h_dentry;
8561 + AuDebugOn(dinfo->di_btop < 0);
8563 + orig_h.dentry = au_hdentry(dinfo, dinfo->di_btop)->hd_dentry;
8564 + orig_h.inode = NULL;
8565 + if (d_is_positive(orig_h.dentry)) {
8566 + orig_h.inode = d_inode(orig_h.dentry);
8567 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8569 + if (tmp->di_btop >= 0) {
8570 + tmp_h.dentry = au_hdentry(tmp, tmp->di_btop)->hd_dentry;
8571 + if (d_is_positive(tmp_h.dentry)) {
8572 + tmp_h.inode = d_inode(tmp_h.dentry);
8573 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8578 + if (d_really_is_positive(dentry))
8579 + inode = d_inode(dentry);
8580 + if (!orig_h.inode) {
8581 + AuDbg("negative originally\n");
8587 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8588 + AuDebugOn(dinfo->di_bdiropq != -1);
8590 + if (!tmp_h.inode) {
8591 + AuDbg("negative --> negative\n");
8592 + /* should have only one negative lower */
8593 + if (tmp->di_btop >= 0
8594 + && tmp->di_btop < dinfo->di_btop) {
8595 + AuDebugOn(tmp->di_btop != tmp->di_bbot);
8596 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8597 + au_set_h_dptr(dentry, dinfo->di_btop, NULL);
8598 + au_di_cp(dinfo, tmp);
8599 + hd = au_hdentry(tmp, tmp->di_btop);
8600 + au_set_h_dptr(dentry, tmp->di_btop,
8601 + dget(hd->hd_dentry));
8603 + au_dbg_verify_dinode(dentry);
8605 + AuDbg("negative --> positive\n");
8607 + * similar to the behaviour of creating with bypassing
8609 + * unhash it in order to force an error in the
8610 + * succeeding create operation.
8611 + * we should not set S_DEAD here.
8614 + /* au_di_swap(tmp, dinfo); */
8615 + au_dbg_verify_dinode(dentry);
8618 + AuDbg("positive originally\n");
8619 + /* inode may be NULL */
8620 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8621 + if (!tmp_h.inode) {
8622 + AuDbg("positive --> negative\n");
8623 + /* or bypassing aufs */
8625 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_btop)
8626 + dinfo->di_bwh = tmp->di_bwh;
8628 + err = au_refresh_hinode_self(inode);
8629 + au_dbg_verify_dinode(dentry);
8630 + } else if (orig_h.mode == tmp_h.mode) {
8631 + AuDbg("positive --> positive, same type\n");
8632 + if (!S_ISDIR(orig_h.mode)
8633 + && dinfo->di_btop > tmp->di_btop) {
8635 + * similar to the behaviour of removing and
8640 + err = au_refresh_hinode_self(inode);
8641 + au_dbg_verify_dinode(dentry);
8643 + /* fill empty slots */
8644 + if (dinfo->di_btop > tmp->di_btop)
8645 + dinfo->di_btop = tmp->di_btop;
8646 + if (dinfo->di_bbot < tmp->di_bbot)
8647 + dinfo->di_bbot = tmp->di_bbot;
8648 + dinfo->di_bwh = tmp->di_bwh;
8649 + dinfo->di_bdiropq = tmp->di_bdiropq;
8650 + bbot = dinfo->di_bbot;
8651 + bindex = tmp->di_btop;
8652 + hd = au_hdentry(tmp, bindex);
8653 + for (; bindex <= bbot; bindex++, hd++) {
8654 + if (au_h_dptr(dentry, bindex))
8656 + h_dentry = hd->hd_dentry;
8659 + AuDebugOn(d_is_negative(h_dentry));
8660 + h_inode = d_inode(h_dentry);
8661 + AuDebugOn(orig_h.mode
8662 + != (h_inode->i_mode
8664 + au_set_h_dptr(dentry, bindex,
8668 + err = au_refresh_hinode(inode, dentry);
8669 + au_dbg_verify_dinode(dentry);
8672 + AuDbg("positive --> positive, different type\n");
8673 + /* similar to the behaviour of removing and creating */
8676 + err = au_refresh_hinode_self(inode);
8677 + au_dbg_verify_dinode(dentry);
8685 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8687 + const struct dentry_operations *dop
8688 + = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8689 + static const unsigned int mask
8690 + = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8692 + BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8694 + if (dentry->d_op == dop)
8697 + AuDbg("%pd\n", dentry);
8698 + spin_lock(&dentry->d_lock);
8699 + if (dop == &aufs_dop)
8700 + dentry->d_flags |= mask;
8702 + dentry->d_flags &= ~mask;
8703 + dentry->d_op = dop;
8704 + spin_unlock(&dentry->d_lock);
8707 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8709 + int err, ebrange, nbr;
8710 + unsigned int sigen;
8711 + struct au_dinfo *dinfo, *tmp;
8712 + struct super_block *sb;
8713 + struct inode *inode;
8715 + DiMustWriteLock(dentry);
8716 + AuDebugOn(IS_ROOT(dentry));
8717 + AuDebugOn(d_really_is_negative(parent));
8719 + sb = dentry->d_sb;
8720 + sigen = au_sigen(sb);
8721 + err = au_digen_test(parent, sigen);
8722 + if (unlikely(err))
8725 + nbr = au_sbbot(sb) + 1;
8726 + dinfo = au_di(dentry);
8727 + err = au_di_realloc(dinfo, nbr, /*may_shrink*/0);
8728 + if (unlikely(err))
8730 + ebrange = au_dbrange_test(dentry);
8732 + ebrange = au_do_refresh_hdentry(dentry, parent);
8734 + if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8735 + AuDebugOn(au_dbtop(dentry) < 0 && au_dbbot(dentry) >= 0);
8736 + if (d_really_is_positive(dentry)) {
8737 + inode = d_inode(dentry);
8738 + err = au_refresh_hinode_self(inode);
8740 + au_dbg_verify_dinode(dentry);
8742 + goto out_dgen; /* success */
8746 + /* temporary dinfo */
8747 + AuDbgDentry(dentry);
8749 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8750 + if (unlikely(!tmp))
8752 + au_di_swap(tmp, dinfo);
8753 + /* returns the number of positive dentries */
8755 + * if current working dir is removed, it returns an error.
8756 + * but the dentry is legal.
8758 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
8759 + AuDbgDentry(dentry);
8760 + au_di_swap(tmp, dinfo);
8761 + if (err == -ENOENT)
8764 + /* compare/refresh by dinfo */
8765 + AuDbgDentry(dentry);
8766 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8767 + au_dbg_verify_dinode(dentry);
8770 + au_di_realloc(dinfo, nbr, /*may_shrink*/1); /* harmless if err */
8771 + au_rw_write_unlock(&tmp->di_rwsem);
8773 + if (unlikely(err))
8777 + au_update_digen(dentry);
8779 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8780 + AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8781 + AuDbgDentry(dentry);
8787 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8788 + struct dentry *dentry, aufs_bindex_t bindex)
8793 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8796 + AuDbg("b%d\n", bindex);
8798 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8799 + * due to whiteout and branch permission.
8801 + flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8802 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
8803 + /* it may return tri-state */
8804 + valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8806 + if (unlikely(valid < 0))
8816 +/* todo: remove this */
8817 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8818 + unsigned int flags, int do_udba, int dirren)
8821 + umode_t mode, h_mode;
8822 + aufs_bindex_t bindex, btail, btop, ibs, ibe;
8823 + unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8824 + struct inode *h_inode, *h_cached_inode;
8825 + struct dentry *h_dentry;
8826 + struct qstr *name, *h_name;
8833 + unhashed = !!d_unhashed(dentry);
8834 + is_root = !!IS_ROOT(dentry);
8835 + name = &dentry->d_name;
8836 + tmpfile = au_di(dentry)->di_tmpfile;
8839 + * Theoretically, REVAL test should be unnecessary in case of
8841 + * But {fs,i}notify doesn't fire some necessary events,
8842 + * IN_ATTRIB for atime/nlink/pageio
8843 + * Let's do REVAL test too.
8845 + if (do_udba && inode) {
8846 + mode = (inode->i_mode & S_IFMT);
8847 + plus = (inode->i_nlink > 0);
8848 + ibs = au_ibtop(inode);
8849 + ibe = au_ibbot(inode);
8852 + btop = au_dbtop(dentry);
8854 + if (inode && S_ISDIR(inode->i_mode))
8855 + btail = au_dbtaildir(dentry);
8856 + for (bindex = btop; bindex <= btail; bindex++) {
8857 + h_dentry = au_h_dptr(dentry, bindex);
8861 + AuDbg("b%d, %pd\n", bindex, h_dentry);
8862 + h_nfs = !!au_test_nfs(h_dentry->d_sb);
8863 + spin_lock(&h_dentry->d_lock);
8864 + h_name = &h_dentry->d_name;
8865 + if (unlikely(do_udba
8868 + && (unhashed != !!d_unhashed(h_dentry)
8869 + || (!tmpfile && !dirren
8870 + && !au_qstreq(name, h_name))
8873 + && !(flags & LOOKUP_OPEN)
8874 + && (h_dentry->d_flags
8875 + & DCACHE_NFSFS_RENAMED)))
8879 + h_unhashed = d_unhashed(h_dentry);
8880 + spin_unlock(&h_dentry->d_lock);
8881 + AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8882 + unhashed, h_unhashed, dentry, h_dentry);
8885 + spin_unlock(&h_dentry->d_lock);
8887 + err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8888 + if (unlikely(err))
8889 + /* do not goto err, to keep the errno */
8892 + /* todo: plink too? */
8897 + if (unlikely(!!inode != d_is_positive(h_dentry)))
8901 + if (d_is_positive(h_dentry))
8902 + h_inode = d_inode(h_dentry);
8905 + h_cached_inode = h_inode;
8907 + h_mode = (h_inode->i_mode & S_IFMT);
8908 + h_plus = (h_inode->i_nlink > 0);
8910 + if (inode && ibs <= bindex && bindex <= ibe)
8911 + h_cached_inode = au_h_iptr(inode, bindex);
8914 + if (unlikely(plus != h_plus && !tmpfile))
8917 + if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8919 + && !IS_ROOT(h_dentry)
8920 + && unhashed != d_unhashed(h_dentry)))
8923 + if (unlikely(mode != h_mode
8924 + || h_cached_inode != h_inode))
8937 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8938 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8941 + struct dentry *parent;
8943 + if (!au_digen_test(dentry, sigen))
8946 + parent = dget_parent(dentry);
8947 + di_read_lock_parent(parent, AuLock_IR);
8948 + AuDebugOn(au_digen_test(parent, sigen));
8949 + au_dbg_verify_gen(parent, sigen);
8950 + err = au_refresh_dentry(dentry, parent);
8951 + di_read_unlock(parent, AuLock_IR);
8957 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8960 + struct dentry *d, *parent;
8962 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8963 + return simple_reval_dpath(dentry, sigen);
8965 + /* slow loop, keep it simple and stupid */
8966 + /* cf: au_cpup_dirs() */
8969 + while (au_digen_test(dentry, sigen)) {
8973 + parent = dget_parent(d);
8974 + if (!au_digen_test(parent, sigen))
8980 + di_write_lock_child2(d);
8982 + /* someone might update our dentry while we were sleeping */
8983 + if (au_digen_test(d, sigen)) {
8985 + * todo: consolidate with simple_reval_dpath(),
8986 + * do_refresh() and au_reval_for_attr().
8988 + di_read_lock_parent(parent, AuLock_IR);
8989 + err = au_refresh_dentry(d, parent);
8990 + di_read_unlock(parent, AuLock_IR);
8994 + di_write_unlock(d);
8996 + if (unlikely(err))
9004 + * if valid returns 1, otherwise 0.
9006 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
9009 + unsigned int sigen;
9010 + unsigned char do_udba, dirren;
9011 + struct super_block *sb;
9012 + struct inode *inode;
9014 + /* todo: support rcu-walk? */
9015 + if (flags & LOOKUP_RCU)
9019 + if (unlikely(!au_di(dentry)))
9023 + sb = dentry->d_sb;
9026 + * i_mutex of parent dir may be held,
9027 + * but we should not return 'invalid' due to busy.
9029 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
9030 + if (unlikely(err)) {
9036 + if (d_really_is_positive(dentry))
9037 + inode = d_inode(dentry);
9038 + if (unlikely(inode && au_is_bad_inode(inode))) {
9043 + if (unlikely(au_dbrange_test(dentry))) {
9049 + sigen = au_sigen(sb);
9050 + if (au_digen_test(dentry, sigen)) {
9051 + AuDebugOn(IS_ROOT(dentry));
9052 + err = au_reval_dpath(dentry, sigen);
9053 + if (unlikely(err)) {
9058 + di_downgrade_lock(dentry, AuLock_IR);
9061 + if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
9063 + && !(inode->i_state && I_LINKABLE)
9064 + && (IS_DEADDIR(inode) || !inode->i_nlink)) {
9069 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
9070 + if (do_udba && inode) {
9071 + aufs_bindex_t btop = au_ibtop(inode);
9072 + struct inode *h_inode;
9075 + h_inode = au_h_iptr(inode, btop);
9076 + if (h_inode && au_test_higen(inode, h_inode)) {
9083 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
9084 + err = h_d_revalidate(dentry, inode, flags, do_udba, dirren);
9085 + if (unlikely(!err && do_udba && au_dbtop(dentry) < 0)) {
9087 + AuDbg("both of real entry and whiteout found, %p, err %d\n",
9093 + di_downgrade_lock(dentry, AuLock_IR);
9095 + aufs_read_unlock(dentry, AuLock_IR);
9100 + AuDbg("%pd invalid, %d\n", dentry, valid);
9106 +static void aufs_d_release(struct dentry *dentry)
9108 + if (au_di(dentry)) {
9109 + au_di_fin(dentry);
9110 + au_hn_di_reinit(dentry);
9114 +const struct dentry_operations aufs_dop = {
9115 + .d_revalidate = aufs_d_revalidate,
9116 + .d_weak_revalidate = aufs_d_revalidate,
9117 + .d_release = aufs_d_release
9120 +/* aufs_dop without d_revalidate */
9121 +const struct dentry_operations aufs_dop_noreval = {
9122 + .d_release = aufs_d_release
9124 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
9125 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
9126 +++ linux/fs/aufs/dentry.h 2022-03-21 14:49:05.719966344 +0100
9128 +/* SPDX-License-Identifier: GPL-2.0 */
9130 + * Copyright (C) 2005-2021 Junjiro R. Okajima
9132 + * This program, aufs is free software; you can redistribute it and/or modify
9133 + * it under the terms of the GNU General Public License as published by
9134 + * the Free Software Foundation; either version 2 of the License, or
9135 + * (at your option) any later version.
9137 + * This program is distributed in the hope that it will be useful,
9138 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9139 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9140 + * GNU General Public License for more details.
9142 + * You should have received a copy of the GNU General Public License
9143 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9147 + * lookup and dentry operations
9150 +#ifndef __AUFS_DENTRY_H__
9151 +#define __AUFS_DENTRY_H__
9155 +#include <linux/dcache.h>
9156 +#include "dirren.h"
9159 +struct au_hdentry {
9160 + struct dentry *hd_dentry;
9161 + aufs_bindex_t hd_id;
9165 + atomic_t di_generation;
9167 + struct au_rwsem di_rwsem;
9168 + aufs_bindex_t di_btop, di_bbot, di_bwh, di_bdiropq;
9169 + unsigned char di_tmpfile; /* to allow the different name */
9170 + struct au_hdentry *di_hdentry;
9171 + struct rcu_head rcu;
9172 +} ____cacheline_aligned_in_smp;
9174 +/* ---------------------------------------------------------------------- */
9176 +/* flags for au_lkup_dentry() */
9177 +#define AuLkup_ALLOW_NEG 1
9178 +#define AuLkup_IGNORE_PERM (1 << 1)
9179 +#define AuLkup_DIRREN (1 << 2)
9180 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
9181 +#define au_fset_lkup(flags, name) \
9182 + do { (flags) |= AuLkup_##name; } while (0)
9183 +#define au_fclr_lkup(flags, name) \
9184 + do { (flags) &= ~AuLkup_##name; } while (0)
9186 +#ifndef CONFIG_AUFS_DIRREN
9187 +#undef AuLkup_DIRREN
9188 +#define AuLkup_DIRREN 0
9191 +struct au_do_lookup_args {
9192 + unsigned int flags;
9194 + struct qstr whname, *name;
9195 + struct au_dr_lookup dirren;
9198 +/* ---------------------------------------------------------------------- */
9201 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
9203 +struct dentry *au_sio_lkup_one(struct user_namespace *userns, struct qstr *name,
9204 + struct path *ppath);
9205 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
9206 + struct dentry *h_parent, struct au_branch *br);
9208 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
9209 + unsigned int flags);
9210 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
9211 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
9212 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
9213 +void au_refresh_dop(struct dentry *dentry, int force_reval);
9216 +void au_di_init_once(void *_di);
9217 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
9218 +void au_di_free(struct au_dinfo *dinfo);
9219 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
9220 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
9221 +int au_di_init(struct dentry *dentry);
9222 +void au_di_fin(struct dentry *dentry);
9223 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink);
9225 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
9226 +void di_read_unlock(struct dentry *d, int flags);
9227 +void di_downgrade_lock(struct dentry *d, int flags);
9228 +void di_write_lock(struct dentry *d, unsigned int lsc);
9229 +void di_write_unlock(struct dentry *d);
9230 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
9231 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
9232 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
9234 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
9235 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
9236 +aufs_bindex_t au_dbtail(struct dentry *dentry);
9237 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
9239 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9240 + struct dentry *h_dentry);
9241 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
9242 +int au_dbrange_test(struct dentry *dentry);
9243 +void au_update_digen(struct dentry *dentry);
9244 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
9245 +void au_update_dbtop(struct dentry *dentry);
9246 +void au_update_dbbot(struct dentry *dentry);
9247 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
9249 +/* ---------------------------------------------------------------------- */
9251 +static inline struct au_dinfo *au_di(struct dentry *dentry)
9253 + return dentry->d_fsdata;
9256 +/* ---------------------------------------------------------------------- */
9258 +/* lock subclass for dinfo */
9260 + AuLsc_DI_CHILD, /* child first */
9261 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
9262 + AuLsc_DI_CHILD3, /* copyup dirs */
9266 + AuLsc_DI_TMP /* temp for replacing dinfo */
9270 + * di_read_lock_child, di_write_lock_child,
9271 + * di_read_lock_child2, di_write_lock_child2,
9272 + * di_read_lock_child3, di_write_lock_child3,
9273 + * di_read_lock_parent, di_write_lock_parent,
9274 + * di_read_lock_parent2, di_write_lock_parent2,
9275 + * di_read_lock_parent3, di_write_lock_parent3,
9277 +#define AuReadLockFunc(name, lsc) \
9278 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
9279 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
9281 +#define AuWriteLockFunc(name, lsc) \
9282 +static inline void di_write_lock_##name(struct dentry *d) \
9283 +{ di_write_lock(d, AuLsc_DI_##lsc); }
9285 +#define AuRWLockFuncs(name, lsc) \
9286 + AuReadLockFunc(name, lsc) \
9287 + AuWriteLockFunc(name, lsc)
9289 +AuRWLockFuncs(child, CHILD);
9290 +AuRWLockFuncs(child2, CHILD2);
9291 +AuRWLockFuncs(child3, CHILD3);
9292 +AuRWLockFuncs(parent, PARENT);
9293 +AuRWLockFuncs(parent2, PARENT2);
9294 +AuRWLockFuncs(parent3, PARENT3);
9296 +#undef AuReadLockFunc
9297 +#undef AuWriteLockFunc
9298 +#undef AuRWLockFuncs
9300 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
9301 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
9302 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
9304 +/* ---------------------------------------------------------------------- */
9306 +/* todo: memory barrier? */
9307 +static inline unsigned int au_digen(struct dentry *d)
9309 + return atomic_read(&au_di(d)->di_generation);
9312 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
9314 + hdentry->hd_dentry = NULL;
9317 +static inline struct au_hdentry *au_hdentry(struct au_dinfo *di,
9318 + aufs_bindex_t bindex)
9320 + return di->di_hdentry + bindex;
9323 +static inline void au_hdput(struct au_hdentry *hd)
9326 + dput(hd->hd_dentry);
9329 +static inline aufs_bindex_t au_dbtop(struct dentry *dentry)
9331 + DiMustAnyLock(dentry);
9332 + return au_di(dentry)->di_btop;
9335 +static inline aufs_bindex_t au_dbbot(struct dentry *dentry)
9337 + DiMustAnyLock(dentry);
9338 + return au_di(dentry)->di_bbot;
9341 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
9343 + DiMustAnyLock(dentry);
9344 + return au_di(dentry)->di_bwh;
9347 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
9349 + DiMustAnyLock(dentry);
9350 + return au_di(dentry)->di_bdiropq;
9353 +/* todo: hard/soft set? */
9354 +static inline void au_set_dbtop(struct dentry *dentry, aufs_bindex_t bindex)
9356 + DiMustWriteLock(dentry);
9357 + au_di(dentry)->di_btop = bindex;
9360 +static inline void au_set_dbbot(struct dentry *dentry, aufs_bindex_t bindex)
9362 + DiMustWriteLock(dentry);
9363 + au_di(dentry)->di_bbot = bindex;
9366 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
9368 + DiMustWriteLock(dentry);
9369 + /* dbwh can be outside of btop - bbot range */
9370 + au_di(dentry)->di_bwh = bindex;
9373 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
9375 + DiMustWriteLock(dentry);
9376 + au_di(dentry)->di_bdiropq = bindex;
9379 +/* ---------------------------------------------------------------------- */
9381 +#ifdef CONFIG_AUFS_HNOTIFY
9382 +static inline void au_digen_dec(struct dentry *d)
9384 + atomic_dec(&au_di(d)->di_generation);
9387 +static inline void au_hn_di_reinit(struct dentry *dentry)
9389 + dentry->d_fsdata = NULL;
9392 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
9393 +#endif /* CONFIG_AUFS_HNOTIFY */
9395 +#endif /* __KERNEL__ */
9396 +#endif /* __AUFS_DENTRY_H__ */
9397 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
9398 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
9399 +++ linux/fs/aufs/dinfo.c 2022-03-21 14:49:05.719966344 +0100
9401 +// SPDX-License-Identifier: GPL-2.0
9403 + * Copyright (C) 2005-2021 Junjiro R. Okajima
9405 + * This program, aufs is free software; you can redistribute it and/or modify
9406 + * it under the terms of the GNU General Public License as published by
9407 + * the Free Software Foundation; either version 2 of the License, or
9408 + * (at your option) any later version.
9410 + * This program is distributed in the hope that it will be useful,
9411 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9412 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9413 + * GNU General Public License for more details.
9415 + * You should have received a copy of the GNU General Public License
9416 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9420 + * dentry private data
9425 +void au_di_init_once(void *_dinfo)
9427 + struct au_dinfo *dinfo = _dinfo;
9429 + au_rw_init(&dinfo->di_rwsem);
9432 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
9434 + struct au_dinfo *dinfo;
9437 + dinfo = au_cache_alloc_dinfo();
9438 + if (unlikely(!dinfo))
9441 + nbr = au_sbbot(sb) + 1;
9444 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
9445 + if (dinfo->di_hdentry) {
9446 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
9447 + dinfo->di_btop = -1;
9448 + dinfo->di_bbot = -1;
9449 + dinfo->di_bwh = -1;
9450 + dinfo->di_bdiropq = -1;
9451 + dinfo->di_tmpfile = 0;
9452 + for (i = 0; i < nbr; i++)
9453 + dinfo->di_hdentry[i].hd_id = -1;
9457 + au_cache_free_dinfo(dinfo);
9464 +void au_di_free(struct au_dinfo *dinfo)
9466 + struct au_hdentry *p;
9467 + aufs_bindex_t bbot, bindex;
9469 + /* dentry may not be revalidated */
9470 + bindex = dinfo->di_btop;
9471 + if (bindex >= 0) {
9472 + bbot = dinfo->di_bbot;
9473 + p = au_hdentry(dinfo, bindex);
9474 + while (bindex++ <= bbot)
9477 + au_kfree_try_rcu(dinfo->di_hdentry);
9478 + au_cache_free_dinfo(dinfo);
9481 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
9483 + struct au_hdentry *p;
9486 + AuRwMustWriteLock(&a->di_rwsem);
9487 + AuRwMustWriteLock(&b->di_rwsem);
9489 +#define DiSwap(v, name) \
9491 + v = a->di_##name; \
9492 + a->di_##name = b->di_##name; \
9493 + b->di_##name = v; \
9496 + DiSwap(p, hdentry);
9500 + DiSwap(bi, bdiropq);
9506 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
9508 + AuRwMustWriteLock(&dst->di_rwsem);
9509 + AuRwMustWriteLock(&src->di_rwsem);
9511 + dst->di_btop = src->di_btop;
9512 + dst->di_bbot = src->di_bbot;
9513 + dst->di_bwh = src->di_bwh;
9514 + dst->di_bdiropq = src->di_bdiropq;
9518 +int au_di_init(struct dentry *dentry)
9521 + struct super_block *sb;
9522 + struct au_dinfo *dinfo;
9525 + sb = dentry->d_sb;
9526 + dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
9528 + atomic_set(&dinfo->di_generation, au_sigen(sb));
9529 + /* smp_mb(); */ /* atomic_set */
9530 + dentry->d_fsdata = dinfo;
9537 +void au_di_fin(struct dentry *dentry)
9539 + struct au_dinfo *dinfo;
9541 + dinfo = au_di(dentry);
9542 + AuRwDestroy(&dinfo->di_rwsem);
9543 + au_di_free(dinfo);
9546 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink)
9549 + struct au_hdentry *hdp;
9551 + AuRwMustWriteLock(&dinfo->di_rwsem);
9554 + sz = sizeof(*hdp) * (dinfo->di_bbot + 1);
9556 + sz = sizeof(*hdp);
9557 + hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS,
9560 + dinfo->di_hdentry = hdp;
9567 +/* ---------------------------------------------------------------------- */
9569 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
9572 + case AuLsc_DI_CHILD:
9573 + ii_write_lock_child(inode);
9575 + case AuLsc_DI_CHILD2:
9576 + ii_write_lock_child2(inode);
9578 + case AuLsc_DI_CHILD3:
9579 + ii_write_lock_child3(inode);
9581 + case AuLsc_DI_PARENT:
9582 + ii_write_lock_parent(inode);
9584 + case AuLsc_DI_PARENT2:
9585 + ii_write_lock_parent2(inode);
9587 + case AuLsc_DI_PARENT3:
9588 + ii_write_lock_parent3(inode);
9595 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9598 + case AuLsc_DI_CHILD:
9599 + ii_read_lock_child(inode);
9601 + case AuLsc_DI_CHILD2:
9602 + ii_read_lock_child2(inode);
9604 + case AuLsc_DI_CHILD3:
9605 + ii_read_lock_child3(inode);
9607 + case AuLsc_DI_PARENT:
9608 + ii_read_lock_parent(inode);
9610 + case AuLsc_DI_PARENT2:
9611 + ii_read_lock_parent2(inode);
9613 + case AuLsc_DI_PARENT3:
9614 + ii_read_lock_parent3(inode);
9621 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9623 + struct inode *inode;
9625 + au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9626 + if (d_really_is_positive(d)) {
9627 + inode = d_inode(d);
9628 + if (au_ftest_lock(flags, IW))
9629 + do_ii_write_lock(inode, lsc);
9630 + else if (au_ftest_lock(flags, IR))
9631 + do_ii_read_lock(inode, lsc);
9635 +void di_read_unlock(struct dentry *d, int flags)
9637 + struct inode *inode;
9639 + if (d_really_is_positive(d)) {
9640 + inode = d_inode(d);
9641 + if (au_ftest_lock(flags, IW)) {
9642 + au_dbg_verify_dinode(d);
9643 + ii_write_unlock(inode);
9644 + } else if (au_ftest_lock(flags, IR)) {
9645 + au_dbg_verify_dinode(d);
9646 + ii_read_unlock(inode);
9649 + au_rw_read_unlock(&au_di(d)->di_rwsem);
9652 +void di_downgrade_lock(struct dentry *d, int flags)
9654 + if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9655 + ii_downgrade_lock(d_inode(d));
9656 + au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9659 +void di_write_lock(struct dentry *d, unsigned int lsc)
9661 + au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9662 + if (d_really_is_positive(d))
9663 + do_ii_write_lock(d_inode(d), lsc);
9666 +void di_write_unlock(struct dentry *d)
9668 + au_dbg_verify_dinode(d);
9669 + if (d_really_is_positive(d))
9670 + ii_write_unlock(d_inode(d));
9671 + au_rw_write_unlock(&au_di(d)->di_rwsem);
9674 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9676 + AuDebugOn(d1 == d2
9677 + || d_inode(d1) == d_inode(d2)
9678 + || d1->d_sb != d2->d_sb);
9680 + if ((isdir && au_test_subdir(d1, d2))
9682 + di_write_lock_child(d1);
9683 + di_write_lock_child2(d2);
9685 + di_write_lock_child(d2);
9686 + di_write_lock_child2(d1);
9690 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9692 + AuDebugOn(d1 == d2
9693 + || d_inode(d1) == d_inode(d2)
9694 + || d1->d_sb != d2->d_sb);
9696 + if ((isdir && au_test_subdir(d1, d2))
9698 + di_write_lock_parent(d1);
9699 + di_write_lock_parent2(d2);
9701 + di_write_lock_parent(d2);
9702 + di_write_lock_parent2(d1);
9706 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9708 + di_write_unlock(d1);
9709 + if (d_inode(d1) == d_inode(d2))
9710 + au_rw_write_unlock(&au_di(d2)->di_rwsem);
9712 + di_write_unlock(d2);
9715 +/* ---------------------------------------------------------------------- */
9717 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9721 + DiMustAnyLock(dentry);
9723 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
9725 + AuDebugOn(bindex < 0);
9726 + d = au_hdentry(au_di(dentry), bindex)->hd_dentry;
9727 + AuDebugOn(d && au_dcount(d) <= 0);
9732 + * extended version of au_h_dptr().
9733 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9736 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9738 + struct dentry *h_dentry;
9739 + struct inode *inode, *h_inode;
9741 + AuDebugOn(d_really_is_negative(dentry));
9744 + if (au_dbtop(dentry) <= bindex
9745 + && bindex <= au_dbbot(dentry))
9746 + h_dentry = au_h_dptr(dentry, bindex);
9747 + if (h_dentry && !au_d_linkable(h_dentry)) {
9749 + goto out; /* success */
9752 + inode = d_inode(dentry);
9753 + AuDebugOn(bindex < au_ibtop(inode));
9754 + AuDebugOn(au_ibbot(inode) < bindex);
9755 + h_inode = au_h_iptr(inode, bindex);
9756 + h_dentry = d_find_alias(h_inode);
9758 + if (!IS_ERR(h_dentry)) {
9759 + if (!au_d_linkable(h_dentry))
9760 + goto out; /* success */
9766 + if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9767 + h_dentry = au_plink_lkup(inode, bindex);
9768 + AuDebugOn(!h_dentry);
9769 + if (!IS_ERR(h_dentry)) {
9770 + if (!au_d_hashed_positive(h_dentry))
9771 + goto out; /* success */
9778 + AuDbgDentry(h_dentry);
9782 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9784 + aufs_bindex_t bbot, bwh;
9786 + bbot = au_dbbot(dentry);
9788 + bwh = au_dbwh(dentry);
9791 + if (0 < bwh && bwh < bbot)
9797 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9799 + aufs_bindex_t bbot, bopq;
9801 + bbot = au_dbtail(dentry);
9803 + bopq = au_dbdiropq(dentry);
9804 + if (0 <= bopq && bopq < bbot)
9810 +/* ---------------------------------------------------------------------- */
9812 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9813 + struct dentry *h_dentry)
9815 + struct au_dinfo *dinfo;
9816 + struct au_hdentry *hd;
9817 + struct au_branch *br;
9819 + DiMustWriteLock(dentry);
9821 + dinfo = au_di(dentry);
9822 + hd = au_hdentry(dinfo, bindex);
9824 + hd->hd_dentry = h_dentry;
9826 + br = au_sbr(dentry->d_sb, bindex);
9827 + hd->hd_id = br->br_id;
9831 +int au_dbrange_test(struct dentry *dentry)
9834 + aufs_bindex_t btop, bbot;
9837 + btop = au_dbtop(dentry);
9838 + bbot = au_dbbot(dentry);
9840 + AuDebugOn(bbot < 0 && btop > bbot);
9843 + AuDebugOn(bbot >= 0);
9849 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9854 + if (unlikely(au_digen(dentry) != sigen
9855 + || au_iigen_test(d_inode(dentry), sigen)))
9861 +void au_update_digen(struct dentry *dentry)
9863 + atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9864 + /* smp_mb(); */ /* atomic_set */
9867 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9869 + struct au_dinfo *dinfo;
9870 + struct dentry *h_d;
9871 + struct au_hdentry *hdp;
9872 + aufs_bindex_t bindex, bbot;
9874 + DiMustWriteLock(dentry);
9876 + dinfo = au_di(dentry);
9877 + if (!dinfo || dinfo->di_btop < 0)
9880 + if (do_put_zero) {
9881 + bbot = dinfo->di_bbot;
9882 + bindex = dinfo->di_btop;
9883 + hdp = au_hdentry(dinfo, bindex);
9884 + for (; bindex <= bbot; bindex++, hdp++) {
9885 + h_d = hdp->hd_dentry;
9886 + if (h_d && d_is_negative(h_d))
9887 + au_set_h_dptr(dentry, bindex, NULL);
9891 + dinfo->di_btop = 0;
9892 + hdp = au_hdentry(dinfo, dinfo->di_btop);
9893 + for (; dinfo->di_btop <= dinfo->di_bbot; dinfo->di_btop++, hdp++)
9894 + if (hdp->hd_dentry)
9896 + if (dinfo->di_btop > dinfo->di_bbot) {
9897 + dinfo->di_btop = -1;
9898 + dinfo->di_bbot = -1;
9902 + hdp = au_hdentry(dinfo, dinfo->di_bbot);
9903 + for (; dinfo->di_bbot >= 0; dinfo->di_bbot--, hdp--)
9904 + if (hdp->hd_dentry)
9906 + AuDebugOn(dinfo->di_btop > dinfo->di_bbot || dinfo->di_bbot < 0);
9909 +void au_update_dbtop(struct dentry *dentry)
9911 + aufs_bindex_t bindex, bbot;
9912 + struct dentry *h_dentry;
9914 + bbot = au_dbbot(dentry);
9915 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
9916 + h_dentry = au_h_dptr(dentry, bindex);
9919 + if (d_is_positive(h_dentry)) {
9920 + au_set_dbtop(dentry, bindex);
9923 + au_set_h_dptr(dentry, bindex, NULL);
9927 +void au_update_dbbot(struct dentry *dentry)
9929 + aufs_bindex_t bindex, btop;
9930 + struct dentry *h_dentry;
9932 + btop = au_dbtop(dentry);
9933 + for (bindex = au_dbbot(dentry); bindex >= btop; bindex--) {
9934 + h_dentry = au_h_dptr(dentry, bindex);
9937 + if (d_is_positive(h_dentry)) {
9938 + au_set_dbbot(dentry, bindex);
9941 + au_set_h_dptr(dentry, bindex, NULL);
9945 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9947 + aufs_bindex_t bindex, bbot;
9949 + bbot = au_dbbot(dentry);
9950 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++)
9951 + if (au_h_dptr(dentry, bindex) == h_dentry)
9955 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9956 --- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
9957 +++ linux/fs/aufs/dir.c 2022-03-21 14:49:05.719966344 +0100
9959 +// SPDX-License-Identifier: GPL-2.0
9961 + * Copyright (C) 2005-2021 Junjiro R. Okajima
9963 + * This program, aufs is free software; you can redistribute it and/or modify
9964 + * it under the terms of the GNU General Public License as published by
9965 + * the Free Software Foundation; either version 2 of the License, or
9966 + * (at your option) any later version.
9968 + * This program is distributed in the hope that it will be useful,
9969 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9970 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9971 + * GNU General Public License for more details.
9973 + * You should have received a copy of the GNU General Public License
9974 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9978 + * directory operations
9981 +#include <linux/fs_stack.h>
9982 +#include <linux/iversion.h>
9985 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
9987 + unsigned int nlink;
9989 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9991 + nlink = dir->i_nlink;
9992 + nlink += h_dir->i_nlink - 2;
9993 + if (h_dir->i_nlink < 2)
9995 + smp_mb(); /* for i_nlink */
9996 + /* 0 can happen in revaliding */
9997 + set_nlink(dir, nlink);
10000 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
10002 + unsigned int nlink;
10004 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10006 + nlink = dir->i_nlink;
10007 + nlink -= h_dir->i_nlink - 2;
10008 + if (h_dir->i_nlink < 2)
10010 + smp_mb(); /* for i_nlink */
10011 + /* nlink == 0 means the branch-fs is broken */
10012 + set_nlink(dir, nlink);
10015 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
10018 + aufs_bindex_t bindex, bbot;
10019 + struct file *h_file;
10020 + struct dentry *h_dentry;
10024 + AuDebugOn(!d_is_dir(file->f_path.dentry));
10026 + bbot = au_fbbot_dir(file);
10027 + for (bindex = au_fbtop(file);
10028 + bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10030 + h_file = au_hf_dir(file, bindex);
10031 + if (h_file && file_inode(h_file))
10032 + sz += vfsub_f_size_read(h_file);
10035 + AuDebugOn(!dentry);
10036 + AuDebugOn(!d_is_dir(dentry));
10038 + bbot = au_dbtaildir(dentry);
10039 + for (bindex = au_dbtop(dentry);
10040 + bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10042 + h_dentry = au_h_dptr(dentry, bindex);
10043 + if (h_dentry && d_is_positive(h_dentry))
10044 + sz += i_size_read(d_inode(h_dentry));
10047 + if (sz < KMALLOC_MAX_SIZE)
10048 + sz = roundup_pow_of_two(sz);
10049 + if (sz > KMALLOC_MAX_SIZE)
10050 + sz = KMALLOC_MAX_SIZE;
10051 + else if (sz < NAME_MAX) {
10052 + BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
10053 + sz = AUFS_RDBLK_DEF;
10058 +struct au_dir_ts_arg {
10059 + struct dentry *dentry;
10060 + aufs_bindex_t brid;
10063 +static void au_do_dir_ts(void *arg)
10065 + struct au_dir_ts_arg *a = arg;
10066 + struct au_dtime dt;
10067 + struct path h_path;
10068 + struct inode *dir, *h_dir;
10069 + struct super_block *sb;
10070 + struct au_branch *br;
10071 + struct au_hinode *hdir;
10073 + aufs_bindex_t btop, bindex;
10075 + sb = a->dentry->d_sb;
10076 + if (d_really_is_negative(a->dentry))
10078 + /* no dir->i_mutex lock */
10079 + aufs_read_lock(a->dentry, AuLock_DW); /* noflush */
10081 + dir = d_inode(a->dentry);
10082 + btop = au_ibtop(dir);
10083 + bindex = au_br_index(sb, a->brid);
10084 + if (bindex < btop)
10087 + br = au_sbr(sb, bindex);
10088 + h_path.dentry = au_h_dptr(a->dentry, bindex);
10089 + if (!h_path.dentry)
10091 + h_path.mnt = au_br_mnt(br);
10092 + au_dtime_store(&dt, a->dentry, &h_path);
10094 + br = au_sbr(sb, btop);
10095 + if (!au_br_writable(br->br_perm))
10097 + h_path.dentry = au_h_dptr(a->dentry, btop);
10098 + h_path.mnt = au_br_mnt(br);
10099 + err = vfsub_mnt_want_write(h_path.mnt);
10102 + hdir = au_hi(dir, btop);
10103 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
10104 + h_dir = au_h_iptr(dir, btop);
10105 + if (h_dir->i_nlink
10106 + && timespec64_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
10107 + dt.dt_h_path = h_path;
10108 + au_dtime_revert(&dt);
10110 + au_hn_inode_unlock(hdir);
10111 + vfsub_mnt_drop_write(h_path.mnt);
10112 + au_cpup_attr_timesizes(dir);
10115 + aufs_read_unlock(a->dentry, AuLock_DW);
10118 + au_nwt_done(&au_sbi(sb)->si_nowait);
10119 + au_kfree_try_rcu(arg);
10122 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
10124 + int perm, wkq_err;
10125 + aufs_bindex_t btop;
10126 + struct au_dir_ts_arg *arg;
10127 + struct dentry *dentry;
10128 + struct super_block *sb;
10132 + dentry = d_find_any_alias(dir);
10133 + AuDebugOn(!dentry);
10134 + sb = dentry->d_sb;
10135 + btop = au_ibtop(dir);
10136 + if (btop == bindex) {
10137 + au_cpup_attr_timesizes(dir);
10141 + perm = au_sbr_perm(sb, btop);
10142 + if (!au_br_writable(perm))
10145 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
10149 + arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
10150 + arg->brid = au_sbr_id(sb, bindex);
10151 + wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
10152 + if (unlikely(wkq_err)) {
10153 + pr_err("wkq %d\n", wkq_err);
10155 + au_kfree_try_rcu(arg);
10162 +/* ---------------------------------------------------------------------- */
10164 +static int reopen_dir(struct file *file)
10167 + unsigned int flags;
10168 + aufs_bindex_t bindex, btail, btop;
10169 + struct dentry *dentry, *h_dentry;
10170 + struct file *h_file;
10172 + /* open all lower dirs */
10173 + dentry = file->f_path.dentry;
10174 + btop = au_dbtop(dentry);
10175 + for (bindex = au_fbtop(file); bindex < btop; bindex++)
10176 + au_set_h_fptr(file, bindex, NULL);
10177 + au_set_fbtop(file, btop);
10179 + btail = au_dbtaildir(dentry);
10180 + for (bindex = au_fbbot_dir(file); btail < bindex; bindex--)
10181 + au_set_h_fptr(file, bindex, NULL);
10182 + au_set_fbbot_dir(file, btail);
10184 + flags = vfsub_file_flags(file);
10185 + for (bindex = btop; bindex <= btail; bindex++) {
10186 + h_dentry = au_h_dptr(dentry, bindex);
10189 + h_file = au_hf_dir(file, bindex);
10193 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10194 + err = PTR_ERR(h_file);
10195 + if (IS_ERR(h_file))
10196 + goto out; /* close all? */
10197 + au_set_h_fptr(file, bindex, h_file);
10199 + au_update_figen(file);
10200 + /* todo: necessary? */
10201 + /* file->f_ra = h_file->f_ra; */
10208 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
10211 + aufs_bindex_t bindex, btail;
10212 + struct dentry *dentry, *h_dentry;
10213 + struct vfsmount *mnt;
10215 + FiMustWriteLock(file);
10216 + AuDebugOn(h_file);
10219 + mnt = file->f_path.mnt;
10220 + dentry = file->f_path.dentry;
10221 + file->f_version = inode_query_iversion(d_inode(dentry));
10222 + bindex = au_dbtop(dentry);
10223 + au_set_fbtop(file, bindex);
10224 + btail = au_dbtaildir(dentry);
10225 + au_set_fbbot_dir(file, btail);
10226 + for (; !err && bindex <= btail; bindex++) {
10227 + h_dentry = au_h_dptr(dentry, bindex);
10231 + err = vfsub_test_mntns(mnt, h_dentry->d_sb);
10232 + if (unlikely(err))
10234 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10235 + if (IS_ERR(h_file)) {
10236 + err = PTR_ERR(h_file);
10239 + au_set_h_fptr(file, bindex, h_file);
10241 + au_update_figen(file);
10242 + /* todo: necessary? */
10243 + /* file->f_ra = h_file->f_ra; */
10245 + return 0; /* success */
10248 + for (bindex = au_fbtop(file); bindex <= btail; bindex++)
10249 + au_set_h_fptr(file, bindex, NULL);
10250 + au_set_fbtop(file, -1);
10251 + au_set_fbbot_dir(file, -1);
10256 +static int aufs_open_dir(struct inode *inode __maybe_unused,
10257 + struct file *file)
10260 + struct super_block *sb;
10261 + struct au_fidir *fidir;
10264 + sb = file->f_path.dentry->d_sb;
10265 + si_read_lock(sb, AuLock_FLUSH);
10266 + fidir = au_fidir_alloc(sb);
10268 + struct au_do_open_args args = {
10269 + .open = do_open_dir,
10272 + err = au_do_open(file, &args);
10273 + if (unlikely(err))
10274 + au_kfree_rcu(fidir);
10276 + si_read_unlock(sb);
10280 +static int aufs_release_dir(struct inode *inode __maybe_unused,
10281 + struct file *file)
10283 + struct au_vdir *vdir_cache;
10284 + struct au_finfo *finfo;
10285 + struct au_fidir *fidir;
10286 + struct au_hfile *hf;
10287 + aufs_bindex_t bindex, bbot;
10289 + finfo = au_fi(file);
10290 + fidir = finfo->fi_hdir;
10292 + au_hbl_del(&finfo->fi_hlist,
10293 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
10294 + vdir_cache = fidir->fd_vdir_cache; /* lock-free */
10296 + au_vdir_free(vdir_cache);
10298 + bindex = finfo->fi_btop;
10299 + if (bindex >= 0) {
10300 + hf = fidir->fd_hfile + bindex;
10302 + * calls fput() instead of filp_close(),
10303 + * since no dnotify or lock for the lower file.
10305 + bbot = fidir->fd_bbot;
10306 + for (; bindex <= bbot; bindex++, hf++)
10308 + au_hfput(hf, /*execed*/0);
10310 + au_kfree_rcu(fidir);
10311 + finfo->fi_hdir = NULL;
10313 + au_finfo_fin(file);
10317 +/* ---------------------------------------------------------------------- */
10319 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
10322 + aufs_bindex_t bindex, bbot;
10323 + struct file *h_file;
10326 + bbot = au_fbbot_dir(file);
10327 + for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10328 + h_file = au_hf_dir(file, bindex);
10330 + err = vfsub_flush(h_file, id);
10335 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
10337 + return au_do_flush(file, id, au_do_flush_dir);
10340 +/* ---------------------------------------------------------------------- */
10342 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
10345 + aufs_bindex_t bbot, bindex;
10346 + struct inode *inode;
10347 + struct super_block *sb;
10350 + sb = dentry->d_sb;
10351 + inode = d_inode(dentry);
10352 + IMustLock(inode);
10353 + bbot = au_dbbot(dentry);
10354 + for (bindex = au_dbtop(dentry); !err && bindex <= bbot; bindex++) {
10355 + struct path h_path;
10357 + if (au_test_ro(sb, bindex, inode))
10359 + h_path.dentry = au_h_dptr(dentry, bindex);
10360 + if (!h_path.dentry)
10363 + h_path.mnt = au_sbr_mnt(sb, bindex);
10364 + err = vfsub_fsync(NULL, &h_path, datasync);
10370 +static int au_do_fsync_dir(struct file *file, int datasync)
10373 + aufs_bindex_t bbot, bindex;
10374 + struct file *h_file;
10375 + struct super_block *sb;
10376 + struct inode *inode;
10378 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10379 + if (unlikely(err))
10382 + inode = file_inode(file);
10383 + sb = inode->i_sb;
10384 + bbot = au_fbbot_dir(file);
10385 + for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10386 + h_file = au_hf_dir(file, bindex);
10387 + if (!h_file || au_test_ro(sb, bindex, inode))
10390 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
10398 + * @file may be NULL
10400 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
10404 + struct dentry *dentry;
10405 + struct inode *inode;
10406 + struct super_block *sb;
10409 + dentry = file->f_path.dentry;
10410 + inode = d_inode(dentry);
10411 + inode_lock(inode);
10412 + sb = dentry->d_sb;
10413 + si_noflush_read_lock(sb);
10415 + err = au_do_fsync_dir(file, datasync);
10417 + di_write_lock_child(dentry);
10418 + err = au_do_fsync_dir_no_file(dentry, datasync);
10420 + au_cpup_attr_timesizes(inode);
10421 + di_write_unlock(dentry);
10423 + fi_write_unlock(file);
10425 + si_read_unlock(sb);
10426 + inode_unlock(inode);
10430 +/* ---------------------------------------------------------------------- */
10432 +static int aufs_iterate_shared(struct file *file, struct dir_context *ctx)
10435 + struct dentry *dentry;
10436 + struct inode *inode, *h_inode;
10437 + struct super_block *sb;
10439 + AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
10441 + dentry = file->f_path.dentry;
10442 + inode = d_inode(dentry);
10443 + IMustLock(inode);
10445 + sb = dentry->d_sb;
10446 + si_read_lock(sb, AuLock_FLUSH);
10447 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10448 + if (unlikely(err))
10450 + err = au_alive_dir(dentry);
10452 + err = au_vdir_init(file);
10453 + di_downgrade_lock(dentry, AuLock_IR);
10454 + if (unlikely(err))
10457 + h_inode = au_h_iptr(inode, au_ibtop(inode));
10458 + if (!au_test_nfsd()) {
10459 + err = au_vdir_fill_de(file, ctx);
10460 + fsstack_copy_attr_atime(inode, h_inode);
10463 + * nfsd filldir may call lookup_one_len(), vfs_getattr(),
10464 + * encode_fh() and others.
10466 + atomic_inc(&h_inode->i_count);
10467 + di_read_unlock(dentry, AuLock_IR);
10468 + si_read_unlock(sb);
10469 + err = au_vdir_fill_de(file, ctx);
10470 + fsstack_copy_attr_atime(inode, h_inode);
10471 + fi_write_unlock(file);
10479 + di_read_unlock(dentry, AuLock_IR);
10480 + fi_write_unlock(file);
10482 + si_read_unlock(sb);
10486 +/* ---------------------------------------------------------------------- */
10488 +#define AuTestEmpty_WHONLY 1
10489 +#define AuTestEmpty_CALLED (1 << 1)
10490 +#define AuTestEmpty_SHWH (1 << 2)
10491 +#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
10492 +#define au_fset_testempty(flags, name) \
10493 + do { (flags) |= AuTestEmpty_##name; } while (0)
10494 +#define au_fclr_testempty(flags, name) \
10495 + do { (flags) &= ~AuTestEmpty_##name; } while (0)
10497 +#ifndef CONFIG_AUFS_SHWH
10498 +#undef AuTestEmpty_SHWH
10499 +#define AuTestEmpty_SHWH 0
10502 +struct test_empty_arg {
10503 + struct dir_context ctx;
10504 + struct au_nhash *whlist;
10505 + unsigned int flags;
10507 + aufs_bindex_t bindex;
10510 +static int test_empty_cb(struct dir_context *ctx, const char *__name,
10511 + int namelen, loff_t offset __maybe_unused, u64 ino,
10512 + unsigned int d_type)
10514 + struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
10516 + char *name = (void *)__name;
10519 + au_fset_testempty(arg->flags, CALLED);
10521 + if (name[0] == '.'
10522 + && (namelen == 1 || (name[1] == '.' && namelen == 2)))
10523 + goto out; /* success */
10525 + if (namelen <= AUFS_WH_PFX_LEN
10526 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
10527 + if (au_ftest_testempty(arg->flags, WHONLY)
10528 + && !au_nhash_test_known_wh(arg->whlist, name, namelen))
10529 + arg->err = -ENOTEMPTY;
10533 + name += AUFS_WH_PFX_LEN;
10534 + namelen -= AUFS_WH_PFX_LEN;
10535 + if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
10536 + arg->err = au_nhash_append_wh
10537 + (arg->whlist, name, namelen, ino, d_type, arg->bindex,
10538 + au_ftest_testempty(arg->flags, SHWH));
10542 + AuTraceErr(arg->err);
10546 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10549 + struct file *h_file;
10550 + struct au_branch *br;
10552 + h_file = au_h_open(dentry, arg->bindex,
10553 + O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
10554 + /*file*/NULL, /*force_wr*/0);
10555 + err = PTR_ERR(h_file);
10556 + if (IS_ERR(h_file))
10560 + if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
10561 + && !file_inode(h_file)->i_nlink)
10566 + au_fclr_testempty(arg->flags, CALLED);
10568 + err = vfsub_iterate_dir(h_file, &arg->ctx);
10571 + } while (!err && au_ftest_testempty(arg->flags, CALLED));
10575 + br = au_sbr(dentry->d_sb, arg->bindex);
10576 + au_lcnt_dec(&br->br_nfiles);
10581 +struct do_test_empty_args {
10583 + struct dentry *dentry;
10584 + struct test_empty_arg *arg;
10587 +static void call_do_test_empty(void *args)
10589 + struct do_test_empty_args *a = args;
10590 + *a->errp = do_test_empty(a->dentry, a->arg);
10593 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10595 + int err, wkq_err;
10596 + struct dentry *h_dentry;
10597 + struct inode *h_inode;
10598 + struct user_namespace *h_userns;
10600 + h_userns = au_sbr_userns(dentry->d_sb, arg->bindex);
10601 + h_dentry = au_h_dptr(dentry, arg->bindex);
10602 + h_inode = d_inode(h_dentry);
10603 + /* todo: i_mode changes anytime? */
10604 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
10605 + err = au_test_h_perm_sio(h_userns, h_inode, MAY_EXEC | MAY_READ);
10606 + inode_unlock_shared(h_inode);
10608 + err = do_test_empty(dentry, arg);
10610 + struct do_test_empty_args args = {
10612 + .dentry = dentry,
10615 + unsigned int flags = arg->flags;
10617 + wkq_err = au_wkq_wait(call_do_test_empty, &args);
10618 + if (unlikely(wkq_err))
10620 + arg->flags = flags;
10626 +int au_test_empty_lower(struct dentry *dentry)
10629 + unsigned int rdhash;
10630 + aufs_bindex_t bindex, btop, btail;
10631 + struct au_nhash whlist;
10632 + struct test_empty_arg arg = {
10634 + .actor = test_empty_cb
10637 + int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10639 + SiMustAnyLock(dentry->d_sb);
10641 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10643 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10644 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10645 + if (unlikely(err))
10649 + arg.whlist = &whlist;
10650 + btop = au_dbtop(dentry);
10651 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10652 + au_fset_testempty(arg.flags, SHWH);
10653 + test_empty = do_test_empty;
10654 + if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10655 + test_empty = sio_test_empty;
10656 + arg.bindex = btop;
10657 + err = test_empty(dentry, &arg);
10658 + if (unlikely(err))
10661 + au_fset_testempty(arg.flags, WHONLY);
10662 + btail = au_dbtaildir(dentry);
10663 + for (bindex = btop + 1; !err && bindex <= btail; bindex++) {
10664 + struct dentry *h_dentry;
10666 + h_dentry = au_h_dptr(dentry, bindex);
10667 + if (h_dentry && d_is_positive(h_dentry)) {
10668 + arg.bindex = bindex;
10669 + err = test_empty(dentry, &arg);
10674 + au_nhash_wh_free(&whlist);
10679 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10682 + struct test_empty_arg arg = {
10684 + .actor = test_empty_cb
10687 + aufs_bindex_t bindex, btail;
10690 + arg.whlist = whlist;
10691 + arg.flags = AuTestEmpty_WHONLY;
10692 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10693 + au_fset_testempty(arg.flags, SHWH);
10694 + btail = au_dbtaildir(dentry);
10695 + for (bindex = au_dbtop(dentry); !err && bindex <= btail; bindex++) {
10696 + struct dentry *h_dentry;
10698 + h_dentry = au_h_dptr(dentry, bindex);
10699 + if (h_dentry && d_is_positive(h_dentry)) {
10700 + arg.bindex = bindex;
10701 + err = sio_test_empty(dentry, &arg);
10708 +/* ---------------------------------------------------------------------- */
10710 +const struct file_operations aufs_dir_fop = {
10711 + .owner = THIS_MODULE,
10712 + .llseek = default_llseek,
10713 + .read = generic_read_dir,
10714 + .iterate_shared = aufs_iterate_shared,
10715 + .unlocked_ioctl = aufs_ioctl_dir,
10716 +#ifdef CONFIG_COMPAT
10717 + .compat_ioctl = aufs_compat_ioctl_dir,
10719 + .open = aufs_open_dir,
10720 + .release = aufs_release_dir,
10721 + .flush = aufs_flush_dir,
10722 + .fsync = aufs_fsync_dir
10724 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10725 --- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
10726 +++ linux/fs/aufs/dir.h 2022-03-21 14:49:05.719966344 +0100
10728 +/* SPDX-License-Identifier: GPL-2.0 */
10730 + * Copyright (C) 2005-2021 Junjiro R. Okajima
10732 + * This program, aufs is free software; you can redistribute it and/or modify
10733 + * it under the terms of the GNU General Public License as published by
10734 + * the Free Software Foundation; either version 2 of the License, or
10735 + * (at your option) any later version.
10737 + * This program is distributed in the hope that it will be useful,
10738 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10739 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10740 + * GNU General Public License for more details.
10742 + * You should have received a copy of the GNU General Public License
10743 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10747 + * directory operations
10750 +#ifndef __AUFS_DIR_H__
10751 +#define __AUFS_DIR_H__
10755 +#include <linux/fs.h>
10757 +/* ---------------------------------------------------------------------- */
10759 +/* need to be faster and smaller */
10762 + unsigned int nh_num;
10763 + struct hlist_head *nh_head;
10766 +struct au_vdir_destr {
10767 + unsigned char len;
10768 + unsigned char name[];
10771 +struct au_vdir_dehstr {
10772 + struct hlist_node hash;
10773 + struct au_vdir_destr *str;
10774 + struct rcu_head rcu;
10775 +} ____cacheline_aligned_in_smp;
10777 +struct au_vdir_de {
10779 + unsigned char de_type;
10780 + /* caution: packed */
10781 + struct au_vdir_destr de_str;
10784 +struct au_vdir_wh {
10785 + struct hlist_node wh_hash;
10786 +#ifdef CONFIG_AUFS_SHWH
10788 + aufs_bindex_t wh_bindex;
10789 + unsigned char wh_type;
10791 + aufs_bindex_t wh_bindex;
10793 + /* caution: packed */
10794 + struct au_vdir_destr wh_str;
10797 +union au_vdir_deblk_p {
10798 + unsigned char *deblk;
10799 + struct au_vdir_de *de;
10803 + unsigned char **vd_deblk;
10804 + unsigned long vd_nblk;
10806 + unsigned long ul;
10807 + union au_vdir_deblk_p p;
10811 + unsigned int vd_deblk_sz;
10812 + unsigned long vd_jiffy;
10813 + struct rcu_head rcu;
10814 +} ____cacheline_aligned_in_smp;
10816 +/* ---------------------------------------------------------------------- */
10819 +extern const struct file_operations aufs_dir_fop;
10820 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10821 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10822 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10823 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10824 +int au_test_empty_lower(struct dentry *dentry);
10825 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10828 +unsigned int au_rdhash_est(loff_t sz);
10829 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10830 +void au_nhash_wh_free(struct au_nhash *whlist);
10831 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10833 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10834 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10835 + unsigned int d_type, aufs_bindex_t bindex,
10836 + unsigned char shwh);
10837 +void au_vdir_free(struct au_vdir *vdir);
10838 +int au_vdir_init(struct file *file);
10839 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10842 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10844 +#ifdef CONFIG_AUFS_RDU
10846 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10847 +#ifdef CONFIG_COMPAT
10848 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10849 + unsigned long arg);
10852 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10853 + unsigned int cmd, unsigned long arg)
10854 +#ifdef CONFIG_COMPAT
10855 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10856 + unsigned int cmd, unsigned long arg)
10860 +#endif /* __KERNEL__ */
10861 +#endif /* __AUFS_DIR_H__ */
10862 diff -urN /usr/share/empty/fs/aufs/dirren.c linux/fs/aufs/dirren.c
10863 --- /usr/share/empty/fs/aufs/dirren.c 1970-01-01 01:00:00.000000000 +0100
10864 +++ linux/fs/aufs/dirren.c 2022-03-21 14:49:05.723299677 +0100
10866 +// SPDX-License-Identifier: GPL-2.0
10868 + * Copyright (C) 2017-2021 Junjiro R. Okajima
10870 + * This program, aufs is free software; you can redistribute it and/or modify
10871 + * it under the terms of the GNU General Public License as published by
10872 + * the Free Software Foundation; either version 2 of the License, or
10873 + * (at your option) any later version.
10875 + * This program is distributed in the hope that it will be useful,
10876 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10877 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10878 + * GNU General Public License for more details.
10880 + * You should have received a copy of the GNU General Public License
10881 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10885 + * special handling in renaming a directory
10886 + * in order to support looking-up the before-renamed name on the lower readonly
10890 +#include <linux/byteorder/generic.h>
10893 +static void au_dr_hino_del(struct au_dr_br *dr, struct au_dr_hino *ent)
10897 + idx = au_dr_ihash(ent->dr_h_ino);
10898 + au_hbl_del(&ent->dr_hnode, dr->dr_h_ino + idx);
10901 +static int au_dr_hino_test_empty(struct au_dr_br *dr)
10904 + struct hlist_bl_head *hbl;
10907 + for (i = 0; ret && i < AuDirren_NHASH; i++) {
10908 + hbl = dr->dr_h_ino + i;
10909 + hlist_bl_lock(hbl);
10910 + ret &= hlist_bl_empty(hbl);
10911 + hlist_bl_unlock(hbl);
10917 +static struct au_dr_hino *au_dr_hino_find(struct au_dr_br *dr, ino_t ino)
10919 + struct au_dr_hino *found, *ent;
10920 + struct hlist_bl_head *hbl;
10921 + struct hlist_bl_node *pos;
10925 + idx = au_dr_ihash(ino);
10926 + hbl = dr->dr_h_ino + idx;
10927 + hlist_bl_lock(hbl);
10928 + hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10929 + if (ent->dr_h_ino == ino) {
10933 + hlist_bl_unlock(hbl);
10938 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t ino,
10939 + struct au_dr_hino *add_ent)
10942 + struct hlist_bl_head *hbl;
10943 + struct hlist_bl_node *pos;
10944 + struct au_dr_hino *ent;
10947 + idx = au_dr_ihash(ino);
10948 + hbl = dr->dr_h_ino + idx;
10949 +#if 0 /* debug print */
10951 + struct hlist_bl_node *tmp;
10953 + hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10954 + AuDbg("hi%llu\n", (unsigned long long)ent->dr_h_ino);
10957 + hlist_bl_lock(hbl);
10958 + hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10959 + if (ent->dr_h_ino == ino) {
10963 + if (!found && add_ent)
10964 + hlist_bl_add_head(&add_ent->dr_hnode, hbl);
10965 + hlist_bl_unlock(hbl);
10967 + if (!found && add_ent)
10968 + AuDbg("i%llu added\n", (unsigned long long)add_ent->dr_h_ino);
10973 +void au_dr_hino_free(struct au_dr_br *dr)
10976 + struct hlist_bl_head *hbl;
10977 + struct hlist_bl_node *pos, *tmp;
10978 + struct au_dr_hino *ent;
10980 + /* SiMustWriteLock(sb); */
10982 + for (i = 0; i < AuDirren_NHASH; i++) {
10983 + hbl = dr->dr_h_ino + i;
10984 + /* no spinlock since sbinfo must be write-locked */
10985 + hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
10986 + au_kfree_rcu(ent);
10987 + INIT_HLIST_BL_HEAD(hbl);
10991 +/* returns the number of inodes or an error */
10992 +static int au_dr_hino_store(struct super_block *sb, struct au_branch *br,
10993 + struct file *hinofile)
10997 + loff_t pos, oldsize;
10999 + struct inode *hinoinode;
11000 + struct hlist_bl_head *hbl;
11001 + struct hlist_bl_node *n1, *n2;
11002 + struct au_dr_hino *ent;
11004 + SiMustWriteLock(sb);
11005 + AuDebugOn(!au_br_writable(br->br_perm));
11007 + hinoinode = file_inode(hinofile);
11008 + oldsize = i_size_read(hinoinode);
11012 + hbl = br->br_dirren.dr_h_ino;
11013 + for (i = 0; !err && i < AuDirren_NHASH; i++, hbl++) {
11014 + /* no bit-lock since sbinfo must be write-locked */
11015 + hlist_bl_for_each_entry_safe(ent, n1, n2, hbl, dr_hnode) {
11016 + AuDbg("hi%llu, %pD2\n",
11017 + (unsigned long long)ent->dr_h_ino, hinofile);
11018 + u64 = cpu_to_be64(ent->dr_h_ino);
11019 + ssz = vfsub_write_k(hinofile, &u64, sizeof(u64), &pos);
11020 + if (ssz == sizeof(u64))
11023 + /* write error */
11024 + pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11031 + /* regardless the error */
11032 + if (pos < oldsize) {
11033 + err = vfsub_trunc(&hinofile->f_path, pos, /*attr*/0, hinofile);
11041 +static int au_dr_hino_load(struct au_dr_br *dr, struct file *hinofile)
11048 + struct au_dr_hino *ent;
11049 + struct inode *hinoinode;
11050 + struct hlist_bl_head *hbl;
11054 + hbl = dr->dr_h_ino;
11055 + hinoinode = file_inode(hinofile);
11056 + sz = i_size_read(hinoinode);
11057 + AuDebugOn(sz % sizeof(u64));
11058 + n = sz / sizeof(u64);
11060 + ssz = vfsub_read_k(hinofile, &u64, sizeof(u64), &pos);
11061 + if (unlikely(ssz != sizeof(u64))) {
11062 + pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11069 + ent = kmalloc(sizeof(*ent), GFP_NOFS);
11075 + ent->dr_h_ino = be64_to_cpu((__force __be64)u64);
11076 + AuDbg("hi%llu, %pD2\n",
11077 + (unsigned long long)ent->dr_h_ino, hinofile);
11078 + hidx = au_dr_ihash(ent->dr_h_ino);
11079 + au_hbl_add(&ent->dr_hnode, hbl + hidx);
11081 + goto out; /* success */
11084 + au_dr_hino_free(dr);
11091 + * @bindex/@br is a switch to distinguish whether suspending hnotify or not.
11092 + * @path is a switch to distinguish load and store.
11094 +static int au_dr_hino(struct super_block *sb, aufs_bindex_t bindex,
11095 + struct au_branch *br, const struct path *path)
11098 + unsigned char load, suspend;
11099 + struct file *hinofile;
11100 + struct au_hinode *hdir;
11101 + struct inode *dir, *delegated;
11102 + struct path hinopath;
11103 + struct qstr hinoname = QSTR_INIT(AUFS_WH_DR_BRHINO,
11104 + sizeof(AUFS_WH_DR_BRHINO) - 1);
11106 + AuDebugOn(bindex < 0 && !br);
11107 + AuDebugOn(bindex >= 0 && br);
11112 + br = au_sbr(sb, bindex);
11115 + path = &br->br_path;
11116 + AuDebugOn(!au_br_writable(br->br_perm));
11117 + if (unlikely(!au_br_writable(br->br_perm)))
11123 + dir = d_inode(sb->s_root);
11124 + hdir = au_hinode(au_ii(dir), bindex);
11125 + dir = hdir->hi_inode;
11126 + au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11128 + dir = d_inode(path->dentry);
11129 + inode_lock_nested(dir, AuLsc_I_CHILD);
11131 + hinopath.mnt = path->mnt;
11132 + hinopath.dentry = vfsub_lkup_one(&hinoname, (struct path *)path);
11133 + err = PTR_ERR(hinopath.dentry);
11134 + if (IS_ERR(hinopath.dentry))
11138 + flags = O_RDONLY;
11140 + if (d_is_negative(hinopath.dentry))
11141 + goto out_dput; /* success */
11143 + if (au_dr_hino_test_empty(&br->br_dirren)) {
11144 + if (d_is_positive(hinopath.dentry)) {
11145 + delegated = NULL;
11146 + err = vfsub_unlink(dir, &hinopath, &delegated,
11149 + if (unlikely(err))
11150 + pr_err("ignored err %d, %pd2\n",
11151 + err, hinopath.dentry);
11152 + if (unlikely(err == -EWOULDBLOCK))
11157 + } else if (!d_is_positive(hinopath.dentry)) {
11158 + err = vfsub_create(dir, &hinopath, 0600,
11159 + /*want_excl*/false);
11161 + if (unlikely(err))
11164 + flags = O_WRONLY;
11166 + hinofile = vfsub_dentry_open(&hinopath, flags);
11168 + au_hn_inode_unlock(hdir);
11170 + inode_unlock(dir);
11171 + dput(hinopath.dentry);
11172 + AuTraceErrPtr(hinofile);
11173 + if (IS_ERR(hinofile)) {
11174 + err = PTR_ERR(hinofile);
11179 + err = au_dr_hino_load(&br->br_dirren, hinofile);
11181 + err = au_dr_hino_store(sb, br, hinofile);
11186 + dput(hinopath.dentry);
11189 + au_hn_inode_unlock(hdir);
11191 + inode_unlock(dir);
11197 +/* ---------------------------------------------------------------------- */
11199 +static int au_dr_brid_init(struct au_dr_brid *brid, const struct path *path)
11202 + struct kstatfs kstfs;
11204 + struct dentry *dentry;
11205 + struct super_block *sb;
11207 + err = vfs_statfs((void *)path, &kstfs);
11209 + if (unlikely(err))
11212 + /* todo: support for UUID */
11214 + if (kstfs.f_fsid.val[0] || kstfs.f_fsid.val[1]) {
11215 + brid->type = AuBrid_FSID;
11216 + brid->fsid = kstfs.f_fsid;
11218 + dentry = path->dentry;
11219 + sb = dentry->d_sb;
11222 + brid->type = AuBrid_DEV;
11231 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
11232 + const struct path *path)
11235 + struct au_dr_br *dr;
11236 + struct hlist_bl_head *hbl;
11238 + dr = &br->br_dirren;
11239 + hbl = dr->dr_h_ino;
11240 + for (i = 0; i < AuDirren_NHASH; i++, hbl++)
11241 + INIT_HLIST_BL_HEAD(hbl);
11243 + err = au_dr_brid_init(&dr->dr_brid, path);
11244 + if (unlikely(err))
11247 + if (au_opt_test(au_mntflags(sb), DIRREN))
11248 + err = au_dr_hino(sb, /*bindex*/-1, br, path);
11255 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br)
11260 + if (au_br_writable(br->br_perm))
11261 + err = au_dr_hino(sb, /*bindex*/-1, br, /*path*/NULL);
11263 + au_dr_hino_free(&br->br_dirren);
11268 +/* ---------------------------------------------------------------------- */
11270 +static int au_brid_str(struct au_dr_brid *brid, struct inode *h_inode,
11271 + char *buf, size_t sz)
11274 + unsigned int major, minor;
11278 + err = snprintf(p, sz, "%d_", brid->type);
11279 + AuDebugOn(err > sz);
11282 + switch (brid->type) {
11283 + case AuBrid_Unset:
11285 + case AuBrid_UUID:
11286 + err = snprintf(p, sz, "%pU", brid->uuid.b);
11288 + case AuBrid_FSID:
11289 + err = snprintf(p, sz, "%08x-%08x",
11290 + brid->fsid.val[0], brid->fsid.val[1]);
11293 + major = MAJOR(brid->dev);
11294 + minor = MINOR(brid->dev);
11295 + if (major <= 0xff && minor <= 0xff)
11296 + err = snprintf(p, sz, "%02x%02x", major, minor);
11298 + err = snprintf(p, sz, "%03x:%05x", major, minor);
11301 + AuDebugOn(err > sz);
11304 + err = snprintf(p, sz, "_%llu", (unsigned long long)h_inode->i_ino);
11305 + AuDebugOn(err > sz);
11312 +static int au_drinfo_name(struct au_branch *br, char *name, int len)
11315 + struct dentry *br_dentry;
11316 + struct inode *br_inode;
11318 + br_dentry = au_br_dentry(br);
11319 + br_inode = d_inode(br_dentry);
11320 + rlen = au_brid_str(&br->br_dirren.dr_brid, br_inode, name, len);
11321 + AuDebugOn(rlen >= AUFS_DIRREN_ENV_VAL_SZ);
11322 + AuDebugOn(rlen > len);
11327 +/* ---------------------------------------------------------------------- */
11330 + * from the given @h_dentry, construct drinfo at @*fdata.
11331 + * when the size of @*fdata is not enough, reallocate and return new @fdata and
11334 +static int au_drinfo_construct(struct au_drinfo_fdata **fdata,
11335 + struct dentry *h_dentry,
11336 + unsigned char *allocated)
11339 + struct au_drinfo_fdata *f, *p;
11340 + struct au_drinfo *drinfo;
11341 + struct inode *h_inode;
11342 + struct qstr *qname;
11346 + h_inode = d_inode(h_dentry);
11347 + qname = &h_dentry->d_name;
11348 + drinfo = &f->drinfo;
11349 + drinfo->ino = (__force uint64_t)cpu_to_be64(h_inode->i_ino);
11350 + drinfo->oldnamelen = qname->len;
11351 + if (*allocated < sizeof(*f) + qname->len) {
11352 + v = roundup_pow_of_two(*allocated + qname->len);
11353 + p = au_krealloc(f, v, GFP_NOFS, /*may_shrink*/0);
11354 + if (unlikely(!p)) {
11362 + drinfo = &f->drinfo;
11364 + memcpy(drinfo->oldname, qname->name, qname->len);
11365 + AuDbg("i%llu, %.*s\n",
11366 + be64_to_cpu((__force __be64)drinfo->ino), drinfo->oldnamelen,
11367 + drinfo->oldname);
11374 +/* callers have to free the return value */
11375 +static struct au_drinfo *au_drinfo_read_k(struct file *file, ino_t h_ino)
11377 + struct au_drinfo *ret, *drinfo;
11378 + struct au_drinfo_fdata fdata;
11383 + ret = ERR_PTR(-EIO);
11385 + ssz = vfsub_read_k(file, &fdata, sizeof(fdata), &pos);
11386 + if (unlikely(ssz != sizeof(fdata))) {
11387 + AuIOErr("ssz %zd, %u, %pD2\n",
11388 + ssz, (unsigned int)sizeof(fdata), file);
11392 + fdata.magic = ntohl((__force __be32)fdata.magic);
11393 + switch (fdata.magic) {
11394 + case AUFS_DRINFO_MAGIC_V1:
11397 + AuIOErr("magic-num 0x%x, 0x%x, %pD2\n",
11398 + fdata.magic, AUFS_DRINFO_MAGIC_V1, file);
11402 + drinfo = &fdata.drinfo;
11403 + len = drinfo->oldnamelen;
11405 + AuIOErr("broken drinfo %pD2\n", file);
11410 + drinfo->ino = be64_to_cpu((__force __be64)drinfo->ino);
11411 + if (unlikely(h_ino && drinfo->ino != h_ino)) {
11412 + AuDbg("ignored i%llu, i%llu, %pD2\n",
11413 + (unsigned long long)drinfo->ino,
11414 + (unsigned long long)h_ino, file);
11415 + goto out; /* success */
11418 + ret = kmalloc(sizeof(*ret) + len, GFP_NOFS);
11419 + if (unlikely(!ret)) {
11420 + ret = ERR_PTR(-ENOMEM);
11421 + AuTraceErrPtr(ret);
11426 + ssz = vfsub_read_k(file, (void *)ret->oldname, len, &pos);
11427 + if (unlikely(ssz != len)) {
11428 + au_kfree_rcu(ret);
11429 + ret = ERR_PTR(-EIO);
11430 + AuIOErr("ssz %zd, %u, %pD2\n", ssz, len, file);
11434 + AuDbg("oldname %.*s\n", ret->oldnamelen, ret->oldname);
11440 +/* ---------------------------------------------------------------------- */
11442 +/* in order to be revertible */
11443 +struct au_drinfo_rev_elm {
11445 + struct dentry *info_dentry;
11446 + struct au_drinfo *info_last;
11449 +struct au_drinfo_rev {
11450 + unsigned char already;
11451 + aufs_bindex_t nelm;
11452 + struct au_drinfo_rev_elm elm[];
11455 +/* todo: isn't it too large? */
11456 +struct au_drinfo_store {
11457 + struct path h_ppath;
11458 + struct dentry *h_dentry;
11459 + struct au_drinfo_fdata *fdata;
11460 + char *infoname; /* inside of whname, just after PFX */
11461 + char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ];
11462 + aufs_bindex_t btgt, btail;
11463 + unsigned char no_sio,
11464 + allocated, /* current size of *fdata */
11465 + infonamelen, /* room size for p */
11466 + whnamelen, /* length of the generated name */
11467 + renameback; /* renamed back */
11470 +/* on rename(2) error, the caller should revert it using @elm */
11471 +static int au_drinfo_do_store(struct au_drinfo_store *w,
11472 + struct au_drinfo_rev_elm *elm)
11477 + struct path infopath = {
11478 + .mnt = w->h_ppath.mnt
11480 + struct inode *h_dir, *h_inode, *delegated;
11481 + struct file *infofile;
11482 + struct qstr *qname;
11485 + && memcmp(elm, page_address(ZERO_PAGE(0)), sizeof(*elm)));
11487 + infopath.dentry = vfsub_lookup_one_len(w->whname, &w->h_ppath,
11489 + AuTraceErrPtr(infopath.dentry);
11490 + if (IS_ERR(infopath.dentry)) {
11491 + err = PTR_ERR(infopath.dentry);
11496 + h_dir = d_inode(w->h_ppath.dentry);
11497 + if (elm && d_is_negative(infopath.dentry)) {
11498 + err = vfsub_create(h_dir, &infopath, 0600, /*want_excl*/true);
11500 + if (unlikely(err))
11502 + elm->created = 1;
11503 + elm->info_dentry = dget(infopath.dentry);
11506 + infofile = vfsub_dentry_open(&infopath, O_RDWR);
11507 + AuTraceErrPtr(infofile);
11508 + if (IS_ERR(infofile)) {
11509 + err = PTR_ERR(infofile);
11513 + h_inode = d_inode(infopath.dentry);
11514 + if (elm && i_size_read(h_inode)) {
11515 + h_inode = d_inode(w->h_dentry);
11516 + elm->info_last = au_drinfo_read_k(infofile, h_inode->i_ino);
11517 + AuTraceErrPtr(elm->info_last);
11518 + if (IS_ERR(elm->info_last)) {
11519 + err = PTR_ERR(elm->info_last);
11520 + elm->info_last = NULL;
11521 + AuDebugOn(elm->info_dentry);
11526 + if (elm && w->renameback) {
11527 + delegated = NULL;
11528 + err = vfsub_unlink(h_dir, &infopath, &delegated, /*force*/0);
11530 + if (unlikely(err == -EWOULDBLOCK))
11536 + qname = &w->h_dentry->d_name;
11537 + len = sizeof(*w->fdata) + qname->len;
11539 + len = sizeof(*w->fdata) + w->fdata->drinfo.oldnamelen;
11540 + ssz = vfsub_write_k(infofile, w->fdata, len, &pos);
11541 + if (ssz == len) {
11542 + AuDbg("hi%llu, %.*s\n", w->fdata->drinfo.ino,
11543 + w->fdata->drinfo.oldnamelen, w->fdata->drinfo.oldname);
11544 + goto out_fput; /* success */
11549 + /* the caller should revert it using @elm */
11555 + dput(infopath.dentry);
11561 +struct au_call_drinfo_do_store_args {
11563 + struct au_drinfo_store *w;
11564 + struct au_drinfo_rev_elm *elm;
11567 +static void au_call_drinfo_do_store(void *args)
11569 + struct au_call_drinfo_do_store_args *a = args;
11571 + *a->errp = au_drinfo_do_store(a->w, a->elm);
11574 +static int au_drinfo_store_sio(struct au_drinfo_store *w,
11575 + struct au_drinfo_rev_elm *elm)
11577 + int err, wkq_err;
11580 + err = au_drinfo_do_store(w, elm);
11582 + struct au_call_drinfo_do_store_args a = {
11587 + wkq_err = au_wkq_wait(au_call_drinfo_do_store, &a);
11588 + if (unlikely(wkq_err))
11596 +static int au_drinfo_store_work_init(struct au_drinfo_store *w,
11597 + aufs_bindex_t btgt)
11601 + memset(w, 0, sizeof(*w));
11602 + w->allocated = roundup_pow_of_two(sizeof(*w->fdata) + 40);
11603 + strcpy(w->whname, AUFS_WH_DR_INFO_PFX);
11604 + w->infoname = w->whname + sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11605 + w->infonamelen = sizeof(w->whname) - sizeof(AUFS_WH_DR_INFO_PFX);
11607 + w->no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
11610 + w->fdata = kcalloc(1, w->allocated, GFP_NOFS);
11611 + if (unlikely(!w->fdata)) {
11615 + w->fdata->magic = (__force uint32_t)htonl(AUFS_DRINFO_MAGIC_V1);
11622 +static void au_drinfo_store_work_fin(struct au_drinfo_store *w)
11624 + au_kfree_rcu(w->fdata);
11627 +static void au_drinfo_store_rev(struct au_drinfo_rev *rev,
11628 + struct au_drinfo_store *w)
11630 + struct au_drinfo_rev_elm *elm;
11631 + struct inode *h_dir, *delegated;
11633 + struct path infopath = {
11634 + .mnt = w->h_ppath.mnt
11637 + h_dir = d_inode(w->h_ppath.dentry);
11638 + IMustLock(h_dir);
11642 + for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11643 + AuDebugOn(elm->created && elm->info_last);
11644 + if (elm->created) {
11646 + delegated = NULL;
11647 + infopath.dentry = elm->info_dentry;
11648 + err = vfsub_unlink(h_dir, &infopath, &delegated,
11651 + if (unlikely(err == -EWOULDBLOCK))
11653 + dput(elm->info_dentry);
11654 + } else if (elm->info_last) {
11656 + w->fdata->drinfo = *elm->info_last;
11657 + memcpy(w->fdata->drinfo.oldname,
11658 + elm->info_last->oldname,
11659 + elm->info_last->oldnamelen);
11660 + err = au_drinfo_store_sio(w, /*elm*/NULL);
11661 + au_kfree_rcu(elm->info_last);
11663 + if (unlikely(err))
11664 + AuIOErr("%d, %s\n", err, w->whname);
11665 + /* go on even if err */
11669 +/* caller has to call au_dr_rename_fin() later */
11670 +static int au_drinfo_store(struct dentry *dentry, aufs_bindex_t btgt,
11671 + struct qstr *dst_name, void *_rev)
11673 + int err, sz, nelm;
11674 + aufs_bindex_t bindex, btail;
11675 + struct au_drinfo_store work;
11676 + struct au_drinfo_rev *rev, **p;
11677 + struct au_drinfo_rev_elm *elm;
11678 + struct super_block *sb;
11679 + struct au_branch *br;
11680 + struct au_hinode *hdir;
11682 + err = au_drinfo_store_work_init(&work, btgt);
11684 + if (unlikely(err))
11688 + btail = au_dbtaildir(dentry);
11689 + nelm = btail - btgt;
11690 + sz = sizeof(*rev) + sizeof(*elm) * nelm;
11691 + rev = kcalloc(1, sz, GFP_NOFS);
11692 + if (unlikely(!rev)) {
11696 + rev->nelm = nelm;
11702 + sb = dentry->d_sb;
11703 + work.h_ppath.dentry = au_h_dptr(dentry, btgt);
11704 + work.h_ppath.mnt = au_sbr_mnt(sb, btgt);
11705 + hdir = au_hi(d_inode(dentry), btgt);
11706 + au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11707 + for (bindex = btgt + 1; bindex <= btail; bindex++, elm++) {
11708 + work.h_dentry = au_h_dptr(dentry, bindex);
11709 + if (!work.h_dentry)
11712 + err = au_drinfo_construct(&work.fdata, work.h_dentry,
11713 + &work.allocated);
11715 + if (unlikely(err))
11718 + work.renameback = au_qstreq(&work.h_dentry->d_name, dst_name);
11719 + br = au_sbr(sb, bindex);
11720 + work.whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11721 + work.whnamelen += au_drinfo_name(br, work.infoname,
11722 + work.infonamelen);
11723 + AuDbg("whname %.*s, i%llu, %.*s\n",
11724 + work.whnamelen, work.whname,
11725 + be64_to_cpu((__force __be64)work.fdata->drinfo.ino),
11726 + work.fdata->drinfo.oldnamelen,
11727 + work.fdata->drinfo.oldname);
11729 + err = au_drinfo_store_sio(&work, elm);
11731 + if (unlikely(err))
11734 + if (unlikely(err)) {
11735 + /* revert all drinfo */
11736 + au_drinfo_store_rev(rev, &work);
11737 + au_kfree_try_rcu(rev);
11740 + au_hn_inode_unlock(hdir);
11743 + au_drinfo_store_work_fin(&work);
11748 +/* ---------------------------------------------------------------------- */
11750 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
11751 + struct qstr *dst_name, void *_rev)
11753 + int err, already;
11755 + struct super_block *sb;
11756 + struct au_branch *br;
11757 + struct au_dr_br *dr;
11758 + struct dentry *h_dentry;
11759 + struct inode *h_inode;
11760 + struct au_dr_hino *ent;
11761 + struct au_drinfo_rev *rev, **p;
11763 + AuDbg("bindex %d\n", bindex);
11766 + ent = kmalloc(sizeof(*ent), GFP_NOFS);
11767 + if (unlikely(!ent))
11771 + br = au_sbr(sb, bindex);
11772 + dr = &br->br_dirren;
11773 + h_dentry = au_h_dptr(src, bindex);
11774 + h_inode = d_inode(h_dentry);
11775 + ino = h_inode->i_ino;
11776 + ent->dr_h_ino = ino;
11777 + already = au_dr_hino_test_add(dr, ino, ent);
11778 + AuDbg("b%d, hi%llu, already %d\n",
11779 + bindex, (unsigned long long)ino, already);
11781 + err = au_drinfo_store(src, bindex, dst_name, _rev);
11786 + rev->already = already;
11787 + goto out; /* success */
11792 + au_dr_hino_del(dr, ent);
11793 + au_kfree_rcu(ent);
11800 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11802 + struct au_drinfo_rev *rev;
11803 + struct au_drinfo_rev_elm *elm;
11808 + for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11809 + dput(elm->info_dentry);
11810 + au_kfree_rcu(elm->info_last);
11812 + au_kfree_try_rcu(rev);
11815 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11818 + struct au_drinfo_store work;
11819 + struct au_drinfo_rev *rev = _rev;
11820 + struct super_block *sb;
11821 + struct au_branch *br;
11822 + struct inode *h_inode;
11823 + struct au_dr_br *dr;
11824 + struct au_dr_hino *ent;
11826 + err = au_drinfo_store_work_init(&work, btgt);
11827 + if (unlikely(err))
11831 + br = au_sbr(sb, btgt);
11832 + work.h_ppath.dentry = au_h_dptr(src, btgt);
11833 + work.h_ppath.mnt = au_br_mnt(br);
11834 + au_drinfo_store_rev(rev, &work);
11835 + au_drinfo_store_work_fin(&work);
11836 + if (rev->already)
11839 + dr = &br->br_dirren;
11840 + h_inode = d_inode(work.h_ppath.dentry);
11841 + ent = au_dr_hino_find(dr, h_inode->i_ino);
11843 + au_dr_hino_del(dr, ent);
11844 + au_kfree_rcu(ent);
11847 + au_kfree_try_rcu(rev);
11848 + if (unlikely(err))
11849 + pr_err("failed to remove dirren info\n");
11852 +/* ---------------------------------------------------------------------- */
11854 +static struct au_drinfo *au_drinfo_do_load(struct path *h_ppath,
11855 + char *whname, int whnamelen,
11856 + struct dentry **info_dentry)
11858 + struct au_drinfo *drinfo;
11860 + struct inode *h_dir;
11861 + struct path infopath;
11864 + AuDbg("%pd/%.*s\n", h_ppath->dentry, whnamelen, whname);
11866 + *info_dentry = NULL;
11869 + h_dir = d_inode(h_ppath->dentry);
11870 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
11871 + infopath.dentry = vfsub_lookup_one_len(whname, h_ppath, whnamelen);
11872 + if (IS_ERR(infopath.dentry)) {
11873 + drinfo = (void *)infopath.dentry;
11877 + if (d_is_negative(infopath.dentry))
11878 + goto out_dput; /* success */
11880 + infopath.mnt = h_ppath->mnt;
11881 + f = vfsub_dentry_open(&infopath, O_RDONLY);
11882 + inode_unlock_shared(h_dir);
11885 + drinfo = (void *)f;
11889 + drinfo = au_drinfo_read_k(f, /*h_ino*/0);
11890 + if (IS_ERR_OR_NULL(drinfo))
11893 + AuDbg("oldname %.*s\n", drinfo->oldnamelen, drinfo->oldname);
11894 + *info_dentry = dget(infopath.dentry); /* keep it alive */
11899 + dput(infopath.dentry);
11902 + inode_unlock_shared(h_dir);
11903 + AuTraceErrPtr(drinfo);
11907 +struct au_drinfo_do_load_args {
11908 + struct au_drinfo **drinfop;
11909 + struct path *h_ppath;
11912 + struct dentry **info_dentry;
11915 +static void au_call_drinfo_do_load(void *args)
11917 + struct au_drinfo_do_load_args *a = args;
11919 + *a->drinfop = au_drinfo_do_load(a->h_ppath, a->whname, a->whnamelen,
11923 +struct au_drinfo_load {
11924 + struct path h_ppath;
11925 + struct qstr *qname;
11926 + unsigned char no_sio;
11928 + aufs_bindex_t ninfo;
11929 + struct au_drinfo **drinfo;
11932 +static int au_drinfo_load(struct au_drinfo_load *w, aufs_bindex_t bindex,
11933 + struct au_branch *br)
11935 + int err, wkq_err, whnamelen, e;
11936 + char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ]
11937 + = AUFS_WH_DR_INFO_PFX;
11938 + struct au_drinfo *drinfo;
11939 + struct qstr oldname;
11940 + struct inode *h_dir, *delegated;
11941 + struct dentry *info_dentry;
11942 + struct path infopath;
11944 + whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11945 + whnamelen += au_drinfo_name(br, whname + whnamelen,
11946 + sizeof(whname) - whnamelen);
11948 + drinfo = au_drinfo_do_load(&w->h_ppath, whname, whnamelen,
11951 + struct au_drinfo_do_load_args args = {
11952 + .drinfop = &drinfo,
11953 + .h_ppath = &w->h_ppath,
11954 + .whname = whname,
11955 + .whnamelen = whnamelen,
11956 + .info_dentry = &info_dentry
11958 + wkq_err = au_wkq_wait(au_call_drinfo_do_load, &args);
11959 + if (unlikely(wkq_err))
11960 + drinfo = ERR_PTR(wkq_err);
11962 + err = PTR_ERR(drinfo);
11963 + if (IS_ERR_OR_NULL(drinfo))
11967 + oldname.len = drinfo->oldnamelen;
11968 + oldname.name = drinfo->oldname;
11969 + if (au_qstreq(w->qname, &oldname)) {
11970 + /* the name is renamed back */
11971 + au_kfree_rcu(drinfo);
11974 + infopath.dentry = info_dentry;
11975 + infopath.mnt = w->h_ppath.mnt;
11976 + h_dir = d_inode(w->h_ppath.dentry);
11977 + delegated = NULL;
11978 + inode_lock_nested(h_dir, AuLsc_I_PARENT);
11979 + e = vfsub_unlink(h_dir, &infopath, &delegated, !w->no_sio);
11980 + inode_unlock(h_dir);
11982 + AuIOErr("ignored %d, %pd2\n", e, &infopath.dentry);
11983 + if (unlikely(e == -EWOULDBLOCK))
11986 + au_kfree_rcu(w->drinfo[bindex]);
11987 + w->drinfo[bindex] = drinfo;
11988 + dput(info_dentry);
11995 +/* ---------------------------------------------------------------------- */
11997 +static void au_dr_lkup_free(struct au_drinfo **drinfo, int n)
11999 + struct au_drinfo **p = drinfo;
12002 + au_kfree_rcu(*drinfo++);
12003 + au_kfree_try_rcu(p);
12006 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12007 + aufs_bindex_t btgt)
12010 + struct au_drinfo_load w;
12011 + aufs_bindex_t bindex, bbot;
12012 + struct au_branch *br;
12013 + struct inode *h_dir;
12014 + struct au_dr_hino *ent;
12015 + struct super_block *sb;
12017 + AuDbg("%.*s, name %.*s, whname %.*s, b%d\n",
12018 + AuLNPair(&dentry->d_name), AuLNPair(&lkup->dirren.dr_name),
12019 + AuLNPair(&lkup->whname), btgt);
12021 + sb = dentry->d_sb;
12022 + bbot = au_sbbot(sb);
12023 + w.ninfo = bbot + 1;
12024 + if (!lkup->dirren.drinfo) {
12025 + lkup->dirren.drinfo = kcalloc(w.ninfo,
12026 + sizeof(*lkup->dirren.drinfo),
12028 + if (unlikely(!lkup->dirren.drinfo)) {
12032 + lkup->dirren.ninfo = w.ninfo;
12034 + w.drinfo = lkup->dirren.drinfo;
12035 + w.no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
12036 + w.h_ppath.dentry = au_h_dptr(dentry, btgt);
12037 + AuDebugOn(!w.h_ppath.dentry);
12038 + w.h_ppath.mnt = au_sbr_mnt(sb, btgt);
12039 + w.qname = &dentry->d_name;
12042 + for (bindex = btgt + 1; bindex <= bbot; bindex++) {
12043 + br = au_sbr(sb, bindex);
12044 + err = au_drinfo_load(&w, bindex, br);
12045 + if (unlikely(err))
12047 + if (w.drinfo[bindex])
12051 + br = au_sbr(sb, btgt);
12052 + h_dir = d_inode(w.h_ppath.dentry);
12053 + ent = au_dr_hino_find(&br->br_dirren, h_dir->i_ino);
12055 + au_dr_hino_del(&br->br_dirren, ent);
12056 + au_kfree_rcu(ent);
12058 + goto out; /* success */
12061 + au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12062 + lkup->dirren.ninfo = 0;
12063 + lkup->dirren.drinfo = NULL;
12069 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup)
12071 + au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12074 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt)
12077 + struct au_drinfo *drinfo;
12080 + if (!lkup->dirren.drinfo)
12082 + AuDebugOn(lkup->dirren.ninfo <= btgt);
12083 + drinfo = lkup->dirren.drinfo[btgt];
12087 + au_kfree_try_rcu(lkup->whname.name);
12088 + lkup->whname.name = NULL;
12089 + lkup->dirren.dr_name.len = drinfo->oldnamelen;
12090 + lkup->dirren.dr_name.name = drinfo->oldname;
12091 + lkup->name = &lkup->dirren.dr_name;
12092 + err = au_wh_name_alloc(&lkup->whname, lkup->name);
12094 + AuDbg("name %.*s, whname %.*s, b%d\n",
12095 + AuLNPair(lkup->name), AuLNPair(&lkup->whname),
12103 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12107 + struct au_drinfo *drinfo;
12110 + if (!lkup->dirren.drinfo)
12112 + AuDebugOn(lkup->dirren.ninfo <= bindex);
12113 + drinfo = lkup->dirren.drinfo[bindex];
12117 + match = (drinfo->ino == h_ino);
12118 + AuDbg("match %d\n", match);
12124 +/* ---------------------------------------------------------------------- */
12126 +int au_dr_opt_set(struct super_block *sb)
12129 + aufs_bindex_t bindex, bbot;
12130 + struct au_branch *br;
12133 + bbot = au_sbbot(sb);
12134 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
12135 + br = au_sbr(sb, bindex);
12136 + err = au_dr_hino(sb, bindex, /*br*/NULL, &br->br_path);
12142 +int au_dr_opt_flush(struct super_block *sb)
12145 + aufs_bindex_t bindex, bbot;
12146 + struct au_branch *br;
12149 + bbot = au_sbbot(sb);
12150 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
12151 + br = au_sbr(sb, bindex);
12152 + if (au_br_writable(br->br_perm))
12153 + err = au_dr_hino(sb, bindex, /*br*/NULL, /*path*/NULL);
12159 +int au_dr_opt_clr(struct super_block *sb, int no_flush)
12162 + aufs_bindex_t bindex, bbot;
12163 + struct au_branch *br;
12167 + err = au_dr_opt_flush(sb);
12168 + if (unlikely(err))
12172 + bbot = au_sbbot(sb);
12173 + for (bindex = 0; bindex <= bbot; bindex++) {
12174 + br = au_sbr(sb, bindex);
12175 + au_dr_hino_free(&br->br_dirren);
12181 diff -urN /usr/share/empty/fs/aufs/dirren.h linux/fs/aufs/dirren.h
12182 --- /usr/share/empty/fs/aufs/dirren.h 1970-01-01 01:00:00.000000000 +0100
12183 +++ linux/fs/aufs/dirren.h 2022-03-21 14:49:05.723299677 +0100
12185 +/* SPDX-License-Identifier: GPL-2.0 */
12187 + * Copyright (C) 2017-2021 Junjiro R. Okajima
12189 + * This program, aufs is free software; you can redistribute it and/or modify
12190 + * it under the terms of the GNU General Public License as published by
12191 + * the Free Software Foundation; either version 2 of the License, or
12192 + * (at your option) any later version.
12194 + * This program is distributed in the hope that it will be useful,
12195 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12196 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12197 + * GNU General Public License for more details.
12199 + * You should have received a copy of the GNU General Public License
12200 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12204 + * renamed dir info
12207 +#ifndef __AUFS_DIRREN_H__
12208 +#define __AUFS_DIRREN_H__
12212 +#include <linux/dcache.h>
12213 +#include <linux/statfs.h>
12214 +#include <linux/uuid.h>
12217 +#define AuDirren_NHASH 100
12219 +#ifdef CONFIG_AUFS_DIRREN
12220 +enum au_brid_type {
12227 +struct au_dr_brid {
12228 + enum au_brid_type type;
12230 + uuid_t uuid; /* unimplemented yet */
12236 +/* 20 is the max digits length of ulong 64 */
12237 +/* brid-type "_" uuid "_" inum */
12238 +#define AUFS_DIRREN_FNAME_SZ (1 + 1 + UUID_STRING_LEN + 20)
12239 +#define AUFS_DIRREN_ENV_VAL_SZ (AUFS_DIRREN_FNAME_SZ + 1 + 20)
12241 +struct au_dr_hino {
12242 + struct hlist_bl_node dr_hnode;
12247 + struct hlist_bl_head dr_h_ino[AuDirren_NHASH];
12248 + struct au_dr_brid dr_brid;
12251 +struct au_dr_lookup {
12252 + /* dr_name is pointed by struct au_do_lookup_args.name */
12253 + struct qstr dr_name; /* subset of dr_info */
12254 + aufs_bindex_t ninfo;
12255 + struct au_drinfo **drinfo;
12258 +struct au_dr_hino;
12260 +struct au_dr_br { };
12261 +struct au_dr_lookup { };
12264 +/* ---------------------------------------------------------------------- */
12267 +struct au_do_lookup_args;
12269 +#ifdef CONFIG_AUFS_DIRREN
12270 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t h_ino,
12271 + struct au_dr_hino *add_ent);
12272 +void au_dr_hino_free(struct au_dr_br *dr);
12273 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
12274 + const struct path *path);
12275 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br);
12276 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
12277 + struct qstr *dst_name, void *_rev);
12278 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *rev);
12279 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t bindex, void *rev);
12280 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12281 + aufs_bindex_t bindex);
12282 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12283 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12285 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup);
12286 +int au_dr_opt_set(struct super_block *sb);
12287 +int au_dr_opt_flush(struct super_block *sb);
12288 +int au_dr_opt_clr(struct super_block *sb, int no_flush);
12290 +AuStubInt0(au_dr_hino_test_add, struct au_dr_br *dr, ino_t h_ino,
12291 + struct au_dr_hino *add_ent);
12292 +AuStubVoid(au_dr_hino_free, struct au_dr_br *dr);
12293 +AuStubInt0(au_dr_br_init, struct super_block *sb, struct au_branch *br,
12294 + const struct path *path);
12295 +AuStubInt0(au_dr_br_fin, struct super_block *sb, struct au_branch *br);
12296 +AuStubInt0(au_dr_rename, struct dentry *src, aufs_bindex_t bindex,
12297 + struct qstr *dst_name, void *_rev);
12298 +AuStubVoid(au_dr_rename_fin, struct dentry *src, aufs_bindex_t btgt, void *rev);
12299 +AuStubVoid(au_dr_rename_rev, struct dentry *src, aufs_bindex_t bindex,
12301 +AuStubInt0(au_dr_lkup, struct au_do_lookup_args *lkup, struct dentry *dentry,
12302 + aufs_bindex_t bindex);
12303 +AuStubInt0(au_dr_lkup_name, struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12304 +AuStubInt0(au_dr_lkup_h_ino, struct au_do_lookup_args *lkup,
12305 + aufs_bindex_t bindex, ino_t h_ino);
12306 +AuStubVoid(au_dr_lkup_fin, struct au_do_lookup_args *lkup);
12307 +AuStubInt0(au_dr_opt_set, struct super_block *sb);
12308 +AuStubInt0(au_dr_opt_flush, struct super_block *sb);
12309 +AuStubInt0(au_dr_opt_clr, struct super_block *sb, int no_flush);
12312 +/* ---------------------------------------------------------------------- */
12314 +#ifdef CONFIG_AUFS_DIRREN
12315 +static inline int au_dr_ihash(ino_t h_ino)
12317 + return h_ino % AuDirren_NHASH;
12320 +AuStubInt0(au_dr_ihash, ino_t h_ino);
12323 +#endif /* __KERNEL__ */
12324 +#endif /* __AUFS_DIRREN_H__ */
12325 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
12326 --- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
12327 +++ linux/fs/aufs/dynop.c 2022-05-23 09:44:37.726505832 +0200
12329 +// SPDX-License-Identifier: GPL-2.0
12331 + * Copyright (C) 2010-2021 Junjiro R. Okajima
12333 + * This program, aufs is free software; you can redistribute it and/or modify
12334 + * it under the terms of the GNU General Public License as published by
12335 + * the Free Software Foundation; either version 2 of the License, or
12336 + * (at your option) any later version.
12338 + * This program is distributed in the hope that it will be useful,
12339 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12340 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12341 + * GNU General Public License for more details.
12343 + * You should have received a copy of the GNU General Public License
12344 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12348 + * dynamically customizable operations for regular files
12353 +#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
12356 + * How large will these lists be?
12357 + * Usually just a few elements, 20-30 at most for each, I guess.
12359 +static struct hlist_bl_head dynop[AuDyLast];
12361 +static struct au_dykey *dy_gfind_get(struct hlist_bl_head *hbl,
12362 + const void *h_op)
12364 + struct au_dykey *key, *tmp;
12365 + struct hlist_bl_node *pos;
12368 + hlist_bl_lock(hbl);
12369 + hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12370 + if (tmp->dk_op.dy_hop == h_op) {
12371 + if (kref_get_unless_zero(&tmp->dk_kref))
12375 + hlist_bl_unlock(hbl);
12380 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
12382 + struct au_dykey **k, *found;
12383 + const void *h_op = key->dk_op.dy_hop;
12387 + k = br->br_dykey;
12388 + for (i = 0; i < AuBrDynOp; i++)
12390 + if (k[i]->dk_op.dy_hop == h_op) {
12397 + spin_lock(&br->br_dykey_lock);
12398 + for (; i < AuBrDynOp; i++)
12400 + if (k[i]->dk_op.dy_hop == h_op) {
12408 + spin_unlock(&br->br_dykey_lock);
12409 + BUG_ON(i == AuBrDynOp); /* expand the array */
12415 +/* kref_get() if @key is already added */
12416 +static struct au_dykey *dy_gadd(struct hlist_bl_head *hbl, struct au_dykey *key)
12418 + struct au_dykey *tmp, *found;
12419 + struct hlist_bl_node *pos;
12420 + const void *h_op = key->dk_op.dy_hop;
12423 + hlist_bl_lock(hbl);
12424 + hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12425 + if (tmp->dk_op.dy_hop == h_op) {
12426 + if (kref_get_unless_zero(&tmp->dk_kref))
12431 + hlist_bl_add_head(&key->dk_hnode, hbl);
12432 + hlist_bl_unlock(hbl);
12439 +static void dy_free_rcu(struct rcu_head *rcu)
12441 + struct au_dykey *key;
12443 + key = container_of(rcu, struct au_dykey, dk_rcu);
12448 +static void dy_free(struct kref *kref)
12450 + struct au_dykey *key;
12451 + struct hlist_bl_head *hbl;
12453 + key = container_of(kref, struct au_dykey, dk_kref);
12454 + hbl = dynop + key->dk_op.dy_type;
12455 + au_hbl_del(&key->dk_hnode, hbl);
12456 + call_rcu(&key->dk_rcu, dy_free_rcu);
12459 +void au_dy_put(struct au_dykey *key)
12461 + kref_put(&key->dk_kref, dy_free);
12464 +/* ---------------------------------------------------------------------- */
12466 +#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
12468 +#ifdef CONFIG_AUFS_DEBUG
12469 +#define DyDbgDeclare(cnt) unsigned int cnt = 0
12470 +#define DyDbgInc(cnt) do { cnt++; } while (0)
12472 +#define DyDbgDeclare(cnt) do {} while (0)
12473 +#define DyDbgInc(cnt) do {} while (0)
12476 +#define DySet(func, dst, src, h_op, h_sb) do { \
12478 + if (h_op->func) { \
12480 + dst.func = src.func; \
12482 + AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
12486 +#define DySetForce(func, dst, src) do { \
12487 + AuDebugOn(!src.func); \
12489 + dst.func = src.func; \
12492 +#define DySetAop(func) \
12493 + DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
12494 +#define DySetAopForce(func) \
12495 + DySetForce(func, dyaop->da_op, aufs_aop)
12497 +static void dy_aop(struct au_dykey *key, const void *h_op,
12498 + struct super_block *h_sb __maybe_unused)
12500 + struct au_dyaop *dyaop = (void *)key;
12501 + const struct address_space_operations *h_aop = h_op;
12502 + DyDbgDeclare(cnt);
12504 + AuDbg("%s\n", au_sbtype(h_sb));
12506 + DySetAop(writepage);
12507 + DySetAopForce(readpage); /* force */
12508 + DySetAop(writepages);
12509 + DySetAop(dirty_folio);
12510 + DySetAop(invalidate_folio);
12511 + DySetAop(readahead);
12512 + DySetAop(write_begin);
12513 + DySetAop(write_end);
12515 + DySetAop(releasepage);
12516 + DySetAop(freepage);
12517 + /* this one will be changed according to an aufs mount option */
12518 + DySetAop(direct_IO);
12519 + DySetAop(migratepage);
12520 + DySetAop(isolate_page);
12521 + DySetAop(putback_page);
12522 + DySetAop(launder_folio);
12523 + DySetAop(is_partially_uptodate);
12524 + DySetAop(is_dirty_writeback);
12525 + DySetAop(error_remove_page);
12526 + DySetAop(swap_activate);
12527 + DySetAop(swap_deactivate);
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 2022-05-23 09:44:37.729839166 +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_readpage(struct file *file __maybe_unused, struct page *page)
14828 + unlock_page(page);
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, unsigned flags,
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 int aufs_releasepage(struct page *page, gfp_t gfp)
14854 +{ AuUnsupport(); return 0; }
14855 +#if 0 /* called by memory compaction regardless file */
14856 +static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
14857 + struct page *page, enum migrate_mode mode)
14858 +{ AuUnsupport(); return 0; }
14860 +static bool aufs_isolate_page(struct page *page, isolate_mode_t mode)
14861 +{ AuUnsupport(); return true; }
14862 +static void aufs_putback_page(struct page *page)
14863 +{ AuUnsupport(); }
14864 +static int aufs_launder_folio(struct folio *folio)
14865 +{ AuUnsupport(); return 0; }
14866 +static bool aufs_is_partially_uptodate(struct folio *folio, size_t from,
14868 +{ AuUnsupport(); return true; }
14869 +static void aufs_is_dirty_writeback(struct page *page, bool *dirty,
14871 +{ AuUnsupport(); }
14872 +static int aufs_error_remove_page(struct address_space *mapping,
14873 + struct page *page)
14874 +{ AuUnsupport(); return 0; }
14875 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
14877 +{ AuUnsupport(); return 0; }
14878 +static void aufs_swap_deactivate(struct file *file)
14879 +{ AuUnsupport(); }
14880 +#endif /* CONFIG_AUFS_DEBUG */
14882 +const struct address_space_operations aufs_aop = {
14883 + .readpage = aufs_readpage,
14884 + .direct_IO = aufs_direct_IO,
14885 +#ifdef CONFIG_AUFS_DEBUG
14886 + .writepage = aufs_writepage,
14887 + /* no writepages, because of writepage */
14888 + .dirty_folio = aufs_dirty_folio,
14889 + /* no readpages, because of readpage */
14890 + .write_begin = aufs_write_begin,
14891 + .write_end = aufs_write_end,
14892 + /* no bmap, no block device */
14893 + .invalidate_folio = aufs_invalidate_folio,
14894 + .releasepage = aufs_releasepage,
14895 + /* is fallback_migrate_page ok? */
14896 + /* .migratepage = aufs_migratepage, */
14897 + .isolate_page = aufs_isolate_page,
14898 + .putback_page = aufs_putback_page,
14899 + .launder_folio = aufs_launder_folio,
14900 + .is_partially_uptodate = aufs_is_partially_uptodate,
14901 + .is_dirty_writeback = aufs_is_dirty_writeback,
14902 + .error_remove_page = aufs_error_remove_page,
14903 + .swap_activate = aufs_swap_activate,
14904 + .swap_deactivate = aufs_swap_deactivate
14905 +#endif /* CONFIG_AUFS_DEBUG */
14907 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
14908 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
14909 +++ linux/fs/aufs/file.h 2022-03-21 14:49:05.723299677 +0100
14911 +/* SPDX-License-Identifier: GPL-2.0 */
14913 + * Copyright (C) 2005-2021 Junjiro R. Okajima
14915 + * This program, aufs is free software; you can redistribute it and/or modify
14916 + * it under the terms of the GNU General Public License as published by
14917 + * the Free Software Foundation; either version 2 of the License, or
14918 + * (at your option) any later version.
14920 + * This program is distributed in the hope that it will be useful,
14921 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14922 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14923 + * GNU General Public License for more details.
14925 + * You should have received a copy of the GNU General Public License
14926 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14930 + * file operations
14933 +#ifndef __AUFS_FILE_H__
14934 +#define __AUFS_FILE_H__
14938 +#include <linux/file.h>
14939 +#include <linux/fs.h>
14940 +#include <linux/mm_types.h>
14941 +#include <linux/poll.h>
14942 +#include "rwsem.h"
14946 + struct file *hf_file;
14947 + struct au_branch *hf_br;
14952 + aufs_bindex_t fd_bbot;
14953 + aufs_bindex_t fd_nent;
14954 + struct au_vdir *fd_vdir_cache;
14955 + struct au_hfile fd_hfile[];
14958 +static inline int au_fidir_sz(int nent)
14960 + AuDebugOn(nent < 0);
14961 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
14965 + atomic_t fi_generation;
14967 + struct au_rwsem fi_rwsem;
14968 + aufs_bindex_t fi_btop;
14970 + /* do not union them */
14971 + struct { /* for non-dir */
14972 + struct au_hfile fi_htop;
14973 + atomic_t fi_mmapped;
14975 + struct au_fidir *fi_hdir; /* for dir only */
14977 + struct hlist_bl_node fi_hlist;
14978 + struct file *fi_file; /* very ugly */
14979 + struct rcu_head rcu;
14980 +} ____cacheline_aligned_in_smp;
14982 +/* ---------------------------------------------------------------------- */
14985 +extern const struct address_space_operations aufs_aop;
14986 +unsigned int au_file_roflags(unsigned int flags);
14987 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14988 + struct file *file, int force_wr);
14989 +struct au_do_open_args {
14991 + int (*open)(struct file *file, int flags,
14992 + struct file *h_file);
14993 + struct au_fidir *fidir;
14994 + struct file *h_file;
14996 +int au_do_open(struct file *file, struct au_do_open_args *args);
14997 +int au_reopen_nondir(struct file *file);
14999 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
15000 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
15001 + int wlock, unsigned int fi_lsc);
15002 +int au_do_flush(struct file *file, fl_owner_t id,
15003 + int (*flush)(struct file *file, fl_owner_t id));
15006 +#ifdef CONFIG_AUFS_POLL
15007 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt);
15010 +#ifdef CONFIG_AUFS_BR_HFSPLUS
15012 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
15014 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
15015 + struct file *h_file);
15017 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
15018 + aufs_bindex_t bindex, int force_wr)
15019 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
15020 + struct file *h_file);
15024 +extern const struct file_operations aufs_file_fop;
15025 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
15026 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
15027 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc);
15030 +void au_hfput(struct au_hfile *hf, int execed);
15031 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
15032 + struct file *h_file);
15034 +void au_update_figen(struct file *file);
15035 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
15036 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink);
15038 +void au_fi_init_once(void *_fi);
15039 +void au_finfo_fin(struct file *file);
15040 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
15043 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
15044 +#ifdef CONFIG_COMPAT
15045 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15046 + unsigned long arg);
15047 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15048 + unsigned long arg);
15051 +/* ---------------------------------------------------------------------- */
15053 +static inline struct au_finfo *au_fi(struct file *file)
15055 + return file->private_data;
15058 +/* ---------------------------------------------------------------------- */
15060 +#define fi_read_lock(f) au_rw_read_lock(&au_fi(f)->fi_rwsem)
15061 +#define fi_write_lock(f) au_rw_write_lock(&au_fi(f)->fi_rwsem)
15062 +#define fi_read_trylock(f) au_rw_read_trylock(&au_fi(f)->fi_rwsem)
15063 +#define fi_write_trylock(f) au_rw_write_trylock(&au_fi(f)->fi_rwsem)
15065 +#define fi_read_trylock_nested(f) \
15066 + au_rw_read_trylock_nested(&au_fi(f)->fi_rwsem)
15067 +#define fi_write_trylock_nested(f) \
15068 + au_rw_write_trylock_nested(&au_fi(f)->fi_rwsem)
15071 +#define fi_read_unlock(f) au_rw_read_unlock(&au_fi(f)->fi_rwsem)
15072 +#define fi_write_unlock(f) au_rw_write_unlock(&au_fi(f)->fi_rwsem)
15073 +#define fi_downgrade_lock(f) au_rw_dgrade_lock(&au_fi(f)->fi_rwsem)
15075 +/* lock subclass for finfo */
15081 +static inline void fi_read_lock_nested(struct file *f, unsigned int lsc)
15083 + au_rw_read_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15086 +static inline void fi_write_lock_nested(struct file *f, unsigned int lsc)
15088 + au_rw_write_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15092 + * fi_read_lock_1, fi_write_lock_1,
15093 + * fi_read_lock_2, fi_write_lock_2
15095 +#define AuReadLockFunc(name) \
15096 +static inline void fi_read_lock_##name(struct file *f) \
15097 +{ fi_read_lock_nested(f, AuLsc_FI_##name); }
15099 +#define AuWriteLockFunc(name) \
15100 +static inline void fi_write_lock_##name(struct file *f) \
15101 +{ fi_write_lock_nested(f, AuLsc_FI_##name); }
15103 +#define AuRWLockFuncs(name) \
15104 + AuReadLockFunc(name) \
15105 + AuWriteLockFunc(name)
15110 +#undef AuReadLockFunc
15111 +#undef AuWriteLockFunc
15112 +#undef AuRWLockFuncs
15114 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
15115 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
15116 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
15118 +/* ---------------------------------------------------------------------- */
15120 +/* todo: hard/soft set? */
15121 +static inline aufs_bindex_t au_fbtop(struct file *file)
15123 + FiMustAnyLock(file);
15124 + return au_fi(file)->fi_btop;
15127 +static inline aufs_bindex_t au_fbbot_dir(struct file *file)
15129 + FiMustAnyLock(file);
15130 + AuDebugOn(!au_fi(file)->fi_hdir);
15131 + return au_fi(file)->fi_hdir->fd_bbot;
15134 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
15136 + FiMustAnyLock(file);
15137 + AuDebugOn(!au_fi(file)->fi_hdir);
15138 + return au_fi(file)->fi_hdir->fd_vdir_cache;
15141 +static inline void au_set_fbtop(struct file *file, aufs_bindex_t bindex)
15143 + FiMustWriteLock(file);
15144 + au_fi(file)->fi_btop = bindex;
15147 +static inline void au_set_fbbot_dir(struct file *file, aufs_bindex_t bindex)
15149 + FiMustWriteLock(file);
15150 + AuDebugOn(!au_fi(file)->fi_hdir);
15151 + au_fi(file)->fi_hdir->fd_bbot = bindex;
15154 +static inline void au_set_fvdir_cache(struct file *file,
15155 + struct au_vdir *vdir_cache)
15157 + FiMustWriteLock(file);
15158 + AuDebugOn(!au_fi(file)->fi_hdir);
15159 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
15162 +static inline struct file *au_hf_top(struct file *file)
15164 + FiMustAnyLock(file);
15165 + AuDebugOn(au_fi(file)->fi_hdir);
15166 + return au_fi(file)->fi_htop.hf_file;
15169 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
15171 + FiMustAnyLock(file);
15172 + AuDebugOn(!au_fi(file)->fi_hdir);
15173 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
15176 +/* todo: memory barrier? */
15177 +static inline unsigned int au_figen(struct file *f)
15179 + return atomic_read(&au_fi(f)->fi_generation);
15182 +static inline void au_set_mmapped(struct file *f)
15184 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
15186 + pr_warn("fi_mmapped wrapped around\n");
15187 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
15191 +static inline void au_unset_mmapped(struct file *f)
15193 + atomic_dec(&au_fi(f)->fi_mmapped);
15196 +static inline int au_test_mmapped(struct file *f)
15198 + return atomic_read(&au_fi(f)->fi_mmapped);
15201 +/* customize vma->vm_file */
15203 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
15204 + struct file *file)
15208 + f = vma->vm_file;
15210 + vma->vm_file = file;
15215 +#define AuDbgVmRegion(file, vma) do {} while (0)
15217 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15218 + struct file *file)
15220 + au_do_vm_file_reset(vma, file);
15223 +#define AuDbgVmRegion(file, vma) \
15224 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
15226 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15227 + struct file *file)
15231 + au_do_vm_file_reset(vma, file);
15232 + f = vma->vm_region->vm_file;
15234 + vma->vm_region->vm_file = file;
15237 +#endif /* CONFIG_MMU */
15239 +/* handle vma->vm_prfile */
15240 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
15241 + struct file *file)
15244 + vma->vm_prfile = file;
15245 +#ifndef CONFIG_MMU
15247 + vma->vm_region->vm_prfile = file;
15251 +#endif /* __KERNEL__ */
15252 +#endif /* __AUFS_FILE_H__ */
15253 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
15254 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
15255 +++ linux/fs/aufs/finfo.c 2022-03-21 14:49:05.723299677 +0100
15257 +// SPDX-License-Identifier: GPL-2.0
15259 + * Copyright (C) 2005-2021 Junjiro R. Okajima
15261 + * This program, aufs is free software; you can redistribute it and/or modify
15262 + * it under the terms of the GNU General Public License as published by
15263 + * the Free Software Foundation; either version 2 of the License, or
15264 + * (at your option) any later version.
15266 + * This program is distributed in the hope that it will be useful,
15267 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15268 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15269 + * GNU General Public License for more details.
15271 + * You should have received a copy of the GNU General Public License
15272 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15276 + * file private data
15281 +void au_hfput(struct au_hfile *hf, int execed)
15284 + allow_write_access(hf->hf_file);
15285 + fput(hf->hf_file);
15286 + hf->hf_file = NULL;
15287 + au_lcnt_dec(&hf->hf_br->br_nfiles);
15288 + hf->hf_br = NULL;
15291 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
15293 + struct au_finfo *finfo = au_fi(file);
15294 + struct au_hfile *hf;
15295 + struct au_fidir *fidir;
15297 + fidir = finfo->fi_hdir;
15299 + AuDebugOn(finfo->fi_btop != bindex);
15300 + hf = &finfo->fi_htop;
15302 + hf = fidir->fd_hfile + bindex;
15304 + if (hf && hf->hf_file)
15305 + au_hfput(hf, vfsub_file_execed(file));
15307 + FiMustWriteLock(file);
15308 + AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
15309 + hf->hf_file = val;
15310 + hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
15314 +void au_update_figen(struct file *file)
15316 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
15317 + /* smp_mb(); */ /* atomic_set */
15320 +/* ---------------------------------------------------------------------- */
15322 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
15324 + struct au_fidir *fidir;
15327 + nbr = au_sbbot(sb) + 1;
15329 + nbr = 2; /* initial allocate for 2 branches */
15330 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
15332 + fidir->fd_bbot = -1;
15333 + fidir->fd_nent = nbr;
15339 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink)
15342 + struct au_fidir *fidir, *p;
15344 + AuRwMustWriteLock(&finfo->fi_rwsem);
15345 + fidir = finfo->fi_hdir;
15346 + AuDebugOn(!fidir);
15349 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
15350 + GFP_NOFS, may_shrink);
15352 + p->fd_nent = nbr;
15353 + finfo->fi_hdir = p;
15360 +/* ---------------------------------------------------------------------- */
15362 +void au_finfo_fin(struct file *file)
15364 + struct au_finfo *finfo;
15366 + au_lcnt_dec(&au_sbi(file->f_path.dentry->d_sb)->si_nfiles);
15368 + finfo = au_fi(file);
15369 + AuDebugOn(finfo->fi_hdir);
15370 + AuRwDestroy(&finfo->fi_rwsem);
15371 + au_cache_free_finfo(finfo);
15374 +void au_fi_init_once(void *_finfo)
15376 + struct au_finfo *finfo = _finfo;
15378 + au_rw_init(&finfo->fi_rwsem);
15381 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
15384 + struct au_finfo *finfo;
15385 + struct dentry *dentry;
15388 + dentry = file->f_path.dentry;
15389 + finfo = au_cache_alloc_finfo();
15390 + if (unlikely(!finfo))
15394 + au_lcnt_inc(&au_sbi(dentry->d_sb)->si_nfiles);
15395 + au_rw_write_lock(&finfo->fi_rwsem);
15396 + finfo->fi_btop = -1;
15397 + finfo->fi_hdir = fidir;
15398 + atomic_set(&finfo->fi_generation, au_digen(dentry));
15399 + /* smp_mb(); */ /* atomic_set */
15401 + file->private_data = finfo;
15406 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
15407 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
15408 +++ linux/fs/aufs/f_op.c 2022-03-21 14:49:05.723299677 +0100
15410 +// SPDX-License-Identifier: GPL-2.0
15412 + * Copyright (C) 2005-2021 Junjiro R. Okajima
15414 + * This program, aufs is free software; you can redistribute it and/or modify
15415 + * it under the terms of the GNU General Public License as published by
15416 + * the Free Software Foundation; either version 2 of the License, or
15417 + * (at your option) any later version.
15419 + * This program is distributed in the hope that it will be useful,
15420 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15421 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15422 + * GNU General Public License for more details.
15424 + * You should have received a copy of the GNU General Public License
15425 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15429 + * file and vm operations
15432 +#include <linux/aio.h>
15433 +#include <linux/fs_stack.h>
15434 +#include <linux/mman.h>
15435 +#include <linux/security.h>
15438 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
15441 + aufs_bindex_t bindex;
15442 + struct dentry *dentry, *h_dentry;
15443 + struct au_finfo *finfo;
15444 + struct inode *h_inode;
15446 + FiMustWriteLock(file);
15449 + dentry = file->f_path.dentry;
15450 + AuDebugOn(IS_ERR_OR_NULL(dentry));
15451 + finfo = au_fi(file);
15452 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
15453 + atomic_set(&finfo->fi_mmapped, 0);
15454 + bindex = au_dbtop(dentry);
15456 + h_dentry = au_h_dptr(dentry, bindex);
15457 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15458 + if (unlikely(err))
15460 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
15461 + if (IS_ERR(h_file)) {
15462 + err = PTR_ERR(h_file);
15466 + h_dentry = h_file->f_path.dentry;
15467 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15468 + if (unlikely(err))
15470 + /* br ref is already inc-ed */
15473 + if ((flags & __O_TMPFILE)
15474 + && !(flags & O_EXCL)) {
15475 + h_inode = file_inode(h_file);
15476 + spin_lock(&h_inode->i_lock);
15477 + h_inode->i_state |= I_LINKABLE;
15478 + spin_unlock(&h_inode->i_lock);
15480 + au_set_fbtop(file, bindex);
15481 + au_set_h_fptr(file, bindex, h_file);
15482 + au_update_figen(file);
15483 + /* todo: necessary? */
15484 + /* file->f_ra = h_file->f_ra; */
15490 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
15491 + struct file *file)
15494 + struct super_block *sb;
15495 + struct au_do_open_args args = {
15496 + .open = au_do_open_nondir
15499 + AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
15500 + file, vfsub_file_flags(file), file->f_mode);
15502 + sb = file->f_path.dentry->d_sb;
15503 + si_read_lock(sb, AuLock_FLUSH);
15504 + err = au_do_open(file, &args);
15505 + si_read_unlock(sb);
15509 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
15511 + struct au_finfo *finfo;
15512 + aufs_bindex_t bindex;
15514 + finfo = au_fi(file);
15515 + au_hbl_del(&finfo->fi_hlist,
15516 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
15517 + bindex = finfo->fi_btop;
15519 + au_set_h_fptr(file, bindex, NULL);
15521 + au_finfo_fin(file);
15525 +/* ---------------------------------------------------------------------- */
15527 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
15530 + struct file *h_file;
15533 + h_file = au_hf_top(file);
15535 + err = vfsub_flush(h_file, id);
15539 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
15541 + return au_do_flush(file, id, au_do_flush_nondir);
15544 +/* ---------------------------------------------------------------------- */
15546 + * read and write functions acquire [fdi]_rwsem once, but release before
15547 + * mmap_sem. This is because to stop a race condition between mmap(2).
15548 + * Releasing these aufs-rwsem should be safe, no branch-management (by keeping
15549 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
15550 + * read functions after [fdi]_rwsem are released, but it should be harmless.
15553 +/* Callers should call au_read_post() or fput() in the end */
15554 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc)
15556 + struct file *h_file;
15559 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0, lsc);
15561 + di_read_unlock(file->f_path.dentry, AuLock_IR);
15562 + h_file = au_hf_top(file);
15563 + get_file(h_file);
15565 + fi_read_unlock(file);
15567 + h_file = ERR_PTR(err);
15572 +static void au_read_post(struct inode *inode, struct file *h_file)
15574 + /* update without lock, I don't think it a problem */
15575 + fsstack_copy_attr_atime(inode, file_inode(h_file));
15579 +struct au_write_pre {
15581 + unsigned int lsc;
15585 + aufs_bindex_t btop;
15589 + * return with iinfo is write-locked
15590 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
15593 +static struct file *au_write_pre(struct file *file, int do_ready,
15594 + struct au_write_pre *wpre)
15596 + struct file *h_file;
15597 + struct dentry *dentry;
15599 + unsigned int lsc;
15600 + struct au_pin pin;
15605 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1, lsc);
15606 + h_file = ERR_PTR(err);
15607 + if (unlikely(err))
15610 + dentry = file->f_path.dentry;
15612 + err = au_ready_to_write(file, -1, &pin);
15613 + if (unlikely(err)) {
15614 + h_file = ERR_PTR(err);
15615 + di_write_unlock(dentry);
15620 + di_downgrade_lock(dentry, /*flags*/0);
15622 + wpre->btop = au_fbtop(file);
15623 + h_file = au_hf_top(file);
15624 + get_file(h_file);
15626 + wpre->blks = file_inode(h_file)->i_blocks;
15629 + di_read_unlock(dentry, /*flags*/0);
15632 + fi_write_unlock(file);
15637 +static void au_write_post(struct inode *inode, struct file *h_file,
15638 + struct au_write_pre *wpre, ssize_t written)
15640 + struct inode *h_inode;
15642 + au_cpup_attr_timesizes(inode);
15643 + AuDebugOn(au_ibtop(inode) != wpre->btop);
15644 + h_inode = file_inode(h_file);
15645 + inode->i_mode = h_inode->i_mode;
15646 + ii_write_unlock(inode);
15647 + /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
15649 + au_fhsm_wrote(inode->i_sb, wpre->btop,
15650 + /*force*/h_inode->i_blocks > wpre->blks);
15655 + * todo: very ugly
15656 + * it locks both of i_mutex and si_rwsem for read in safe.
15657 + * if the plink maintenance mode continues forever (that is the problem),
15658 + * may loop forever.
15660 +static void au_mtx_and_read_lock(struct inode *inode)
15663 + struct super_block *sb = inode->i_sb;
15666 + inode_lock(inode);
15667 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
15670 + inode_unlock(inode);
15671 + si_read_lock(sb, AuLock_NOPLMW);
15672 + si_read_unlock(sb);
15676 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
15677 + struct iov_iter *iov_iter)
15680 + struct file *file;
15681 + ssize_t (*iter)(struct kiocb *, struct iov_iter *);
15683 + err = security_file_permission(h_file, rw);
15684 + if (unlikely(err))
15687 + err = -ENOSYS; /* the branch doesn't have its ->(read|write)_iter() */
15689 + if (rw == MAY_READ)
15690 + iter = h_file->f_op->read_iter;
15691 + else if (rw == MAY_WRITE)
15692 + iter = h_file->f_op->write_iter;
15694 + file = kio->ki_filp;
15695 + kio->ki_filp = h_file;
15698 + err = iter(kio, iov_iter);
15701 + /* currently there is no such fs */
15703 + kio->ki_filp = file;
15709 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15712 + struct file *file, *h_file;
15713 + struct inode *inode;
15714 + struct super_block *sb;
15716 + file = kio->ki_filp;
15717 + inode = file_inode(file);
15718 + sb = inode->i_sb;
15719 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15721 + h_file = au_read_pre(file, /*keep_fi*/1, /*lsc*/0);
15722 + err = PTR_ERR(h_file);
15723 + if (IS_ERR(h_file))
15726 + if (au_test_loopback_kthread()) {
15727 + au_warn_loopback(h_file->f_path.dentry->d_sb);
15728 + if (file->f_mapping != h_file->f_mapping) {
15729 + file->f_mapping = h_file->f_mapping;
15730 + smp_mb(); /* unnecessary? */
15733 + fi_read_unlock(file);
15735 + err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
15736 + /* todo: necessary? */
15737 + /* file->f_ra = h_file->f_ra; */
15738 + au_read_post(inode, h_file);
15741 + si_read_unlock(sb);
15745 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15748 + struct au_write_pre wpre;
15749 + struct inode *inode;
15750 + struct file *file, *h_file;
15752 + file = kio->ki_filp;
15753 + inode = file_inode(file);
15754 + au_mtx_and_read_lock(inode);
15757 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15758 + err = PTR_ERR(h_file);
15759 + if (IS_ERR(h_file))
15762 + err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
15763 + au_write_post(inode, h_file, &wpre, err);
15766 + si_read_unlock(inode->i_sb);
15767 + inode_unlock(inode);
15772 + * We may be able to remove aufs_splice_{read,write}() since almost all FSes
15773 + * don't have their own .splice_{read,write} implimentations, and they use
15774 + * generic_file_splice_read() and iter_file_splice_write() who can act like the
15775 + * simple converters to f_op->iter_read() and ->iter_write().
15776 + * But we keep our own implementations because some non-mainlined FSes may have
15777 + * their own .splice_{read,write} implimentations and aufs doesn't want to take
15778 + * away an opportunity to co-work with aufs from them.
15780 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
15781 + struct pipe_inode_info *pipe, size_t len,
15782 + unsigned int flags)
15785 + struct file *h_file;
15786 + struct inode *inode;
15787 + struct super_block *sb;
15789 + inode = file_inode(file);
15790 + sb = inode->i_sb;
15791 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15793 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15794 + err = PTR_ERR(h_file);
15795 + if (IS_ERR(h_file))
15798 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
15799 + /* todo: necessary? */
15800 + /* file->f_ra = h_file->f_ra; */
15801 + au_read_post(inode, h_file);
15804 + si_read_unlock(sb);
15809 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
15810 + size_t len, unsigned int flags)
15813 + struct au_write_pre wpre;
15814 + struct inode *inode;
15815 + struct file *h_file;
15817 + inode = file_inode(file);
15818 + au_mtx_and_read_lock(inode);
15821 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15822 + err = PTR_ERR(h_file);
15823 + if (IS_ERR(h_file))
15826 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
15827 + au_write_post(inode, h_file, &wpre, err);
15830 + si_read_unlock(inode->i_sb);
15831 + inode_unlock(inode);
15835 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
15839 + struct au_write_pre wpre;
15840 + struct inode *inode;
15841 + struct file *h_file;
15843 + inode = file_inode(file);
15844 + au_mtx_and_read_lock(inode);
15847 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15848 + err = PTR_ERR(h_file);
15849 + if (IS_ERR(h_file))
15853 + err = vfs_fallocate(h_file, mode, offset, len);
15855 + au_write_post(inode, h_file, &wpre, /*written*/1);
15858 + si_read_unlock(inode->i_sb);
15859 + inode_unlock(inode);
15863 +static ssize_t aufs_copy_file_range(struct file *src, loff_t src_pos,
15864 + struct file *dst, loff_t dst_pos,
15865 + size_t len, unsigned int flags)
15868 + struct au_write_pre wpre;
15869 + enum { SRC, DST };
15871 + struct inode *inode;
15872 + struct file *h_file;
15873 + struct super_block *h_sb;
15875 +#define a_src a[SRC]
15876 +#define a_dst a[DST]
15879 + a_src.inode = file_inode(src);
15880 + if (unlikely(!S_ISREG(a_src.inode->i_mode)))
15882 + a_dst.inode = file_inode(dst);
15883 + if (unlikely(!S_ISREG(a_dst.inode->i_mode)))
15886 + au_mtx_and_read_lock(a_dst.inode);
15888 + * in order to match the order in di_write_lock2_{child,parent}(),
15889 + * use f_path.dentry for this comparison.
15891 + if (src->f_path.dentry < dst->f_path.dentry) {
15892 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_1);
15893 + err = PTR_ERR(a_src.h_file);
15894 + if (IS_ERR(a_src.h_file))
15897 + wpre.lsc = AuLsc_FI_2;
15898 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15899 + err = PTR_ERR(a_dst.h_file);
15900 + if (IS_ERR(a_dst.h_file)) {
15901 + au_read_post(a_src.inode, a_src.h_file);
15905 + wpre.lsc = AuLsc_FI_1;
15906 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15907 + err = PTR_ERR(a_dst.h_file);
15908 + if (IS_ERR(a_dst.h_file))
15911 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_2);
15912 + err = PTR_ERR(a_src.h_file);
15913 + if (IS_ERR(a_src.h_file)) {
15914 + au_write_post(a_dst.inode, a_dst.h_file, &wpre,
15921 + a_src.h_sb = file_inode(a_src.h_file)->i_sb;
15922 + a_dst.h_sb = file_inode(a_dst.h_file)->i_sb;
15923 + if (unlikely(a_src.h_sb != a_dst.h_sb)) {
15929 + err = vfsub_copy_file_range(a_src.h_file, src_pos, a_dst.h_file,
15930 + dst_pos, len, flags);
15933 + au_write_post(a_dst.inode, a_dst.h_file, &wpre, err);
15934 + fi_read_unlock(src);
15935 + au_read_post(a_src.inode, a_src.h_file);
15937 + si_read_unlock(a_dst.inode->i_sb);
15938 + inode_unlock(a_dst.inode);
15945 +/* ---------------------------------------------------------------------- */
15948 + * The locking order around current->mmap_sem.
15949 + * - in most and regular cases
15950 + * file I/O syscall -- aufs_read() or something
15951 + * -- si_rwsem for read -- mmap_sem
15952 + * (Note that [fdi]i_rwsem are released before mmap_sem).
15954 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
15955 + * This AB-BA order is definitely bad, but is not a problem since "si_rwsem for
15956 + * read" allows multiple processes to acquire it and [fdi]i_rwsem are not held
15957 + * in file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
15958 + * It means that when aufs acquires si_rwsem for write, the process should never
15959 + * acquire mmap_sem.
15961 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
15962 + * problem either since any directory is not able to be mmap-ed.
15963 + * The similar scenario is applied to aufs_readlink() too.
15966 +#if 0 /* stop calling security_file_mmap() */
15967 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
15968 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
15970 +static unsigned long au_arch_prot_conv(unsigned long flags)
15972 + /* currently ppc64 only */
15973 +#ifdef CONFIG_PPC64
15974 + /* cf. linux/arch/powerpc/include/asm/mman.h */
15975 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
15976 + return AuConv_VM_PROT(flags, SAO);
15978 + AuDebugOn(arch_calc_vm_prot_bits(-1));
15983 +static unsigned long au_prot_conv(unsigned long flags)
15985 + return AuConv_VM_PROT(flags, READ)
15986 + | AuConv_VM_PROT(flags, WRITE)
15987 + | AuConv_VM_PROT(flags, EXEC)
15988 + | au_arch_prot_conv(flags);
15991 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
15992 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
15994 +static unsigned long au_flag_conv(unsigned long flags)
15996 + return AuConv_VM_MAP(flags, GROWSDOWN)
15997 + | AuConv_VM_MAP(flags, DENYWRITE)
15998 + | AuConv_VM_MAP(flags, LOCKED);
16002 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
16005 + const unsigned char wlock
16006 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
16007 + struct super_block *sb;
16008 + struct file *h_file;
16009 + struct inode *inode;
16011 + AuDbgVmRegion(file, vma);
16013 + inode = file_inode(file);
16014 + sb = inode->i_sb;
16016 + si_read_lock(sb, AuLock_NOPLMW);
16018 + h_file = au_write_pre(file, wlock, /*wpre*/NULL);
16020 + err = PTR_ERR(h_file);
16021 + if (IS_ERR(h_file))
16025 + au_set_mmapped(file);
16026 + au_vm_file_reset(vma, h_file);
16028 + * we cannot call security_mmap_file() here since it may acquire
16029 + * mmap_sem or i_mutex.
16031 + * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
16032 + * au_flag_conv(vma->vm_flags));
16035 + err = call_mmap(h_file, vma);
16037 + au_vm_prfile_set(vma, file);
16038 + fsstack_copy_attr_atime(inode, file_inode(h_file));
16039 + goto out_fput; /* success */
16041 + au_unset_mmapped(file);
16042 + au_vm_file_reset(vma, file);
16046 + ii_write_unlock(inode);
16051 + si_read_unlock(sb);
16057 +/* ---------------------------------------------------------------------- */
16059 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
16063 + struct au_write_pre wpre;
16064 + struct inode *inode;
16065 + struct file *h_file;
16067 + err = 0; /* -EBADF; */ /* posix? */
16068 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
16071 + inode = file_inode(file);
16072 + au_mtx_and_read_lock(inode);
16075 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
16076 + err = PTR_ERR(h_file);
16077 + if (IS_ERR(h_file))
16080 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
16081 + au_write_post(inode, h_file, &wpre, /*written*/0);
16084 + si_read_unlock(inode->i_sb);
16085 + inode_unlock(inode);
16090 +static int aufs_fasync(int fd, struct file *file, int flag)
16093 + struct file *h_file;
16094 + struct super_block *sb;
16096 + sb = file->f_path.dentry->d_sb;
16097 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16099 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16100 + err = PTR_ERR(h_file);
16101 + if (IS_ERR(h_file))
16104 + if (h_file->f_op->fasync)
16105 + err = h_file->f_op->fasync(fd, h_file, flag);
16106 + fput(h_file); /* instead of au_read_post() */
16109 + si_read_unlock(sb);
16113 +static int aufs_setfl(struct file *file, unsigned long arg)
16116 + struct file *h_file;
16117 + struct super_block *sb;
16119 + sb = file->f_path.dentry->d_sb;
16120 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16122 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16123 + err = PTR_ERR(h_file);
16124 + if (IS_ERR(h_file))
16127 + /* stop calling h_file->fasync */
16128 + arg |= vfsub_file_flags(file) & FASYNC;
16129 + err = setfl(/*unused fd*/-1, h_file, arg);
16130 + fput(h_file); /* instead of au_read_post() */
16133 + si_read_unlock(sb);
16137 +/* ---------------------------------------------------------------------- */
16139 +/* no one supports this operation, currently */
16140 +#if 0 /* reserved for future use */
16141 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
16142 + size_t len, loff_t *pos, int more)
16147 +/* ---------------------------------------------------------------------- */
16149 +const struct file_operations aufs_file_fop = {
16150 + .owner = THIS_MODULE,
16152 + .llseek = default_llseek,
16154 + .read_iter = aufs_read_iter,
16155 + .write_iter = aufs_write_iter,
16157 +#ifdef CONFIG_AUFS_POLL
16158 + .poll = aufs_poll,
16160 + .unlocked_ioctl = aufs_ioctl_nondir,
16161 +#ifdef CONFIG_COMPAT
16162 + .compat_ioctl = aufs_compat_ioctl_nondir,
16164 + .mmap = aufs_mmap,
16165 + .open = aufs_open_nondir,
16166 + .flush = aufs_flush_nondir,
16167 + .release = aufs_release_nondir,
16168 + .fsync = aufs_fsync_nondir,
16169 + .fasync = aufs_fasync,
16170 + /* .sendpage = aufs_sendpage, */
16171 + .setfl = aufs_setfl,
16172 + .splice_write = aufs_splice_write,
16173 + .splice_read = aufs_splice_read,
16174 +#if 0 /* reserved for future use */
16175 + .aio_splice_write = aufs_aio_splice_write,
16176 + .aio_splice_read = aufs_aio_splice_read,
16178 + .fallocate = aufs_fallocate,
16179 + .copy_file_range = aufs_copy_file_range
16181 diff -urN /usr/share/empty/fs/aufs/fsctx.c linux/fs/aufs/fsctx.c
16182 --- /usr/share/empty/fs/aufs/fsctx.c 1970-01-01 01:00:00.000000000 +0100
16183 +++ linux/fs/aufs/fsctx.c 2022-03-21 14:49:05.723299677 +0100
16185 +// SPDX-License-Identifier: GPL-2.0
16187 + * Copyright (C) 2022 Junjiro R. Okajima
16189 + * This program, aufs is free software; you can redistribute it and/or modify
16190 + * it under the terms of the GNU General Public License as published by
16191 + * the Free Software Foundation; either version 2 of the License, or
16192 + * (at your option) any later version.
16194 + * This program is distributed in the hope that it will be useful,
16195 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16196 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16197 + * GNU General Public License for more details.
16199 + * You should have received a copy of the GNU General Public License
16200 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16204 + * fs context, aka new mount api
16207 +#include <linux/fs_context.h>
16210 +struct au_fsctx_opts {
16211 + aufs_bindex_t bindex;
16212 + unsigned char skipped;
16213 + struct au_opt *opt, *opt_tail;
16214 + struct super_block *sb;
16215 + struct au_sbinfo *sbinfo;
16216 + struct au_opts opts;
16219 +/* stop extra interpretation of errno in mount(8), and strange error messages */
16220 +static int cvt_err(int err)
16234 +static int au_fsctx_reconfigure(struct fs_context *fc)
16237 + unsigned int mntflags;
16238 + struct dentry *root;
16239 + struct super_block *sb;
16240 + struct inode *inode;
16241 + struct au_fsctx_opts *a = fc->fs_private;
16243 + AuDbg("fc %p\n", fc);
16247 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16249 + di_write_lock_child(root);
16250 + err = au_opts_verify(sb, fc->sb_flags, /*pending*/0);
16251 + aufs_write_unlock(root);
16254 + inode = d_inode(root);
16255 + inode_lock(inode);
16256 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16257 + if (unlikely(err))
16259 + di_write_lock_child(root);
16261 + /* au_opts_remount() may return an error */
16262 + err = au_opts_remount(sb, &a->opts);
16264 + if (au_ftest_opts(a->opts.flags, REFRESH))
16265 + au_remount_refresh(sb, au_ftest_opts(a->opts.flags,
16268 + if (au_ftest_opts(a->opts.flags, REFRESH_DYAOP)) {
16269 + mntflags = au_mntflags(sb);
16270 + do_dx = !!au_opt_test(mntflags, DIO);
16271 + au_dy_arefresh(do_dx);
16274 + au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
16275 + aufs_write_unlock(root);
16278 + inode_unlock(inode);
16279 + err = cvt_err(err);
16285 +/* ---------------------------------------------------------------------- */
16287 +static int au_fsctx_fill_super(struct super_block *sb, struct fs_context *fc)
16290 + struct au_fsctx_opts *a = fc->fs_private;
16291 + struct au_sbinfo *sbinfo = a->sbinfo;
16292 + struct dentry *root;
16293 + struct inode *inode;
16295 + sbinfo->si_sb = sb;
16296 + sb->s_fs_info = sbinfo;
16297 + kobject_get(&sbinfo->si_kobj);
16299 + __si_write_lock(sb);
16301 + au_sbilist_add(sb);
16303 + /* all timestamps always follow the ones on the branch */
16304 + sb->s_flags |= SB_NOATIME | SB_NODIRATIME;
16305 + sb->s_flags |= SB_I_VERSION; /* do we really need this? */
16306 + sb->s_op = &aufs_sop;
16307 + sb->s_d_op = &aufs_dop;
16308 + sb->s_magic = AUFS_SUPER_MAGIC;
16309 + sb->s_maxbytes = 0;
16310 + sb->s_stack_depth = 1;
16311 + au_export_init(sb);
16312 + au_xattr_init(sb);
16314 + err = au_alloc_root(sb);
16315 + if (unlikely(err)) {
16316 + si_write_unlock(sb);
16319 + root = sb->s_root;
16320 + inode = d_inode(root);
16321 + ii_write_lock_parent(inode);
16322 + aufs_write_unlock(root);
16324 + /* lock vfs_inode first, then aufs. */
16325 + inode_lock(inode);
16326 + aufs_write_lock(root);
16327 + err = au_opts_mount(sb, &a->opts);
16329 + if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
16330 + sb->s_d_op = &aufs_dop_noreval;
16331 + /* infofc(fc, "%ps", sb->s_d_op); */
16332 + pr_info("%ps\n", sb->s_d_op);
16333 + au_refresh_dop(root, /*force_reval*/0);
16334 + sbinfo->si_iop_array = aufs_iop_nogetattr;
16335 + au_refresh_iop(inode, /*force_getattr*/0);
16337 + aufs_write_unlock(root);
16338 + inode_unlock(inode);
16340 + goto out; /* success */
16343 + sb->s_root = NULL;
16346 + if (unlikely(err))
16347 + kobject_put(&sbinfo->si_kobj);
16349 + err = cvt_err(err);
16354 +static int au_fsctx_get_tree(struct fs_context *fc)
16358 + AuDbg("fc %p\n", fc);
16359 + err = get_tree_nodev(fc, au_fsctx_fill_super);
16365 +/* ---------------------------------------------------------------------- */
16367 +static void au_fsctx_dump(struct au_opts *opts)
16369 +#ifdef CONFIG_AUFS_DEBUG
16370 + /* reduce stack space */
16372 + struct au_opt_add *add;
16373 + struct au_opt_del *del;
16374 + struct au_opt_mod *mod;
16375 + struct au_opt_xino *xino;
16376 + struct au_opt_xino_itrunc *xino_itrunc;
16377 + struct au_opt_wbr_create *create;
16379 + struct au_opt *opt;
16382 + while (opt->type != Opt_tail) {
16383 + switch (opt->type) {
16385 + u.add = &opt->add;
16386 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
16387 + u.add->bindex, u.add->pathname, u.add->perm,
16388 + u.add->path.dentry);
16393 + u.del = &opt->del;
16394 + AuDbg("del {%s, %p}\n",
16395 + u.del->pathname, u.del->h_path.dentry);
16400 + u.mod = &opt->mod;
16401 + AuDbg("mod {%s, 0x%x, %p}\n",
16402 + u.mod->path, u.mod->perm, u.mod->h_root);
16405 + u.add = &opt->add;
16406 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
16407 + u.add->bindex, u.add->pathname, u.add->perm,
16408 + u.add->path.dentry);
16410 + case Opt_prepend:
16411 + u.add = &opt->add;
16412 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
16413 + u.add->bindex, u.add->pathname, u.add->perm,
16414 + u.add->path.dentry);
16418 + AuDbg("dirwh %d\n", opt->dirwh);
16420 + case Opt_rdcache:
16421 + AuDbg("rdcache %d\n", opt->rdcache);
16424 + AuDbg("rdblk %d\n", opt->rdblk);
16427 + AuDbg("rdhash %u\n", opt->rdhash);
16431 + u.xino = &opt->xino;
16432 + AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
16435 +#define au_fsctx_TF(name) \
16436 + case Opt_##name: \
16440 + AuLabel(no##name); \
16443 + /* simple true/false flag */
16444 + au_fsctx_TF(trunc_xino);
16445 + au_fsctx_TF(trunc_xib);
16446 + au_fsctx_TF(dirperm1);
16447 + au_fsctx_TF(plink);
16448 + au_fsctx_TF(shwh);
16449 + au_fsctx_TF(dio);
16450 + au_fsctx_TF(warn_perm);
16451 + au_fsctx_TF(verbose);
16452 + au_fsctx_TF(sum);
16453 + au_fsctx_TF(dirren);
16454 + au_fsctx_TF(acl);
16455 +#undef au_fsctx_TF
16457 + case Opt_trunc_xino_path:
16459 + case Opt_itrunc_xino:
16460 + u.xino_itrunc = &opt->xino_itrunc;
16461 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
16467 + case Opt_list_plink:
16468 + AuLabel(list_plink);
16471 + AuDbg("udba %d, %s\n",
16472 + opt->udba, au_optstr_udba(opt->udba));
16474 + case Opt_diropq_a:
16475 + AuLabel(diropq_a);
16477 + case Opt_diropq_w:
16478 + AuLabel(diropq_w);
16483 + case Opt_wbr_create:
16484 + u.create = &opt->wbr_create;
16485 + AuDbg("create %d, %s\n", u.create->wbr_create,
16486 + au_optstr_wbr_create(u.create->wbr_create));
16487 + switch (u.create->wbr_create) {
16488 + case AuWbrCreate_MFSV:
16490 + case AuWbrCreate_PMFSV:
16491 + AuDbg("%d sec\n", u.create->mfs_second);
16493 + case AuWbrCreate_MFSRR:
16495 + case AuWbrCreate_TDMFS:
16496 + AuDbg("%llu watermark\n",
16497 + u.create->mfsrr_watermark);
16499 + case AuWbrCreate_MFSRRV:
16501 + case AuWbrCreate_TDMFSV:
16503 + case AuWbrCreate_PMFSRRV:
16504 + AuDbg("%llu watermark, %d sec\n",
16505 + u.create->mfsrr_watermark,
16506 + u.create->mfs_second);
16510 + case Opt_wbr_copyup:
16511 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
16512 + au_optstr_wbr_copyup(opt->wbr_copyup));
16514 + case Opt_fhsm_sec:
16515 + AuDbg("fhsm_sec %u\n", opt->fhsm_second);
16519 + AuDbg("type %d\n", opt->type);
16527 +/* ---------------------------------------------------------------------- */
16530 + * For conditionally compiled mount options.
16531 + * Instead of fsparam_flag_no(), use this macro to distinguish ignore_silent.
16533 +#define au_ignore_flag(name, action) \
16534 + fsparam_flag(name, action), \
16535 + fsparam_flag("no" name, Opt_ignore_silent)
16537 +const struct fs_parameter_spec aufs_fsctx_paramspec[] = {
16538 + fsparam_string("br", Opt_br),
16540 + /* "add=%d:%s" or "ins=%d:%s" */
16541 + fsparam_string("add", Opt_add),
16542 + fsparam_string("ins", Opt_add),
16543 + fsparam_path("append", Opt_append),
16544 + fsparam_path("prepend", Opt_prepend),
16546 + fsparam_path("del", Opt_del),
16547 + /* fsparam_s32("idel", Opt_idel), */
16548 + fsparam_path("mod", Opt_mod),
16549 + /* fsparam_string("imod", Opt_imod), */
16551 + fsparam_s32("dirwh", Opt_dirwh),
16553 + fsparam_path("xino", Opt_xino),
16554 + fsparam_flag("noxino", Opt_noxino),
16555 + fsparam_flag_no("trunc_xino", Opt_trunc_xino),
16556 + /* "trunc_xino_v=%d:%d" */
16557 + /* fsparam_string("trunc_xino_v", Opt_trunc_xino_v), */
16558 + fsparam_path("trunc_xino", Opt_trunc_xino_path),
16559 + fsparam_s32("itrunc_xino", Opt_itrunc_xino),
16560 + /* fsparam_path("zxino", Opt_zxino), */
16561 + fsparam_flag_no("trunc_xib", Opt_trunc_xib),
16563 +#ifdef CONFIG_PROC_FS
16564 + fsparam_flag_no("plink", Opt_plink),
16566 + au_ignore_flag("plink", Opt_ignore),
16569 +#ifdef CONFIG_AUFS_DEBUG
16570 + fsparam_flag("list_plink", Opt_list_plink),
16573 + fsparam_string("udba", Opt_udba),
16575 + fsparam_flag_no("dio", Opt_dio),
16577 +#ifdef CONFIG_AUFS_DIRREN
16578 + fsparam_flag_no("dirren", Opt_dirren),
16580 + au_ignore_flag("dirren", Opt_ignore),
16583 +#ifdef CONFIG_AUFS_FHSM
16584 + fsparam_s32("fhsm_sec", Opt_fhsm_sec),
16586 + fsparam_s32("fhsm_sec", Opt_ignore),
16589 + /* always | a | whiteouted | w */
16590 + fsparam_string("diropq", Opt_diropq),
16592 + fsparam_flag_no("warn_perm", Opt_warn_perm),
16594 +#ifdef CONFIG_AUFS_SHWH
16595 + fsparam_flag_no("shwh", Opt_shwh),
16597 + au_ignore_flag("shwh", Opt_err),
16600 + fsparam_flag_no("dirperm1", Opt_dirperm1),
16602 + fsparam_flag_no("verbose", Opt_verbose),
16603 + fsparam_flag("v", Opt_verbose),
16604 + fsparam_flag("quiet", Opt_noverbose),
16605 + fsparam_flag("q", Opt_noverbose),
16606 + /* user-space may handle this */
16607 + fsparam_flag("silent", Opt_noverbose),
16609 + fsparam_flag_no("sum", Opt_sum),
16610 + fsparam_flag("wsum", Opt_wsum),
16612 + fsparam_s32("rdcache", Opt_rdcache),
16613 + /* "def" or s32 */
16614 + fsparam_string("rdblk", Opt_rdblk),
16615 + /* "def" or s32 */
16616 + fsparam_string("rdhash", Opt_rdhash),
16618 + fsparam_string("create", Opt_wbr_create),
16619 + fsparam_string("create_policy", Opt_wbr_create),
16620 + fsparam_string("cpup", Opt_wbr_copyup),
16621 + fsparam_string("copyup", Opt_wbr_copyup),
16622 + fsparam_string("copyup_policy", Opt_wbr_copyup),
16624 + /* generic VFS flag */
16625 +#ifdef CONFIG_FS_POSIX_ACL
16626 + fsparam_flag_no("acl", Opt_acl),
16628 + au_ignore_flag("acl"),
16631 + /* internal use for the scripts */
16632 + fsparam_string("si", Opt_ignore_silent),
16634 + /* obsoleted, keep them temporary */
16635 + fsparam_flag("nodlgt", Opt_ignore_silent),
16636 + fsparam_flag("clean_plink", Opt_ignore),
16637 + fsparam_string("dirs", Opt_br),
16638 + fsparam_u32("debug", Opt_ignore),
16639 + /* "whiteout" or "all" */
16640 + fsparam_string("delete", Opt_ignore),
16641 + fsparam_string("imap", Opt_ignore),
16643 + /* temporary workaround, due to old mount(8)? */
16644 + fsparam_flag("relatime", Opt_ignore_silent),
16649 +static int au_fsctx_parse_do_add(struct fs_context *fc, struct au_opt *opt,
16650 + char *brspec, size_t speclen,
16651 + aufs_bindex_t bindex)
16656 + AuDbg("brspec %s\n", brspec);
16660 + speclen = strlen(brspec);
16661 + /* will be freed by au_fsctx_free() */
16662 + p = kmemdup_nul(brspec, speclen, GFP_NOFS);
16663 + if (unlikely(!p)) {
16664 + errorfc(fc, "failed in %s", brspec);
16667 + err = au_opt_add(opt, p, fc->sb_flags, bindex);
16674 +static int au_fsctx_parse_br(struct fs_context *fc, char *brspec)
16678 + struct au_fsctx_opts *a = fc->fs_private;
16679 + struct au_opt *opt = a->opt;
16680 + aufs_bindex_t bindex = a->bindex;
16682 + AuDbg("brspec %s\n", brspec);
16685 + while ((p = strsep(&brspec, ":")) && *p) {
16686 + err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, bindex);
16688 + if (unlikely(err))
16692 + if (unlikely(opt > a->opt_tail)) {
16698 + opt->type = Opt_tail;
16701 + a->bindex = bindex;
16708 +static int au_fsctx_parse_add(struct fs_context *fc, char *addspec)
16712 + struct au_fsctx_opts *a = fc->fs_private;
16713 + struct au_opt *opt = a->opt;
16716 + p = strchr(addspec, ':');
16717 + if (unlikely(!p)) {
16718 + errorfc(fc, "bad arg in %s", addspec);
16722 + err = kstrtoint(addspec, 0, &n);
16723 + if (unlikely(err)) {
16724 + errorfc(fc, "bad integer in %s", addspec);
16727 + AuDbg("n %d\n", n);
16728 + err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, n);
16735 +static int au_fsctx_parse_del(struct fs_context *fc, struct au_opt_del *del,
16736 + struct fs_parameter *param)
16741 + /* will be freed by au_fsctx_free() */
16742 + del->pathname = kmemdup_nul(param->string, param->size, GFP_NOFS);
16743 + if (unlikely(!del->pathname))
16745 + AuDbg("del %s\n", del->pathname);
16746 + err = vfsub_kern_path(del->pathname, AuOpt_LkupDirFlags, &del->h_path);
16747 + if (unlikely(err))
16748 + errorfc(fc, "lookup failed %s (%d)", del->pathname, err);
16755 +#if 0 /* reserved for future use */
16756 +static int au_fsctx_parse_idel(struct fs_context *fc, struct au_opt_del *del,
16757 + aufs_bindex_t bindex)
16760 + struct super_block *sb;
16761 + struct dentry *root;
16762 + struct au_fsctx_opts *a = fc->fs_private;
16768 + root = sb->s_root;
16769 + aufs_read_lock(root, AuLock_FLUSH);
16770 + if (bindex < 0 || au_sbbot(sb) < bindex) {
16771 + errorfc(fc, "out of bounds, %d", bindex);
16776 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
16777 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
16780 + aufs_read_unlock(root, !AuLock_IR);
16786 +static int au_fsctx_parse_mod(struct fs_context *fc, struct au_opt_mod *mod,
16787 + struct fs_parameter *param)
16790 + struct path path;
16794 + /* will be freed by au_fsctx_free() */
16795 + mod->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16796 + if (unlikely(!mod->path))
16800 + p = strchr(mod->path, '=');
16801 + if (unlikely(!p)) {
16802 + errorfc(fc, "no permission %s", mod->path);
16807 + err = vfsub_kern_path(mod->path, AuOpt_LkupDirFlags, &path);
16808 + if (unlikely(err)) {
16809 + errorfc(fc, "lookup failed %s (%d)", mod->path, err);
16813 + mod->perm = au_br_perm_val(p);
16814 + AuDbg("mod path %s, perm 0x%x, %s", mod->path, mod->perm, p);
16815 + mod->h_root = dget(path.dentry);
16823 +#if 0 /* reserved for future use */
16824 +static int au_fsctx_parse_imod(struct fs_context *fc, struct au_opt_mod *mod,
16829 + struct super_block *sb;
16830 + struct dentry *root;
16831 + struct au_fsctx_opts *a = fc->fs_private;
16837 + p = strchr(ibrspec, ':');
16838 + if (unlikely(!p)) {
16839 + errorfc(fc, "no index, %s", ibrspec);
16843 + err = kstrtoint(ibrspec, 0, &n);
16844 + if (unlikely(err)) {
16845 + errorfc(fc, "bad integer in %s", ibrspec);
16848 + AuDbg("n %d\n", n);
16850 + root = sb->s_root;
16851 + aufs_read_lock(root, AuLock_FLUSH);
16852 + if (n < 0 || au_sbbot(sb) < n) {
16853 + errorfc(fc, "out of bounds, %d", bindex);
16858 + mod->perm = au_br_perm_val(p);
16859 + AuDbg("mod path %s, perm 0x%x, %s\n",
16860 + mod->path, mod->perm, p);
16861 + mod->h_root = dget(au_h_dptr(root, bindex));
16864 + aufs_read_unlock(root, !AuLock_IR);
16871 +static int au_fsctx_parse_xino(struct fs_context *fc,
16872 + struct au_opt_xino *xino,
16873 + struct fs_parameter *param)
16876 + struct au_fsctx_opts *a = fc->fs_private;
16879 + /* will be freed by au_opts_free() */
16880 + xino->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16881 + if (unlikely(!xino->path))
16883 + AuDbg("path %s\n", xino->path);
16885 + xino->file = au_xino_create(a->sb, xino->path, /*silent*/0,
16887 + err = PTR_ERR(xino->file);
16888 + if (IS_ERR(xino->file)) {
16889 + xino->file = NULL;
16894 + if (unlikely(a->sb && xino->file->f_path.dentry->d_sb == a->sb)) {
16896 + errorfc(fc, "%s must be outside", xino->path);
16905 +int au_fsctx_parse_xino_itrunc_path(struct fs_context *fc,
16906 + struct au_opt_xino_itrunc *xino_itrunc,
16910 + aufs_bindex_t bbot, bindex;
16911 + struct path path;
16912 + struct dentry *root;
16913 + struct au_fsctx_opts *a = fc->fs_private;
16915 + AuDebugOn(!a->sb);
16917 + err = vfsub_kern_path(pathname, AuOpt_LkupDirFlags, &path);
16918 + if (unlikely(err)) {
16919 + errorfc(fc, "lookup failed %s (%d)", pathname, err);
16923 + xino_itrunc->bindex = -1;
16924 + root = a->sb->s_root;
16925 + aufs_read_lock(root, AuLock_FLUSH);
16926 + bbot = au_sbbot(a->sb);
16927 + for (bindex = 0; bindex <= bbot; bindex++) {
16928 + if (au_h_dptr(root, bindex) == path.dentry) {
16929 + xino_itrunc->bindex = bindex;
16933 + aufs_read_unlock(root, !AuLock_IR);
16936 + if (unlikely(xino_itrunc->bindex < 0)) {
16938 + errorfc(fc, "no such branch %s", pathname);
16946 +static int au_fsctx_parse_xino_itrunc(struct fs_context *fc,
16947 + struct au_opt_xino_itrunc *xino_itrunc,
16948 + unsigned int bindex)
16951 + aufs_bindex_t bbot;
16952 + struct super_block *sb;
16953 + struct au_fsctx_opts *a = fc->fs_private;
16959 + si_noflush_read_lock(sb);
16960 + bbot = au_sbbot(sb);
16961 + si_read_unlock(sb);
16962 + if (bindex <= bbot)
16963 + xino_itrunc->bindex = bindex;
16966 + errorfc(fc, "out of bounds, %u", bindex);
16973 +static int au_fsctx_parse_param(struct fs_context *fc, struct fs_parameter *param)
16976 + struct fs_parse_result result;
16977 + struct au_fsctx_opts *a = fc->fs_private;
16978 + struct au_opt *opt = a->opt;
16980 + AuDbg("fc %p, param {key %s, string %s}\n",
16981 + fc, param->key, param->string);
16982 + err = fs_parse(fc, aufs_fsctx_paramspec, param, &result);
16983 + if (unlikely(err < 0))
16986 + AuDbg("token %d, res{negated %d, uint64 %llu}\n",
16987 + token, result.negated, result.uint_64);
16993 + err = au_fsctx_parse_br(fc, param->string);
16996 + err = au_fsctx_parse_add(fc, param->string);
16999 + err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17000 + /*dummy bindex*/1);
17002 + case Opt_prepend:
17003 + err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17008 + err = au_fsctx_parse_del(fc, &opt->del, param);
17010 +#if 0 /* reserved for future use */
17017 + del->pathname = "(indexed)";
17018 + err = au_opts_parse_idel(fc, &opt->del, result.uint_32);
17023 + err = au_fsctx_parse_mod(fc, &opt->mod, param);
17025 +#ifdef IMOD /* reserved for future use */
17032 + u.mod->path = "(indexed)";
17033 + err = au_opts_parse_imod(fc, &opt->mod, param->string);
17038 + err = au_fsctx_parse_xino(fc, &opt->xino, param);
17040 + case Opt_trunc_xino_path:
17042 + errorfc(fc, "no such branch %s", param->string);
17045 + err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17049 + case Opt_trunc_xino_v:
17055 + err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17059 + case Opt_itrunc_xino:
17061 + errorfc(fc, "out of bounds %s", param->string);
17064 + err = au_fsctx_parse_xino_itrunc(fc, &opt->xino_itrunc,
17070 + opt->dirwh = result.int_32;
17073 + case Opt_rdcache:
17074 + if (unlikely(result.int_32 > AUFS_RDCACHE_MAX)) {
17075 + errorfc(fc, "rdcache must be smaller than %d",
17076 + AUFS_RDCACHE_MAX);
17080 + opt->rdcache = result.int_32;
17085 + opt->rdblk = AUFS_RDBLK_DEF;
17086 + if (!strcmp(param->string, "def"))
17089 + err = kstrtoint(param->string, 0, &result.int_32);
17090 + if (unlikely(err)) {
17091 + errorfc(fc, "bad value in %s", param->key);
17095 + if (unlikely(result.int_32 < 0
17096 + || result.int_32 > KMALLOC_MAX_SIZE)) {
17097 + errorfc(fc, "bad value in %s", param->key);
17100 + if (unlikely(result.int_32 && result.int_32 < NAME_MAX)) {
17101 + errorfc(fc, "rdblk must be larger than %d", NAME_MAX);
17105 + opt->rdblk = result.int_32;
17110 + opt->rdhash = AUFS_RDHASH_DEF;
17111 + if (!strcmp(param->string, "def"))
17114 + err = kstrtoint(param->string, 0, &result.int_32);
17115 + if (unlikely(err)) {
17116 + errorfc(fc, "bad value in %s", param->key);
17119 + /* how about zero? */
17120 + if (result.int_32 < 0
17121 + || result.int_32 * sizeof(struct hlist_head)
17122 + > KMALLOC_MAX_SIZE) {
17124 + errorfc(fc, "bad integer in %s", param->key);
17127 + opt->rdhash = result.int_32;
17132 + * As other options, fs/aufs/opts.c can handle these strings by
17133 + * match_token(). But "diropq=" is deprecated now and will
17134 + * never have other value. So simple strcmp() is enough here.
17136 + if (!strcmp(param->string, "a") ||
17137 + !strcmp(param->string, "always")) {
17139 + opt->type = Opt_diropq_a;
17140 + } else if (!strcmp(param->string, "w") ||
17141 + !strcmp(param->string, "whiteouted")) {
17143 + opt->type = Opt_diropq_w;
17145 + errorfc(fc, "unknown value %s", param->string);
17149 + opt->udba = au_udba_val(param->string);
17150 + if (opt->udba >= 0)
17153 + errorf(fc, "wrong value, %s", param->string);
17156 + case Opt_wbr_create:
17157 + opt->wbr_create.wbr_create
17158 + = au_wbr_create_val(param->string, &opt->wbr_create);
17159 + if (opt->wbr_create.wbr_create >= 0)
17162 + errorf(fc, "wrong value, %s", param->key);
17165 + case Opt_wbr_copyup:
17166 + opt->wbr_copyup = au_wbr_copyup_val(param->string);
17167 + if (opt->wbr_copyup >= 0)
17170 + errorfc(fc, "wrong value, %s", param->key);
17173 + case Opt_fhsm_sec:
17174 + if (unlikely(result.int_32 < 0)) {
17175 + errorfc(fc, "bad integer in %s\n", param->key);
17180 + opt->fhsm_second = result.int_32;
17182 + warnfc(fc, "ignored %s %s", param->key, param->string);
17185 + /* simple true/false flag */
17186 +#define au_fsctx_TF(name) \
17187 + case Opt_##name: \
17189 + opt->tf = !result.negated; \
17191 + au_fsctx_TF(trunc_xino);
17192 + au_fsctx_TF(trunc_xib);
17193 + au_fsctx_TF(dirperm1);
17194 + au_fsctx_TF(plink);
17195 + au_fsctx_TF(shwh);
17196 + au_fsctx_TF(dio);
17197 + au_fsctx_TF(warn_perm);
17198 + au_fsctx_TF(verbose);
17199 + au_fsctx_TF(sum);
17200 + au_fsctx_TF(dirren);
17201 + au_fsctx_TF(acl);
17202 +#undef au_fsctx_TF
17204 + case Opt_noverbose:
17206 + opt->type = Opt_verbose;
17212 + case Opt_list_plink:
17219 + warnfc(fc, "ignored %s", param->key);
17221 + case Opt_ignore_silent:
17230 + if (unlikely(err))
17238 + case Opt_noverbose:
17243 + opt->type = token;
17247 + if (unlikely(opt > a->opt_tail)) {
17251 + opt->type = Opt_tail;
17259 + * these options accept both 'name=val' and 'name:val' form.
17260 + * some accept optional '=' in its value.
17261 + * eg. br:/br1=rw:/br2=ro and br=/br1=rw:/br2=ro
17263 +static inline unsigned int is_colonopt(char *str)
17265 +#define do_test(name) \
17266 + if (!strncmp(str, name ":", sizeof(name))) \
17267 + return sizeof(name) - 1;
17271 + do_test("append");
17272 + do_test("prepend");
17274 + /* do_test("idel"); */
17276 + /* do_test("imod"); */
17282 +static int au_fsctx_parse_monolithic(struct fs_context *fc, void *data)
17287 + struct au_fsctx_opts *a = fc->fs_private;
17290 + AuDbg("str %s\n", str);
17292 + u = is_colonopt(str);
17295 + str = strchr(str, ',');
17301 + AuDbg("str %s\n", str);
17303 + err = generic_parse_monolithic(fc, str);
17305 + au_fsctx_dump(&a->opts);
17310 +/* ---------------------------------------------------------------------- */
17312 +static void au_fsctx_opts_free(struct au_opts *opts)
17314 + struct au_opt *opt;
17317 + while (opt->type != Opt_tail) {
17318 + switch (opt->type) {
17323 + case Opt_prepend:
17324 + kfree(opt->add.pathname);
17325 + path_put(&opt->add.path);
17328 + kfree(opt->del.pathname);
17331 + path_put(&opt->del.h_path);
17334 + kfree(opt->mod.path);
17337 + dput(opt->mod.h_root);
17340 + kfree(opt->xino.path);
17341 + fput(opt->xino.file);
17348 +static void au_fsctx_free(struct fs_context *fc)
17350 + struct au_fsctx_opts *a = fc->fs_private;
17352 + /* fs_type=%p, root=%pD */
17353 + AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17354 + fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17356 + kobject_put(&a->sbinfo->si_kobj);
17357 + au_fsctx_opts_free(&a->opts);
17358 + free_page((unsigned long)a->opts.opt);
17362 +static const struct fs_context_operations au_fsctx_ops = {
17363 + .free = au_fsctx_free,
17364 + .parse_param = au_fsctx_parse_param,
17365 + .parse_monolithic = au_fsctx_parse_monolithic,
17366 + .get_tree = au_fsctx_get_tree,
17367 + .reconfigure = au_fsctx_reconfigure
17369 + * nfs4 requires ->dup()? No.
17370 + * I don't know what is this ->dup() for.
17374 +int aufs_fsctx_init(struct fs_context *fc)
17377 + struct au_fsctx_opts *a;
17379 + /* fs_type=%p, root=%pD */
17380 + AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17381 + fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17383 + /* they will be freed by au_fsctx_free() */
17385 + a = kzalloc(sizeof(*a), GFP_NOFS);
17386 + if (unlikely(!a))
17389 + a->opts.opt = (void *)__get_free_page(GFP_NOFS);
17390 + if (unlikely(!a->opts.opt))
17392 + a->opt = a->opts.opt;
17393 + a->opt->type = Opt_tail;
17394 + a->opts.max_opt = PAGE_SIZE / sizeof(*a->opts.opt);
17395 + a->opt_tail = a->opt + a->opts.max_opt - 1;
17396 + a->opts.sb_flags = fc->sb_flags;
17400 + AuDebugOn(fc->purpose != FS_CONTEXT_FOR_RECONFIGURE);
17401 + a->opts.flags = AuOpts_REMOUNT;
17402 + a->sb = fc->root->d_sb;
17403 + a->sbinfo = au_sbi(a->sb);
17404 + kobject_get(&a->sbinfo->si_kobj);
17406 + a->sbinfo = au_si_alloc(a->sb);
17407 + AuDebugOn(!a->sbinfo);
17408 + err = PTR_ERR(a->sbinfo);
17409 + if (IS_ERR(a->sbinfo))
17411 + au_rw_write_unlock(&a->sbinfo->si_rwsem);
17415 + fc->fs_private = a;
17416 + fc->ops = &au_fsctx_ops;
17417 + goto out; /* success */
17420 + free_page((unsigned long)a->opts.opt);
17427 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
17428 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
17429 +++ linux/fs/aufs/fstype.h 2022-03-21 14:49:05.723299677 +0100
17431 +/* SPDX-License-Identifier: GPL-2.0 */
17433 + * Copyright (C) 2005-2021 Junjiro R. Okajima
17435 + * This program, aufs is free software; you can redistribute it and/or modify
17436 + * it under the terms of the GNU General Public License as published by
17437 + * the Free Software Foundation; either version 2 of the License, or
17438 + * (at your option) any later version.
17440 + * This program is distributed in the hope that it will be useful,
17441 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17442 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17443 + * GNU General Public License for more details.
17445 + * You should have received a copy of the GNU General Public License
17446 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17450 + * judging filesystem type
17453 +#ifndef __AUFS_FSTYPE_H__
17454 +#define __AUFS_FSTYPE_H__
17458 +#include <linux/fs.h>
17459 +#include <linux/magic.h>
17460 +#include <linux/nfs_fs.h>
17461 +#include <linux/romfs_fs.h>
17463 +static inline int au_test_aufs(struct super_block *sb)
17465 + return sb->s_magic == AUFS_SUPER_MAGIC;
17468 +static inline const char *au_sbtype(struct super_block *sb)
17470 + return sb->s_type->name;
17473 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
17475 +#if IS_ENABLED(CONFIG_ISO9660_FS)
17476 + return sb->s_magic == ISOFS_SUPER_MAGIC;
17482 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
17484 +#if IS_ENABLED(CONFIG_ROMFS_FS)
17485 + return sb->s_magic == ROMFS_MAGIC;
17491 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
17493 +#if IS_ENABLED(CONFIG_CRAMFS)
17494 + return sb->s_magic == CRAMFS_MAGIC;
17499 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
17501 +#if IS_ENABLED(CONFIG_NFS_FS)
17502 + return sb->s_magic == NFS_SUPER_MAGIC;
17508 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
17510 +#if IS_ENABLED(CONFIG_FUSE_FS)
17511 + return sb->s_magic == FUSE_SUPER_MAGIC;
17517 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
17519 +#if IS_ENABLED(CONFIG_XFS_FS)
17520 + return sb->s_magic == XFS_SB_MAGIC;
17526 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
17528 +#ifdef CONFIG_TMPFS
17529 + return sb->s_magic == TMPFS_MAGIC;
17535 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
17537 +#if IS_ENABLED(CONFIG_ECRYPT_FS)
17538 + return !strcmp(au_sbtype(sb), "ecryptfs");
17544 +static inline int au_test_ramfs(struct super_block *sb)
17546 + return sb->s_magic == RAMFS_MAGIC;
17549 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
17551 +#if IS_ENABLED(CONFIG_UBIFS_FS)
17552 + return sb->s_magic == UBIFS_SUPER_MAGIC;
17558 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
17560 +#ifdef CONFIG_PROC_FS
17561 + return sb->s_magic == PROC_SUPER_MAGIC;
17567 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
17569 +#ifdef CONFIG_SYSFS
17570 + return sb->s_magic == SYSFS_MAGIC;
17576 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
17578 +#if IS_ENABLED(CONFIG_CONFIGFS_FS)
17579 + return sb->s_magic == CONFIGFS_MAGIC;
17585 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
17587 +#if IS_ENABLED(CONFIG_MINIX_FS)
17588 + return sb->s_magic == MINIX3_SUPER_MAGIC
17589 + || sb->s_magic == MINIX2_SUPER_MAGIC
17590 + || sb->s_magic == MINIX2_SUPER_MAGIC2
17591 + || sb->s_magic == MINIX_SUPER_MAGIC
17592 + || sb->s_magic == MINIX_SUPER_MAGIC2;
17598 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
17600 +#if IS_ENABLED(CONFIG_FAT_FS)
17601 + return sb->s_magic == MSDOS_SUPER_MAGIC;
17607 +static inline int au_test_msdos(struct super_block *sb)
17609 + return au_test_fat(sb);
17612 +static inline int au_test_vfat(struct super_block *sb)
17614 + return au_test_fat(sb);
17617 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
17619 +#ifdef CONFIG_SECURITYFS
17620 + return sb->s_magic == SECURITYFS_MAGIC;
17626 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
17628 +#if IS_ENABLED(CONFIG_SQUASHFS)
17629 + return sb->s_magic == SQUASHFS_MAGIC;
17635 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
17637 +#if IS_ENABLED(CONFIG_BTRFS_FS)
17638 + return sb->s_magic == BTRFS_SUPER_MAGIC;
17644 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
17646 +#if IS_ENABLED(CONFIG_XENFS)
17647 + return sb->s_magic == XENFS_SUPER_MAGIC;
17653 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
17655 +#ifdef CONFIG_DEBUG_FS
17656 + return sb->s_magic == DEBUGFS_MAGIC;
17662 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
17664 +#if IS_ENABLED(CONFIG_NILFS)
17665 + return sb->s_magic == NILFS_SUPER_MAGIC;
17671 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
17673 +#if IS_ENABLED(CONFIG_HFSPLUS_FS)
17674 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
17680 +/* ---------------------------------------------------------------------- */
17682 + * they can't be an aufs branch.
17684 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
17687 +#ifndef CONFIG_AUFS_BR_RAMFS
17688 + au_test_ramfs(sb) ||
17690 + au_test_procfs(sb)
17691 + || au_test_sysfs(sb)
17692 + || au_test_configfs(sb)
17693 + || au_test_debugfs(sb)
17694 + || au_test_securityfs(sb)
17695 + || au_test_xenfs(sb)
17696 + || au_test_ecryptfs(sb)
17697 + /* || !strcmp(au_sbtype(sb), "unionfs") */
17698 + || au_test_aufs(sb); /* will be supported in next version */
17701 +static inline int au_test_fs_remote(struct super_block *sb)
17703 + return !au_test_tmpfs(sb)
17704 +#ifdef CONFIG_AUFS_BR_RAMFS
17705 + && !au_test_ramfs(sb)
17707 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
17710 +/* ---------------------------------------------------------------------- */
17713 + * Note: these functions (below) are created after reading ->getattr() in all
17714 + * filesystems under linux/fs. it means we have to do so in every update...
17718 + * some filesystems require getattr to refresh the inode attributes before
17720 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
17721 + * and leave the work for d_revalidate()
17723 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
17725 + return au_test_nfs(sb)
17726 + || au_test_fuse(sb)
17727 + /* || au_test_btrfs(sb) */ /* untested */
17732 + * filesystems which don't maintain i_size or i_blocks.
17734 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
17736 + return au_test_xfs(sb)
17737 + || au_test_btrfs(sb)
17738 + || au_test_ubifs(sb)
17739 + || au_test_hfsplus(sb) /* maintained, but incorrect */
17740 + /* || au_test_minix(sb) */ /* untested */
17745 + * filesystems which don't store the correct value in some of their inode
17748 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
17750 + return au_test_fs_bad_iattr_size(sb)
17751 + || au_test_fat(sb)
17752 + || au_test_msdos(sb)
17753 + || au_test_vfat(sb);
17756 +/* they don't check i_nlink in link(2) */
17757 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
17759 + return au_test_tmpfs(sb)
17760 +#ifdef CONFIG_AUFS_BR_RAMFS
17761 + || au_test_ramfs(sb)
17763 + || au_test_ubifs(sb)
17764 + || au_test_hfsplus(sb);
17768 + * filesystems which sets S_NOATIME and S_NOCMTIME.
17770 +static inline int au_test_fs_notime(struct super_block *sb)
17772 + return au_test_nfs(sb)
17773 + || au_test_fuse(sb)
17774 + || au_test_ubifs(sb)
17778 +/* temporary support for i#1 in cramfs */
17779 +static inline int au_test_fs_unique_ino(struct inode *inode)
17781 + if (au_test_cramfs(inode->i_sb))
17782 + return inode->i_ino != 1;
17786 +/* ---------------------------------------------------------------------- */
17789 + * the filesystem where the xino files placed must support i/o after unlink and
17790 + * maintain i_size and i_blocks.
17792 +static inline int au_test_fs_bad_xino(struct super_block *sb)
17794 + return au_test_fs_remote(sb)
17795 + || au_test_fs_bad_iattr_size(sb)
17796 + /* don't want unnecessary work for xino */
17797 + || au_test_aufs(sb)
17798 + || au_test_ecryptfs(sb)
17799 + || au_test_nilfs(sb);
17802 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
17804 + return au_test_tmpfs(sb)
17805 + || au_test_ramfs(sb);
17809 + * test if the @sb is real-readonly.
17811 +static inline int au_test_fs_rr(struct super_block *sb)
17813 + return au_test_squashfs(sb)
17814 + || au_test_iso9660(sb)
17815 + || au_test_cramfs(sb)
17816 + || au_test_romfs(sb);
17820 + * test if the @inode is nfs with 'noacl' option
17821 + * NFS always sets SB_POSIXACL regardless its mount option 'noacl.'
17823 +static inline int au_test_nfs_noacl(struct inode *inode)
17825 + return au_test_nfs(inode->i_sb)
17826 + /* && IS_POSIXACL(inode) */
17827 + && !nfs_server_capable(inode, NFS_CAP_ACLS);
17830 +#endif /* __KERNEL__ */
17831 +#endif /* __AUFS_FSTYPE_H__ */
17832 diff -urN /usr/share/empty/fs/aufs/hbl.h linux/fs/aufs/hbl.h
17833 --- /usr/share/empty/fs/aufs/hbl.h 1970-01-01 01:00:00.000000000 +0100
17834 +++ linux/fs/aufs/hbl.h 2022-03-21 14:49:05.723299677 +0100
17836 +/* SPDX-License-Identifier: GPL-2.0 */
17838 + * Copyright (C) 2017-2021 Junjiro R. Okajima
17840 + * This program, aufs is free software; you can redistribute it and/or modify
17841 + * it under the terms of the GNU General Public License as published by
17842 + * the Free Software Foundation; either version 2 of the License, or
17843 + * (at your option) any later version.
17845 + * This program is distributed in the hope that it will be useful,
17846 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17847 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17848 + * GNU General Public License for more details.
17850 + * You should have received a copy of the GNU General Public License
17851 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17855 + * helpers for hlist_bl.h
17858 +#ifndef __AUFS_HBL_H__
17859 +#define __AUFS_HBL_H__
17863 +#include <linux/list_bl.h>
17865 +static inline void au_hbl_add(struct hlist_bl_node *node,
17866 + struct hlist_bl_head *hbl)
17868 + hlist_bl_lock(hbl);
17869 + hlist_bl_add_head(node, hbl);
17870 + hlist_bl_unlock(hbl);
17873 +static inline void au_hbl_del(struct hlist_bl_node *node,
17874 + struct hlist_bl_head *hbl)
17876 + hlist_bl_lock(hbl);
17877 + hlist_bl_del(node);
17878 + hlist_bl_unlock(hbl);
17881 +#define au_hbl_for_each(pos, head) \
17882 + for (pos = hlist_bl_first(head); \
17886 +static inline unsigned long au_hbl_count(struct hlist_bl_head *hbl)
17888 + unsigned long cnt;
17889 + struct hlist_bl_node *pos;
17892 + hlist_bl_lock(hbl);
17893 + au_hbl_for_each(pos, hbl)
17895 + hlist_bl_unlock(hbl);
17899 +#endif /* __KERNEL__ */
17900 +#endif /* __AUFS_HBL_H__ */
17901 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
17902 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
17903 +++ linux/fs/aufs/hfsnotify.c 2022-03-21 14:49:05.723299677 +0100
17905 +// SPDX-License-Identifier: GPL-2.0
17907 + * Copyright (C) 2005-2021 Junjiro R. Okajima
17909 + * This program, aufs is free software; you can redistribute it and/or modify
17910 + * it under the terms of the GNU General Public License as published by
17911 + * the Free Software Foundation; either version 2 of the License, or
17912 + * (at your option) any later version.
17914 + * This program is distributed in the hope that it will be useful,
17915 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17916 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17917 + * GNU General Public License for more details.
17919 + * You should have received a copy of the GNU General Public License
17920 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17924 + * fsnotify for the lower directories
17929 +/* FS_IN_IGNORED is unnecessary */
17930 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
17931 + | FS_CREATE | FS_EVENT_ON_CHILD);
17932 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
17933 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
17935 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
17937 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
17939 + /* AuDbg("here\n"); */
17940 + au_cache_free_hnotify(hn);
17941 + smp_mb__before_atomic(); /* for atomic64_dec */
17942 + if (atomic64_dec_and_test(&au_hfsn_ifree))
17943 + wake_up(&au_hfsn_wq);
17946 +static int au_hfsn_alloc(struct au_hinode *hinode)
17949 + struct au_hnotify *hn;
17950 + struct super_block *sb;
17951 + struct au_branch *br;
17952 + struct fsnotify_mark *mark;
17953 + aufs_bindex_t bindex;
17955 + hn = hinode->hi_notify;
17956 + sb = hn->hn_aufs_inode->i_sb;
17957 + bindex = au_br_index(sb, hinode->hi_id);
17958 + br = au_sbr(sb, bindex);
17959 + AuDebugOn(!br->br_hfsn);
17961 + mark = &hn->hn_mark;
17962 + fsnotify_init_mark(mark, br->br_hfsn->hfsn_group);
17963 + mark->mask = AuHfsnMask;
17965 + * by udba rename or rmdir, aufs assign a new inode to the known
17966 + * h_inode, so specify 1 to allow dups.
17969 + err = fsnotify_add_inode_mark(mark, hinode->hi_inode, /*allow_dups*/1);
17975 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
17977 + struct fsnotify_mark *mark;
17978 + unsigned long long ull;
17979 + struct fsnotify_group *group;
17981 + ull = atomic64_inc_return(&au_hfsn_ifree);
17984 + mark = &hn->hn_mark;
17985 + spin_lock(&mark->lock);
17986 + group = mark->group;
17987 + fsnotify_get_group(group);
17988 + spin_unlock(&mark->lock);
17990 + fsnotify_destroy_mark(mark, group);
17991 + fsnotify_put_mark(mark);
17992 + fsnotify_put_group(group);
17995 + /* free hn by myself */
17999 +/* ---------------------------------------------------------------------- */
18001 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
18003 + struct fsnotify_mark *mark;
18005 + mark = &hinode->hi_notify->hn_mark;
18006 + spin_lock(&mark->lock);
18008 + AuDebugOn(mark->mask & AuHfsnMask);
18009 + mark->mask |= AuHfsnMask;
18011 + AuDebugOn(!(mark->mask & AuHfsnMask));
18012 + mark->mask &= ~AuHfsnMask;
18014 + spin_unlock(&mark->lock);
18015 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
18018 +/* ---------------------------------------------------------------------- */
18020 +/* #define AuDbgHnotify */
18021 +#ifdef AuDbgHnotify
18022 +static char *au_hfsn_name(u32 mask)
18024 +#ifdef CONFIG_AUFS_DEBUG
18025 +#define test_ret(flag) \
18027 + if (mask & flag) \
18030 + test_ret(FS_ACCESS);
18031 + test_ret(FS_MODIFY);
18032 + test_ret(FS_ATTRIB);
18033 + test_ret(FS_CLOSE_WRITE);
18034 + test_ret(FS_CLOSE_NOWRITE);
18035 + test_ret(FS_OPEN);
18036 + test_ret(FS_MOVED_FROM);
18037 + test_ret(FS_MOVED_TO);
18038 + test_ret(FS_CREATE);
18039 + test_ret(FS_DELETE);
18040 + test_ret(FS_DELETE_SELF);
18041 + test_ret(FS_MOVE_SELF);
18042 + test_ret(FS_UNMOUNT);
18043 + test_ret(FS_Q_OVERFLOW);
18044 + test_ret(FS_IN_IGNORED);
18045 + test_ret(FS_ISDIR);
18046 + test_ret(FS_IN_ONESHOT);
18047 + test_ret(FS_EVENT_ON_CHILD);
18056 +/* ---------------------------------------------------------------------- */
18058 +static void au_hfsn_free_group(struct fsnotify_group *group)
18060 + struct au_br_hfsnotify *hfsn = group->private;
18062 + /* AuDbg("here\n"); */
18063 + au_kfree_try_rcu(hfsn);
18066 +static int au_hfsn_handle_event(struct fsnotify_group *group,
18067 + u32 mask, const void *data, int data_type,
18068 + struct inode *dir,
18069 + const struct qstr *file_name, u32 cookie,
18070 + struct fsnotify_iter_info *iter_info)
18073 + struct au_hnotify *hnotify;
18074 + struct inode *h_dir, *h_inode;
18075 + struct fsnotify_mark *inode_mark;
18077 + AuDebugOn(!(data_type == FSNOTIFY_EVENT_INODE
18078 + || data_type == FSNOTIFY_EVENT_DENTRY));
18081 + /* if FS_UNMOUNT happens, there must be another bug */
18082 + AuDebugOn(mask & FS_UNMOUNT);
18083 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
18088 +#ifdef AuDbgHnotify
18090 + if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
18091 + || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
18092 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
18093 + h_dir->i_ino, mask, au_hfsn_name(mask),
18094 + AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
18095 + /* WARN_ON(1); */
18100 + inode_mark = fsnotify_iter_inode_mark(iter_info);
18101 + AuDebugOn(!inode_mark);
18102 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
18103 + err = au_hnotify(h_dir, hnotify, mask, file_name, h_inode);
18109 +static struct fsnotify_ops au_hfsn_ops = {
18110 + .handle_event = au_hfsn_handle_event,
18111 + .free_group_priv = au_hfsn_free_group,
18112 + .free_mark = au_hfsn_free_mark
18115 +/* ---------------------------------------------------------------------- */
18117 +static void au_hfsn_fin_br(struct au_branch *br)
18119 + struct au_br_hfsnotify *hfsn;
18121 + hfsn = br->br_hfsn;
18124 + fsnotify_put_group(hfsn->hfsn_group);
18129 +static int au_hfsn_init_br(struct au_branch *br, int perm)
18132 + struct fsnotify_group *group;
18133 + struct au_br_hfsnotify *hfsn;
18136 + br->br_hfsn = NULL;
18137 + if (!au_br_hnotifyable(perm))
18141 + hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
18142 + if (unlikely(!hfsn))
18146 + group = fsnotify_alloc_group(&au_hfsn_ops);
18147 + if (IS_ERR(group)) {
18148 + err = PTR_ERR(group);
18149 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
18153 + group->private = hfsn;
18154 + hfsn->hfsn_group = group;
18155 + br->br_hfsn = hfsn;
18156 + goto out; /* success */
18159 + au_kfree_try_rcu(hfsn);
18164 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
18169 + if (!br->br_hfsn)
18170 + err = au_hfsn_init_br(br, perm);
18175 +/* ---------------------------------------------------------------------- */
18177 +static void au_hfsn_fin(void)
18179 + AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
18180 + wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
18183 +const struct au_hnotify_op au_hnotify_op = {
18184 + .ctl = au_hfsn_ctl,
18185 + .alloc = au_hfsn_alloc,
18186 + .free = au_hfsn_free,
18188 + .fin = au_hfsn_fin,
18190 + .reset_br = au_hfsn_reset_br,
18191 + .fin_br = au_hfsn_fin_br,
18192 + .init_br = au_hfsn_init_br
18194 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
18195 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
18196 +++ linux/fs/aufs/hfsplus.c 2022-03-21 14:49:05.723299677 +0100
18198 +// SPDX-License-Identifier: GPL-2.0
18200 + * Copyright (C) 2010-2021 Junjiro R. Okajima
18202 + * This program, aufs is free software; you can redistribute it and/or modify
18203 + * it under the terms of the GNU General Public License as published by
18204 + * the Free Software Foundation; either version 2 of the License, or
18205 + * (at your option) any later version.
18207 + * This program is distributed in the hope that it will be useful,
18208 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18209 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18210 + * GNU General Public License for more details.
18212 + * You should have received a copy of the GNU General Public License
18213 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18217 + * special support for filesystems which acquires an inode mutex
18218 + * at final closing a file, eg, hfsplus.
18220 + * This trick is very simple and stupid, just to open the file before really
18221 + * necessary open to tell hfsplus that this is not the final closing.
18222 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
18223 + * and au_h_open_post() after releasing it.
18228 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
18231 + struct file *h_file;
18232 + struct dentry *h_dentry;
18234 + h_dentry = au_h_dptr(dentry, bindex);
18235 + AuDebugOn(!h_dentry);
18236 + AuDebugOn(d_is_negative(h_dentry));
18239 + if (au_test_hfsplus(h_dentry->d_sb)
18240 + && d_is_reg(h_dentry))
18241 + h_file = au_h_open(dentry, bindex,
18242 + O_RDONLY | O_NOATIME | O_LARGEFILE,
18243 + /*file*/NULL, force_wr);
18247 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
18248 + struct file *h_file)
18250 + struct au_branch *br;
18254 + br = au_sbr(dentry->d_sb, bindex);
18255 + au_lcnt_dec(&br->br_nfiles);
18258 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
18259 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
18260 +++ linux/fs/aufs/hnotify.c 2022-03-21 14:49:05.723299677 +0100
18262 +// SPDX-License-Identifier: GPL-2.0
18264 + * Copyright (C) 2005-2021 Junjiro R. Okajima
18266 + * This program, aufs is free software; you can redistribute it and/or modify
18267 + * it under the terms of the GNU General Public License as published by
18268 + * the Free Software Foundation; either version 2 of the License, or
18269 + * (at your option) any later version.
18271 + * This program is distributed in the hope that it will be useful,
18272 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18273 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18274 + * GNU General Public License for more details.
18276 + * You should have received a copy of the GNU General Public License
18277 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18281 + * abstraction to notify the direct changes on lower directories
18284 +/* #include <linux/iversion.h> */
18287 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
18290 + struct au_hnotify *hn;
18293 + hn = au_cache_alloc_hnotify();
18295 + hn->hn_aufs_inode = inode;
18296 + hinode->hi_notify = hn;
18297 + err = au_hnotify_op.alloc(hinode);
18299 + if (unlikely(err)) {
18300 + hinode->hi_notify = NULL;
18301 + au_cache_free_hnotify(hn);
18303 + * The upper dir was removed by udba, but the same named
18304 + * dir left. In this case, aufs assigns a new inode
18305 + * number and set the monitor again.
18306 + * For the lower dir, the old monitor is still left.
18308 + if (err == -EEXIST)
18317 +void au_hn_free(struct au_hinode *hinode)
18319 + struct au_hnotify *hn;
18321 + hn = hinode->hi_notify;
18323 + hinode->hi_notify = NULL;
18324 + if (au_hnotify_op.free(hinode, hn))
18325 + au_cache_free_hnotify(hn);
18329 +/* ---------------------------------------------------------------------- */
18331 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
18333 + if (hinode->hi_notify)
18334 + au_hnotify_op.ctl(hinode, do_set);
18337 +void au_hn_reset(struct inode *inode, unsigned int flags)
18339 + aufs_bindex_t bindex, bbot;
18340 + struct inode *hi;
18341 + struct dentry *iwhdentry;
18343 + bbot = au_ibbot(inode);
18344 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
18345 + hi = au_h_iptr(inode, bindex);
18349 + /* inode_lock_nested(hi, AuLsc_I_CHILD); */
18350 + iwhdentry = au_hi_wh(inode, bindex);
18354 + au_set_h_iptr(inode, bindex, NULL, 0);
18355 + au_set_h_iptr(inode, bindex, au_igrab(hi),
18356 + flags & ~AuHi_XINO);
18359 + /* inode_unlock(hi); */
18363 +/* ---------------------------------------------------------------------- */
18365 +static int hn_xino(struct inode *inode, struct inode *h_inode)
18368 + aufs_bindex_t bindex, bbot, bfound, btop;
18369 + struct inode *h_i;
18372 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18373 + pr_warn("branch root dir was changed\n");
18378 + bbot = au_ibbot(inode);
18379 + btop = au_ibtop(inode);
18380 +#if 0 /* reserved for future use */
18381 + if (bindex == bbot) {
18382 + /* keep this ino in rename case */
18386 + for (bindex = btop; bindex <= bbot; bindex++)
18387 + if (au_h_iptr(inode, bindex) == h_inode) {
18394 + for (bindex = btop; bindex <= bbot; bindex++) {
18395 + h_i = au_h_iptr(inode, bindex);
18399 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
18400 + /* ignore this error */
18401 + /* bad action? */
18404 + /* children inode number will be broken */
18411 +static int hn_gen_tree(struct dentry *dentry)
18413 + int err, i, j, ndentry;
18414 + struct au_dcsub_pages dpages;
18415 + struct au_dpage *dpage;
18416 + struct dentry **dentries;
18418 + err = au_dpages_init(&dpages, GFP_NOFS);
18419 + if (unlikely(err))
18421 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
18422 + if (unlikely(err))
18425 + for (i = 0; i < dpages.ndpage; i++) {
18426 + dpage = dpages.dpages + i;
18427 + dentries = dpage->dentries;
18428 + ndentry = dpage->ndentry;
18429 + for (j = 0; j < ndentry; j++) {
18430 + struct dentry *d;
18437 + if (d_really_is_positive(d))
18438 + /* todo: reset children xino?
18439 + cached children only? */
18440 + au_iigen_dec(d_inode(d));
18445 + au_dpages_free(&dpages);
18451 + * return 0 if processed.
18453 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
18454 + const unsigned int isdir)
18457 + struct dentry *d;
18458 + struct qstr *dname;
18461 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18462 + pr_warn("branch root dir was changed\n");
18468 + AuDebugOn(!name);
18469 + au_iigen_dec(inode);
18470 + spin_lock(&inode->i_lock);
18471 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
18472 + spin_lock(&d->d_lock);
18473 + dname = &d->d_name;
18474 + if (dname->len != nlen
18475 + && memcmp(dname->name, name, nlen)) {
18476 + spin_unlock(&d->d_lock);
18481 + spin_unlock(&d->d_lock);
18484 + spin_unlock(&inode->i_lock);
18486 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
18487 + d = d_find_any_alias(inode);
18489 + au_iigen_dec(inode);
18493 + spin_lock(&d->d_lock);
18494 + dname = &d->d_name;
18495 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
18496 + spin_unlock(&d->d_lock);
18497 + err = hn_gen_tree(d);
18498 + spin_lock(&d->d_lock);
18500 + spin_unlock(&d->d_lock);
18509 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
18513 + if (IS_ROOT(dentry)) {
18514 + pr_warn("branch root dir was changed\n");
18520 + au_digen_dec(dentry);
18521 + if (d_really_is_positive(dentry))
18522 + au_iigen_dec(d_inode(dentry));
18524 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
18525 + if (d_really_is_positive(dentry))
18526 + err = hn_gen_tree(dentry);
18533 +/* ---------------------------------------------------------------------- */
18535 +/* hnotify job flags */
18536 +#define AuHnJob_XINO0 1
18537 +#define AuHnJob_GEN (1 << 1)
18538 +#define AuHnJob_DIRENT (1 << 2)
18539 +#define AuHnJob_ISDIR (1 << 3)
18540 +#define AuHnJob_TRYXINO0 (1 << 4)
18541 +#define AuHnJob_MNTPNT (1 << 5)
18542 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
18543 +#define au_fset_hnjob(flags, name) \
18544 + do { (flags) |= AuHnJob_##name; } while (0)
18545 +#define au_fclr_hnjob(flags, name) \
18546 + do { (flags) &= ~AuHnJob_##name; } while (0)
18554 +struct au_hnotify_args {
18555 + struct inode *h_dir, *dir, *h_child_inode;
18557 + unsigned int flags[AuHnLast];
18558 + unsigned int h_child_nlen;
18559 + char h_child_name[];
18562 +struct hn_job_args {
18563 + unsigned int flags;
18564 + struct inode *inode, *h_inode, *dir, *h_dir;
18565 + struct dentry *dentry;
18570 +static int hn_job(struct hn_job_args *a)
18572 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
18576 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
18577 + hn_xino(a->inode, a->h_inode); /* ignore this error */
18579 + if (au_ftest_hnjob(a->flags, TRYXINO0)
18582 + inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
18583 + if (!a->h_inode->i_nlink
18584 + && !(a->h_inode->i_state & I_LINKABLE))
18585 + hn_xino(a->inode, a->h_inode); /* ignore this error */
18586 + inode_unlock_shared(a->h_inode);
18589 + /* make the generation obsolete */
18590 + if (au_ftest_hnjob(a->flags, GEN)) {
18593 + e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
18595 + if (e && a->dentry)
18596 + hn_gen_by_name(a->dentry, isdir);
18597 + /* ignore this error */
18600 + /* make dir entries obsolete */
18601 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
18602 + struct au_vdir *vdir;
18604 + vdir = au_ivdir(a->inode);
18606 + vdir->vd_jiffy = 0;
18607 + /* IMustLock(a->inode); */
18608 + /* inode_inc_iversion(a->inode); */
18611 + /* can do nothing but warn */
18612 + if (au_ftest_hnjob(a->flags, MNTPNT)
18614 + && d_mountpoint(a->dentry))
18615 + pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
18620 +/* ---------------------------------------------------------------------- */
18622 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
18623 + struct inode *dir)
18625 + struct dentry *dentry, *d, *parent;
18626 + struct qstr *dname;
18628 + parent = d_find_any_alias(dir);
18633 + spin_lock(&parent->d_lock);
18634 + list_for_each_entry(d, &parent->d_subdirs, d_child) {
18635 + /* AuDbg("%pd\n", d); */
18636 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
18637 + dname = &d->d_name;
18638 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
18639 + goto cont_unlock;
18643 + goto cont_unlock;
18644 + if (au_dcount(d) > 0) {
18645 + dentry = dget_dlock(d);
18646 + spin_unlock(&d->d_lock);
18651 + spin_unlock(&d->d_lock);
18653 + spin_unlock(&parent->d_lock);
18657 + di_write_lock_child(dentry);
18662 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
18663 + aufs_bindex_t bindex, ino_t h_ino)
18665 + struct inode *inode;
18670 + err = au_xino_read(sb, bindex, h_ino, &ino);
18672 + inode = ilookup(sb, ino);
18676 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18677 + pr_warn("wrong root branch\n");
18683 + ii_write_lock_child(inode);
18689 +static void au_hn_bh(void *_args)
18691 + struct au_hnotify_args *a = _args;
18692 + struct super_block *sb;
18693 + aufs_bindex_t bindex, bbot, bfound;
18694 + unsigned char xino, try_iput;
18696 + struct inode *inode;
18698 + struct hn_job_args args;
18699 + struct dentry *dentry;
18700 + struct au_sbinfo *sbinfo;
18702 + AuDebugOn(!_args);
18703 + AuDebugOn(!a->h_dir);
18704 + AuDebugOn(!a->dir);
18705 + AuDebugOn(!a->mask);
18706 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
18707 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
18708 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
18713 + * do not lock a->dir->i_mutex here
18714 + * because of d_revalidate() may cause a deadlock.
18716 + sb = a->dir->i_sb;
18718 + sbinfo = au_sbi(sb);
18719 + AuDebugOn(!sbinfo);
18720 + si_write_lock(sb, AuLock_NOPLMW);
18722 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
18723 + switch (a->mask & FS_EVENTS_POSS_ON_CHILD) {
18724 + case FS_MOVED_FROM:
18725 + case FS_MOVED_TO:
18726 + AuWarn1("DIRREN with UDBA may not work correctly "
18727 + "for the direct rename(2)\n");
18730 + ii_read_lock_parent(a->dir);
18732 + bbot = au_ibbot(a->dir);
18733 + for (bindex = au_ibtop(a->dir); bindex <= bbot; bindex++)
18734 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
18738 + ii_read_unlock(a->dir);
18739 + if (unlikely(bfound < 0))
18742 + xino = !!au_opt_test(au_mntflags(sb), XINO);
18744 + if (a->h_child_inode)
18745 + h_ino = a->h_child_inode->i_ino;
18747 + if (a->h_child_nlen
18748 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
18749 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
18750 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
18753 + if (dentry && d_really_is_positive(dentry))
18754 + inode = d_inode(dentry);
18755 + if (xino && !inode && h_ino
18756 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
18757 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
18758 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
18759 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
18763 + args.flags = a->flags[AuHn_CHILD];
18764 + args.dentry = dentry;
18765 + args.inode = inode;
18766 + args.h_inode = a->h_child_inode;
18767 + args.dir = a->dir;
18768 + args.h_dir = a->h_dir;
18769 + args.h_name = a->h_child_name;
18770 + args.h_nlen = a->h_child_nlen;
18771 + err = hn_job(&args);
18773 + if (au_di(dentry))
18774 + di_write_unlock(dentry);
18777 + if (inode && try_iput) {
18778 + ii_write_unlock(inode);
18782 + ii_write_lock_parent(a->dir);
18783 + args.flags = a->flags[AuHn_PARENT];
18784 + args.dentry = NULL;
18785 + args.inode = a->dir;
18786 + args.h_inode = a->h_dir;
18788 + args.h_dir = NULL;
18789 + args.h_name = NULL;
18791 + err = hn_job(&args);
18792 + ii_write_unlock(a->dir);
18795 + iput(a->h_child_inode);
18798 + si_write_unlock(sb);
18799 + au_nwt_done(&sbinfo->si_nowait);
18803 +/* ---------------------------------------------------------------------- */
18805 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
18806 + const struct qstr *h_child_qstr, struct inode *h_child_inode)
18809 + unsigned int flags[AuHnLast], f;
18810 + unsigned char isdir, isroot, wh;
18811 + struct inode *dir;
18812 + struct au_hnotify_args *args;
18813 + char *p, *h_child_name;
18816 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
18817 + dir = igrab(hnotify->hn_aufs_inode);
18821 + isroot = (dir->i_ino == AUFS_ROOT_INO);
18823 + h_child_name = (void *)h_child_qstr->name;
18824 + len = h_child_qstr->len;
18825 + if (h_child_name) {
18826 + if (len > AUFS_WH_PFX_LEN
18827 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
18828 + h_child_name += AUFS_WH_PFX_LEN;
18829 + len -= AUFS_WH_PFX_LEN;
18835 + if (h_child_inode)
18836 + isdir = !!S_ISDIR(h_child_inode->i_mode);
18837 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
18838 + flags[AuHn_CHILD] = 0;
18840 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
18841 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
18842 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
18843 + switch (mask & ALL_FSNOTIFY_DIRENT_EVENTS) {
18844 + case FS_MOVED_FROM:
18845 + case FS_MOVED_TO:
18846 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
18847 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18850 + AuDebugOn(!h_child_name);
18855 + * aufs never be able to get this child inode.
18856 + * revalidation should be in d_revalidate()
18857 + * by checking i_nlink, i_generation or d_unhashed().
18859 + AuDebugOn(!h_child_name);
18860 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
18861 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18869 + h_child_inode = NULL;
18872 + /* iput() and kfree() will be called in au_hnotify() */
18873 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
18874 + if (unlikely(!args)) {
18875 + AuErr1("no memory\n");
18879 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
18880 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
18881 + args->mask = mask;
18883 + args->h_dir = igrab(h_dir);
18884 + if (h_child_inode)
18885 + h_child_inode = igrab(h_child_inode); /* can be NULL */
18886 + args->h_child_inode = h_child_inode;
18887 + args->h_child_nlen = len;
18889 + p = (void *)args;
18890 + p += sizeof(*args);
18891 + memcpy(p, h_child_name, len);
18895 + /* NFS fires the event for silly-renamed one from kworker */
18897 + if (!dir->i_nlink
18898 + || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
18900 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
18901 + if (unlikely(err)) {
18902 + pr_err("wkq %d\n", err);
18903 + iput(args->h_child_inode);
18904 + iput(args->h_dir);
18906 + au_kfree_rcu(args);
18913 +/* ---------------------------------------------------------------------- */
18915 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
18919 + AuDebugOn(!(udba & AuOptMask_UDBA));
18922 + if (au_hnotify_op.reset_br)
18923 + err = au_hnotify_op.reset_br(udba, br, perm);
18928 +int au_hnotify_init_br(struct au_branch *br, int perm)
18933 + if (au_hnotify_op.init_br)
18934 + err = au_hnotify_op.init_br(br, perm);
18939 +void au_hnotify_fin_br(struct au_branch *br)
18941 + if (au_hnotify_op.fin_br)
18942 + au_hnotify_op.fin_br(br);
18945 +static void au_hn_destroy_cache(void)
18947 + kmem_cache_destroy(au_cache[AuCache_HNOTIFY]);
18948 + au_cache[AuCache_HNOTIFY] = NULL;
18951 +int __init au_hnotify_init(void)
18956 + au_cache[AuCache_HNOTIFY] = AuCache(au_hnotify);
18957 + if (au_cache[AuCache_HNOTIFY]) {
18959 + if (au_hnotify_op.init)
18960 + err = au_hnotify_op.init();
18961 + if (unlikely(err))
18962 + au_hn_destroy_cache();
18968 +void au_hnotify_fin(void)
18970 + if (au_hnotify_op.fin)
18971 + au_hnotify_op.fin();
18973 + /* cf. au_cache_fin() */
18974 + if (au_cache[AuCache_HNOTIFY])
18975 + au_hn_destroy_cache();
18977 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
18978 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
18979 +++ linux/fs/aufs/iinfo.c 2022-03-21 14:49:05.726633010 +0100
18981 +// SPDX-License-Identifier: GPL-2.0
18983 + * Copyright (C) 2005-2021 Junjiro R. Okajima
18985 + * This program, aufs is free software; you can redistribute it and/or modify
18986 + * it under the terms of the GNU General Public License as published by
18987 + * the Free Software Foundation; either version 2 of the License, or
18988 + * (at your option) any later version.
18990 + * This program is distributed in the hope that it will be useful,
18991 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18992 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18993 + * GNU General Public License for more details.
18995 + * You should have received a copy of the GNU General Public License
18996 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19000 + * inode private data
19005 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
19007 + struct inode *h_inode;
19008 + struct au_hinode *hinode;
19010 + IiMustAnyLock(inode);
19012 + hinode = au_hinode(au_ii(inode), bindex);
19013 + h_inode = hinode->hi_inode;
19014 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19018 +/* todo: hard/soft set? */
19019 +void au_hiput(struct au_hinode *hinode)
19021 + au_hn_free(hinode);
19022 + dput(hinode->hi_whdentry);
19023 + iput(hinode->hi_inode);
19026 +unsigned int au_hi_flags(struct inode *inode, int isdir)
19028 + unsigned int flags;
19029 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
19032 + if (au_opt_test(mnt_flags, XINO))
19033 + au_fset_hi(flags, XINO);
19034 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
19035 + au_fset_hi(flags, HNOTIFY);
19039 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
19040 + struct inode *h_inode, unsigned int flags)
19042 + struct au_hinode *hinode;
19043 + struct inode *hi;
19044 + struct au_iinfo *iinfo = au_ii(inode);
19046 + IiMustWriteLock(inode);
19048 + hinode = au_hinode(iinfo, bindex);
19049 + hi = hinode->hi_inode;
19050 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19053 + au_hiput(hinode);
19054 + hinode->hi_inode = h_inode;
19057 + struct super_block *sb = inode->i_sb;
19058 + struct au_branch *br;
19060 + AuDebugOn(inode->i_mode
19061 + && (h_inode->i_mode & S_IFMT)
19062 + != (inode->i_mode & S_IFMT));
19063 + if (bindex == iinfo->ii_btop)
19064 + au_cpup_igen(inode, h_inode);
19065 + br = au_sbr(sb, bindex);
19066 + hinode->hi_id = br->br_id;
19067 + if (au_ftest_hi(flags, XINO)) {
19068 + err = au_xino_write(sb, bindex, h_inode->i_ino,
19070 + if (unlikely(err))
19071 + AuIOErr1("failed au_xino_write() %d\n", err);
19074 + if (au_ftest_hi(flags, HNOTIFY)
19075 + && au_br_hnotifyable(br->br_perm)) {
19076 + err = au_hn_alloc(hinode, inode);
19077 + if (unlikely(err))
19078 + AuIOErr1("au_hn_alloc() %d\n", err);
19083 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
19084 + struct dentry *h_wh)
19086 + struct au_hinode *hinode;
19088 + IiMustWriteLock(inode);
19090 + hinode = au_hinode(au_ii(inode), bindex);
19091 + AuDebugOn(hinode->hi_whdentry);
19092 + hinode->hi_whdentry = h_wh;
19095 +void au_update_iigen(struct inode *inode, int half)
19097 + struct au_iinfo *iinfo;
19098 + struct au_iigen *iigen;
19099 + unsigned int sigen;
19101 + sigen = au_sigen(inode->i_sb);
19102 + iinfo = au_ii(inode);
19103 + iigen = &iinfo->ii_generation;
19104 + spin_lock(&iigen->ig_spin);
19105 + iigen->ig_generation = sigen;
19107 + au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
19109 + au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
19110 + spin_unlock(&iigen->ig_spin);
19113 +/* it may be called at remount time, too */
19114 +void au_update_ibrange(struct inode *inode, int do_put_zero)
19116 + struct au_iinfo *iinfo;
19117 + aufs_bindex_t bindex, bbot;
19119 + AuDebugOn(au_is_bad_inode(inode));
19120 + IiMustWriteLock(inode);
19122 + iinfo = au_ii(inode);
19123 + if (do_put_zero && iinfo->ii_btop >= 0) {
19124 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19126 + struct inode *h_i;
19128 + h_i = au_hinode(iinfo, bindex)->hi_inode;
19131 + && !(h_i->i_state & I_LINKABLE))
19132 + au_set_h_iptr(inode, bindex, NULL, 0);
19136 + iinfo->ii_btop = -1;
19137 + iinfo->ii_bbot = -1;
19138 + bbot = au_sbbot(inode->i_sb);
19139 + for (bindex = 0; bindex <= bbot; bindex++)
19140 + if (au_hinode(iinfo, bindex)->hi_inode) {
19141 + iinfo->ii_btop = bindex;
19144 + if (iinfo->ii_btop >= 0)
19145 + for (bindex = bbot; bindex >= iinfo->ii_btop; bindex--)
19146 + if (au_hinode(iinfo, bindex)->hi_inode) {
19147 + iinfo->ii_bbot = bindex;
19150 + AuDebugOn(iinfo->ii_btop > iinfo->ii_bbot);
19153 +/* ---------------------------------------------------------------------- */
19155 +void au_icntnr_init_once(void *_c)
19157 + struct au_icntnr *c = _c;
19158 + struct au_iinfo *iinfo = &c->iinfo;
19160 + spin_lock_init(&iinfo->ii_generation.ig_spin);
19161 + au_rw_init(&iinfo->ii_rwsem);
19162 + inode_init_once(&c->vfs_inode);
19165 +void au_hinode_init(struct au_hinode *hinode)
19167 + hinode->hi_inode = NULL;
19168 + hinode->hi_id = -1;
19169 + au_hn_init(hinode);
19170 + hinode->hi_whdentry = NULL;
19173 +int au_iinfo_init(struct inode *inode)
19175 + struct au_iinfo *iinfo;
19176 + struct super_block *sb;
19177 + struct au_hinode *hi;
19180 + sb = inode->i_sb;
19181 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19182 + nbr = au_sbbot(sb) + 1;
19183 + if (unlikely(nbr <= 0))
19185 + hi = kmalloc_array(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
19187 + au_lcnt_inc(&au_sbi(sb)->si_ninodes);
19189 + iinfo->ii_hinode = hi;
19190 + for (i = 0; i < nbr; i++, hi++)
19191 + au_hinode_init(hi);
19193 + iinfo->ii_generation.ig_generation = au_sigen(sb);
19194 + iinfo->ii_btop = -1;
19195 + iinfo->ii_bbot = -1;
19196 + iinfo->ii_vdir = NULL;
19202 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink)
19205 + struct au_hinode *hip;
19207 + AuRwMustWriteLock(&iinfo->ii_rwsem);
19210 + hip = au_krealloc(iinfo->ii_hinode, sizeof(*hip) * nbr, GFP_NOFS,
19213 + iinfo->ii_hinode = hip;
19214 + i = iinfo->ii_bbot + 1;
19216 + for (; i < nbr; i++, hip++)
19217 + au_hinode_init(hip);
19224 +void au_iinfo_fin(struct inode *inode)
19226 + struct au_iinfo *iinfo;
19227 + struct au_hinode *hi;
19228 + struct super_block *sb;
19229 + aufs_bindex_t bindex, bbot;
19230 + const unsigned char unlinked = !inode->i_nlink;
19232 + AuDebugOn(au_is_bad_inode(inode));
19234 + sb = inode->i_sb;
19235 + au_lcnt_dec(&au_sbi(sb)->si_ninodes);
19236 + if (si_pid_test(sb))
19237 + au_xino_delete_inode(inode, unlinked);
19240 + * it is safe to hide the dependency between sbinfo and
19244 + si_noflush_read_lock(sb);
19245 + au_xino_delete_inode(inode, unlinked);
19246 + si_read_unlock(sb);
19250 + iinfo = au_ii(inode);
19251 + if (iinfo->ii_vdir)
19252 + au_vdir_free(iinfo->ii_vdir);
19254 + bindex = iinfo->ii_btop;
19255 + if (bindex >= 0) {
19256 + hi = au_hinode(iinfo, bindex);
19257 + bbot = iinfo->ii_bbot;
19258 + while (bindex++ <= bbot) {
19259 + if (hi->hi_inode)
19264 + au_kfree_rcu(iinfo->ii_hinode);
19265 + AuRwDestroy(&iinfo->ii_rwsem);
19267 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
19268 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
19269 +++ linux/fs/aufs/inode.c 2022-03-21 14:49:05.726633010 +0100
19271 +// SPDX-License-Identifier: GPL-2.0
19273 + * Copyright (C) 2005-2021 Junjiro R. Okajima
19275 + * This program, aufs is free software; you can redistribute it and/or modify
19276 + * it under the terms of the GNU General Public License as published by
19277 + * the Free Software Foundation; either version 2 of the License, or
19278 + * (at your option) any later version.
19280 + * This program is distributed in the hope that it will be useful,
19281 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19282 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19283 + * GNU General Public License for more details.
19285 + * You should have received a copy of the GNU General Public License
19286 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19290 + * inode functions
19293 +#include <linux/iversion.h>
19296 +struct inode *au_igrab(struct inode *inode)
19299 + AuDebugOn(!atomic_read(&inode->i_count));
19305 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
19307 + au_cpup_attr_all(inode, /*force*/0);
19308 + au_update_iigen(inode, /*half*/1);
19310 + inode_inc_iversion(inode);
19313 +static int au_ii_refresh(struct inode *inode, int *update)
19317 + aufs_bindex_t bindex, new_bindex;
19318 + struct super_block *sb;
19319 + struct au_iinfo *iinfo;
19320 + struct au_hinode *p, *q, tmp;
19322 + AuDebugOn(au_is_bad_inode(inode));
19323 + IiMustWriteLock(inode);
19326 + sb = inode->i_sb;
19327 + nbr = au_sbbot(sb) + 1;
19328 + type = inode->i_mode & S_IFMT;
19329 + iinfo = au_ii(inode);
19330 + err = au_hinode_realloc(iinfo, nbr, /*may_shrink*/0);
19331 + if (unlikely(err))
19334 + AuDebugOn(iinfo->ii_btop < 0);
19335 + p = au_hinode(iinfo, iinfo->ii_btop);
19336 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19338 + if (!p->hi_inode)
19341 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
19342 + new_bindex = au_br_index(sb, p->hi_id);
19343 + if (new_bindex == bindex)
19346 + if (new_bindex < 0) {
19349 + p->hi_inode = NULL;
19353 + if (new_bindex < iinfo->ii_btop)
19354 + iinfo->ii_btop = new_bindex;
19355 + if (iinfo->ii_bbot < new_bindex)
19356 + iinfo->ii_bbot = new_bindex;
19357 + /* swap two lower inode, and loop again */
19358 + q = au_hinode(iinfo, new_bindex);
19362 + if (tmp.hi_inode) {
19367 + au_update_ibrange(inode, /*do_put_zero*/0);
19368 + au_hinode_realloc(iinfo, nbr, /*may_shrink*/1); /* harmless if err */
19369 + e = au_dy_irefresh(inode);
19370 + if (unlikely(e && !err))
19378 +void au_refresh_iop(struct inode *inode, int force_getattr)
19381 + struct au_sbinfo *sbi = au_sbi(inode->i_sb);
19382 + const struct inode_operations *iop
19383 + = force_getattr ? aufs_iop : sbi->si_iop_array;
19385 + if (inode->i_op == iop)
19388 + switch (inode->i_mode & S_IFMT) {
19390 + type = AuIop_DIR;
19393 + type = AuIop_SYMLINK;
19396 + type = AuIop_OTHER;
19400 + inode->i_op = iop + type;
19401 + /* unnecessary smp_wmb() */
19404 +int au_refresh_hinode_self(struct inode *inode)
19408 + err = au_ii_refresh(inode, &update);
19410 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
19416 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
19418 + int err, e, update;
19419 + unsigned int flags;
19421 + aufs_bindex_t bindex, bbot;
19422 + unsigned char isdir;
19423 + struct au_hinode *p;
19424 + struct au_iinfo *iinfo;
19426 + err = au_ii_refresh(inode, &update);
19427 + if (unlikely(err))
19431 + iinfo = au_ii(inode);
19432 + p = au_hinode(iinfo, iinfo->ii_btop);
19433 + mode = (inode->i_mode & S_IFMT);
19434 + isdir = S_ISDIR(mode);
19435 + flags = au_hi_flags(inode, isdir);
19436 + bbot = au_dbbot(dentry);
19437 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
19438 + struct inode *h_i, *h_inode;
19439 + struct dentry *h_d;
19441 + h_d = au_h_dptr(dentry, bindex);
19442 + if (!h_d || d_is_negative(h_d))
19445 + h_inode = d_inode(h_d);
19446 + AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
19447 + if (iinfo->ii_btop <= bindex && bindex <= iinfo->ii_bbot) {
19448 + h_i = au_h_iptr(inode, bindex);
19450 + if (h_i == h_inode)
19456 + if (bindex < iinfo->ii_btop)
19457 + iinfo->ii_btop = bindex;
19458 + if (iinfo->ii_bbot < bindex)
19459 + iinfo->ii_bbot = bindex;
19460 + au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
19463 + au_update_ibrange(inode, /*do_put_zero*/0);
19464 + e = au_dy_irefresh(inode);
19465 + if (unlikely(e && !err))
19468 + au_refresh_hinode_attr(inode, update && isdir);
19475 +static int set_inode(struct inode *inode, struct dentry *dentry)
19478 + unsigned int flags;
19480 + aufs_bindex_t bindex, btop, btail;
19481 + unsigned char isdir;
19482 + struct dentry *h_dentry;
19483 + struct inode *h_inode;
19484 + struct au_iinfo *iinfo;
19485 + const struct inode_operations *iop;
19487 + IiMustWriteLock(inode);
19491 + iop = au_sbi(inode->i_sb)->si_iop_array;
19492 + btop = au_dbtop(dentry);
19493 + h_dentry = au_h_dptr(dentry, btop);
19494 + h_inode = d_inode(h_dentry);
19495 + mode = h_inode->i_mode;
19496 + switch (mode & S_IFMT) {
19498 + btail = au_dbtail(dentry);
19499 + inode->i_op = iop + AuIop_OTHER;
19500 + inode->i_fop = &aufs_file_fop;
19501 + err = au_dy_iaop(inode, btop, h_inode);
19502 + if (unlikely(err))
19507 + btail = au_dbtaildir(dentry);
19508 + inode->i_op = iop + AuIop_DIR;
19509 + inode->i_fop = &aufs_dir_fop;
19512 + btail = au_dbtail(dentry);
19513 + inode->i_op = iop + AuIop_SYMLINK;
19519 + btail = au_dbtail(dentry);
19520 + inode->i_op = iop + AuIop_OTHER;
19521 + init_special_inode(inode, mode, h_inode->i_rdev);
19524 + AuIOErr("Unknown file type 0%o\n", mode);
19529 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
19530 + flags = au_hi_flags(inode, isdir);
19531 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
19532 + && au_ftest_hi(flags, HNOTIFY)
19533 + && dentry->d_name.len > AUFS_WH_PFX_LEN
19534 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
19535 + au_fclr_hi(flags, HNOTIFY);
19536 + iinfo = au_ii(inode);
19537 + iinfo->ii_btop = btop;
19538 + iinfo->ii_bbot = btail;
19539 + for (bindex = btop; bindex <= btail; bindex++) {
19540 + h_dentry = au_h_dptr(dentry, bindex);
19542 + au_set_h_iptr(inode, bindex,
19543 + au_igrab(d_inode(h_dentry)), flags);
19545 + au_cpup_attr_all(inode, /*force*/1);
19547 + * to force calling aufs_get_acl() every time,
19548 + * do not call cache_no_acl() for aufs inode.
19556 + * successful returns with iinfo write_locked
19558 + * zero: success, matched
19559 + * plus: no error, but unmatched
19561 +static int reval_inode(struct inode *inode, struct dentry *dentry)
19564 + unsigned int gen, igflags;
19565 + aufs_bindex_t bindex, bbot;
19566 + struct inode *h_inode, *h_dinode;
19567 + struct dentry *h_dentry;
19570 + * before this function, if aufs got any iinfo lock, it must be only
19571 + * one, the parent dir.
19572 + * it can happen by UDBA and the obsoleted inode number.
19575 + if (unlikely(inode->i_ino == parent_ino(dentry)))
19579 + ii_write_lock_new_child(inode);
19580 + h_dentry = au_h_dptr(dentry, au_dbtop(dentry));
19581 + h_dinode = d_inode(h_dentry);
19582 + bbot = au_ibbot(inode);
19583 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
19584 + h_inode = au_h_iptr(inode, bindex);
19585 + if (!h_inode || h_inode != h_dinode)
19589 + gen = au_iigen(inode, &igflags);
19590 + if (gen == au_digen(dentry)
19591 + && !au_ig_ftest(igflags, HALF_REFRESHED))
19594 + /* fully refresh inode using dentry */
19595 + err = au_refresh_hinode(inode, dentry);
19597 + au_update_iigen(inode, /*half*/0);
19601 + if (unlikely(err))
19602 + ii_write_unlock(inode);
19607 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19608 + unsigned int d_type, ino_t *ino)
19611 + const int isnondir = d_type != DT_DIR;
19613 + /* prevent hardlinked inode number from race condition */
19615 + err = au_xinondir_enter(sb, bindex, h_ino, &idx);
19616 + if (unlikely(err))
19620 + err = au_xino_read(sb, bindex, h_ino, ino);
19621 + if (unlikely(err))
19622 + goto out_xinondir;
19626 + *ino = au_xino_new_ino(sb);
19627 + if (unlikely(!*ino))
19628 + goto out_xinondir;
19629 + err = au_xino_write(sb, bindex, h_ino, *ino);
19630 + if (unlikely(err))
19631 + goto out_xinondir;
19635 + if (isnondir && idx >= 0)
19636 + au_xinondir_leave(sb, bindex, h_ino, idx);
19641 +/* successful returns with iinfo write_locked */
19642 +/* todo: return with unlocked? */
19643 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
19645 + struct inode *inode, *h_inode;
19646 + struct dentry *h_dentry;
19647 + struct super_block *sb;
19648 + ino_t h_ino, ino;
19649 + int err, idx, hlinked;
19650 + aufs_bindex_t btop;
19652 + sb = dentry->d_sb;
19653 + btop = au_dbtop(dentry);
19654 + h_dentry = au_h_dptr(dentry, btop);
19655 + h_inode = d_inode(h_dentry);
19656 + h_ino = h_inode->i_ino;
19657 + hlinked = !d_is_dir(h_dentry) && h_inode->i_nlink > 1;
19661 + * stop 'race'-ing between hardlinks under different
19665 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
19666 + inode = ERR_PTR(err);
19667 + if (unlikely(err))
19671 + err = au_xino_read(sb, btop, h_ino, &ino);
19672 + inode = ERR_PTR(err);
19673 + if (unlikely(err))
19674 + goto out_xinondir;
19677 + ino = au_xino_new_ino(sb);
19678 + if (unlikely(!ino)) {
19679 + inode = ERR_PTR(-EIO);
19680 + goto out_xinondir;
19684 + AuDbg("i%lu\n", (unsigned long)ino);
19685 + inode = au_iget_locked(sb, ino);
19686 + err = PTR_ERR(inode);
19687 + if (IS_ERR(inode))
19688 + goto out_xinondir;
19690 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
19691 + if (inode->i_state & I_NEW) {
19692 + ii_write_lock_new_child(inode);
19693 + err = set_inode(inode, dentry);
19695 + unlock_new_inode(inode);
19696 + goto out_xinondir; /* success */
19700 + * iget_failed() calls iput(), but we need to call
19701 + * ii_write_unlock() after iget_failed(). so dirty hack for
19704 + atomic_inc(&inode->i_count);
19705 + iget_failed(inode);
19706 + ii_write_unlock(inode);
19707 + au_xino_write(sb, btop, h_ino, /*ino*/0);
19708 + /* ignore this error */
19710 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
19712 + * horrible race condition between lookup, readdir and copyup
19713 + * (or something).
19715 + if (hlinked && idx >= 0)
19716 + au_xinondir_leave(sb, btop, h_ino, idx);
19717 + err = reval_inode(inode, dentry);
19718 + if (unlikely(err < 0)) {
19723 + goto out; /* success */
19724 + else if (hlinked && idx >= 0) {
19725 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
19726 + if (unlikely(err)) {
19728 + inode = ERR_PTR(err);
19734 + if (unlikely(au_test_fs_unique_ino(h_inode)))
19735 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
19736 + " b%d, %s, %pd, hi%lu, i%lu.\n",
19737 + btop, au_sbtype(h_dentry->d_sb), dentry,
19738 + (unsigned long)h_ino, (unsigned long)ino);
19740 + err = au_xino_write(sb, btop, h_ino, /*ino*/0);
19743 + if (hlinked && idx >= 0)
19744 + au_xinondir_leave(sb, btop, h_ino, idx);
19750 + inode = ERR_PTR(err);
19752 + if (hlinked && idx >= 0)
19753 + au_xinondir_leave(sb, btop, h_ino, idx);
19758 +/* ---------------------------------------------------------------------- */
19760 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19761 + struct inode *inode)
19764 + struct inode *hi;
19766 + err = au_br_rdonly(au_sbr(sb, bindex));
19768 + /* pseudo-link after flushed may happen out of bounds */
19771 + && au_ibtop(inode) <= bindex
19772 + && bindex <= au_ibbot(inode)) {
19774 + * permission check is unnecessary since vfsub routine
19775 + * will be called later
19777 + hi = au_h_iptr(inode, bindex);
19779 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
19785 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19788 + if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
19790 + return inode_permission(h_userns, h_inode, mask);
19793 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19796 + if (au_test_nfs(h_inode->i_sb)
19797 + && (mask & MAY_WRITE)
19798 + && S_ISDIR(h_inode->i_mode))
19799 + mask |= MAY_READ; /* force permission check */
19800 + return au_test_h_perm(h_userns, h_inode, mask);
19802 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
19803 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
19804 +++ linux/fs/aufs/inode.h 2022-03-21 14:49:05.726633010 +0100
19806 +/* SPDX-License-Identifier: GPL-2.0 */
19808 + * Copyright (C) 2005-2021 Junjiro R. Okajima
19810 + * This program, aufs is free software; you can redistribute it and/or modify
19811 + * it under the terms of the GNU General Public License as published by
19812 + * the Free Software Foundation; either version 2 of the License, or
19813 + * (at your option) any later version.
19815 + * This program is distributed in the hope that it will be useful,
19816 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19817 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19818 + * GNU General Public License for more details.
19820 + * You should have received a copy of the GNU General Public License
19821 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19825 + * inode operations
19828 +#ifndef __AUFS_INODE_H__
19829 +#define __AUFS_INODE_H__
19833 +#include <linux/fsnotify.h>
19834 +#include "rwsem.h"
19838 +struct au_hnotify {
19839 +#ifdef CONFIG_AUFS_HNOTIFY
19840 +#ifdef CONFIG_AUFS_HFSNOTIFY
19841 + /* never use fsnotify_add_vfsmount_mark() */
19842 + struct fsnotify_mark hn_mark;
19844 + struct inode *hn_aufs_inode; /* no get/put */
19845 + struct rcu_head rcu;
19847 +} ____cacheline_aligned_in_smp;
19849 +struct au_hinode {
19850 + struct inode *hi_inode;
19851 + aufs_bindex_t hi_id;
19852 +#ifdef CONFIG_AUFS_HNOTIFY
19853 + struct au_hnotify *hi_notify;
19856 + /* reference to the copied-up whiteout with get/put */
19857 + struct dentry *hi_whdentry;
19861 +#define AuIG_HALF_REFRESHED 1
19862 +#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
19863 +#define au_ig_fset(flags, name) \
19864 + do { (flags) |= AuIG_##name; } while (0)
19865 +#define au_ig_fclr(flags, name) \
19866 + do { (flags) &= ~AuIG_##name; } while (0)
19869 + spinlock_t ig_spin;
19870 + __u32 ig_generation, ig_flags;
19875 + struct au_iigen ii_generation;
19876 + struct super_block *ii_hsb1; /* no get/put */
19878 + struct au_rwsem ii_rwsem;
19879 + aufs_bindex_t ii_btop, ii_bbot;
19881 + struct au_hinode *ii_hinode;
19882 + struct au_vdir *ii_vdir;
19885 +struct au_icntnr {
19886 + struct au_iinfo iinfo;
19887 + struct inode vfs_inode;
19888 + struct hlist_bl_node plink;
19889 + struct rcu_head rcu;
19890 +} ____cacheline_aligned_in_smp;
19892 +/* au_pin flags */
19893 +#define AuPin_DI_LOCKED 1
19894 +#define AuPin_MNT_WRITE (1 << 1)
19895 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
19896 +#define au_fset_pin(flags, name) \
19897 + do { (flags) |= AuPin_##name; } while (0)
19898 +#define au_fclr_pin(flags, name) \
19899 + do { (flags) &= ~AuPin_##name; } while (0)
19903 + struct dentry *dentry;
19904 + unsigned int udba;
19905 + unsigned char lsc_di, lsc_hi, flags;
19906 + aufs_bindex_t bindex;
19909 + struct dentry *parent;
19910 + struct au_hinode *hdir;
19911 + struct vfsmount *h_mnt;
19913 + /* temporary unlock/relock for copyup */
19914 + struct dentry *h_dentry, *h_parent;
19915 + struct au_branch *br;
19916 + struct task_struct *task;
19919 +void au_pin_hdir_unlock(struct au_pin *p);
19920 +int au_pin_hdir_lock(struct au_pin *p);
19921 +int au_pin_hdir_relock(struct au_pin *p);
19922 +void au_pin_hdir_acquire_nest(struct au_pin *p);
19923 +void au_pin_hdir_release(struct au_pin *p);
19925 +/* ---------------------------------------------------------------------- */
19927 +static inline struct au_iinfo *au_ii(struct inode *inode)
19929 + BUG_ON(is_bad_inode(inode));
19930 + return &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19933 +/* ---------------------------------------------------------------------- */
19936 +struct inode *au_igrab(struct inode *inode);
19937 +void au_refresh_iop(struct inode *inode, int force_getattr);
19938 +int au_refresh_hinode_self(struct inode *inode);
19939 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
19940 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19941 + unsigned int d_type, ino_t *ino);
19942 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
19943 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19944 + struct inode *inode);
19945 +int au_test_h_perm(struct user_namespace *h_userns, struct inode *h_inode,
19947 +int au_test_h_perm_sio(struct user_namespace *h_userns, struct inode *h_inode,
19950 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
19951 + ino_t h_ino, unsigned int d_type, ino_t *ino)
19953 +#ifdef CONFIG_AUFS_SHWH
19954 + return au_ino(sb, bindex, h_ino, d_type, ino);
19967 +extern struct inode_operations aufs_iop[AuIop_Last], /* not const */
19968 + aufs_iop_nogetattr[AuIop_Last];
19970 +/* au_wr_dir flags */
19971 +#define AuWrDir_ADD_ENTRY 1
19972 +#define AuWrDir_ISDIR (1 << 1)
19973 +#define AuWrDir_TMPFILE (1 << 2)
19974 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
19975 +#define au_fset_wrdir(flags, name) \
19976 + do { (flags) |= AuWrDir_##name; } while (0)
19977 +#define au_fclr_wrdir(flags, name) \
19978 + do { (flags) &= ~AuWrDir_##name; } while (0)
19980 +struct au_wr_dir_args {
19981 + aufs_bindex_t force_btgt;
19982 + unsigned char flags;
19984 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
19985 + struct au_wr_dir_args *args);
19987 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
19988 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
19989 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
19990 + unsigned int udba, unsigned char flags);
19991 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
19992 + unsigned int udba, unsigned char flags) __must_check;
19993 +int au_do_pin(struct au_pin *pin) __must_check;
19994 +void au_unpin(struct au_pin *pin);
19995 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
19997 +#define AuIcpup_DID_CPUP 1
19998 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
19999 +#define au_fset_icpup(flags, name) \
20000 + do { (flags) |= AuIcpup_##name; } while (0)
20001 +#define au_fclr_icpup(flags, name) \
20002 + do { (flags) &= ~AuIcpup_##name; } while (0)
20004 +struct au_icpup_args {
20005 + unsigned char flags;
20006 + unsigned char pin_flags;
20007 + aufs_bindex_t btgt;
20008 + unsigned int udba;
20009 + struct au_pin pin;
20010 + struct path h_path;
20011 + struct inode *h_inode;
20014 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
20015 + struct au_icpup_args *a);
20017 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
20018 + struct path *h_path, int locked);
20021 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20022 + struct dentry *h_parent, int isdir);
20023 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
20024 + struct dentry *dentry, umode_t mode, dev_t dev);
20025 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
20026 + struct dentry *dentry, const char *symname);
20027 +int aufs_create(struct user_namespace *userns, struct inode *dir,
20028 + struct dentry *dentry, umode_t mode, bool want_excl);
20029 +struct vfsub_aopen_args;
20030 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
20031 + struct vfsub_aopen_args *args);
20032 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
20033 + struct dentry *dentry, umode_t mode);
20034 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
20035 + struct dentry *dentry);
20036 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
20037 + struct dentry *dentry, umode_t mode);
20040 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
20041 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
20042 + struct dentry *h_parent, int isdir);
20043 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
20044 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
20047 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
20048 +int aufs_rename(struct user_namespace *userns,
20049 + struct inode *_src_dir, struct dentry *_src_dentry,
20050 + struct inode *_dst_dir, struct dentry *_dst_dentry,
20051 + unsigned int _flags);
20054 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
20055 +void au_hiput(struct au_hinode *hinode);
20056 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
20057 + struct dentry *h_wh);
20058 +unsigned int au_hi_flags(struct inode *inode, int isdir);
20060 +/* hinode flags */
20061 +#define AuHi_XINO 1
20062 +#define AuHi_HNOTIFY (1 << 1)
20063 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
20064 +#define au_fset_hi(flags, name) \
20065 + do { (flags) |= AuHi_##name; } while (0)
20066 +#define au_fclr_hi(flags, name) \
20067 + do { (flags) &= ~AuHi_##name; } while (0)
20069 +#ifndef CONFIG_AUFS_HNOTIFY
20070 +#undef AuHi_HNOTIFY
20071 +#define AuHi_HNOTIFY 0
20074 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
20075 + struct inode *h_inode, unsigned int flags);
20077 +void au_update_iigen(struct inode *inode, int half);
20078 +void au_update_ibrange(struct inode *inode, int do_put_zero);
20080 +void au_icntnr_init_once(void *_c);
20081 +void au_hinode_init(struct au_hinode *hinode);
20082 +int au_iinfo_init(struct inode *inode);
20083 +void au_iinfo_fin(struct inode *inode);
20084 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink);
20086 +#ifdef CONFIG_PROC_FS
20088 +int au_plink_maint(struct super_block *sb, int flags);
20090 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
20091 +int au_plink_maint_enter(struct super_block *sb);
20092 +#ifdef CONFIG_AUFS_DEBUG
20093 +void au_plink_list(struct super_block *sb);
20095 +AuStubVoid(au_plink_list, struct super_block *sb)
20097 +int au_plink_test(struct inode *inode);
20098 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
20099 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
20100 + struct dentry *h_dentry);
20101 +void au_plink_put(struct super_block *sb, int verbose);
20102 +void au_plink_clean(struct super_block *sb, int verbose);
20103 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
20105 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
20106 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
20107 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
20108 +AuStubVoid(au_plink_list, struct super_block *sb);
20109 +AuStubInt0(au_plink_test, struct inode *inode);
20110 +AuStub(struct dentry *, au_plink_lkup, return NULL,
20111 + struct inode *inode, aufs_bindex_t bindex);
20112 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
20113 + struct dentry *h_dentry);
20114 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
20115 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
20116 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
20117 +#endif /* CONFIG_PROC_FS */
20119 +#ifdef CONFIG_AUFS_XATTR
20121 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
20122 + unsigned int verbose);
20123 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
20124 +void au_xattr_init(struct super_block *sb);
20126 +AuStubInt0(au_cpup_xattr, struct path *h_dst, struct path *h_src,
20127 + int ignore_flags, unsigned int verbose);
20128 +AuStubVoid(au_xattr_init, struct super_block *sb);
20131 +#ifdef CONFIG_FS_POSIX_ACL
20132 +struct posix_acl *aufs_get_acl(struct inode *inode, int type, bool rcu);
20133 +int aufs_set_acl(struct user_namespace *userns, struct inode *inode,
20134 + struct posix_acl *acl, int type);
20137 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
20143 +struct au_sxattr {
20147 + const char *name;
20148 + const void *value;
20153 + struct posix_acl *acl;
20158 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
20159 + struct au_sxattr *arg);
20162 +/* ---------------------------------------------------------------------- */
20164 +/* lock subclass for iinfo */
20166 + AuLsc_II_CHILD, /* child first */
20167 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
20168 + AuLsc_II_CHILD3, /* copyup dirs */
20169 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
20170 + AuLsc_II_PARENT2,
20171 + AuLsc_II_PARENT3, /* copyup dirs */
20172 + AuLsc_II_NEW_CHILD
20176 + * ii_read_lock_child, ii_write_lock_child,
20177 + * ii_read_lock_child2, ii_write_lock_child2,
20178 + * ii_read_lock_child3, ii_write_lock_child3,
20179 + * ii_read_lock_parent, ii_write_lock_parent,
20180 + * ii_read_lock_parent2, ii_write_lock_parent2,
20181 + * ii_read_lock_parent3, ii_write_lock_parent3,
20182 + * ii_read_lock_new_child, ii_write_lock_new_child,
20184 +#define AuReadLockFunc(name, lsc) \
20185 +static inline void ii_read_lock_##name(struct inode *i) \
20187 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20190 +#define AuWriteLockFunc(name, lsc) \
20191 +static inline void ii_write_lock_##name(struct inode *i) \
20193 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20196 +#define AuRWLockFuncs(name, lsc) \
20197 + AuReadLockFunc(name, lsc) \
20198 + AuWriteLockFunc(name, lsc)
20200 +AuRWLockFuncs(child, CHILD);
20201 +AuRWLockFuncs(child2, CHILD2);
20202 +AuRWLockFuncs(child3, CHILD3);
20203 +AuRWLockFuncs(parent, PARENT);
20204 +AuRWLockFuncs(parent2, PARENT2);
20205 +AuRWLockFuncs(parent3, PARENT3);
20206 +AuRWLockFuncs(new_child, NEW_CHILD);
20208 +#undef AuReadLockFunc
20209 +#undef AuWriteLockFunc
20210 +#undef AuRWLockFuncs
20212 +#define ii_read_unlock(i) au_rw_read_unlock(&au_ii(i)->ii_rwsem)
20213 +#define ii_write_unlock(i) au_rw_write_unlock(&au_ii(i)->ii_rwsem)
20214 +#define ii_downgrade_lock(i) au_rw_dgrade_lock(&au_ii(i)->ii_rwsem)
20216 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
20217 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
20218 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
20220 +/* ---------------------------------------------------------------------- */
20222 +static inline void au_icntnr_init(struct au_icntnr *c)
20224 +#ifdef CONFIG_AUFS_DEBUG
20225 + c->vfs_inode.i_mode = 0;
20229 +static inline unsigned int au_iigen(struct inode *inode, unsigned int *igflags)
20231 + unsigned int gen;
20232 + struct au_iinfo *iinfo;
20233 + struct au_iigen *iigen;
20235 + iinfo = au_ii(inode);
20236 + iigen = &iinfo->ii_generation;
20237 + spin_lock(&iigen->ig_spin);
20239 + *igflags = iigen->ig_flags;
20240 + gen = iigen->ig_generation;
20241 + spin_unlock(&iigen->ig_spin);
20246 +/* tiny test for inode number */
20247 +/* tmpfs generation is too rough */
20248 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
20250 + struct au_iinfo *iinfo;
20252 + iinfo = au_ii(inode);
20253 + AuRwMustAnyLock(&iinfo->ii_rwsem);
20254 + return !(iinfo->ii_hsb1 == h_inode->i_sb
20255 + && iinfo->ii_higen == h_inode->i_generation);
20258 +static inline void au_iigen_dec(struct inode *inode)
20260 + struct au_iinfo *iinfo;
20261 + struct au_iigen *iigen;
20263 + iinfo = au_ii(inode);
20264 + iigen = &iinfo->ii_generation;
20265 + spin_lock(&iigen->ig_spin);
20266 + iigen->ig_generation--;
20267 + spin_unlock(&iigen->ig_spin);
20270 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
20275 + if (unlikely(inode && au_iigen(inode, NULL) != sigen))
20281 +/* ---------------------------------------------------------------------- */
20283 +static inline struct au_hinode *au_hinode(struct au_iinfo *iinfo,
20284 + aufs_bindex_t bindex)
20286 + return iinfo->ii_hinode + bindex;
20289 +static inline int au_is_bad_inode(struct inode *inode)
20291 + return !!(is_bad_inode(inode) || !au_hinode(au_ii(inode), 0));
20294 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
20295 + aufs_bindex_t bindex)
20297 + IiMustAnyLock(inode);
20298 + return au_hinode(au_ii(inode), bindex)->hi_id;
20301 +static inline aufs_bindex_t au_ibtop(struct inode *inode)
20303 + IiMustAnyLock(inode);
20304 + return au_ii(inode)->ii_btop;
20307 +static inline aufs_bindex_t au_ibbot(struct inode *inode)
20309 + IiMustAnyLock(inode);
20310 + return au_ii(inode)->ii_bbot;
20313 +static inline struct au_vdir *au_ivdir(struct inode *inode)
20315 + IiMustAnyLock(inode);
20316 + return au_ii(inode)->ii_vdir;
20319 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
20321 + IiMustAnyLock(inode);
20322 + return au_hinode(au_ii(inode), bindex)->hi_whdentry;
20325 +static inline void au_set_ibtop(struct inode *inode, aufs_bindex_t bindex)
20327 + IiMustWriteLock(inode);
20328 + au_ii(inode)->ii_btop = bindex;
20331 +static inline void au_set_ibbot(struct inode *inode, aufs_bindex_t bindex)
20333 + IiMustWriteLock(inode);
20334 + au_ii(inode)->ii_bbot = bindex;
20337 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
20339 + IiMustWriteLock(inode);
20340 + au_ii(inode)->ii_vdir = vdir;
20343 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
20345 + IiMustAnyLock(inode);
20346 + return au_hinode(au_ii(inode), bindex);
20349 +/* ---------------------------------------------------------------------- */
20351 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
20354 + return pin->parent;
20358 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
20360 + if (pin && pin->hdir)
20361 + return pin->hdir->hi_inode;
20365 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
20368 + return pin->hdir;
20372 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
20375 + pin->dentry = dentry;
20378 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
20379 + unsigned char lflag)
20383 + au_fset_pin(pin->flags, DI_LOCKED);
20385 + au_fclr_pin(pin->flags, DI_LOCKED);
20389 +#if 0 /* reserved */
20390 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
20393 + dput(pin->parent);
20394 + pin->parent = dget(parent);
20399 +/* ---------------------------------------------------------------------- */
20402 +#ifdef CONFIG_AUFS_HNOTIFY
20403 +struct au_hnotify_op {
20404 + void (*ctl)(struct au_hinode *hinode, int do_set);
20405 + int (*alloc)(struct au_hinode *hinode);
20408 + * if it returns true, the caller should free hinode->hi_notify,
20409 + * otherwise ->free() frees it.
20411 + int (*free)(struct au_hinode *hinode,
20412 + struct au_hnotify *hn) __must_check;
20414 + void (*fin)(void);
20415 + int (*init)(void);
20417 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
20418 + void (*fin_br)(struct au_branch *br);
20419 + int (*init_br)(struct au_branch *br, int perm);
20423 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
20424 +void au_hn_free(struct au_hinode *hinode);
20425 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
20426 +void au_hn_reset(struct inode *inode, unsigned int flags);
20427 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
20428 + const struct qstr *h_child_qstr, struct inode *h_child_inode);
20429 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
20430 +int au_hnotify_init_br(struct au_branch *br, int perm);
20431 +void au_hnotify_fin_br(struct au_branch *br);
20432 +int __init au_hnotify_init(void);
20433 +void au_hnotify_fin(void);
20436 +extern const struct au_hnotify_op au_hnotify_op;
20439 +void au_hn_init(struct au_hinode *hinode)
20441 + hinode->hi_notify = NULL;
20444 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
20446 + return hinode->hi_notify;
20450 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
20451 + struct au_hinode *hinode __maybe_unused,
20452 + struct inode *inode __maybe_unused)
20453 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
20454 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
20455 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
20456 + int do_set __maybe_unused)
20457 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
20458 + unsigned int flags __maybe_unused)
20459 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
20460 + struct au_branch *br __maybe_unused,
20461 + int perm __maybe_unused)
20462 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
20463 + int perm __maybe_unused)
20464 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
20465 +AuStubInt0(__init au_hnotify_init, void)
20466 +AuStubVoid(au_hnotify_fin, void)
20467 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
20468 +#endif /* CONFIG_AUFS_HNOTIFY */
20470 +static inline void au_hn_suspend(struct au_hinode *hdir)
20472 + au_hn_ctl(hdir, /*do_set*/0);
20475 +static inline void au_hn_resume(struct au_hinode *hdir)
20477 + au_hn_ctl(hdir, /*do_set*/1);
20480 +static inline void au_hn_inode_lock(struct au_hinode *hdir)
20482 + inode_lock(hdir->hi_inode);
20483 + au_hn_suspend(hdir);
20486 +static inline void au_hn_inode_lock_nested(struct au_hinode *hdir,
20487 + unsigned int sc __maybe_unused)
20489 + inode_lock_nested(hdir->hi_inode, sc);
20490 + au_hn_suspend(hdir);
20493 +#if 0 /* unused */
20494 +#include "vfsub.h"
20495 +static inline void au_hn_inode_lock_shared_nested(struct au_hinode *hdir,
20498 + inode_lock_shared_nested(hdir->hi_inode, sc);
20499 + au_hn_suspend(hdir);
20503 +static inline void au_hn_inode_unlock(struct au_hinode *hdir)
20505 + au_hn_resume(hdir);
20506 + inode_unlock(hdir->hi_inode);
20509 +#endif /* __KERNEL__ */
20510 +#endif /* __AUFS_INODE_H__ */
20511 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
20512 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
20513 +++ linux/fs/aufs/ioctl.c 2022-03-21 14:49:05.726633010 +0100
20515 +// SPDX-License-Identifier: GPL-2.0
20517 + * Copyright (C) 2005-2021 Junjiro R. Okajima
20519 + * This program, aufs is free software; you can redistribute it and/or modify
20520 + * it under the terms of the GNU General Public License as published by
20521 + * the Free Software Foundation; either version 2 of the License, or
20522 + * (at your option) any later version.
20524 + * This program is distributed in the hope that it will be useful,
20525 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20526 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20527 + * GNU General Public License for more details.
20529 + * You should have received a copy of the GNU General Public License
20530 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20535 + * plink-management and readdir in userspace.
20536 + * assist the pathconf(3) wrapper library.
20538 + * File-based Hierarchical Storage Management.
20541 +#include <linux/compat.h>
20542 +#include <linux/file.h>
20545 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
20548 + aufs_bindex_t wbi, bindex, bbot;
20549 + struct file *h_file;
20550 + struct super_block *sb;
20551 + struct dentry *root;
20552 + struct au_branch *br;
20553 + struct aufs_wbr_fd wbrfd = {
20554 + .oflags = au_dir_roflags,
20557 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
20558 + | O_NOATIME | O_CLOEXEC;
20560 + AuDebugOn(wbrfd.oflags & ~valid);
20563 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
20564 + if (unlikely(err)) {
20570 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
20571 + wbrfd.oflags |= au_dir_roflags;
20572 + AuDbg("0%o\n", wbrfd.oflags);
20573 + if (unlikely(wbrfd.oflags & ~valid))
20577 + fd = get_unused_fd_flags(0);
20579 + if (unlikely(fd < 0))
20582 + h_file = ERR_PTR(-EINVAL);
20585 + sb = path->dentry->d_sb;
20586 + root = sb->s_root;
20587 + aufs_read_lock(root, AuLock_IR);
20588 + bbot = au_sbbot(sb);
20589 + if (wbrfd.brid >= 0) {
20590 + wbi = au_br_index(sb, wbrfd.brid);
20591 + if (unlikely(wbi < 0 || wbi > bbot))
20595 + h_file = ERR_PTR(-ENOENT);
20596 + br = au_sbr(sb, wbi);
20597 + if (!au_br_writable(br->br_perm)) {
20601 + bindex = wbi + 1;
20603 + for (; bindex <= bbot; bindex++) {
20604 + br = au_sbr(sb, bindex);
20605 + if (au_br_writable(br->br_perm)) {
20607 + br = au_sbr(sb, wbi);
20612 + AuDbg("wbi %d\n", wbi);
20614 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
20618 + aufs_read_unlock(root, AuLock_IR);
20619 + err = PTR_ERR(h_file);
20620 + if (IS_ERR(h_file))
20623 + au_lcnt_dec(&br->br_nfiles); /* cf. au_h_open() */
20624 + fd_install(fd, h_file);
20626 + goto out; /* success */
20629 + put_unused_fd(fd);
20635 +/* ---------------------------------------------------------------------- */
20637 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
20640 + struct dentry *dentry;
20643 + case AUFS_CTL_RDU:
20644 + case AUFS_CTL_RDU_INO:
20645 + err = au_rdu_ioctl(file, cmd, arg);
20648 + case AUFS_CTL_WBR_FD:
20649 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
20652 + case AUFS_CTL_IBUSY:
20653 + err = au_ibusy_ioctl(file, arg);
20656 + case AUFS_CTL_BRINFO:
20657 + err = au_brinfo_ioctl(file, arg);
20660 + case AUFS_CTL_FHSM_FD:
20661 + dentry = file->f_path.dentry;
20662 + if (IS_ROOT(dentry))
20663 + err = au_fhsm_fd(dentry->d_sb, arg);
20669 + /* do not call the lower */
20670 + AuDbg("0x%x\n", cmd);
20678 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
20683 + case AUFS_CTL_MVDOWN:
20684 + err = au_mvdown(file->f_path.dentry, (void __user *)arg);
20687 + case AUFS_CTL_WBR_FD:
20688 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
20692 + /* do not call the lower */
20693 + AuDbg("0x%x\n", cmd);
20701 +#ifdef CONFIG_COMPAT
20702 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
20703 + unsigned long arg)
20708 + case AUFS_CTL_RDU:
20709 + case AUFS_CTL_RDU_INO:
20710 + err = au_rdu_compat_ioctl(file, cmd, arg);
20713 + case AUFS_CTL_IBUSY:
20714 + err = au_ibusy_compat_ioctl(file, arg);
20717 + case AUFS_CTL_BRINFO:
20718 + err = au_brinfo_compat_ioctl(file, arg);
20722 + err = aufs_ioctl_dir(file, cmd, arg);
20729 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
20730 + unsigned long arg)
20732 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
20735 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
20736 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
20737 +++ linux/fs/aufs/i_op_add.c 2022-03-21 14:49:05.726633010 +0100
20739 +// SPDX-License-Identifier: GPL-2.0
20741 + * Copyright (C) 2005-2021 Junjiro R. Okajima
20743 + * This program, aufs is free software; you can redistribute it and/or modify
20744 + * it under the terms of the GNU General Public License as published by
20745 + * the Free Software Foundation; either version 2 of the License, or
20746 + * (at your option) any later version.
20748 + * This program is distributed in the hope that it will be useful,
20749 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20750 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20751 + * GNU General Public License for more details.
20753 + * You should have received a copy of the GNU General Public License
20754 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20758 + * inode operations (add entry)
20761 +#include <linux/iversion.h>
20765 + * final procedure of adding a new entry, except link(2).
20766 + * remove whiteout, instantiate, copyup the parent dir's times and size
20767 + * and update version.
20768 + * if it failed, re-create the removed whiteout.
20770 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
20771 + struct dentry *wh_dentry, struct dentry *dentry)
20774 + aufs_bindex_t bwh;
20775 + struct path h_path;
20776 + struct super_block *sb;
20777 + struct inode *inode, *h_dir;
20778 + struct dentry *wh;
20783 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
20784 + IMustLock(h_dir);
20785 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
20786 + bwh = au_dbwh(dentry);
20787 + h_path.dentry = wh_dentry;
20788 + h_path.mnt = au_sbr_mnt(sb, bindex);
20789 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
20791 + if (unlikely(err))
20795 + inode = au_new_inode(dentry, /*must_new*/1);
20796 + if (!IS_ERR(inode)) {
20797 + d_instantiate(dentry, inode);
20798 + dir = d_inode(dentry->d_parent); /* dir inode is locked */
20800 + au_dir_ts(dir, bindex);
20801 + inode_inc_iversion(dir);
20802 + au_fhsm_wrote(sb, bindex, /*force*/0);
20803 + return 0; /* success */
20806 + err = PTR_ERR(inode);
20811 + /* dir inode is locked */
20812 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
20813 + rerr = PTR_ERR(wh);
20814 + if (IS_ERR(wh)) {
20815 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
20816 + dentry, err, rerr);
20825 +static int au_d_may_add(struct dentry *dentry)
20830 + if (unlikely(d_unhashed(dentry)))
20832 + if (unlikely(d_really_is_positive(dentry)))
20838 + * simple tests for the adding inode operations.
20839 + * following the checks in vfs, plus the parent-child relationship.
20841 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20842 + struct dentry *h_parent, int isdir)
20846 + struct dentry *h_dentry;
20847 + struct inode *h_inode;
20849 + err = -ENAMETOOLONG;
20850 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20853 + h_dentry = au_h_dptr(dentry, bindex);
20854 + if (d_really_is_negative(dentry)) {
20856 + if (unlikely(d_is_positive(h_dentry)))
20859 + /* rename(2) case */
20861 + if (unlikely(d_is_negative(h_dentry)))
20863 + h_inode = d_inode(h_dentry);
20864 + if (unlikely(!h_inode->i_nlink))
20867 + h_mode = h_inode->i_mode;
20870 + if (unlikely(S_ISDIR(h_mode)))
20872 + } else if (unlikely(!S_ISDIR(h_mode))) {
20879 + /* expected parent dir is locked */
20880 + if (unlikely(h_parent != h_dentry->d_parent))
20889 + * initial procedure of adding a new entry.
20890 + * prepare writable branch and the parent dir, lock it,
20891 + * and lookup whiteout for the new entry.
20893 +static struct dentry*
20894 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
20895 + struct dentry *src_dentry, struct au_pin *pin,
20896 + struct au_wr_dir_args *wr_dir_args)
20898 + struct dentry *wh_dentry, *h_parent;
20899 + struct super_block *sb;
20900 + struct au_branch *br;
20902 + unsigned int udba;
20903 + aufs_bindex_t bcpup;
20905 + AuDbg("%pd\n", dentry);
20907 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
20909 + wh_dentry = ERR_PTR(err);
20910 + if (unlikely(err < 0))
20913 + sb = dentry->d_sb;
20914 + udba = au_opt_udba(sb);
20915 + err = au_pin(pin, dentry, bcpup, udba,
20916 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20917 + wh_dentry = ERR_PTR(err);
20918 + if (unlikely(err))
20921 + h_parent = au_pinned_h_parent(pin);
20922 + if (udba != AuOpt_UDBA_NONE
20923 + && au_dbtop(dentry) == bcpup)
20924 + err = au_may_add(dentry, bcpup, h_parent,
20925 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
20926 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20927 + err = -ENAMETOOLONG;
20928 + wh_dentry = ERR_PTR(err);
20929 + if (unlikely(err))
20932 + br = au_sbr(sb, bcpup);
20934 + struct path tmp = {
20935 + .dentry = h_parent,
20936 + .mnt = au_br_mnt(br)
20938 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
20941 + wh_dentry = NULL;
20942 + if (bcpup != au_dbwh(dentry))
20943 + goto out; /* success */
20946 + * ENAMETOOLONG here means that if we allowed create such name, then it
20947 + * would not be able to removed in the future. So we don't allow such
20948 + * name here and we don't handle ENAMETOOLONG differently here.
20950 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
20953 + if (IS_ERR(wh_dentry))
20956 + return wh_dentry;
20959 +/* ---------------------------------------------------------------------- */
20961 +enum { Mknod, Symlink, Creat };
20962 +struct simple_arg {
20969 + struct vfsub_aopen_args *aopen;
20972 + const char *symname;
20981 +static int add_simple(struct inode *dir, struct dentry *dentry,
20982 + struct simple_arg *arg)
20985 + aufs_bindex_t btop;
20986 + unsigned char created;
20987 + const unsigned char try_aopen
20988 + = (arg->type == Creat && arg->u.c.try_aopen);
20989 + struct vfsub_aopen_args *aopen = arg->u.c.aopen;
20990 + struct dentry *wh_dentry, *parent;
20991 + struct inode *h_dir;
20992 + struct super_block *sb;
20993 + struct au_branch *br;
20994 + /* to reduce stack size */
20996 + struct au_dtime dt;
20997 + struct au_pin pin;
20998 + struct path h_path;
20999 + struct au_wr_dir_args wr_dir_args;
21002 + AuDbg("%pd\n", dentry);
21006 + a = kmalloc(sizeof(*a), GFP_NOFS);
21007 + if (unlikely(!a))
21009 + a->wr_dir_args.force_btgt = -1;
21010 + a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
21012 + parent = dentry->d_parent; /* dir inode is locked */
21013 + if (!try_aopen) {
21014 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21015 + if (unlikely(err))
21018 + err = au_d_may_add(dentry);
21019 + if (unlikely(err))
21022 + di_write_lock_parent(parent);
21023 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21024 + &a->pin, &a->wr_dir_args);
21025 + err = PTR_ERR(wh_dentry);
21026 + if (IS_ERR(wh_dentry))
21029 + btop = au_dbtop(dentry);
21030 + sb = dentry->d_sb;
21031 + br = au_sbr(sb, btop);
21032 + a->h_path.dentry = au_h_dptr(dentry, btop);
21033 + a->h_path.mnt = au_br_mnt(br);
21034 + h_dir = au_pinned_h_dir(&a->pin);
21035 + switch (arg->type) {
21037 + if (!try_aopen || !h_dir->i_op->atomic_open) {
21038 + err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
21039 + arg->u.c.want_excl);
21041 + if (!err && try_aopen)
21042 + aopen->file->f_mode |= FMODE_CREATED;
21045 + err = vfsub_atomic_open(h_dir, a->h_path.dentry, aopen);
21046 + AuDbg("err %d\n", err);
21047 + AuDbgFile(aopen->file);
21048 + created = err >= 0
21049 + && !!(aopen->file->f_mode & FMODE_CREATED);
21053 + err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
21057 + err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
21064 + if (unlikely(err < 0))
21067 + err = epilog(dir, btop, wh_dentry, dentry);
21069 + goto out_unpin; /* success */
21072 + if (created /* && d_is_positive(a->h_path.dentry) */) {
21073 + /* no delegation since it is just created */
21074 + rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
21077 + AuIOErr("%pd revert failure(%d, %d)\n",
21078 + dentry, err, rerr);
21081 + au_dtime_revert(&a->dt);
21083 + if (try_aopen && h_dir->i_op->atomic_open
21084 + && (aopen->file->f_mode & FMODE_OPENED))
21085 + /* aopen->file is still opened */
21086 + au_lcnt_dec(&aopen->br->br_nfiles);
21089 + au_unpin(&a->pin);
21093 + di_write_unlock(parent);
21095 + if (unlikely(err)) {
21096 + au_update_dbtop(dentry);
21100 + aufs_read_unlock(dentry, AuLock_DW);
21107 +int aufs_mknod(struct user_namespace *userns, struct inode *dir,
21108 + struct dentry *dentry, umode_t mode, dev_t dev)
21110 + struct simple_arg arg = {
21117 + return add_simple(dir, dentry, &arg);
21120 +int aufs_symlink(struct user_namespace *userns, struct inode *dir,
21121 + struct dentry *dentry, const char *symname)
21123 + struct simple_arg arg = {
21125 + .u.s.symname = symname
21127 + return add_simple(dir, dentry, &arg);
21130 +int aufs_create(struct user_namespace *userns, struct inode *dir,
21131 + struct dentry *dentry, umode_t mode, bool want_excl)
21133 + struct simple_arg arg = {
21137 + .want_excl = want_excl
21140 + return add_simple(dir, dentry, &arg);
21143 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
21144 + struct vfsub_aopen_args *aopen_args)
21146 + struct simple_arg arg = {
21149 + .mode = aopen_args->create_mode,
21150 + .want_excl = aopen_args->open_flag & O_EXCL,
21151 + .try_aopen = true,
21152 + .aopen = aopen_args
21155 + return add_simple(dir, dentry, &arg);
21158 +int aufs_tmpfile(struct user_namespace *userns, struct inode *dir,
21159 + struct dentry *dentry, umode_t mode)
21162 + aufs_bindex_t bindex;
21163 + struct super_block *sb;
21164 + struct dentry *parent, *h_parent, *h_dentry;
21165 + struct inode *h_dir, *inode;
21166 + struct vfsmount *h_mnt;
21167 + struct user_namespace *h_userns;
21168 + struct au_wr_dir_args wr_dir_args = {
21169 + .force_btgt = -1,
21170 + .flags = AuWrDir_TMPFILE
21173 + /* copy-up may happen */
21177 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21178 + if (unlikely(err))
21181 + err = au_di_init(dentry);
21182 + if (unlikely(err))
21186 + parent = d_find_any_alias(dir);
21187 + AuDebugOn(!parent);
21188 + di_write_lock_parent(parent);
21189 + if (unlikely(d_inode(parent) != dir))
21192 + err = au_digen_test(parent, au_sigen(sb));
21193 + if (unlikely(err))
21196 + bindex = au_dbtop(parent);
21197 + au_set_dbtop(dentry, bindex);
21198 + au_set_dbbot(dentry, bindex);
21199 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
21201 + if (unlikely(err < 0))
21204 + err = -EOPNOTSUPP;
21205 + h_dir = au_h_iptr(dir, bindex);
21206 + if (unlikely(!h_dir->i_op->tmpfile))
21209 + h_mnt = au_sbr_mnt(sb, bindex);
21210 + err = vfsub_mnt_want_write(h_mnt);
21211 + if (unlikely(err))
21214 + h_userns = mnt_user_ns(h_mnt);
21215 + h_parent = au_h_dptr(parent, bindex);
21216 + h_dentry = vfs_tmpfile(h_userns, h_parent, mode, /*open_flag*/0);
21217 + if (IS_ERR(h_dentry)) {
21218 + err = PTR_ERR(h_dentry);
21222 + au_set_dbtop(dentry, bindex);
21223 + au_set_dbbot(dentry, bindex);
21224 + au_set_h_dptr(dentry, bindex, dget(h_dentry));
21225 + inode = au_new_inode(dentry, /*must_new*/1);
21226 + if (IS_ERR(inode)) {
21227 + err = PTR_ERR(inode);
21228 + au_set_h_dptr(dentry, bindex, NULL);
21229 + au_set_dbtop(dentry, -1);
21230 + au_set_dbbot(dentry, -1);
21232 + if (!inode->i_nlink)
21233 + set_nlink(inode, 1);
21234 + d_tmpfile(dentry, inode);
21235 + au_di(dentry)->di_tmpfile = 1;
21237 + /* update without i_mutex */
21238 + if (au_ibtop(dir) == au_dbtop(dentry))
21239 + au_cpup_attr_timesizes(dir);
21244 + vfsub_mnt_drop_write(h_mnt);
21246 + di_write_unlock(parent);
21248 + di_write_unlock(dentry);
21249 + if (unlikely(err)) {
21250 + au_di_fin(dentry);
21251 + dentry->d_fsdata = NULL;
21254 + si_read_unlock(sb);
21256 + inode_unlock(dir);
21260 +/* ---------------------------------------------------------------------- */
21262 +struct au_link_args {
21263 + aufs_bindex_t bdst, bsrc;
21264 + struct au_pin pin;
21265 + struct path h_path;
21266 + struct dentry *src_parent, *parent;
21269 +static int au_cpup_before_link(struct dentry *src_dentry,
21270 + struct au_link_args *a)
21273 + struct dentry *h_src_dentry;
21274 + struct au_cp_generic cpg = {
21275 + .dentry = src_dentry,
21280 + .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
21283 + di_read_lock_parent(a->src_parent, AuLock_IR);
21284 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
21285 + if (unlikely(err))
21288 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
21289 + err = au_pin(&a->pin, src_dentry, a->bdst,
21290 + au_opt_udba(src_dentry->d_sb),
21291 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21292 + if (unlikely(err))
21295 + err = au_sio_cpup_simple(&cpg);
21296 + au_unpin(&a->pin);
21299 + di_read_unlock(a->src_parent, AuLock_IR);
21303 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
21304 + struct au_link_args *a)
21307 + unsigned char plink;
21308 + aufs_bindex_t bbot;
21309 + struct dentry *h_src_dentry;
21310 + struct inode *h_inode, *inode, *delegated;
21311 + struct super_block *sb;
21312 + struct file *h_file;
21316 + sb = src_dentry->d_sb;
21317 + inode = d_inode(src_dentry);
21318 + if (au_ibtop(inode) <= a->bdst)
21319 + h_inode = au_h_iptr(inode, a->bdst);
21320 + if (!h_inode || !h_inode->i_nlink) {
21321 + /* copyup src_dentry as the name of dentry. */
21322 + bbot = au_dbbot(dentry);
21323 + if (bbot < a->bsrc)
21324 + au_set_dbbot(dentry, a->bsrc);
21325 + au_set_h_dptr(dentry, a->bsrc,
21326 + dget(au_h_dptr(src_dentry, a->bsrc)));
21327 + dget(a->h_path.dentry);
21328 + au_set_h_dptr(dentry, a->bdst, NULL);
21329 + AuDbg("temporary d_inode...\n");
21330 + spin_lock(&dentry->d_lock);
21331 + dentry->d_inode = d_inode(src_dentry); /* tmp */
21332 + spin_unlock(&dentry->d_lock);
21333 + h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
21334 + if (IS_ERR(h_file))
21335 + err = PTR_ERR(h_file);
21337 + struct au_cp_generic cpg = {
21338 + .dentry = dentry,
21343 + .flags = AuCpup_KEEPLINO
21345 + err = au_sio_cpup_simple(&cpg);
21346 + au_h_open_post(dentry, a->bsrc, h_file);
21348 + dput(a->h_path.dentry);
21349 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21351 + au_set_h_dptr(dentry, a->bdst,
21352 + a->h_path.dentry);
21354 + spin_lock(&dentry->d_lock);
21355 + dentry->d_inode = NULL; /* restore */
21356 + spin_unlock(&dentry->d_lock);
21357 + AuDbg("temporary d_inode...done\n");
21358 + au_set_h_dptr(dentry, a->bsrc, NULL);
21359 + au_set_dbbot(dentry, bbot);
21361 + /* the inode of src_dentry already exists on a.bdst branch */
21362 + h_src_dentry = d_find_alias(h_inode);
21363 + if (!h_src_dentry && au_plink_test(inode)) {
21365 + h_src_dentry = au_plink_lkup(inode, a->bdst);
21366 + err = PTR_ERR(h_src_dentry);
21367 + if (IS_ERR(h_src_dentry))
21370 + if (unlikely(d_is_negative(h_src_dentry))) {
21371 + dput(h_src_dentry);
21372 + h_src_dentry = NULL;
21376 + if (h_src_dentry) {
21377 + delegated = NULL;
21378 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21379 + &a->h_path, &delegated);
21380 + if (unlikely(err == -EWOULDBLOCK)) {
21381 + pr_warn("cannot retry for NFSv4 delegation"
21382 + " for an internal link\n");
21385 + dput(h_src_dentry);
21387 + AuIOErr("no dentry found for hi%lu on b%d\n",
21388 + h_inode->i_ino, a->bdst);
21393 + if (!err && !plink)
21394 + au_plink_append(inode, a->bdst, a->h_path.dentry);
21401 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
21402 + struct dentry *dentry)
21405 + struct au_dtime dt;
21406 + struct au_link_args *a;
21407 + struct dentry *wh_dentry, *h_src_dentry;
21408 + struct inode *inode, *delegated;
21409 + struct super_block *sb;
21410 + struct au_wr_dir_args wr_dir_args = {
21411 + /* .force_btgt = -1, */
21412 + .flags = AuWrDir_ADD_ENTRY
21416 + inode = d_inode(src_dentry);
21417 + IMustLock(inode);
21420 + a = kzalloc(sizeof(*a), GFP_NOFS);
21421 + if (unlikely(!a))
21424 + a->parent = dentry->d_parent; /* dir inode is locked */
21425 + err = aufs_read_and_write_lock2(dentry, src_dentry,
21426 + AuLock_NOPLM | AuLock_GEN);
21427 + if (unlikely(err))
21429 + err = au_d_linkable(src_dentry);
21430 + if (unlikely(err))
21432 + err = au_d_may_add(dentry);
21433 + if (unlikely(err))
21436 + a->src_parent = dget_parent(src_dentry);
21437 + wr_dir_args.force_btgt = au_ibtop(inode);
21439 + di_write_lock_parent(a->parent);
21440 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
21441 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
21443 + err = PTR_ERR(wh_dentry);
21444 + if (IS_ERR(wh_dentry))
21448 + sb = dentry->d_sb;
21449 + a->bdst = au_dbtop(dentry);
21450 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21451 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
21452 + a->bsrc = au_ibtop(inode);
21453 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21454 + if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
21455 + h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
21456 + if (!h_src_dentry) {
21457 + a->bsrc = au_dbtop(src_dentry);
21458 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21459 + AuDebugOn(!h_src_dentry);
21460 + } else if (IS_ERR(h_src_dentry)) {
21461 + err = PTR_ERR(h_src_dentry);
21466 + * aufs doesn't touch the credential so
21467 + * security_dentry_create_files_as() is unnecessary.
21469 + if (au_opt_test(au_mntflags(sb), PLINK)) {
21470 + if (a->bdst < a->bsrc
21471 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
21472 + err = au_cpup_or_link(src_dentry, dentry, a);
21474 + delegated = NULL;
21475 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21476 + &a->h_path, &delegated);
21477 + if (unlikely(err == -EWOULDBLOCK)) {
21478 + pr_warn("cannot retry for NFSv4 delegation"
21479 + " for an internal link\n");
21483 + dput(h_src_dentry);
21486 + * copyup src_dentry to the branch we process,
21487 + * and then link(2) to it.
21489 + dput(h_src_dentry);
21490 + if (a->bdst < a->bsrc
21491 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
21492 + au_unpin(&a->pin);
21493 + di_write_unlock(a->parent);
21494 + err = au_cpup_before_link(src_dentry, a);
21495 + di_write_lock_parent(a->parent);
21497 + err = au_pin(&a->pin, dentry, a->bdst,
21499 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21500 + if (unlikely(err))
21504 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
21506 + if (h_src_dentry && d_is_positive(h_src_dentry)) {
21507 + delegated = NULL;
21508 + err = vfsub_link(h_src_dentry,
21509 + au_pinned_h_dir(&a->pin),
21510 + &a->h_path, &delegated);
21511 + if (unlikely(err == -EWOULDBLOCK)) {
21512 + pr_warn("cannot retry"
21513 + " for NFSv4 delegation"
21514 + " for an internal link\n");
21520 + if (unlikely(err))
21524 + a->h_path.dentry = wh_dentry;
21525 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
21527 + if (unlikely(err))
21531 + au_dir_ts(dir, a->bdst);
21532 + inode_inc_iversion(dir);
21533 + inc_nlink(inode);
21534 + inode->i_ctime = dir->i_ctime;
21535 + d_instantiate(dentry, au_igrab(inode));
21536 + if (d_unhashed(a->h_path.dentry))
21537 + /* some filesystem calls d_drop() */
21539 + /* some filesystems consume an inode even hardlink */
21540 + au_fhsm_wrote(sb, a->bdst, /*force*/0);
21541 + goto out_unpin; /* success */
21544 + /* no delegation since it is just created */
21545 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
21546 + /*delegated*/NULL, /*force*/0);
21547 + if (unlikely(rerr)) {
21548 + AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
21551 + au_dtime_revert(&dt);
21553 + au_unpin(&a->pin);
21557 + di_write_unlock(a->parent);
21558 + dput(a->src_parent);
21560 + if (unlikely(err)) {
21561 + au_update_dbtop(dentry);
21564 + aufs_read_and_write_unlock2(dentry, src_dentry);
21572 +int aufs_mkdir(struct user_namespace *userns, struct inode *dir,
21573 + struct dentry *dentry, umode_t mode)
21576 + aufs_bindex_t bindex;
21577 + unsigned char diropq;
21578 + struct path h_path;
21579 + struct dentry *wh_dentry, *parent, *opq_dentry;
21580 + struct inode *h_inode;
21581 + struct super_block *sb;
21583 + struct au_pin pin;
21584 + struct au_dtime dt;
21585 + } *a; /* reduce the stack usage */
21586 + struct au_wr_dir_args wr_dir_args = {
21587 + .force_btgt = -1,
21588 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
21594 + a = kmalloc(sizeof(*a), GFP_NOFS);
21595 + if (unlikely(!a))
21598 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21599 + if (unlikely(err))
21601 + err = au_d_may_add(dentry);
21602 + if (unlikely(err))
21605 + parent = dentry->d_parent; /* dir inode is locked */
21606 + di_write_lock_parent(parent);
21607 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21608 + &a->pin, &wr_dir_args);
21609 + err = PTR_ERR(wh_dentry);
21610 + if (IS_ERR(wh_dentry))
21613 + sb = dentry->d_sb;
21614 + bindex = au_dbtop(dentry);
21615 + h_path.dentry = au_h_dptr(dentry, bindex);
21616 + h_path.mnt = au_sbr_mnt(sb, bindex);
21617 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
21618 + if (unlikely(err))
21621 + /* make the dir opaque */
21623 + h_inode = d_inode(h_path.dentry);
21625 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
21626 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
21627 + opq_dentry = au_diropq_create(dentry, bindex);
21628 + inode_unlock(h_inode);
21629 + err = PTR_ERR(opq_dentry);
21630 + if (IS_ERR(opq_dentry))
21632 + dput(opq_dentry);
21636 + err = epilog(dir, bindex, wh_dentry, dentry);
21639 + goto out_unpin; /* success */
21644 + AuLabel(revert opq);
21645 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
21646 + rerr = au_diropq_remove(dentry, bindex);
21647 + inode_unlock(h_inode);
21649 + AuIOErr("%pd reverting diropq failed(%d, %d)\n",
21650 + dentry, err, rerr);
21656 + AuLabel(revert dir);
21657 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
21659 + AuIOErr("%pd reverting dir failed(%d, %d)\n",
21660 + dentry, err, rerr);
21663 + au_dtime_revert(&a->dt);
21665 + au_unpin(&a->pin);
21668 + di_write_unlock(parent);
21670 + if (unlikely(err)) {
21671 + au_update_dbtop(dentry);
21674 + aufs_read_unlock(dentry, AuLock_DW);
21680 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
21681 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
21682 +++ linux/fs/aufs/i_op.c 2022-03-21 14:49:05.726633010 +0100
21684 +// SPDX-License-Identifier: GPL-2.0
21686 + * Copyright (C) 2005-2021 Junjiro R. Okajima
21688 + * This program, aufs is free software; you can redistribute it and/or modify
21689 + * it under the terms of the GNU General Public License as published by
21690 + * the Free Software Foundation; either version 2 of the License, or
21691 + * (at your option) any later version.
21693 + * This program is distributed in the hope that it will be useful,
21694 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21695 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21696 + * GNU General Public License for more details.
21698 + * You should have received a copy of the GNU General Public License
21699 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21703 + * inode operations (except add/del/rename)
21706 +#include <linux/device_cgroup.h>
21707 +#include <linux/fs_stack.h>
21708 +#include <linux/iversion.h>
21709 +#include <linux/security.h>
21712 +static int h_permission(struct inode *h_inode, int mask,
21713 + struct path *h_path, int brperm)
21716 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21717 + struct user_namespace *h_userns;
21720 + if (write_mask && IS_IMMUTABLE(h_inode))
21724 + if (((mask & MAY_EXEC)
21725 + && S_ISREG(h_inode->i_mode)
21726 + && (path_noexec(h_path)
21727 + || !(h_inode->i_mode & 0111))))
21731 + * - skip the lower fs test in the case of write to ro branch.
21732 + * - nfs dir permission write check is optimized, but a policy for
21733 + * link/rename requires a real check.
21734 + * - nfs always sets SB_POSIXACL regardless its mount option 'noacl.'
21735 + * in this case, generic_permission() returns -EOPNOTSUPP.
21737 + h_userns = mnt_user_ns(h_path->mnt);
21738 + if ((write_mask && !au_br_writable(brperm))
21739 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
21740 + && write_mask && !(mask & MAY_READ))
21741 + || !h_inode->i_op->permission) {
21742 + /* AuLabel(generic_permission); */
21743 + /* AuDbg("get_acl %ps\n", h_inode->i_op->get_acl); */
21744 + err = generic_permission(h_userns, h_inode, mask);
21745 + if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
21746 + err = h_inode->i_op->permission(h_userns, h_inode,
21750 + /* AuLabel(h_inode->permission); */
21751 + err = h_inode->i_op->permission(h_userns, h_inode, mask);
21756 + err = devcgroup_inode_permission(h_inode, mask);
21758 + err = security_inode_permission(h_inode, mask);
21764 +static int aufs_permission(struct user_namespace *userns, struct inode *inode,
21768 + aufs_bindex_t bindex, bbot;
21769 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
21770 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21771 + struct inode *h_inode;
21772 + struct super_block *sb;
21773 + struct au_branch *br;
21775 + /* todo: support rcu-walk? */
21776 + if (mask & MAY_NOT_BLOCK)
21779 + sb = inode->i_sb;
21780 + si_read_lock(sb, AuLock_FLUSH);
21781 + ii_read_lock_child(inode);
21782 +#if 0 /* reserved for future use */
21784 + * This test may be rather 'too much' since the test is essentially done
21785 + * in the aufs_lookup(). Theoretically it is possible that the inode
21786 + * generation doesn't match to the superblock's here. But it isn't a
21787 + * big deal I suppose.
21789 + err = au_iigen_test(inode, au_sigen(sb));
21790 + if (unlikely(err))
21796 + || au_opt_test(au_mntflags(sb), DIRPERM1)) {
21797 + err = au_busy_or_stale();
21798 + h_inode = au_h_iptr(inode, au_ibtop(inode));
21799 + if (unlikely(!h_inode
21800 + || (h_inode->i_mode & S_IFMT)
21801 + != (inode->i_mode & S_IFMT)))
21805 + bindex = au_ibtop(inode);
21806 + br = au_sbr(sb, bindex);
21807 + err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
21810 + && !special_file(h_inode->i_mode)) {
21811 + /* test whether the upper writable branch exists */
21813 + for (; bindex >= 0; bindex--)
21814 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
21822 + /* non-write to dir */
21824 + bbot = au_ibbot(inode);
21825 + for (bindex = au_ibtop(inode); !err && bindex <= bbot; bindex++) {
21826 + h_inode = au_h_iptr(inode, bindex);
21828 + err = au_busy_or_stale();
21829 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
21832 + br = au_sbr(sb, bindex);
21833 + err = h_permission(h_inode, mask, &br->br_path,
21839 + ii_read_unlock(inode);
21840 + si_read_unlock(sb);
21844 +/* ---------------------------------------------------------------------- */
21846 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
21847 + unsigned int flags)
21849 + struct dentry *ret, *parent;
21850 + struct inode *inode;
21851 + struct super_block *sb;
21852 + int err, npositive;
21856 + /* todo: support rcu-walk? */
21857 + ret = ERR_PTR(-ECHILD);
21858 + if (flags & LOOKUP_RCU)
21861 + ret = ERR_PTR(-ENAMETOOLONG);
21862 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
21866 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21867 + ret = ERR_PTR(err);
21868 + if (unlikely(err))
21871 + err = au_di_init(dentry);
21872 + ret = ERR_PTR(err);
21873 + if (unlikely(err))
21877 + npositive = 0; /* suppress a warning */
21878 + parent = dentry->d_parent; /* dir inode is locked */
21879 + di_read_lock_parent(parent, AuLock_IR);
21880 + err = au_alive_dir(parent);
21882 + err = au_digen_test(parent, au_sigen(sb));
21884 + /* regardless LOOKUP_CREATE, always ALLOW_NEG */
21885 + npositive = au_lkup_dentry(dentry, au_dbtop(parent),
21886 + AuLkup_ALLOW_NEG);
21889 + di_read_unlock(parent, AuLock_IR);
21890 + ret = ERR_PTR(err);
21891 + if (unlikely(err < 0))
21895 + inode = au_new_inode(dentry, /*must_new*/0);
21896 + if (IS_ERR(inode)) {
21897 + ret = (void *)inode;
21904 + atomic_inc(&inode->i_count);
21905 + ret = d_splice_alias(inode, dentry);
21906 +#if 0 /* reserved for future use */
21907 + if (unlikely(d_need_lookup(dentry))) {
21908 + spin_lock(&dentry->d_lock);
21909 + dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
21910 + spin_unlock(&dentry->d_lock);
21914 + if (!IS_ERR(ret)) {
21916 + if (ret && ret != dentry)
21917 + ii_write_unlock(inode);
21919 + ii_write_unlock(inode);
21926 + di_write_unlock(dentry);
21928 + si_read_unlock(sb);
21933 +/* ---------------------------------------------------------------------- */
21936 + * very dirty and complicated aufs ->atomic_open().
21937 + * aufs_atomic_open()
21938 + * + au_aopen_or_create()
21940 + * + vfsub_atomic_open()
21941 + * + branch fs ->atomic_open()
21942 + * may call the actual 'open' for h_file
21943 + * + inc br_nfiles only if opened
21944 + * + au_aopen_no_open() or au_aopen_do_open()
21946 + * au_aopen_do_open()
21947 + * + finish_open()
21948 + * + au_do_aopen()
21949 + * + au_do_open() the body of all 'open'
21950 + * + au_do_open_nondir()
21951 + * set the passed h_file
21953 + * au_aopen_no_open()
21954 + * + finish_no_open()
21957 +struct aopen_node {
21958 + struct hlist_bl_node hblist;
21959 + struct file *file, *h_file;
21962 +static int au_do_aopen(struct inode *inode, struct file *file)
21964 + struct hlist_bl_head *aopen;
21965 + struct hlist_bl_node *pos;
21966 + struct aopen_node *node;
21967 + struct au_do_open_args args = {
21969 + .open = au_do_open_nondir
21972 + aopen = &au_sbi(inode->i_sb)->si_aopen;
21973 + hlist_bl_lock(aopen);
21974 + hlist_bl_for_each_entry(node, pos, aopen, hblist)
21975 + if (node->file == file) {
21976 + args.h_file = node->h_file;
21979 + hlist_bl_unlock(aopen);
21980 + /* AuDebugOn(!args.h_file); */
21982 + return au_do_open(file, &args);
21985 +static int au_aopen_do_open(struct file *file, struct dentry *dentry,
21986 + struct aopen_node *aopen_node)
21989 + struct hlist_bl_head *aopen;
21992 + aopen = &au_sbi(dentry->d_sb)->si_aopen;
21993 + au_hbl_add(&aopen_node->hblist, aopen);
21994 + err = finish_open(file, dentry, au_do_aopen);
21995 + au_hbl_del(&aopen_node->hblist, aopen);
21996 + /* AuDbgFile(file); */
21997 + AuDbg("%pd%s%s\n", dentry,
21998 + (file->f_mode & FMODE_CREATED) ? " created" : "",
21999 + (file->f_mode & FMODE_OPENED) ? " opened" : "");
22005 +static int au_aopen_no_open(struct file *file, struct dentry *dentry)
22011 + err = finish_no_open(file, dentry);
22017 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
22018 + struct file *file, unsigned int open_flag,
22019 + umode_t create_mode)
22021 + int err, did_open;
22022 + unsigned int lkup_flags;
22023 + aufs_bindex_t bindex;
22024 + struct super_block *sb;
22025 + struct dentry *parent, *d;
22026 + struct vfsub_aopen_args args = {
22027 + .open_flag = open_flag,
22028 + .create_mode = create_mode
22030 + struct aopen_node aopen_node = {
22035 + AuDbg("open_flag 0%o\n", open_flag);
22036 + AuDbgDentry(dentry);
22039 + if (!au_di(dentry)) {
22040 + lkup_flags = LOOKUP_OPEN;
22041 + if (open_flag & O_CREAT)
22042 + lkup_flags |= LOOKUP_CREATE;
22043 + d = aufs_lookup(dir, dentry, lkup_flags);
22045 + err = PTR_ERR(d);
22050 + * obsoleted dentry found.
22051 + * another error will be returned later.
22057 + AuDbgDentry(dentry);
22060 + if (d_is_positive(dentry)
22061 + || d_unhashed(dentry)
22062 + || d_unlinked(dentry)
22063 + || !(open_flag & O_CREAT)) {
22064 + err = au_aopen_no_open(file, dentry);
22065 + goto out; /* success */
22068 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
22069 + if (unlikely(err))
22072 + sb = dentry->d_sb;
22073 + parent = dentry->d_parent; /* dir is locked */
22074 + di_write_lock_parent(parent);
22075 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
22076 + if (unlikely(err < 0))
22079 + AuDbgDentry(dentry);
22080 + if (d_is_positive(dentry)) {
22081 + err = au_aopen_no_open(file, dentry);
22082 + goto out_parent; /* success */
22085 + args.file = alloc_empty_file(file->f_flags, current_cred());
22086 + err = PTR_ERR(args.file);
22087 + if (IS_ERR(args.file))
22090 + bindex = au_dbtop(dentry);
22091 + err = au_aopen_or_create(dir, dentry, &args);
22093 + AuDbgFile(args.file);
22094 + file->f_mode = args.file->f_mode & ~FMODE_OPENED;
22095 + did_open = !!(args.file->f_mode & FMODE_OPENED);
22098 + args.file = NULL;
22100 + di_write_unlock(parent);
22101 + di_write_unlock(dentry);
22102 + if (unlikely(err < 0)) {
22109 + err = au_aopen_no_open(file, dentry);
22111 + aopen_node.h_file = args.file;
22112 + err = au_aopen_do_open(file, dentry, &aopen_node);
22114 + if (unlikely(err < 0)) {
22118 + au_lcnt_dec(&args.br->br_nfiles);
22120 + goto out_sb; /* success */
22123 + di_write_unlock(parent);
22124 + di_write_unlock(dentry);
22126 + si_read_unlock(sb);
22134 +/* ---------------------------------------------------------------------- */
22136 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
22137 + const unsigned char add_entry, aufs_bindex_t bcpup,
22138 + aufs_bindex_t btop)
22141 + struct dentry *h_parent;
22142 + struct inode *h_dir;
22145 + IMustLock(d_inode(parent));
22147 + di_write_lock_parent(parent);
22150 + if (!au_h_dptr(parent, bcpup)) {
22151 + if (btop > bcpup)
22152 + err = au_cpup_dirs(dentry, bcpup);
22153 + else if (btop < bcpup)
22154 + err = au_cpdown_dirs(dentry, bcpup);
22158 + if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
22159 + h_parent = au_h_dptr(parent, bcpup);
22160 + h_dir = d_inode(h_parent);
22161 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
22162 + err = au_lkup_neg(dentry, bcpup, /*wh*/0);
22163 + /* todo: no unlock here */
22164 + inode_unlock_shared(h_dir);
22166 + AuDbg("bcpup %d\n", bcpup);
22168 + if (d_really_is_negative(dentry))
22169 + au_set_h_dptr(dentry, btop, NULL);
22170 + au_update_dbrange(dentry, /*do_put_zero*/0);
22175 + di_write_unlock(parent);
22177 + err = bcpup; /* success */
22184 + * decide the branch and the parent dir where we will create a new entry.
22185 + * returns new bindex or an error.
22186 + * copyup the parent dir if needed.
22188 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
22189 + struct au_wr_dir_args *args)
22192 + unsigned int flags;
22193 + aufs_bindex_t bcpup, btop, src_btop;
22194 + const unsigned char add_entry
22195 + = au_ftest_wrdir(args->flags, ADD_ENTRY)
22196 + | au_ftest_wrdir(args->flags, TMPFILE);
22197 + struct super_block *sb;
22198 + struct dentry *parent;
22199 + struct au_sbinfo *sbinfo;
22201 + sb = dentry->d_sb;
22202 + sbinfo = au_sbi(sb);
22203 + parent = dget_parent(dentry);
22204 + btop = au_dbtop(dentry);
22206 + if (args->force_btgt < 0) {
22207 + if (src_dentry) {
22208 + src_btop = au_dbtop(src_dentry);
22209 + if (src_btop < btop)
22210 + bcpup = src_btop;
22211 + } else if (add_entry) {
22213 + if (au_ftest_wrdir(args->flags, ISDIR))
22214 + au_fset_wbr(flags, DIR);
22215 + err = AuWbrCreate(sbinfo, dentry, flags);
22219 + if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
22221 + err = AuWbrCopyup(sbinfo, dentry);
22223 + if (!IS_ROOT(dentry)) {
22224 + di_read_lock_parent(parent, !AuLock_IR);
22225 + err = AuWbrCopyup(sbinfo, dentry);
22226 + di_read_unlock(parent, !AuLock_IR);
22228 + err = AuWbrCopyup(sbinfo, dentry);
22231 + if (unlikely(err < 0))
22235 + bcpup = args->force_btgt;
22236 + AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
22239 + AuDbg("btop %d, bcpup %d\n", btop, bcpup);
22241 + if (bcpup == btop)
22242 + goto out; /* success */
22244 + /* copyup the new parent into the branch we process */
22245 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, btop);
22247 + if (d_really_is_negative(dentry)) {
22248 + au_set_h_dptr(dentry, btop, NULL);
22249 + au_set_dbtop(dentry, bcpup);
22250 + au_set_dbbot(dentry, bcpup);
22252 + AuDebugOn(add_entry
22253 + && !au_ftest_wrdir(args->flags, TMPFILE)
22254 + && !au_h_dptr(dentry, bcpup));
22262 +/* ---------------------------------------------------------------------- */
22264 +void au_pin_hdir_unlock(struct au_pin *p)
22267 + au_hn_inode_unlock(p->hdir);
22270 +int au_pin_hdir_lock(struct au_pin *p)
22278 + /* even if an error happens later, keep this lock */
22279 + au_hn_inode_lock_nested(p->hdir, p->lsc_hi);
22282 + if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
22287 + err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
22288 + p->h_parent, p->br);
22294 +int au_pin_hdir_relock(struct au_pin *p)
22297 + struct inode *h_i;
22298 + struct dentry *h_d[] = {
22303 + err = au_pin_hdir_lock(p);
22304 + if (unlikely(err))
22307 + for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
22310 + if (d_is_positive(h_d[i])) {
22311 + h_i = d_inode(h_d[i]);
22312 + err = !h_i->i_nlink;
22320 +static void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
22322 + atomic_long_set(&p->hdir->hi_inode->i_rwsem.owner, (long)task);
22325 +void au_pin_hdir_acquire_nest(struct au_pin *p)
22328 + rwsem_acquire_nest(&p->hdir->hi_inode->i_rwsem.dep_map,
22329 + p->lsc_hi, 0, NULL, _RET_IP_);
22330 + au_pin_hdir_set_owner(p, current);
22334 +void au_pin_hdir_release(struct au_pin *p)
22337 + au_pin_hdir_set_owner(p, p->task);
22338 + rwsem_release(&p->hdir->hi_inode->i_rwsem.dep_map, _RET_IP_);
22342 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
22344 + if (pin && pin->parent)
22345 + return au_h_dptr(pin->parent, pin->bindex);
22349 +void au_unpin(struct au_pin *p)
22352 + au_pin_hdir_unlock(p);
22353 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
22354 + vfsub_mnt_drop_write(p->h_mnt);
22358 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22359 + di_read_unlock(p->parent, AuLock_IR);
22360 + iput(p->hdir->hi_inode);
22362 + p->parent = NULL;
22365 + /* do not clear p->task */
22368 +int au_do_pin(struct au_pin *p)
22371 + struct super_block *sb;
22372 + struct inode *h_dir;
22375 + sb = p->dentry->d_sb;
22376 + p->br = au_sbr(sb, p->bindex);
22377 + if (IS_ROOT(p->dentry)) {
22378 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
22379 + p->h_mnt = au_br_mnt(p->br);
22380 + err = vfsub_mnt_want_write(p->h_mnt);
22381 + if (unlikely(err)) {
22382 + au_fclr_pin(p->flags, MNT_WRITE);
22389 + p->h_dentry = NULL;
22390 + if (p->bindex <= au_dbbot(p->dentry))
22391 + p->h_dentry = au_h_dptr(p->dentry, p->bindex);
22393 + p->parent = dget_parent(p->dentry);
22394 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22395 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
22398 + p->h_parent = au_h_dptr(p->parent, p->bindex);
22399 + p->hdir = au_hi(d_inode(p->parent), p->bindex);
22401 + h_dir = p->hdir->hi_inode;
22405 + * if DI_LOCKED is not set, then p->parent may be different
22406 + * and h_parent can be NULL.
22408 + if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
22410 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22411 + di_read_unlock(p->parent, AuLock_IR);
22413 + p->parent = NULL;
22417 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
22418 + p->h_mnt = au_br_mnt(p->br);
22419 + err = vfsub_mnt_want_write(p->h_mnt);
22420 + if (unlikely(err)) {
22421 + au_fclr_pin(p->flags, MNT_WRITE);
22422 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22423 + di_read_unlock(p->parent, AuLock_IR);
22425 + p->parent = NULL;
22431 + err = au_pin_hdir_lock(p);
22433 + goto out; /* success */
22438 + pr_err("err %d\n", err);
22439 + err = au_busy_or_stale();
22444 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
22445 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
22446 + unsigned int udba, unsigned char flags)
22448 + p->dentry = dentry;
22450 + p->lsc_di = lsc_di;
22451 + p->lsc_hi = lsc_hi;
22452 + p->flags = flags;
22453 + p->bindex = bindex;
22455 + p->parent = NULL;
22459 + p->h_dentry = NULL;
22460 + p->h_parent = NULL;
22462 + p->task = current;
22465 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
22466 + unsigned int udba, unsigned char flags)
22468 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
22470 + return au_do_pin(pin);
22473 +/* ---------------------------------------------------------------------- */
22476 + * ->setattr() and ->getattr() are called in various cases.
22477 + * chmod, stat: dentry is revalidated.
22478 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
22480 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
22482 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
22483 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
22486 + struct dentry *parent;
22489 + if (au_digen_test(dentry, sigen)) {
22490 + parent = dget_parent(dentry);
22491 + di_read_lock_parent(parent, AuLock_IR);
22492 + err = au_refresh_dentry(dentry, parent);
22493 + di_read_unlock(parent, AuLock_IR);
22501 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
22502 + struct au_icpup_args *a)
22506 + aufs_bindex_t btop, ibtop;
22507 + struct dentry *hi_wh, *parent;
22508 + struct inode *inode;
22509 + struct au_wr_dir_args wr_dir_args = {
22510 + .force_btgt = -1,
22514 + if (d_is_dir(dentry))
22515 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
22516 + /* plink or hi_wh() case */
22517 + btop = au_dbtop(dentry);
22518 + inode = d_inode(dentry);
22519 + ibtop = au_ibtop(inode);
22520 + if (btop != ibtop && !au_test_ro(inode->i_sb, ibtop, inode))
22521 + wr_dir_args.force_btgt = ibtop;
22522 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
22523 + if (unlikely(err < 0))
22527 + au_fset_icpup(a->flags, DID_CPUP);
22530 + a->pin_flags = AuPin_MNT_WRITE;
22532 + if (!IS_ROOT(dentry)) {
22533 + au_fset_pin(a->pin_flags, DI_LOCKED);
22534 + parent = dget_parent(dentry);
22535 + di_write_lock_parent(parent);
22538 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
22539 + if (unlikely(err))
22543 + a->h_path.dentry = au_h_dptr(dentry, btop);
22544 + a->h_inode = d_inode(a->h_path.dentry);
22545 + if (ia && (ia->ia_valid & ATTR_SIZE)) {
22546 + inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
22547 + if (ia->ia_size < i_size_read(a->h_inode))
22548 + sz = ia->ia_size;
22549 + inode_unlock_shared(a->h_inode);
22553 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
22554 + hi_wh = au_hi_wh(inode, a->btgt);
22556 + struct au_cp_generic cpg = {
22557 + .dentry = dentry,
22563 + err = au_sio_cpup_wh(&cpg, /*file*/NULL);
22564 + if (unlikely(err))
22566 + hi_wh = au_hi_wh(inode, a->btgt);
22567 + /* todo: revalidate hi_wh? */
22572 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
22573 + di_downgrade_lock(parent, AuLock_IR);
22577 + if (!au_ftest_icpup(a->flags, DID_CPUP))
22578 + goto out; /* success */
22580 + if (!d_unhashed(dentry)) {
22581 + struct au_cp_generic cpg = {
22582 + .dentry = dentry,
22587 + .flags = AuCpup_DTIME | AuCpup_HOPEN
22589 + err = au_sio_cpup_simple(&cpg);
22591 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22592 + } else if (!hi_wh)
22593 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22595 + a->h_path.dentry = hi_wh; /* do not dget here */
22598 + a->h_inode = d_inode(a->h_path.dentry);
22600 + goto out; /* success */
22601 + au_unpin(&a->pin);
22604 + di_write_unlock(parent);
22609 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22613 +static int aufs_setattr(struct user_namespace *userns, struct dentry *dentry,
22614 + struct iattr *ia)
22617 + struct inode *inode, *delegated;
22618 + struct super_block *sb;
22619 + struct file *file;
22620 + struct au_icpup_args *a;
22621 + struct user_namespace *h_userns;
22623 + inode = d_inode(dentry);
22624 + IMustLock(inode);
22626 + err = setattr_prepare(userns, dentry, ia);
22627 + if (unlikely(err))
22631 + a = kzalloc(sizeof(*a), GFP_NOFS);
22632 + if (unlikely(!a))
22635 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
22636 + ia->ia_valid &= ~ATTR_MODE;
22639 + sb = dentry->d_sb;
22640 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22641 + if (unlikely(err))
22644 + if (ia->ia_valid & ATTR_FILE) {
22645 + /* currently ftruncate(2) only */
22646 + AuDebugOn(!d_is_reg(dentry));
22647 + file = ia->ia_file;
22648 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1,
22650 + if (unlikely(err))
22652 + ia->ia_file = au_hf_top(file);
22653 + a->udba = AuOpt_UDBA_NONE;
22655 + /* fchmod() doesn't pass ia_file */
22656 + a->udba = au_opt_udba(sb);
22657 + di_write_lock_child(dentry);
22658 + /* no d_unlinked(), to set UDBA_NONE for root */
22659 + if (d_unhashed(dentry))
22660 + a->udba = AuOpt_UDBA_NONE;
22661 + if (a->udba != AuOpt_UDBA_NONE) {
22662 + AuDebugOn(IS_ROOT(dentry));
22663 + err = au_reval_for_attr(dentry, au_sigen(sb));
22664 + if (unlikely(err))
22669 + err = au_pin_and_icpup(dentry, ia, a);
22670 + if (unlikely(err < 0))
22672 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
22673 + ia->ia_file = NULL;
22674 + ia->ia_valid &= ~ATTR_FILE;
22677 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
22678 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
22679 + == (ATTR_MODE | ATTR_CTIME)) {
22680 + err = security_path_chmod(&a->h_path, ia->ia_mode);
22681 + if (unlikely(err))
22683 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
22684 + && (ia->ia_valid & ATTR_CTIME)) {
22685 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
22686 + if (unlikely(err))
22690 + if (ia->ia_valid & ATTR_SIZE) {
22693 + if (ia->ia_size < i_size_read(inode))
22695 + truncate_setsize(inode, ia->ia_size);
22698 + if (ia->ia_valid & ATTR_FILE)
22700 + inode_unlock(a->h_inode);
22701 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
22702 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22704 + delegated = NULL;
22706 + err = vfsub_notify_change(&a->h_path, ia, &delegated);
22708 + err = break_deleg_wait(&delegated);
22716 + * regardless aufs 'acl' option setting.
22717 + * why don't all acl-aware fs call this func from their ->setattr()?
22719 + if (!err && (ia->ia_valid & ATTR_MODE)) {
22720 + h_userns = mnt_user_ns(a->h_path.mnt);
22721 + err = vfsub_acl_chmod(h_userns, a->h_inode, ia->ia_mode);
22724 + au_cpup_attr_changeable(inode);
22727 + inode_unlock(a->h_inode);
22728 + au_unpin(&a->pin);
22729 + if (unlikely(err))
22730 + au_update_dbtop(dentry);
22732 + di_write_unlock(dentry);
22734 + fi_write_unlock(file);
22735 + ia->ia_file = file;
22736 + ia->ia_valid |= ATTR_FILE;
22739 + si_read_unlock(sb);
22747 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
22748 +static int au_h_path_to_set_attr(struct dentry *dentry,
22749 + struct au_icpup_args *a, struct path *h_path)
22752 + struct super_block *sb;
22754 + sb = dentry->d_sb;
22755 + a->udba = au_opt_udba(sb);
22756 + /* no d_unlinked(), to set UDBA_NONE for root */
22757 + if (d_unhashed(dentry))
22758 + a->udba = AuOpt_UDBA_NONE;
22759 + if (a->udba != AuOpt_UDBA_NONE) {
22760 + AuDebugOn(IS_ROOT(dentry));
22761 + err = au_reval_for_attr(dentry, au_sigen(sb));
22762 + if (unlikely(err))
22765 + err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
22766 + if (unlikely(err < 0))
22769 + h_path->dentry = a->h_path.dentry;
22770 + h_path->mnt = au_sbr_mnt(sb, a->btgt);
22776 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
22777 + struct au_sxattr *arg)
22780 + struct path h_path;
22781 + struct super_block *sb;
22782 + struct au_icpup_args *a;
22783 + struct inode *h_inode;
22784 + struct user_namespace *h_userns;
22786 + IMustLock(inode);
22789 + a = kzalloc(sizeof(*a), GFP_NOFS);
22790 + if (unlikely(!a))
22793 + sb = dentry->d_sb;
22794 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22795 + if (unlikely(err))
22798 + h_path.dentry = NULL; /* silence gcc */
22799 + di_write_lock_child(dentry);
22800 + err = au_h_path_to_set_attr(dentry, a, &h_path);
22801 + if (unlikely(err))
22803 + h_userns = mnt_user_ns(h_path.mnt);
22805 + inode_unlock(a->h_inode);
22806 + switch (arg->type) {
22807 + case AU_XATTR_SET:
22808 + AuDebugOn(d_is_negative(h_path.dentry));
22809 + err = vfsub_setxattr(h_userns, h_path.dentry,
22810 + arg->u.set.name, arg->u.set.value,
22811 + arg->u.set.size, arg->u.set.flags);
22814 + err = -EOPNOTSUPP;
22815 + h_inode = d_inode(h_path.dentry);
22816 + if (h_inode->i_op->set_acl) {
22817 + /* this will call posix_acl_update_mode */
22818 + err = h_inode->i_op->set_acl(h_userns, h_inode,
22819 + arg->u.acl_set.acl,
22820 + arg->u.acl_set.type);
22825 + au_cpup_attr_timesizes(inode);
22827 + au_unpin(&a->pin);
22828 + if (unlikely(err))
22829 + au_update_dbtop(dentry);
22832 + di_write_unlock(dentry);
22833 + si_read_unlock(sb);
22842 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
22843 + unsigned int nlink)
22847 + inode->i_mode = st->mode;
22848 + /* don't i_[ug]id_write() here */
22849 + inode->i_uid = st->uid;
22850 + inode->i_gid = st->gid;
22851 + inode->i_atime = st->atime;
22852 + inode->i_mtime = st->mtime;
22853 + inode->i_ctime = st->ctime;
22855 + au_cpup_attr_nlink(inode, /*force*/0);
22856 + if (S_ISDIR(inode->i_mode)) {
22857 + n = inode->i_nlink;
22860 + smp_mb(); /* for i_nlink */
22861 + /* 0 can happen */
22862 + set_nlink(inode, n);
22865 + spin_lock(&inode->i_lock);
22866 + inode->i_blocks = st->blocks;
22867 + i_size_write(inode, st->size);
22868 + spin_unlock(&inode->i_lock);
22872 + * common routine for aufs_getattr() and au_getxattr().
22873 + * returns zero or negative (an error).
22874 + * @dentry will be read-locked in success.
22876 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
22877 + struct path *h_path, int locked)
22880 + unsigned int mnt_flags, sigen;
22881 + unsigned char udba_none;
22882 + aufs_bindex_t bindex;
22883 + struct super_block *sb, *h_sb;
22885 + h_path->mnt = NULL;
22886 + h_path->dentry = NULL;
22889 + sb = dentry->d_sb;
22890 + mnt_flags = au_mntflags(sb);
22891 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
22893 + if (unlikely(locked))
22894 + goto body; /* skip locking dinfo */
22896 + /* support fstat(2) */
22897 + if (!d_unlinked(dentry) && !udba_none) {
22898 + sigen = au_sigen(sb);
22899 + err = au_digen_test(dentry, sigen);
22901 + di_read_lock_child(dentry, AuLock_IR);
22902 + err = au_dbrange_test(dentry);
22903 + if (unlikely(err)) {
22904 + di_read_unlock(dentry, AuLock_IR);
22908 + AuDebugOn(IS_ROOT(dentry));
22909 + di_write_lock_child(dentry);
22910 + err = au_dbrange_test(dentry);
22912 + err = au_reval_for_attr(dentry, sigen);
22914 + di_downgrade_lock(dentry, AuLock_IR);
22916 + di_write_unlock(dentry);
22921 + di_read_lock_child(dentry, AuLock_IR);
22925 + inode = d_inode(dentry);
22926 + if (unlikely(!inode))
22929 + bindex = au_ibtop(inode);
22930 + h_path->mnt = au_sbr_mnt(sb, bindex);
22931 + h_sb = h_path->mnt->mnt_sb;
22933 + && !au_test_fs_bad_iattr(h_sb)
22935 + goto out; /* success */
22937 + if (au_dbtop(dentry) == bindex)
22938 + h_path->dentry = au_h_dptr(dentry, bindex);
22939 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
22940 + h_path->dentry = au_plink_lkup(inode, bindex);
22941 + if (IS_ERR(h_path->dentry))
22942 + /* pretending success */
22943 + h_path->dentry = NULL;
22945 + dput(h_path->dentry);
22952 +static int aufs_getattr(struct user_namespace *userns, const struct path *path,
22953 + struct kstat *st, u32 request, unsigned int query)
22956 + unsigned char positive;
22957 + struct path h_path;
22958 + struct dentry *dentry;
22959 + struct inode *inode;
22960 + struct super_block *sb;
22962 + dentry = path->dentry;
22963 + inode = d_inode(dentry);
22964 + sb = dentry->d_sb;
22965 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22966 + if (unlikely(err))
22968 + err = au_h_path_getattr(dentry, /*inode*/NULL, /*force*/0, &h_path,
22970 + if (unlikely(err))
22972 + if (unlikely(!h_path.dentry))
22973 + /* illegally overlapped or something */
22974 + goto out_fill; /* pretending success */
22976 + positive = d_is_positive(h_path.dentry);
22978 + /* no vfsub version */
22979 + err = vfs_getattr(&h_path, st, request, query);
22982 + au_refresh_iattr(inode, st,
22983 + d_inode(h_path.dentry)->i_nlink);
22984 + goto out_fill; /* success */
22990 + generic_fillattr(userns, inode, st);
22992 + di_read_unlock(dentry, AuLock_IR);
22994 + si_read_unlock(sb);
23000 +/* ---------------------------------------------------------------------- */
23002 +static const char *aufs_get_link(struct dentry *dentry, struct inode *inode,
23003 + struct delayed_call *done)
23006 + struct dentry *h_dentry;
23007 + struct inode *h_inode;
23009 + aufs_bindex_t bindex;
23011 + ret = NULL; /* suppress a warning */
23016 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
23017 + if (unlikely(err))
23020 + err = au_d_hashed_positive(dentry);
23021 + if (unlikely(err))
23025 + inode = d_inode(dentry);
23026 + bindex = au_ibtop(inode);
23027 + h_inode = au_h_iptr(inode, bindex);
23028 + if (unlikely(!h_inode->i_op->get_link))
23033 + if (au_dbtop(dentry) <= bindex) {
23034 + h_dentry = au_h_dptr(dentry, bindex);
23039 + h_dentry = d_find_any_alias(h_inode);
23040 + if (IS_ERR(h_dentry)) {
23041 + err = PTR_ERR(h_dentry);
23045 + if (unlikely(!h_dentry))
23049 + AuDbg("%ps\n", h_inode->i_op->get_link);
23050 + AuDbgDentry(h_dentry);
23051 + ret = vfs_get_link(h_dentry, done);
23054 + err = PTR_ERR(ret);
23057 + aufs_read_unlock(dentry, AuLock_IR);
23059 + if (unlikely(err))
23060 + ret = ERR_PTR(err);
23061 + AuTraceErrPtr(ret);
23065 +/* ---------------------------------------------------------------------- */
23067 +static int au_is_special(struct inode *inode)
23069 + return (inode->i_mode & (S_IFBLK | S_IFCHR | S_IFIFO | S_IFSOCK));
23072 +static int aufs_update_time(struct inode *inode, struct timespec64 *ts,
23076 + aufs_bindex_t bindex;
23077 + struct super_block *sb;
23078 + struct inode *h_inode;
23079 + struct vfsmount *h_mnt;
23081 + sb = inode->i_sb;
23082 + WARN_ONCE((flags & S_ATIME) && !IS_NOATIME(inode),
23083 + "unexpected s_flags 0x%lx", sb->s_flags);
23085 + /* mmap_sem might be acquired already, cf. aufs_mmap() */
23087 + si_read_lock(sb, AuLock_FLUSH);
23088 + ii_write_lock_child(inode);
23091 + bindex = au_ibtop(inode);
23092 + h_inode = au_h_iptr(inode, bindex);
23093 + if (!au_test_ro(sb, bindex, inode)) {
23094 + h_mnt = au_sbr_mnt(sb, bindex);
23095 + err = vfsub_mnt_want_write(h_mnt);
23097 + err = vfsub_update_time(h_inode, ts, flags);
23098 + vfsub_mnt_drop_write(h_mnt);
23100 + } else if (au_is_special(h_inode)) {
23102 + * Never copy-up here.
23103 + * These special files may already be opened and used for
23104 + * communicating. If we copied it up, then the communication
23105 + * would be corrupted.
23107 + AuWarn1("timestamps for i%lu are ignored "
23108 + "since it is on readonly branch (hi%lu).\n",
23109 + inode->i_ino, h_inode->i_ino);
23110 + } else if (flags & ~S_ATIME) {
23112 + AuIOErr1("unexpected flags 0x%x\n", flags);
23117 + au_cpup_attr_timesizes(inode);
23118 + ii_write_unlock(inode);
23119 + si_read_unlock(sb);
23122 + if (!err && (flags & S_VERSION))
23123 + inode_inc_iversion(inode);
23128 +/* ---------------------------------------------------------------------- */
23130 +/* no getattr version will be set by module.c:aufs_init() */
23131 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
23133 + [AuIop_SYMLINK] = {
23134 + .permission = aufs_permission,
23135 +#ifdef CONFIG_FS_POSIX_ACL
23136 + .get_acl = aufs_get_acl,
23137 + .set_acl = aufs_set_acl, /* unsupport for symlink? */
23140 + .setattr = aufs_setattr,
23141 + .getattr = aufs_getattr,
23143 +#ifdef CONFIG_AUFS_XATTR
23144 + .listxattr = aufs_listxattr,
23147 + .get_link = aufs_get_link,
23149 + /* .update_time = aufs_update_time */
23152 + .create = aufs_create,
23153 + .lookup = aufs_lookup,
23154 + .link = aufs_link,
23155 + .unlink = aufs_unlink,
23156 + .symlink = aufs_symlink,
23157 + .mkdir = aufs_mkdir,
23158 + .rmdir = aufs_rmdir,
23159 + .mknod = aufs_mknod,
23160 + .rename = aufs_rename,
23162 + .permission = aufs_permission,
23163 +#ifdef CONFIG_FS_POSIX_ACL
23164 + .get_acl = aufs_get_acl,
23165 + .set_acl = aufs_set_acl,
23168 + .setattr = aufs_setattr,
23169 + .getattr = aufs_getattr,
23171 +#ifdef CONFIG_AUFS_XATTR
23172 + .listxattr = aufs_listxattr,
23175 + .update_time = aufs_update_time,
23176 + .atomic_open = aufs_atomic_open,
23177 + .tmpfile = aufs_tmpfile
23179 + [AuIop_OTHER] = {
23180 + .permission = aufs_permission,
23181 +#ifdef CONFIG_FS_POSIX_ACL
23182 + .get_acl = aufs_get_acl,
23183 + .set_acl = aufs_set_acl,
23186 + .setattr = aufs_setattr,
23187 + .getattr = aufs_getattr,
23189 +#ifdef CONFIG_AUFS_XATTR
23190 + .listxattr = aufs_listxattr,
23193 + .update_time = aufs_update_time
23196 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
23197 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
23198 +++ linux/fs/aufs/i_op_del.c 2022-03-21 14:49:05.726633010 +0100
23200 +// SPDX-License-Identifier: GPL-2.0
23202 + * Copyright (C) 2005-2021 Junjiro R. Okajima
23204 + * This program, aufs is free software; you can redistribute it and/or modify
23205 + * it under the terms of the GNU General Public License as published by
23206 + * the Free Software Foundation; either version 2 of the License, or
23207 + * (at your option) any later version.
23209 + * This program is distributed in the hope that it will be useful,
23210 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23211 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23212 + * GNU General Public License for more details.
23214 + * You should have received a copy of the GNU General Public License
23215 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23219 + * inode operations (del entry)
23222 +#include <linux/iversion.h>
23226 + * decide if a new whiteout for @dentry is necessary or not.
23227 + * when it is necessary, prepare the parent dir for the upper branch whose
23228 + * branch index is @bcpup for creation. the actual creation of the whiteout will
23229 + * be done by caller.
23231 + * 0: wh is unnecessary
23232 + * plus: wh is necessary
23235 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
23237 + int need_wh, err;
23238 + aufs_bindex_t btop;
23239 + struct super_block *sb;
23241 + sb = dentry->d_sb;
23242 + btop = au_dbtop(dentry);
23243 + if (*bcpup < 0) {
23245 + if (au_test_ro(sb, btop, d_inode(dentry))) {
23246 + err = AuWbrCopyup(au_sbi(sb), dentry);
23248 + if (unlikely(err < 0))
23252 + AuDebugOn(btop < *bcpup
23253 + || au_test_ro(sb, *bcpup, d_inode(dentry)));
23254 + AuDbg("bcpup %d, btop %d\n", *bcpup, btop);
23256 + if (*bcpup != btop) {
23257 + err = au_cpup_dirs(dentry, *bcpup);
23258 + if (unlikely(err))
23262 + struct au_dinfo *dinfo, *tmp;
23264 + need_wh = -ENOMEM;
23265 + dinfo = au_di(dentry);
23266 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
23268 + au_di_cp(tmp, dinfo);
23269 + au_di_swap(tmp, dinfo);
23270 + /* returns the number of positive dentries */
23271 + need_wh = au_lkup_dentry(dentry, btop + 1,
23272 + /* AuLkup_IGNORE_PERM */ 0);
23273 + au_di_swap(tmp, dinfo);
23274 + au_rw_write_unlock(&tmp->di_rwsem);
23278 + AuDbg("need_wh %d\n", need_wh);
23286 + * simple tests for the del-entry operations.
23287 + * following the checks in vfs, plus the parent-child relationship.
23289 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
23290 + struct dentry *h_parent, int isdir)
23294 + struct dentry *h_dentry, *h_latest;
23295 + struct inode *h_inode;
23296 + struct path h_ppath;
23297 + struct super_block *sb;
23298 + struct au_branch *br;
23299 + struct user_namespace *h_userns;
23301 + h_dentry = au_h_dptr(dentry, bindex);
23302 + if (d_really_is_positive(dentry)) {
23304 + if (unlikely(d_is_negative(h_dentry)))
23306 + h_inode = d_inode(h_dentry);
23307 + if (unlikely(!h_inode->i_nlink))
23310 + h_mode = h_inode->i_mode;
23313 + if (unlikely(S_ISDIR(h_mode)))
23315 + } else if (unlikely(!S_ISDIR(h_mode))) {
23320 + /* rename(2) case */
23322 + if (unlikely(d_is_positive(h_dentry)))
23327 + /* expected parent dir is locked */
23328 + if (unlikely(h_parent != h_dentry->d_parent))
23333 + * rmdir a dir may break the consistency on some filesystem.
23334 + * let's try heavy test.
23337 + sb = dentry->d_sb;
23338 + br = au_sbr(sb, bindex);
23339 + h_userns = au_br_userns(br);
23340 + if (unlikely(!au_opt_test(au_mntflags(sb), DIRPERM1)
23341 + && au_test_h_perm(h_userns, d_inode(h_parent),
23342 + MAY_EXEC | MAY_WRITE)))
23345 + h_ppath.dentry = h_parent;
23346 + h_ppath.mnt = au_br_mnt(br);
23347 + h_latest = au_sio_lkup_one(h_userns, &dentry->d_name, &h_ppath);
23349 + if (IS_ERR(h_latest))
23351 + if (h_latest == h_dentry)
23360 + * decide the branch where we operate for @dentry. the branch index will be set
23361 + * @rbcpup. after deciding it, 'pin' it and store the timestamps of the parent
23362 + * dir for reverting.
23363 + * when a new whiteout is necessary, create it.
23365 +static struct dentry*
23366 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
23367 + struct au_dtime *dt, struct au_pin *pin)
23369 + struct dentry *wh_dentry;
23370 + struct super_block *sb;
23371 + struct path h_path;
23372 + int err, need_wh;
23373 + unsigned int udba;
23374 + aufs_bindex_t bcpup;
23376 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
23377 + wh_dentry = ERR_PTR(need_wh);
23378 + if (unlikely(need_wh < 0))
23381 + sb = dentry->d_sb;
23382 + udba = au_opt_udba(sb);
23384 + err = au_pin(pin, dentry, bcpup, udba,
23385 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23386 + wh_dentry = ERR_PTR(err);
23387 + if (unlikely(err))
23390 + h_path.dentry = au_pinned_h_parent(pin);
23391 + if (udba != AuOpt_UDBA_NONE
23392 + && au_dbtop(dentry) == bcpup) {
23393 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
23394 + wh_dentry = ERR_PTR(err);
23395 + if (unlikely(err))
23399 + h_path.mnt = au_sbr_mnt(sb, bcpup);
23400 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
23401 + wh_dentry = NULL;
23403 + goto out; /* success, no need to create whiteout */
23405 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
23406 + if (IS_ERR(wh_dentry))
23409 + /* returns with the parent is locked and wh_dentry is dget-ed */
23410 + goto out; /* success */
23415 + return wh_dentry;
23419 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
23420 + * in order to be revertible and save time for removing many child whiteouts
23422 + * returns 1 when there are too many child whiteout and caller should remove
23423 + * them asynchronously. returns 0 when the number of children is enough small to
23424 + * remove now or the branch fs is a remote fs.
23425 + * otherwise return an error.
23427 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
23428 + struct au_nhash *whlist, struct inode *dir)
23430 + int rmdir_later, err, dirwh;
23431 + struct dentry *h_dentry;
23432 + struct super_block *sb;
23433 + struct inode *inode;
23435 + sb = dentry->d_sb;
23436 + SiMustAnyLock(sb);
23437 + h_dentry = au_h_dptr(dentry, bindex);
23438 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
23439 + if (unlikely(err))
23442 + /* stop monitoring */
23443 + inode = d_inode(dentry);
23444 + au_hn_free(au_hi(inode, bindex));
23446 + if (!au_test_fs_remote(h_dentry->d_sb)) {
23447 + dirwh = au_sbi(sb)->si_dirwh;
23448 + rmdir_later = (dirwh <= 1);
23449 + if (!rmdir_later)
23450 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
23453 + return rmdir_later;
23456 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
23457 + if (unlikely(err)) {
23458 + AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
23459 + h_dentry, bindex, err);
23469 + * final procedure for deleting a entry.
23470 + * maintain dentry and iattr.
23472 +static void epilog(struct inode *dir, struct dentry *dentry,
23473 + aufs_bindex_t bindex)
23475 + struct inode *inode;
23477 + inode = d_inode(dentry);
23479 + inode->i_ctime = dir->i_ctime;
23481 + au_dir_ts(dir, bindex);
23482 + inode_inc_iversion(dir);
23486 + * when an error happened, remove the created whiteout and revert everything.
23488 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
23489 + aufs_bindex_t bwh, struct dentry *wh_dentry,
23490 + struct dentry *dentry, struct au_dtime *dt)
23493 + struct path h_path = {
23494 + .dentry = wh_dentry,
23495 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
23498 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
23500 + au_set_dbwh(dentry, bwh);
23501 + au_dtime_revert(dt);
23505 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
23509 +/* ---------------------------------------------------------------------- */
23511 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
23514 + aufs_bindex_t bwh, bindex, btop;
23515 + struct inode *inode, *h_dir, *delegated;
23516 + struct dentry *parent, *wh_dentry;
23517 + /* to reduce stack size */
23519 + struct au_dtime dt;
23520 + struct au_pin pin;
23521 + struct path h_path;
23527 + a = kmalloc(sizeof(*a), GFP_NOFS);
23528 + if (unlikely(!a))
23531 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
23532 + if (unlikely(err))
23534 + err = au_d_hashed_positive(dentry);
23535 + if (unlikely(err))
23537 + inode = d_inode(dentry);
23538 + IMustLock(inode);
23540 + if (unlikely(d_is_dir(dentry)))
23541 + goto out_unlock; /* possible? */
23543 + btop = au_dbtop(dentry);
23544 + bwh = au_dbwh(dentry);
23546 + parent = dentry->d_parent; /* dir inode is locked */
23547 + di_write_lock_parent(parent);
23548 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
23550 + err = PTR_ERR(wh_dentry);
23551 + if (IS_ERR(wh_dentry))
23554 + a->h_path.mnt = au_sbr_mnt(dentry->d_sb, btop);
23555 + a->h_path.dentry = au_h_dptr(dentry, btop);
23556 + dget(a->h_path.dentry);
23557 + if (bindex == btop) {
23558 + h_dir = au_pinned_h_dir(&a->pin);
23559 + delegated = NULL;
23560 + err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
23561 + if (unlikely(err == -EWOULDBLOCK)) {
23562 + pr_warn("cannot retry for NFSv4 delegation"
23563 + " for an internal unlink\n");
23567 + /* dir inode is locked */
23568 + h_dir = d_inode(wh_dentry->d_parent);
23569 + IMustLock(h_dir);
23574 + vfsub_drop_nlink(inode);
23575 + epilog(dir, dentry, bindex);
23577 + /* update target timestamps */
23578 + if (bindex == btop) {
23579 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
23581 + inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
23583 + /* todo: this timestamp may be reverted later */
23584 + inode->i_ctime = h_dir->i_ctime;
23585 + goto out_unpin; /* success */
23592 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23599 + au_unpin(&a->pin);
23601 + dput(a->h_path.dentry);
23603 + di_write_unlock(parent);
23605 + aufs_read_unlock(dentry, AuLock_DW);
23612 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
23614 + int err, rmdir_later;
23615 + aufs_bindex_t bwh, bindex, btop;
23616 + struct inode *inode;
23617 + struct dentry *parent, *wh_dentry, *h_dentry;
23618 + struct au_whtmp_rmdir *args;
23619 + /* to reduce stack size */
23621 + struct au_dtime dt;
23622 + struct au_pin pin;
23628 + a = kmalloc(sizeof(*a), GFP_NOFS);
23629 + if (unlikely(!a))
23632 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
23633 + if (unlikely(err))
23635 + err = au_alive_dir(dentry);
23636 + if (unlikely(err))
23638 + inode = d_inode(dentry);
23639 + IMustLock(inode);
23641 + if (unlikely(!d_is_dir(dentry)))
23642 + goto out_unlock; /* possible? */
23645 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
23646 + if (unlikely(!args))
23649 + parent = dentry->d_parent; /* dir inode is locked */
23650 + di_write_lock_parent(parent);
23651 + err = au_test_empty(dentry, &args->whlist);
23652 + if (unlikely(err))
23655 + btop = au_dbtop(dentry);
23656 + bwh = au_dbwh(dentry);
23658 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
23660 + err = PTR_ERR(wh_dentry);
23661 + if (IS_ERR(wh_dentry))
23664 + h_dentry = au_h_dptr(dentry, btop);
23667 + if (bindex == btop) {
23668 + err = renwh_and_rmdir(dentry, btop, &args->whlist, dir);
23670 + rmdir_later = err;
23674 + /* stop monitoring */
23675 + au_hn_free(au_hi(inode, btop));
23677 + /* dir inode is locked */
23678 + IMustLock(d_inode(wh_dentry->d_parent));
23683 + vfsub_dead_dir(inode);
23684 + au_set_dbdiropq(dentry, -1);
23685 + epilog(dir, dentry, bindex);
23687 + if (rmdir_later) {
23688 + au_whtmp_kick_rmdir(dir, btop, h_dentry, args);
23692 + goto out_unpin; /* success */
23700 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23707 + au_unpin(&a->pin);
23711 + di_write_unlock(parent);
23713 + au_whtmp_rmdir_free(args);
23715 + aufs_read_unlock(dentry, AuLock_DW);
23722 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
23723 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
23724 +++ linux/fs/aufs/i_op_ren.c 2022-03-21 14:49:05.726633010 +0100
23726 +// SPDX-License-Identifier: GPL-2.0
23728 + * Copyright (C) 2005-2021 Junjiro R. Okajima
23730 + * This program, aufs is free software; you can redistribute it and/or modify
23731 + * it under the terms of the GNU General Public License as published by
23732 + * the Free Software Foundation; either version 2 of the License, or
23733 + * (at your option) any later version.
23735 + * This program is distributed in the hope that it will be useful,
23736 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23737 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23738 + * GNU General Public License for more details.
23740 + * You should have received a copy of the GNU General Public License
23741 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23745 + * inode operation (rename entry)
23746 + * todo: this is crazy monster
23749 +#include <linux/iversion.h>
23752 +enum { AuSRC, AuDST, AuSrcDst };
23753 +enum { AuPARENT, AuCHILD, AuParentChild };
23755 +#define AuRen_ISDIR_SRC 1
23756 +#define AuRen_ISDIR_DST (1 << 1)
23757 +#define AuRen_ISSAMEDIR (1 << 2)
23758 +#define AuRen_WHSRC (1 << 3)
23759 +#define AuRen_WHDST (1 << 4)
23760 +#define AuRen_MNT_WRITE (1 << 5)
23761 +#define AuRen_DT_DSTDIR (1 << 6)
23762 +#define AuRen_DIROPQ_SRC (1 << 7)
23763 +#define AuRen_DIROPQ_DST (1 << 8)
23764 +#define AuRen_DIRREN (1 << 9)
23765 +#define AuRen_DROPPED_SRC (1 << 10)
23766 +#define AuRen_DROPPED_DST (1 << 11)
23767 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
23768 +#define au_fset_ren(flags, name) \
23769 + do { (flags) |= AuRen_##name; } while (0)
23770 +#define au_fclr_ren(flags, name) \
23771 + do { (flags) &= ~AuRen_##name; } while (0)
23773 +#ifndef CONFIG_AUFS_DIRREN
23774 +#undef AuRen_DIRREN
23775 +#define AuRen_DIRREN 0
23778 +struct au_ren_args {
23780 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
23782 + struct inode *dir, *inode;
23783 + struct au_hinode *hdir, *hinode;
23784 + struct au_dtime dt[AuParentChild];
23785 + aufs_bindex_t btop, bdiropq;
23788 +#define src_dentry sd[AuSRC].dentry
23789 +#define src_dir sd[AuSRC].dir
23790 +#define src_inode sd[AuSRC].inode
23791 +#define src_h_dentry sd[AuSRC].h_dentry
23792 +#define src_parent sd[AuSRC].parent
23793 +#define src_h_parent sd[AuSRC].h_parent
23794 +#define src_wh_dentry sd[AuSRC].wh_dentry
23795 +#define src_hdir sd[AuSRC].hdir
23796 +#define src_hinode sd[AuSRC].hinode
23797 +#define src_h_dir sd[AuSRC].hdir->hi_inode
23798 +#define src_dt sd[AuSRC].dt
23799 +#define src_btop sd[AuSRC].btop
23800 +#define src_bdiropq sd[AuSRC].bdiropq
23802 +#define dst_dentry sd[AuDST].dentry
23803 +#define dst_dir sd[AuDST].dir
23804 +#define dst_inode sd[AuDST].inode
23805 +#define dst_h_dentry sd[AuDST].h_dentry
23806 +#define dst_parent sd[AuDST].parent
23807 +#define dst_h_parent sd[AuDST].h_parent
23808 +#define dst_wh_dentry sd[AuDST].wh_dentry
23809 +#define dst_hdir sd[AuDST].hdir
23810 +#define dst_hinode sd[AuDST].hinode
23811 +#define dst_h_dir sd[AuDST].hdir->hi_inode
23812 +#define dst_dt sd[AuDST].dt
23813 +#define dst_btop sd[AuDST].btop
23814 +#define dst_bdiropq sd[AuDST].bdiropq
23816 + struct dentry *h_trap;
23817 + struct au_branch *br;
23818 + struct path h_path;
23819 + struct au_nhash whlist;
23820 + aufs_bindex_t btgt, src_bwh;
23823 + unsigned short auren_flags;
23824 + unsigned char flags; /* syscall parameter */
23825 + unsigned char exchange;
23828 + struct au_whtmp_rmdir *thargs;
23829 + struct dentry *h_dst;
23830 + struct au_hinode *h_root;
23833 +/* ---------------------------------------------------------------------- */
23836 + * functions for reverting.
23837 + * when an error happened in a single rename systemcall, we should revert
23838 + * everything as if nothing happened.
23839 + * we don't need to revert the copied-up/down the parent dir since they are
23843 +#define RevertFailure(fmt, ...) do { \
23844 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
23845 + ##__VA_ARGS__, err, rerr); \
23849 +static void au_ren_do_rev_diropq(int err, struct au_ren_args *a, int idx)
23852 + struct dentry *d;
23853 +#define src_or_dst(member) a->sd[idx].member
23855 + d = src_or_dst(dentry); /* {src,dst}_dentry */
23856 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
23857 + rerr = au_diropq_remove(d, a->btgt);
23858 + au_hn_inode_unlock(src_or_dst(hinode));
23859 + au_set_dbdiropq(d, src_or_dst(bdiropq));
23861 + RevertFailure("remove diropq %pd", d);
23863 +#undef src_or_dst_
23866 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
23868 + if (au_ftest_ren(a->auren_flags, DIROPQ_SRC))
23869 + au_ren_do_rev_diropq(err, a, AuSRC);
23870 + if (au_ftest_ren(a->auren_flags, DIROPQ_DST))
23871 + au_ren_do_rev_diropq(err, a, AuDST);
23874 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
23877 + struct inode *delegated;
23878 + struct path h_ppath = {
23879 + .dentry = a->src_h_parent,
23880 + .mnt = a->h_path.mnt
23883 + a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name, &h_ppath);
23884 + rerr = PTR_ERR(a->h_path.dentry);
23885 + if (IS_ERR(a->h_path.dentry)) {
23886 + RevertFailure("lkup one %pd", a->src_dentry);
23890 + delegated = NULL;
23891 + rerr = vfsub_rename(a->dst_h_dir,
23892 + au_h_dptr(a->src_dentry, a->btgt),
23893 + a->src_h_dir, &a->h_path, &delegated, a->flags);
23894 + if (unlikely(rerr == -EWOULDBLOCK)) {
23895 + pr_warn("cannot retry for NFSv4 delegation"
23896 + " for an internal rename\n");
23899 + d_drop(a->h_path.dentry);
23900 + dput(a->h_path.dentry);
23901 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
23903 + RevertFailure("rename %pd", a->src_dentry);
23906 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
23909 + struct inode *delegated;
23910 + struct path h_ppath = {
23911 + .dentry = a->dst_h_parent,
23912 + .mnt = a->h_path.mnt
23915 + a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name, &h_ppath);
23916 + rerr = PTR_ERR(a->h_path.dentry);
23917 + if (IS_ERR(a->h_path.dentry)) {
23918 + RevertFailure("lkup one %pd", a->dst_dentry);
23921 + if (d_is_positive(a->h_path.dentry)) {
23922 + d_drop(a->h_path.dentry);
23923 + dput(a->h_path.dentry);
23927 + delegated = NULL;
23928 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
23929 + &delegated, a->flags);
23930 + if (unlikely(rerr == -EWOULDBLOCK)) {
23931 + pr_warn("cannot retry for NFSv4 delegation"
23932 + " for an internal rename\n");
23935 + d_drop(a->h_path.dentry);
23936 + dput(a->h_path.dentry);
23938 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
23940 + RevertFailure("rename %pd", a->h_dst);
23943 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
23947 + a->h_path.dentry = a->src_wh_dentry;
23948 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
23949 + au_set_dbwh(a->src_dentry, a->src_bwh);
23951 + RevertFailure("unlink %pd", a->src_wh_dentry);
23953 +#undef RevertFailure
23955 +/* ---------------------------------------------------------------------- */
23958 + * when we have to copyup the renaming entry, do it with the rename-target name
23959 + * in order to minimize the cost (the later actual rename is unnecessary).
23960 + * otherwise rename it on the target branch.
23962 +static int au_ren_or_cpup(struct au_ren_args *a)
23965 + struct dentry *d;
23966 + struct inode *delegated;
23968 + d = a->src_dentry;
23969 + if (au_dbtop(d) == a->btgt) {
23970 + a->h_path.dentry = a->dst_h_dentry;
23971 + AuDebugOn(au_dbtop(d) != a->btgt);
23972 + delegated = NULL;
23973 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
23974 + a->dst_h_dir, &a->h_path, &delegated,
23976 + if (unlikely(err == -EWOULDBLOCK)) {
23977 + pr_warn("cannot retry for NFSv4 delegation"
23978 + " for an internal rename\n");
23984 + if (!err && a->h_dst)
23985 + /* it will be set to dinfo later */
23991 +/* cf. aufs_rmdir() */
23992 +static int au_ren_del_whtmp(struct au_ren_args *a)
23995 + struct inode *dir;
23997 + dir = a->dst_dir;
23998 + SiMustAnyLock(dir->i_sb);
23999 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
24000 + au_sbi(dir->i_sb)->si_dirwh)
24001 + || au_test_fs_remote(a->h_dst->d_sb)) {
24002 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
24003 + if (unlikely(err))
24004 + pr_warn("failed removing whtmp dir %pd (%d), "
24005 + "ignored.\n", a->h_dst, err);
24007 + au_nhash_wh_free(&a->thargs->whlist);
24008 + a->thargs->whlist = a->whlist;
24009 + a->whlist.nh_num = 0;
24010 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
24012 + a->thargs = NULL;
24018 +/* make it 'opaque' dir. */
24019 +static int au_ren_do_diropq(struct au_ren_args *a, int idx)
24022 + struct dentry *d, *diropq;
24023 +#define src_or_dst(member) a->sd[idx].member
24026 + d = src_or_dst(dentry); /* {src,dst}_dentry */
24027 + src_or_dst(bdiropq) = au_dbdiropq(d);
24028 + src_or_dst(hinode) = au_hi(src_or_dst(inode), a->btgt);
24029 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
24030 + diropq = au_diropq_create(d, a->btgt);
24031 + au_hn_inode_unlock(src_or_dst(hinode));
24032 + if (IS_ERR(diropq))
24033 + err = PTR_ERR(diropq);
24037 +#undef src_or_dst_
24041 +static int au_ren_diropq(struct au_ren_args *a)
24044 + unsigned char always;
24045 + struct dentry *d;
24048 + d = a->dst_dentry; /* already renamed on the branch */
24049 + always = !!au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ);
24050 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24051 + && !au_ftest_ren(a->auren_flags, DIRREN)
24052 + && a->btgt != au_dbdiropq(a->src_dentry)
24053 + && (a->dst_wh_dentry
24054 + || a->btgt <= au_dbdiropq(d)
24055 + /* hide the lower to keep xino */
24056 + /* the lowers may not be a dir, but we hide them anyway */
24057 + || a->btgt < au_dbbot(d)
24060 + err = au_ren_do_diropq(a, AuSRC);
24061 + if (unlikely(err))
24063 + au_fset_ren(a->auren_flags, DIROPQ_SRC);
24065 + if (!a->exchange)
24066 + goto out; /* success */
24068 + d = a->src_dentry; /* already renamed on the branch */
24069 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24070 + && a->btgt != au_dbdiropq(a->dst_dentry)
24071 + && (a->btgt < au_dbdiropq(d)
24072 + || a->btgt < au_dbbot(d)
24074 + AuDbgDentry(a->src_dentry);
24075 + AuDbgDentry(a->dst_dentry);
24076 + err = au_ren_do_diropq(a, AuDST);
24077 + if (unlikely(err))
24078 + goto out_rev_src;
24079 + au_fset_ren(a->auren_flags, DIROPQ_DST);
24081 + goto out; /* success */
24084 + AuDbg("err %d, reverting src\n", err);
24085 + au_ren_rev_diropq(err, a);
24090 +static int do_rename(struct au_ren_args *a)
24093 + struct dentry *d, *h_d;
24095 + if (!a->exchange) {
24096 + /* prepare workqueue args for asynchronous rmdir */
24097 + h_d = a->dst_h_dentry;
24098 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24099 + /* && !au_ftest_ren(a->auren_flags, DIRREN) */
24100 + && d_is_positive(h_d)) {
24102 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb,
24104 + if (unlikely(!a->thargs))
24106 + a->h_dst = dget(h_d);
24109 + /* create whiteout for src_dentry */
24110 + if (au_ftest_ren(a->auren_flags, WHSRC)) {
24111 + a->src_bwh = au_dbwh(a->src_dentry);
24112 + AuDebugOn(a->src_bwh >= 0);
24113 + a->src_wh_dentry = au_wh_create(a->src_dentry, a->btgt,
24114 + a->src_h_parent);
24115 + err = PTR_ERR(a->src_wh_dentry);
24116 + if (IS_ERR(a->src_wh_dentry))
24120 + /* lookup whiteout for dentry */
24121 + if (au_ftest_ren(a->auren_flags, WHDST)) {
24122 + h_d = au_wh_lkup(a->dst_h_parent,
24123 + &a->dst_dentry->d_name, a->br);
24124 + err = PTR_ERR(h_d);
24127 + if (d_is_negative(h_d))
24130 + a->dst_wh_dentry = h_d;
24133 + /* rename dentry to tmpwh */
24135 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
24136 + if (unlikely(err))
24139 + d = a->dst_dentry;
24140 + au_set_h_dptr(d, a->btgt, NULL);
24141 + err = au_lkup_neg(d, a->btgt, /*wh*/0);
24142 + if (unlikely(err))
24144 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
24148 + BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_btop != a->btgt);
24149 +#if 0 /* debugging */
24150 + BUG_ON(!au_ftest_ren(a->auren_flags, DIRREN)
24151 + && d_is_positive(a->dst_h_dentry)
24152 + && a->src_btop != a->btgt);
24155 + /* rename by vfs_rename or cpup */
24156 + err = au_ren_or_cpup(a);
24157 + if (unlikely(err))
24158 + /* leave the copied-up one */
24161 + /* make dir opaque */
24162 + err = au_ren_diropq(a);
24163 + if (unlikely(err))
24166 + /* update target timestamps */
24167 + if (a->exchange) {
24168 + AuDebugOn(au_dbtop(a->dst_dentry) != a->btgt);
24169 + a->h_path.dentry = au_h_dptr(a->dst_dentry, a->btgt);
24170 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24171 + a->dst_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24173 + AuDebugOn(au_dbtop(a->src_dentry) != a->btgt);
24174 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
24175 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24176 + a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24178 + if (!a->exchange) {
24179 + /* remove whiteout for dentry */
24180 + if (a->dst_wh_dentry) {
24181 + a->h_path.dentry = a->dst_wh_dentry;
24182 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
24184 + if (unlikely(err))
24188 + /* remove whtmp */
24190 + au_ren_del_whtmp(a); /* ignore this error */
24192 + au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
24195 + goto out_success;
24198 + au_ren_rev_diropq(err, a);
24200 + au_ren_rev_rename(err, a);
24204 + au_ren_rev_whtmp(err, a);
24206 + dput(a->dst_wh_dentry);
24207 + a->dst_wh_dentry = NULL;
24209 + if (a->src_wh_dentry)
24210 + au_ren_rev_whsrc(err, a);
24212 + dput(a->src_wh_dentry);
24213 + dput(a->dst_wh_dentry);
24217 + au_whtmp_rmdir_free(a->thargs);
24218 + a->thargs = NULL;
24224 +/* ---------------------------------------------------------------------- */
24227 + * test if @dentry dir can be rename destination or not.
24228 + * success means, it is a logically empty dir.
24230 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
24232 + return au_test_empty(dentry, whlist);
24236 + * test if @a->src_dentry dir can be rename source or not.
24237 + * if it can, return 0.
24239 + * - it is a logically empty dir.
24240 + * - or, it exists on writable branch and has no children including whiteouts
24241 + * on the lower branch unless DIRREN is on.
24243 +static int may_rename_srcdir(struct au_ren_args *a)
24246 + unsigned int rdhash;
24247 + aufs_bindex_t btop, btgt;
24248 + struct dentry *dentry;
24249 + struct super_block *sb;
24250 + struct au_sbinfo *sbinfo;
24252 + dentry = a->src_dentry;
24253 + sb = dentry->d_sb;
24254 + sbinfo = au_sbi(sb);
24255 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
24256 + au_fset_ren(a->auren_flags, DIRREN);
24259 + btop = au_dbtop(dentry);
24260 + if (btop != btgt) {
24261 + struct au_nhash whlist;
24263 + SiMustAnyLock(sb);
24264 + rdhash = sbinfo->si_rdhash;
24266 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
24268 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
24269 + if (unlikely(err))
24271 + err = au_test_empty(dentry, &whlist);
24272 + au_nhash_wh_free(&whlist);
24276 + if (btop == au_dbtaildir(dentry))
24277 + return 0; /* success */
24279 + err = au_test_empty_lower(dentry);
24282 + if (err == -ENOTEMPTY) {
24283 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24286 + AuWarn1("renaming dir who has child(ren) on multiple "
24287 + "branches, is not supported\n");
24294 +/* side effect: sets whlist and h_dentry */
24295 +static int au_ren_may_dir(struct au_ren_args *a)
24298 + unsigned int rdhash;
24299 + struct dentry *d;
24301 + d = a->dst_dentry;
24302 + SiMustAnyLock(d->d_sb);
24305 + if (au_ftest_ren(a->auren_flags, ISDIR_DST) && a->dst_inode) {
24306 + rdhash = au_sbi(d->d_sb)->si_rdhash;
24308 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
24309 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
24310 + if (unlikely(err))
24313 + if (!a->exchange) {
24314 + au_set_dbtop(d, a->dst_btop);
24315 + err = may_rename_dstdir(d, &a->whlist);
24316 + au_set_dbtop(d, a->btgt);
24318 + err = may_rename_srcdir(a);
24320 + a->dst_h_dentry = au_h_dptr(d, au_dbtop(d));
24321 + if (unlikely(err))
24324 + d = a->src_dentry;
24325 + a->src_h_dentry = au_h_dptr(d, au_dbtop(d));
24326 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24327 + err = may_rename_srcdir(a);
24328 + if (unlikely(err)) {
24329 + au_nhash_wh_free(&a->whlist);
24330 + a->whlist.nh_num = 0;
24337 +/* ---------------------------------------------------------------------- */
24340 + * simple tests for rename.
24341 + * following the checks in vfs, plus the parent-child relationship.
24343 +static int au_may_ren(struct au_ren_args *a)
24346 + struct inode *h_inode;
24348 + if (a->src_btop == a->btgt) {
24349 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
24350 + au_ftest_ren(a->auren_flags, ISDIR_SRC));
24351 + if (unlikely(err))
24354 + if (unlikely(a->src_h_dentry == a->h_trap))
24359 + if (a->dst_btop != a->btgt)
24362 + err = -ENOTEMPTY;
24363 + if (unlikely(a->dst_h_dentry == a->h_trap))
24367 + isdir = !!au_ftest_ren(a->auren_flags, ISDIR_DST);
24368 + if (d_really_is_negative(a->dst_dentry)) {
24369 + if (d_is_negative(a->dst_h_dentry))
24370 + err = au_may_add(a->dst_dentry, a->btgt,
24371 + a->dst_h_parent, isdir);
24373 + if (unlikely(d_is_negative(a->dst_h_dentry)))
24375 + h_inode = d_inode(a->dst_h_dentry);
24376 + if (h_inode->i_nlink)
24377 + err = au_may_del(a->dst_dentry, a->btgt,
24378 + a->dst_h_parent, isdir);
24382 + if (unlikely(err == -ENOENT || err == -EEXIST))
24388 +/* ---------------------------------------------------------------------- */
24393 + * - src_dir and dir by lock_rename()
24394 + * - inode if exists
24397 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
24399 + * + di_write_lock2_child()
24400 + * + di_write_lock_child()
24401 + * + ii_write_lock_child()
24402 + * + di_write_lock_child2()
24403 + * + ii_write_lock_child2()
24404 + * + src_parent and parent
24405 + * + di_write_lock_parent()
24406 + * + ii_write_lock_parent()
24407 + * + di_write_lock_parent2()
24408 + * + ii_write_lock_parent2()
24409 + * + lower src_dir and dir by vfsub_lock_rename()
24410 + * + verify the every relationships between child and parent. if any
24411 + * of them failed, unlock all and return -EBUSY.
24413 +static void au_ren_unlock(struct au_ren_args *a)
24415 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
24416 + a->dst_h_parent, a->dst_hdir);
24417 + if (au_ftest_ren(a->auren_flags, DIRREN)
24419 + au_hn_inode_unlock(a->h_root);
24420 + if (au_ftest_ren(a->auren_flags, MNT_WRITE))
24421 + vfsub_mnt_drop_write(au_br_mnt(a->br));
24424 +static int au_ren_lock(struct au_ren_args *a)
24427 + unsigned int udba;
24430 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
24431 + a->src_hdir = au_hi(a->src_dir, a->btgt);
24432 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
24433 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
24435 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
24436 + if (unlikely(err))
24438 + au_fset_ren(a->auren_flags, MNT_WRITE);
24439 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24440 + struct dentry *root;
24441 + struct inode *dir;
24444 + * sbinfo is already locked, so this ii_read_lock is
24445 + * unnecessary. but our debugging feature checks it.
24447 + root = a->src_inode->i_sb->s_root;
24448 + if (root != a->src_parent && root != a->dst_parent) {
24449 + dir = d_inode(root);
24450 + ii_read_lock_parent3(dir);
24451 + a->h_root = au_hi(dir, a->btgt);
24452 + ii_read_unlock(dir);
24453 + au_hn_inode_lock_nested(a->h_root, AuLsc_I_PARENT3);
24456 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
24457 + a->dst_h_parent, a->dst_hdir);
24458 + udba = au_opt_udba(a->src_dentry->d_sb);
24459 + if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
24460 + || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
24461 + err = au_busy_or_stale();
24462 + if (!err && au_dbtop(a->src_dentry) == a->btgt)
24463 + err = au_h_verify(a->src_h_dentry, udba,
24464 + d_inode(a->src_h_parent), a->src_h_parent,
24466 + if (!err && au_dbtop(a->dst_dentry) == a->btgt)
24467 + err = au_h_verify(a->dst_h_dentry, udba,
24468 + d_inode(a->dst_h_parent), a->dst_h_parent,
24471 + goto out; /* success */
24473 + err = au_busy_or_stale();
24474 + au_ren_unlock(a);
24480 +/* ---------------------------------------------------------------------- */
24482 +static void au_ren_refresh_dir(struct au_ren_args *a)
24484 + struct inode *dir;
24486 + dir = a->dst_dir;
24487 + inode_inc_iversion(dir);
24488 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24489 + /* is this updating defined in POSIX? */
24490 + au_cpup_attr_timesizes(a->src_inode);
24491 + au_cpup_attr_nlink(dir, /*force*/1);
24493 + au_dir_ts(dir, a->btgt);
24495 + if (a->exchange) {
24496 + dir = a->src_dir;
24497 + inode_inc_iversion(dir);
24498 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24499 + /* is this updating defined in POSIX? */
24500 + au_cpup_attr_timesizes(a->dst_inode);
24501 + au_cpup_attr_nlink(dir, /*force*/1);
24503 + au_dir_ts(dir, a->btgt);
24506 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24509 + dir = a->src_dir;
24510 + inode_inc_iversion(dir);
24511 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC))
24512 + au_cpup_attr_nlink(dir, /*force*/1);
24513 + au_dir_ts(dir, a->btgt);
24516 +static void au_ren_refresh(struct au_ren_args *a)
24518 + aufs_bindex_t bbot, bindex;
24519 + struct dentry *d, *h_d;
24520 + struct inode *i, *h_i;
24521 + struct super_block *sb;
24523 + d = a->dst_dentry;
24526 + /* already dget-ed by au_ren_or_cpup() */
24527 + au_set_h_dptr(d, a->btgt, a->h_dst);
24529 + i = a->dst_inode;
24531 + if (!a->exchange) {
24532 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST))
24533 + vfsub_drop_nlink(i);
24535 + vfsub_dead_dir(i);
24536 + au_cpup_attr_timesizes(i);
24538 + au_update_dbrange(d, /*do_put_zero*/1);
24540 + au_cpup_attr_nlink(i, /*force*/1);
24543 + for (bindex = au_dbtop(d); bindex < bbot; bindex++)
24544 + au_set_h_dptr(d, bindex, NULL);
24545 + bbot = au_dbbot(d);
24546 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++)
24547 + au_set_h_dptr(d, bindex, NULL);
24548 + au_update_dbrange(d, /*do_put_zero*/0);
24552 + || au_ftest_ren(a->auren_flags, DIRREN)) {
24553 + d_drop(a->src_dentry);
24554 + if (au_ftest_ren(a->auren_flags, DIRREN))
24555 + au_set_dbwh(a->src_dentry, -1);
24559 + d = a->src_dentry;
24560 + au_set_dbwh(d, -1);
24561 + bbot = au_dbbot(d);
24562 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24563 + h_d = au_h_dptr(d, bindex);
24565 + au_set_h_dptr(d, bindex, NULL);
24567 + au_set_dbbot(d, a->btgt);
24570 + i = a->src_inode;
24571 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
24572 + return; /* success */
24574 + bbot = au_ibbot(i);
24575 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24576 + h_i = au_h_iptr(i, bindex);
24578 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
24579 + /* ignore this error */
24580 + au_set_h_iptr(i, bindex, NULL, 0);
24583 + au_set_ibbot(i, a->btgt);
24586 +/* ---------------------------------------------------------------------- */
24588 +/* mainly for link(2) and rename(2) */
24589 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
24591 + aufs_bindex_t bdiropq, bwh;
24592 + struct dentry *parent;
24593 + struct au_branch *br;
24595 + parent = dentry->d_parent;
24596 + IMustLock(d_inode(parent)); /* dir is locked */
24598 + bdiropq = au_dbdiropq(parent);
24599 + bwh = au_dbwh(dentry);
24600 + br = au_sbr(dentry->d_sb, btgt);
24601 + if (au_br_rdonly(br)
24602 + || (0 <= bdiropq && bdiropq < btgt)
24603 + || (0 <= bwh && bwh < btgt))
24606 + AuDbg("btgt %d\n", btgt);
24610 +/* sets src_btop, dst_btop and btgt */
24611 +static int au_ren_wbr(struct au_ren_args *a)
24614 + struct au_wr_dir_args wr_dir_args = {
24615 + /* .force_btgt = -1, */
24616 + .flags = AuWrDir_ADD_ENTRY
24619 + a->src_btop = au_dbtop(a->src_dentry);
24620 + a->dst_btop = au_dbtop(a->dst_dentry);
24621 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24622 + || au_ftest_ren(a->auren_flags, ISDIR_DST))
24623 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
24624 + wr_dir_args.force_btgt = a->src_btop;
24625 + if (a->dst_inode && a->dst_btop < a->src_btop)
24626 + wr_dir_args.force_btgt = a->dst_btop;
24627 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
24628 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
24631 + au_update_dbtop(a->dst_dentry);
24636 +static void au_ren_dt(struct au_ren_args *a)
24638 + a->h_path.dentry = a->src_h_parent;
24639 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
24640 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR)) {
24641 + a->h_path.dentry = a->dst_h_parent;
24642 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
24645 + au_fclr_ren(a->auren_flags, DT_DSTDIR);
24646 + if (!au_ftest_ren(a->auren_flags, ISDIR_SRC)
24650 + a->h_path.dentry = a->src_h_dentry;
24651 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
24652 + if (d_is_positive(a->dst_h_dentry)) {
24653 + au_fset_ren(a->auren_flags, DT_DSTDIR);
24654 + a->h_path.dentry = a->dst_h_dentry;
24655 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
24659 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
24661 + struct dentry *h_d;
24662 + struct inode *h_inode;
24664 + au_dtime_revert(a->src_dt + AuPARENT);
24665 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR))
24666 + au_dtime_revert(a->dst_dt + AuPARENT);
24668 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC) && err != -EIO) {
24669 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
24670 + h_inode = d_inode(h_d);
24671 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
24672 + au_dtime_revert(a->src_dt + AuCHILD);
24673 + inode_unlock(h_inode);
24675 + if (au_ftest_ren(a->auren_flags, DT_DSTDIR)) {
24676 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
24677 + h_inode = d_inode(h_d);
24678 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
24679 + au_dtime_revert(a->dst_dt + AuCHILD);
24680 + inode_unlock(h_inode);
24685 +/* ---------------------------------------------------------------------- */
24687 +int aufs_rename(struct user_namespace *userns,
24688 + struct inode *_src_dir, struct dentry *_src_dentry,
24689 + struct inode *_dst_dir, struct dentry *_dst_dentry,
24690 + unsigned int _flags)
24692 + int err, lock_flags;
24694 + /* reduce stack space */
24695 + struct au_ren_args *a;
24696 + struct au_pin pin;
24698 + AuDbg("%pd, %pd, 0x%x\n", _src_dentry, _dst_dentry, _flags);
24699 + IMustLock(_src_dir);
24700 + IMustLock(_dst_dir);
24703 + if (unlikely(_flags & RENAME_WHITEOUT))
24707 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
24708 + a = kzalloc(sizeof(*a), GFP_NOFS);
24709 + if (unlikely(!a))
24712 + a->flags = _flags;
24713 + BUILD_BUG_ON(sizeof(a->exchange) == sizeof(u8)
24714 + && RENAME_EXCHANGE > U8_MAX);
24715 + a->exchange = _flags & RENAME_EXCHANGE;
24716 + a->src_dir = _src_dir;
24717 + a->src_dentry = _src_dentry;
24718 + a->src_inode = NULL;
24719 + if (d_really_is_positive(a->src_dentry))
24720 + a->src_inode = d_inode(a->src_dentry);
24721 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
24722 + a->dst_dir = _dst_dir;
24723 + a->dst_dentry = _dst_dentry;
24724 + a->dst_inode = NULL;
24725 + if (d_really_is_positive(a->dst_dentry))
24726 + a->dst_inode = d_inode(a->dst_dentry);
24727 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
24728 + if (a->dst_inode) {
24730 + * if EXCHANGE && src is non-dir && dst is dir,
24731 + * dst is not locked.
24733 + /* IMustLock(a->dst_inode); */
24734 + au_igrab(a->dst_inode);
24738 + lock_flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
24739 + if (d_is_dir(a->src_dentry)) {
24740 + au_fset_ren(a->auren_flags, ISDIR_SRC);
24741 + if (unlikely(!a->exchange
24742 + && d_really_is_positive(a->dst_dentry)
24743 + && !d_is_dir(a->dst_dentry)))
24745 + lock_flags |= AuLock_DIRS;
24747 + if (a->dst_inode && d_is_dir(a->dst_dentry)) {
24748 + au_fset_ren(a->auren_flags, ISDIR_DST);
24749 + if (unlikely(!a->exchange
24750 + && d_really_is_positive(a->src_dentry)
24751 + && !d_is_dir(a->src_dentry)))
24753 + lock_flags |= AuLock_DIRS;
24755 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
24757 + if (unlikely(err))
24760 + err = au_d_hashed_positive(a->src_dentry);
24761 + if (unlikely(err))
24764 + if (a->dst_inode) {
24766 + * If it is a dir, VFS unhash it before this
24767 + * function. It means we cannot rely upon d_unhashed().
24769 + if (unlikely(!a->dst_inode->i_nlink))
24771 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24772 + err = au_d_hashed_positive(a->dst_dentry);
24773 + if (unlikely(err && !a->exchange))
24775 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
24777 + } else if (unlikely(d_unhashed(a->dst_dentry)))
24781 + * is it possible?
24782 + * yes, it happened (in linux-3.3-rcN) but I don't know why.
24783 + * there may exist a problem somewhere else.
24786 + if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
24789 + au_fset_ren(a->auren_flags, ISSAMEDIR); /* temporary */
24790 + di_write_lock_parent(a->dst_parent);
24792 + /* which branch we process */
24793 + err = au_ren_wbr(a);
24794 + if (unlikely(err < 0))
24796 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
24797 + a->h_path.mnt = au_br_mnt(a->br);
24799 + /* are they available to be renamed */
24800 + err = au_ren_may_dir(a);
24801 + if (unlikely(err))
24802 + goto out_children;
24804 + /* prepare the writable parent dir on the same branch */
24805 + if (a->dst_btop == a->btgt) {
24806 + au_fset_ren(a->auren_flags, WHDST);
24808 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
24809 + if (unlikely(err))
24810 + goto out_children;
24814 + if (!a->exchange) {
24815 + if (a->src_dir != a->dst_dir) {
24817 + * this temporary unlock is safe,
24818 + * because both dir->i_mutex are locked.
24820 + di_write_unlock(a->dst_parent);
24821 + di_write_lock_parent(a->src_parent);
24822 + err = au_wr_dir_need_wh(a->src_dentry,
24823 + au_ftest_ren(a->auren_flags,
24826 + di_write_unlock(a->src_parent);
24827 + di_write_lock2_parent(a->src_parent, a->dst_parent,
24829 + au_fclr_ren(a->auren_flags, ISSAMEDIR);
24831 + err = au_wr_dir_need_wh(a->src_dentry,
24832 + au_ftest_ren(a->auren_flags,
24836 + if (unlikely(err < 0))
24837 + goto out_children;
24839 + au_fset_ren(a->auren_flags, WHSRC);
24842 + if (a->src_btop != a->btgt) {
24843 + err = au_pin(&pin, a->src_dentry, a->btgt,
24844 + au_opt_udba(a->src_dentry->d_sb),
24845 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24847 + struct au_cp_generic cpg = {
24848 + .dentry = a->src_dentry,
24850 + .bsrc = a->src_btop,
24853 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24855 + AuDebugOn(au_dbtop(a->src_dentry) != a->src_btop);
24856 + err = au_sio_cpup_simple(&cpg);
24859 + if (unlikely(err))
24860 + goto out_children;
24861 + a->src_btop = a->btgt;
24862 + a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
24863 + if (!a->exchange)
24864 + au_fset_ren(a->auren_flags, WHSRC);
24868 + if (a->exchange && a->dst_inode
24869 + && a->dst_btop != a->btgt) {
24870 + err = au_pin(&pin, a->dst_dentry, a->btgt,
24871 + au_opt_udba(a->dst_dentry->d_sb),
24872 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24874 + struct au_cp_generic cpg = {
24875 + .dentry = a->dst_dentry,
24877 + .bsrc = a->dst_btop,
24880 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24882 + err = au_sio_cpup_simple(&cpg);
24885 + if (unlikely(err))
24886 + goto out_children;
24887 + a->dst_btop = a->btgt;
24888 + a->dst_h_dentry = au_h_dptr(a->dst_dentry, a->btgt);
24891 + /* lock them all */
24892 + err = au_ren_lock(a);
24893 + if (unlikely(err))
24894 + /* leave the copied-up one */
24895 + goto out_children;
24897 + if (!a->exchange) {
24898 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
24899 + err = au_may_ren(a);
24900 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
24901 + err = -ENAMETOOLONG;
24902 + if (unlikely(err))
24906 + /* store timestamps to be revertible */
24909 + /* store dirren info */
24910 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24911 + err = au_dr_rename(a->src_dentry, a->btgt,
24912 + &a->dst_dentry->d_name, &rev);
24914 + if (unlikely(err))
24919 + err = do_rename(a);
24920 + if (unlikely(err))
24923 + if (au_ftest_ren(a->auren_flags, DIRREN))
24924 + au_dr_rename_fin(a->src_dentry, a->btgt, rev);
24926 + /* update dir attributes */
24927 + au_ren_refresh_dir(a);
24929 + /* dput/iput all lower dentries */
24930 + au_ren_refresh(a);
24932 + goto out_hdir; /* success */
24935 + if (au_ftest_ren(a->auren_flags, DIRREN))
24936 + au_dr_rename_rev(a->src_dentry, a->btgt, rev);
24938 + au_ren_rev_dt(err, a);
24940 + au_ren_unlock(a);
24942 + au_nhash_wh_free(&a->whlist);
24943 + if (err && a->dst_inode && a->dst_btop != a->btgt) {
24944 + AuDbg("btop %d, btgt %d\n", a->dst_btop, a->btgt);
24945 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
24946 + au_set_dbtop(a->dst_dentry, a->dst_btop);
24950 + if (d_unhashed(a->src_dentry))
24951 + au_fset_ren(a->auren_flags, DROPPED_SRC);
24952 + if (d_unhashed(a->dst_dentry))
24953 + au_fset_ren(a->auren_flags, DROPPED_DST);
24954 + if (!a->exchange)
24955 + d_move(a->src_dentry, a->dst_dentry);
24957 + d_exchange(a->src_dentry, a->dst_dentry);
24958 + if (au_ftest_ren(a->auren_flags, DROPPED_DST))
24959 + d_drop(a->dst_dentry);
24961 + if (au_ftest_ren(a->auren_flags, DROPPED_SRC))
24962 + d_drop(a->src_dentry);
24964 + au_update_dbtop(a->dst_dentry);
24965 + if (!a->dst_inode)
24966 + d_drop(a->dst_dentry);
24968 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24969 + di_write_unlock(a->dst_parent);
24971 + di_write_unlock2(a->src_parent, a->dst_parent);
24973 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
24975 + iput(a->dst_inode);
24977 + au_whtmp_rmdir_free(a->thargs);
24983 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
24984 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
24985 +++ linux/fs/aufs/Kconfig 2022-03-21 14:49:05.716633010 +0100
24987 +# SPDX-License-Identifier: GPL-2.0
24989 + tristate "Aufs (Advanced multi layered unification filesystem) support"
24991 + Aufs is a stackable unification filesystem such as Unionfs,
24992 + which unifies several directories and provides a merged single
24994 + In the early days, aufs was entirely re-designed and
24995 + re-implemented Unionfs Version 1.x series. Introducing many
24996 + original ideas, approaches and improvements, it becomes totally
24997 + different from Unionfs while keeping the basic features.
25001 + prompt "Maximum number of branches"
25002 + default AUFS_BRANCH_MAX_127
25004 + Specifies the maximum number of branches (or member directories)
25005 + in a single aufs. The larger value consumes more system
25006 + resources and has a minor impact to performance.
25007 +config AUFS_BRANCH_MAX_127
25010 + Specifies the maximum number of branches (or member directories)
25011 + in a single aufs. The larger value consumes more system
25012 + resources and has a minor impact to performance.
25013 +config AUFS_BRANCH_MAX_511
25016 + Specifies the maximum number of branches (or member directories)
25017 + in a single aufs. The larger value consumes more system
25018 + resources and has a minor impact to performance.
25019 +config AUFS_BRANCH_MAX_1023
25022 + Specifies the maximum number of branches (or member directories)
25023 + in a single aufs. The larger value consumes more system
25024 + resources and has a minor impact to performance.
25025 +config AUFS_BRANCH_MAX_32767
25028 + Specifies the maximum number of branches (or member directories)
25029 + in a single aufs. The larger value consumes more system
25030 + resources and has a minor impact to performance.
25033 +config AUFS_SBILIST
25035 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
25038 + Automatic configuration for internal use.
25039 + When aufs supports Magic SysRq or /proc, enabled automatically.
25041 +config AUFS_HNOTIFY
25042 + bool "Detect direct branch access (bypassing aufs)"
25044 + If you want to modify files on branches directly, eg. bypassing aufs,
25045 + and want aufs to detect the changes of them fully, then enable this
25046 + option and use 'udba=notify' mount option.
25047 + Currently there is only one available configuration, "fsnotify".
25048 + It will have a negative impact to the performance.
25049 + See detail in aufs.5.
25052 + prompt "method" if AUFS_HNOTIFY
25053 + default AUFS_HFSNOTIFY
25054 +config AUFS_HFSNOTIFY
25059 +config AUFS_EXPORT
25060 + bool "NFS-exportable aufs"
25061 + depends on EXPORTFS
25063 + If you want to export your mounted aufs via NFS, then enable this
25064 + option. There are several requirements for this configuration.
25065 + See detail in aufs.5.
25067 +config AUFS_INO_T_64
25069 + depends on AUFS_EXPORT
25070 + depends on 64BIT && !(ALPHA || S390)
25073 + Automatic configuration for internal use.
25074 + /* typedef unsigned long/int __kernel_ino_t */
25075 + /* alpha and s390x are int */
25078 + bool "support for XATTR/EA (including Security Labels)"
25080 + If your branch fs supports XATTR/EA and you want to make them
25081 + available in aufs too, then enable this opsion and specify the
25082 + branch attributes for EA.
25083 + See detail in aufs.5.
25086 + bool "File-based Hierarchical Storage Management"
25088 + Hierarchical Storage Management (or HSM) is a well-known feature
25089 + in the storage world. Aufs provides this feature as file-based.
25090 + with multiple branches.
25091 + These multiple branches are prioritized, ie. the topmost one
25092 + should be the fastest drive and be used heavily.
25095 + bool "Readdir in userspace"
25097 + Aufs has two methods to provide a merged view for a directory,
25098 + by a user-space library and by kernel-space natively. The latter
25099 + is always enabled but sometimes large and slow.
25100 + If you enable this option, install the library in aufs2-util
25101 + package, and set some environment variables for your readdir(3),
25102 + then the work will be handled in user-space which generally
25103 + shows better performance in most cases.
25104 + See detail in aufs.5.
25106 +config AUFS_DIRREN
25107 + bool "Workaround for rename(2)-ing a directory"
25109 + By default, aufs returns EXDEV error in renameing a dir who has
25110 + his child on the lower branch, since it is a bad idea to issue
25111 + rename(2) internally for every lower branch. But user may not
25112 + accept this behaviour. So here is a workaround to allow such
25113 + rename(2) and store some extra infromation on the writable
25114 + branch. Obviously this costs high (and I don't like it).
25115 + To use this feature, you need to enable this configuration AND
25116 + to specify the mount option `dirren.'
25117 + See details in aufs.5 and the design documents.
25120 + bool "Show whiteouts"
25122 + If you want to make the whiteouts in aufs visible, then enable
25123 + this option and specify 'shwh' mount option. Although it may
25124 + sounds like philosophy or something, but in technically it
25125 + simply shows the name of whiteout with keeping its behaviour.
25127 +config AUFS_BR_RAMFS
25128 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
25130 + If you want to use ramfs as an aufs branch fs, then enable this
25131 + option. Generally tmpfs is recommended.
25132 + Aufs prohibited them to be a branch fs by default, because
25133 + initramfs becomes unusable after switch_root or something
25134 + generally. If you sets initramfs as an aufs branch and boot your
25135 + system by switch_root, you will meet a problem easily since the
25136 + files in initramfs may be inaccessible.
25137 + Unless you are going to use ramfs as an aufs branch fs without
25138 + switch_root or something, leave it N.
25140 +config AUFS_BR_FUSE
25141 + bool "Fuse fs as an aufs branch"
25142 + depends on FUSE_FS
25145 + If you want to use fuse-based userspace filesystem as an aufs
25146 + branch fs, then enable this option.
25147 + It implements the internal poll(2) operation which is
25148 + implemented by fuse only (curretnly).
25153 + Automatic configuration for internal use.
25155 +config AUFS_BR_HFSPLUS
25156 + bool "Hfsplus as an aufs branch"
25157 + depends on HFSPLUS_FS
25160 + If you want to use hfsplus fs as an aufs branch fs, then enable
25161 + this option. This option introduces a small overhead at
25162 + copying-up a file on hfsplus.
25164 +config AUFS_BDEV_LOOP
25166 + depends on BLK_DEV_LOOP
25169 + Automatic configuration for internal use.
25170 + Convert =[ym] into =y.
25173 + bool "Debug aufs"
25175 + Enable this to compile aufs internal debug code.
25176 + It will have a negative impact to the performance.
25178 +config AUFS_MAGIC_SYSRQ
25180 + depends on AUFS_DEBUG && MAGIC_SYSRQ
25183 + Automatic configuration for internal use.
25184 + When aufs supports Magic SysRq, enabled automatically.
25186 diff -urN /usr/share/empty/fs/aufs/lcnt.h linux/fs/aufs/lcnt.h
25187 --- /usr/share/empty/fs/aufs/lcnt.h 1970-01-01 01:00:00.000000000 +0100
25188 +++ linux/fs/aufs/lcnt.h 2022-03-21 14:49:05.726633010 +0100
25190 +/* SPDX-License-Identifier: GPL-2.0 */
25192 + * Copyright (C) 2018-2021 Junjiro R. Okajima
25194 + * This program, aufs is free software; you can redistribute it and/or modify
25195 + * it under the terms of the GNU General Public License as published by
25196 + * the Free Software Foundation; either version 2 of the License, or
25197 + * (at your option) any later version.
25199 + * This program is distributed in the hope that it will be useful,
25200 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25201 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25202 + * GNU General Public License for more details.
25204 + * You should have received a copy of the GNU General Public License
25205 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25209 + * simple long counter wrapper
25212 +#ifndef __AUFS_LCNT_H__
25213 +#define __AUFS_LCNT_H__
25217 +#include "debug.h"
25219 +#define AuLCntATOMIC 1
25220 +#define AuLCntPCPUCNT 2
25222 + * why does percpu_refcount require extra synchronize_rcu()s in
25223 + * au_br_do_free()
25225 +#define AuLCntPCPUREF 3
25227 +/* #define AuLCntChosen AuLCntATOMIC */
25228 +#define AuLCntChosen AuLCntPCPUCNT
25229 +/* #define AuLCntChosen AuLCntPCPUREF */
25231 +#if AuLCntChosen == AuLCntATOMIC
25232 +#include <linux/atomic.h>
25234 +typedef atomic_long_t au_lcnt_t;
25236 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25238 + atomic_long_set(cnt, 0);
25242 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25247 +static inline void au_lcnt_fin(au_lcnt_t *cnt __maybe_unused,
25248 + int do_sync __maybe_unused)
25253 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25255 + atomic_long_inc(cnt);
25258 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25260 + atomic_long_dec(cnt);
25263 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25265 + return atomic_long_read(cnt);
25269 +#if AuLCntChosen == AuLCntPCPUCNT
25270 +#include <linux/percpu_counter.h>
25272 +typedef struct percpu_counter au_lcnt_t;
25274 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25276 + return percpu_counter_init(cnt, 0, GFP_NOFS);
25279 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25284 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync __maybe_unused)
25286 + percpu_counter_destroy(cnt);
25289 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25291 + percpu_counter_inc(cnt);
25294 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25296 + percpu_counter_dec(cnt);
25299 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25303 + n = percpu_counter_sum(cnt);
25305 + if (LONG_MAX != LLONG_MAX
25307 + AuWarn1("%s\n", "wrap-around");
25313 +#if AuLCntChosen == AuLCntPCPUREF
25314 +#include <linux/percpu-refcount.h>
25316 +typedef struct percpu_ref au_lcnt_t;
25318 +static inline int au_lcnt_init(au_lcnt_t *cnt, percpu_ref_func_t *release)
25321 + release = percpu_ref_exit;
25322 + return percpu_ref_init(cnt, release, /*percpu mode*/0, GFP_NOFS);
25325 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25327 + synchronize_rcu();
25330 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync)
25332 + percpu_ref_kill(cnt);
25334 + au_lcnt_wait_for_fin(cnt);
25337 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25339 + percpu_ref_get(cnt);
25342 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25344 + percpu_ref_put(cnt);
25348 + * avoid calling this func as possible.
25350 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev)
25354 + percpu_ref_switch_to_atomic_sync(cnt);
25355 + l = atomic_long_read(&cnt->count);
25357 + percpu_ref_switch_to_percpu(cnt);
25359 + /* percpu_ref is initialized by 1 instead of 0 */
25364 +#ifdef CONFIG_AUFS_DEBUG
25365 +#define AuLCntZero(val) do { \
25368 + AuDbg("%s = %ld\n", #val, l); \
25371 +#define AuLCntZero(val) do {} while (0)
25374 +#endif /* __KERNEL__ */
25375 +#endif /* __AUFS_LCNT_H__ */
25376 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
25377 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
25378 +++ linux/fs/aufs/loop.c 2022-03-21 14:49:05.726633010 +0100
25380 +// SPDX-License-Identifier: GPL-2.0
25382 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25384 + * This program, aufs is free software; you can redistribute it and/or modify
25385 + * it under the terms of the GNU General Public License as published by
25386 + * the Free Software Foundation; either version 2 of the License, or
25387 + * (at your option) any later version.
25389 + * This program is distributed in the hope that it will be useful,
25390 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25391 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25392 + * GNU General Public License for more details.
25394 + * You should have received a copy of the GNU General Public License
25395 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25399 + * support for loopback block device as a branch
25404 +/* added into drivers/block/loop.c */
25405 +static struct file *(*backing_file_func)(struct super_block *sb);
25408 + * test if two lower dentries have overlapping branches.
25410 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
25412 + struct super_block *h_sb;
25413 + struct file *backing_file;
25415 + if (unlikely(!backing_file_func)) {
25416 + /* don't load "loop" module here */
25417 + backing_file_func = symbol_get(loop_backing_file);
25418 + if (unlikely(!backing_file_func))
25419 + /* "loop" module is not loaded */
25423 + h_sb = h_adding->d_sb;
25424 + backing_file = backing_file_func(h_sb);
25425 + if (!backing_file)
25428 + h_adding = backing_file->f_path.dentry;
25430 + * h_adding can be local NFS.
25431 + * in this case aufs cannot detect the loop.
25433 + if (unlikely(h_adding->d_sb == sb))
25435 + return !!au_test_subdir(h_adding, sb->s_root);
25438 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
25439 +int au_test_loopback_kthread(void)
25442 + struct task_struct *tsk = current;
25443 + char c, comm[sizeof(tsk->comm)];
25446 + if (tsk->flags & PF_KTHREAD) {
25447 + get_task_comm(comm, tsk);
25449 + ret = ('0' <= c && c <= '9'
25450 + && !strncmp(comm, "loop", 4));
25456 +/* ---------------------------------------------------------------------- */
25458 +#define au_warn_loopback_step 16
25459 +static int au_warn_loopback_nelem = au_warn_loopback_step;
25460 +static unsigned long *au_warn_loopback_array;
25462 +void au_warn_loopback(struct super_block *h_sb)
25464 + int i, new_nelem;
25465 + unsigned long *a, magic;
25466 + static DEFINE_SPINLOCK(spin);
25468 + magic = h_sb->s_magic;
25469 + spin_lock(&spin);
25470 + a = au_warn_loopback_array;
25471 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
25472 + if (a[i] == magic) {
25473 + spin_unlock(&spin);
25477 + /* h_sb is new to us, print it */
25478 + if (i < au_warn_loopback_nelem) {
25483 + /* expand the array */
25484 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
25485 + a = au_kzrealloc(au_warn_loopback_array,
25486 + au_warn_loopback_nelem * sizeof(unsigned long),
25487 + new_nelem * sizeof(unsigned long), GFP_ATOMIC,
25488 + /*may_shrink*/0);
25490 + au_warn_loopback_nelem = new_nelem;
25491 + au_warn_loopback_array = a;
25496 + spin_unlock(&spin);
25497 + AuWarn1("realloc failed, ignored\n");
25501 + spin_unlock(&spin);
25502 + pr_warn("you may want to try another patch for loopback file "
25503 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
25506 +int au_loopback_init(void)
25509 + struct super_block *sb __maybe_unused;
25511 + BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(*au_warn_loopback_array));
25514 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
25515 + sizeof(unsigned long), GFP_NOFS);
25516 + if (unlikely(!au_warn_loopback_array))
25522 +void au_loopback_fin(void)
25524 + if (backing_file_func)
25525 + symbol_put(loop_backing_file);
25526 + au_kfree_try_rcu(au_warn_loopback_array);
25528 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
25529 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
25530 +++ linux/fs/aufs/loop.h 2022-03-21 14:49:05.726633010 +0100
25532 +/* SPDX-License-Identifier: GPL-2.0 */
25534 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25536 + * This program, aufs is free software; you can redistribute it and/or modify
25537 + * it under the terms of the GNU General Public License as published by
25538 + * the Free Software Foundation; either version 2 of the License, or
25539 + * (at your option) any later version.
25541 + * This program is distributed in the hope that it will be useful,
25542 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25543 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25544 + * GNU General Public License for more details.
25546 + * You should have received a copy of the GNU General Public License
25547 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25551 + * support for loopback mount as a branch
25554 +#ifndef __AUFS_LOOP_H__
25555 +#define __AUFS_LOOP_H__
25560 +struct super_block;
25562 +#ifdef CONFIG_AUFS_BDEV_LOOP
25563 +/* drivers/block/loop.c */
25564 +struct file *loop_backing_file(struct super_block *sb);
25567 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
25568 +int au_test_loopback_kthread(void);
25569 +void au_warn_loopback(struct super_block *h_sb);
25571 +int au_loopback_init(void);
25572 +void au_loopback_fin(void);
25574 +AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
25576 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
25577 + struct dentry *h_adding)
25578 +AuStubInt0(au_test_loopback_kthread, void)
25579 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
25581 +AuStubInt0(au_loopback_init, void)
25582 +AuStubVoid(au_loopback_fin, void)
25583 +#endif /* BLK_DEV_LOOP */
25585 +#endif /* __KERNEL__ */
25586 +#endif /* __AUFS_LOOP_H__ */
25587 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
25588 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
25589 +++ linux/fs/aufs/magic.mk 2022-03-21 14:49:05.726633010 +0100
25591 +# SPDX-License-Identifier: GPL-2.0
25593 +# defined in ${srctree}/fs/fuse/inode.c
25595 +ifdef CONFIG_FUSE_FS
25596 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
25599 +# defined in ${srctree}/fs/xfs/xfs_sb.h
25601 +ifdef CONFIG_XFS_FS
25602 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
25605 +# defined in ${srctree}/fs/configfs/mount.c
25607 +ifdef CONFIG_CONFIGFS_FS
25608 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
25611 +# defined in ${srctree}/fs/ubifs/ubifs.h
25613 +ifdef CONFIG_UBIFS_FS
25614 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
25617 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
25619 +ifdef CONFIG_HFSPLUS_FS
25620 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
25622 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
25623 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
25624 +++ linux/fs/aufs/Makefile 2022-03-21 14:49:05.716633010 +0100
25626 +# SPDX-License-Identifier: GPL-2.0
25628 +include ${src}/magic.mk
25629 +ifeq (${CONFIG_AUFS_FS},m)
25630 +include ${src}/conf.mk
25632 +-include ${src}/priv_def.mk
25634 +# cf. include/linux/kernel.h
25636 +ccflags-y += -DDEBUG
25637 +# sparse requires the full pathname
25639 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
25641 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
25644 +obj-$(CONFIG_AUFS_FS) += aufs.o
25645 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o fsctx.o \
25646 + wkq.o vfsub.o dcsub.o \
25647 + cpup.o whout.o wbr_policy.o \
25648 + dinfo.o dentry.o \
25650 + finfo.o file.o f_op.o \
25652 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
25656 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
25657 +aufs-$(CONFIG_SYSFS) += sysfs.o
25658 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
25659 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
25660 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
25661 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
25662 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
25663 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
25664 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
25665 +aufs-$(CONFIG_AUFS_DIRREN) += dirren.o
25666 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
25667 +aufs-$(CONFIG_AUFS_POLL) += poll.o
25668 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
25669 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
25670 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
25671 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
25672 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
25673 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
25674 +++ linux/fs/aufs/module.c 2022-03-21 14:49:05.726633010 +0100
25676 +// SPDX-License-Identifier: GPL-2.0
25678 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25680 + * This program, aufs is free software; you can redistribute it and/or modify
25681 + * it under the terms of the GNU General Public License as published by
25682 + * the Free Software Foundation; either version 2 of the License, or
25683 + * (at your option) any later version.
25685 + * This program is distributed in the hope that it will be useful,
25686 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25687 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25688 + * GNU General Public License for more details.
25690 + * You should have received a copy of the GNU General Public License
25691 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25695 + * module global variables and operations
25698 +#include <linux/module.h>
25699 +#include <linux/seq_file.h>
25702 +/* shrinkable realloc */
25703 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink)
25711 +#if 0 /* unused */
25718 + AuDebugOn(!new_sz);
25721 + diff = au_kmidx_sub(sz, new_sz);
25727 + /* expand or SLOB */
25728 + p = krealloc(p, new_sz, gfp);
25729 + else if (new_sz < sz && may_shrink) {
25733 + q = kmalloc(new_sz, gfp);
25736 + memcpy(q, p, new_sz);
25737 + au_kfree_try_rcu(p);
25748 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
25751 + p = au_krealloc(p, new_sz, gfp, may_shrink);
25752 + if (p && new_sz > nused)
25753 + memset(p + nused, 0, new_sz - nused);
25757 +/* ---------------------------------------------------------------------- */
25761 +struct kmem_cache *au_cache[AuCache_Last];
25763 +static void au_cache_fin(void)
25768 + * Make sure all delayed rcu free inodes are flushed before we
25773 + /* excluding AuCache_HNOTIFY */
25774 + BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
25775 + for (i = 0; i < AuCache_HNOTIFY; i++) {
25776 + kmem_cache_destroy(au_cache[i]);
25777 + au_cache[i] = NULL;
25781 +static int __init au_cache_init(void)
25783 + au_cache[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
25784 + if (au_cache[AuCache_DINFO])
25785 + /* SLAB_DESTROY_BY_RCU */
25786 + au_cache[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
25787 + au_icntnr_init_once);
25788 + if (au_cache[AuCache_ICNTNR])
25789 + au_cache[AuCache_FINFO] = AuCacheCtor(au_finfo,
25790 + au_fi_init_once);
25791 + if (au_cache[AuCache_FINFO])
25792 + au_cache[AuCache_VDIR] = AuCache(au_vdir);
25793 + if (au_cache[AuCache_VDIR])
25794 + au_cache[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
25795 + if (au_cache[AuCache_DEHSTR])
25802 +/* ---------------------------------------------------------------------- */
25804 +int au_dir_roflags;
25806 +#ifdef CONFIG_AUFS_SBILIST
25808 + * iterate_supers_type() doesn't protect us from
25809 + * remounting (branch management)
25811 +struct hlist_bl_head au_sbilist;
25815 + * functions for module interface.
25817 +MODULE_LICENSE("GPL");
25818 +/* MODULE_LICENSE("GPL v2"); */
25819 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
25820 +MODULE_DESCRIPTION(AUFS_NAME
25821 + " -- Advanced multi layered unification filesystem");
25822 +MODULE_VERSION(AUFS_VERSION);
25823 +MODULE_ALIAS_FS(AUFS_NAME);
25825 +/* this module parameter has no meaning when SYSFS is disabled */
25826 +int sysaufs_brs = 1;
25827 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
25828 +module_param_named(brs, sysaufs_brs, int, 0444);
25830 +/* this module parameter has no meaning when USER_NS is disabled */
25832 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
25833 +module_param_named(allow_userns, au_userns, bool, 0444);
25835 +/* ---------------------------------------------------------------------- */
25837 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
25839 +int au_seq_path(struct seq_file *seq, struct path *path)
25843 + err = seq_path(seq, path, au_esc_chars);
25852 +/* ---------------------------------------------------------------------- */
25854 +static int __init aufs_init(void)
25859 + p = au_esc_chars;
25860 + for (i = 1; i <= ' '; i++)
25866 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
25868 + memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
25869 + for (i = 0; i < AuIop_Last; i++)
25870 + aufs_iop_nogetattr[i].getattr = NULL;
25872 + memset(au_cache, 0, sizeof(au_cache)); /* including hnotify */
25874 + au_sbilist_init();
25875 + sysaufs_brs_init();
25878 + err = sysaufs_init();
25879 + if (unlikely(err))
25881 + err = dbgaufs_init();
25882 + if (unlikely(err))
25883 + goto out_sysaufs;
25884 + err = au_procfs_init();
25885 + if (unlikely(err))
25886 + goto out_dbgaufs;
25887 + err = au_wkq_init();
25888 + if (unlikely(err))
25890 + err = au_loopback_init();
25891 + if (unlikely(err))
25893 + err = au_hnotify_init();
25894 + if (unlikely(err))
25895 + goto out_loopback;
25896 + err = au_sysrq_init();
25897 + if (unlikely(err))
25899 + err = au_cache_init();
25900 + if (unlikely(err))
25903 + aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
25904 + err = register_filesystem(&aufs_fs_type);
25905 + if (unlikely(err))
25908 + /* since we define pr_fmt, call printk directly */
25909 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
25910 + goto out; /* success */
25917 + au_hnotify_fin();
25919 + au_loopback_fin();
25933 +static void __exit aufs_exit(void)
25935 + unregister_filesystem(&aufs_fs_type);
25938 + au_hnotify_fin();
25939 + au_loopback_fin();
25947 +module_init(aufs_init);
25948 +module_exit(aufs_exit);
25949 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
25950 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
25951 +++ linux/fs/aufs/module.h 2022-05-23 09:44:37.729839166 +0200
25953 +/* SPDX-License-Identifier: GPL-2.0 */
25955 + * Copyright (C) 2005-2021 Junjiro R. Okajima
25957 + * This program, aufs is free software; you can redistribute it and/or modify
25958 + * it under the terms of the GNU General Public License as published by
25959 + * the Free Software Foundation; either version 2 of the License, or
25960 + * (at your option) any later version.
25962 + * This program is distributed in the hope that it will be useful,
25963 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25964 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25965 + * GNU General Public License for more details.
25967 + * You should have received a copy of the GNU General Public License
25968 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25972 + * module initialization and module-global
25975 +#ifndef __AUFS_MODULE_H__
25976 +#define __AUFS_MODULE_H__
25980 +#include <linux/slab.h>
25981 +#include "debug.h"
25982 +#include "dentry.h"
25985 +#include "inode.h"
25990 +/* module parameters */
25991 +extern int sysaufs_brs;
25992 +extern bool au_userns;
25994 +/* ---------------------------------------------------------------------- */
25996 +extern int au_dir_roflags;
25998 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink);
25999 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
26003 + * Comparing the size of the object with sizeof(struct rcu_head)
26004 + * case 1: object is always larger
26005 + * --> au_kfree_rcu() or au_kfree_do_rcu()
26006 + * case 2: object is always smaller
26007 + * --> au_kfree_small()
26008 + * case 3: object can be any size
26009 + * --> au_kfree_try_rcu()
26012 +static inline void au_kfree_do_rcu(const void *p)
26015 + struct rcu_head rcu;
26016 + } *a = (void *)p;
26018 + kfree_rcu(a, rcu);
26021 +#define au_kfree_rcu(_p) do { \
26022 + typeof(_p) p = (_p); \
26023 + BUILD_BUG_ON(sizeof(*p) < sizeof(struct rcu_head)); \
26025 + au_kfree_do_rcu(p); \
26028 +#define au_kfree_do_sz_test(sz) (sz >= sizeof(struct rcu_head))
26029 +#define au_kfree_sz_test(p) (p && au_kfree_do_sz_test(ksize(p)))
26031 +static inline void au_kfree_try_rcu(const void *p)
26035 + if (au_kfree_sz_test(p))
26036 + au_kfree_do_rcu(p);
26041 +static inline void au_kfree_small(const void *p)
26045 + AuDebugOn(au_kfree_sz_test(p));
26049 +static inline int au_kmidx_sub(size_t sz, size_t new_sz)
26051 +#ifndef CONFIG_SLOB
26052 + return __kmalloc_index(sz, false) - __kmalloc_index(new_sz, false);
26054 + return -1; /* SLOB is untested */
26058 +int au_seq_path(struct seq_file *seq, struct path *path);
26060 +#ifdef CONFIG_PROC_FS
26062 +int __init au_procfs_init(void);
26063 +void au_procfs_fin(void);
26065 +AuStubInt0(au_procfs_init, void);
26066 +AuStubVoid(au_procfs_fin, void);
26069 +/* ---------------------------------------------------------------------- */
26078 + AuCache_HNOTIFY, /* must be last */
26082 +extern struct kmem_cache *au_cache[AuCache_Last];
26084 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
26085 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
26086 +#define AuCacheCtor(type, ctor) \
26087 + kmem_cache_create(#type, sizeof(struct type), \
26088 + __alignof__(struct type), AuCacheFlags, ctor)
26090 +#define AuCacheFuncAlloc(name, index) \
26091 + static inline struct au_##name *au_cache_alloc_##name(void) \
26092 + { return kmem_cache_alloc(au_cache[AuCache_##index], GFP_NOFS); }
26094 +#define AuCacheFuncs(name, index) \
26095 + static inline void au_cache_free_##name##_norcu(struct au_##name *p) \
26096 + { kmem_cache_free(au_cache[AuCache_##index], p); } \
26098 + static inline void au_cache_free_##name##_rcu_cb(struct rcu_head *rcu) \
26099 + { void *p = rcu; \
26100 + p -= offsetof(struct au_##name, rcu); \
26101 + kmem_cache_free(au_cache[AuCache_##index], p); } \
26102 + static inline void au_cache_free_##name##_rcu(struct au_##name *p) \
26103 + { BUILD_BUG_ON(sizeof(struct au_##name) < sizeof(struct rcu_head)); \
26104 + call_rcu(&p->rcu, au_cache_free_##name##_rcu_cb); } \
26106 + static inline void au_cache_free_##name(struct au_##name *p) \
26107 + { /* au_cache_free_##name##_norcu(p); */ \
26108 + au_cache_free_##name##_rcu(p); }
26110 +AuCacheFuncs(dinfo, DINFO);
26111 +AuCacheFuncAlloc(dinfo, DINFO);
26113 +AuCacheFuncs(icntnr, ICNTNR);
26114 +static inline struct au_icntnr *au_cache_alloc_icntnr(struct super_block *sb)
26115 +{ return alloc_inode_sb(sb, au_cache[AuCache_ICNTNR], GFP_NOFS); }
26117 +AuCacheFuncs(finfo, FINFO);
26118 +AuCacheFuncAlloc(finfo, FINFO);
26120 +AuCacheFuncs(vdir, VDIR);
26121 +AuCacheFuncAlloc(vdir, VDIR);
26123 +AuCacheFuncs(vdir_dehstr, DEHSTR);
26124 +AuCacheFuncAlloc(vdir_dehstr, DEHSTR);
26126 +#ifdef CONFIG_AUFS_HNOTIFY
26127 +AuCacheFuncs(hnotify, HNOTIFY);
26128 +AuCacheFuncAlloc(hnotify, HNOTIFY);
26131 +#endif /* __KERNEL__ */
26132 +#endif /* __AUFS_MODULE_H__ */
26133 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
26134 --- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
26135 +++ linux/fs/aufs/mvdown.c 2022-03-21 14:49:05.729966343 +0100
26137 +// SPDX-License-Identifier: GPL-2.0
26139 + * Copyright (C) 2011-2021 Junjiro R. Okajima
26141 + * This program, aufs is free software; you can redistribute it and/or modify
26142 + * it under the terms of the GNU General Public License as published by
26143 + * the Free Software Foundation; either version 2 of the License, or
26144 + * (at your option) any later version.
26146 + * This program is distributed in the hope that it will be useful,
26147 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26148 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26149 + * GNU General Public License for more details.
26151 + * You should have received a copy of the GNU General Public License
26152 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26156 + * move-down, opposite of copy-up
26161 +struct au_mvd_args {
26163 + struct super_block *h_sb;
26164 + struct dentry *h_parent;
26165 + struct au_hinode *hdir;
26166 + struct inode *h_dir, *h_inode;
26167 + struct au_pin pin;
26168 + } info[AUFS_MVDOWN_NARRAY];
26170 + struct aufs_mvdown mvdown;
26171 + struct dentry *dentry, *parent;
26172 + struct inode *inode, *dir;
26173 + struct super_block *sb;
26174 + aufs_bindex_t bopq, bwh, bfound;
26175 + unsigned char rename_lock;
26178 +#define mvd_errno mvdown.au_errno
26179 +#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
26180 +#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
26181 +#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
26182 +#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
26184 +#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
26185 +#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
26186 +#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
26187 +#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
26188 +#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
26189 +#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
26191 +#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
26192 +#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
26193 +#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
26194 +#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
26195 +#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
26196 +#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
26198 +#define AU_MVD_PR(flag, ...) do { \
26200 + pr_err(__VA_ARGS__); \
26203 +static int find_lower_writable(struct au_mvd_args *a)
26205 + struct super_block *sb;
26206 + aufs_bindex_t bindex, bbot;
26207 + struct au_branch *br;
26210 + bindex = a->mvd_bsrc;
26211 + bbot = au_sbbot(sb);
26212 + if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
26213 + for (bindex++; bindex <= bbot; bindex++) {
26214 + br = au_sbr(sb, bindex);
26215 + if (au_br_fhsm(br->br_perm)
26216 + && !sb_rdonly(au_br_sb(br)))
26219 + else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
26220 + for (bindex++; bindex <= bbot; bindex++) {
26221 + br = au_sbr(sb, bindex);
26222 + if (!au_br_rdonly(br))
26226 + for (bindex++; bindex <= bbot; bindex++) {
26227 + br = au_sbr(sb, bindex);
26228 + if (!sb_rdonly(au_br_sb(br))) {
26229 + if (au_br_rdonly(br))
26231 + |= AUFS_MVDOWN_ROLOWER_R;
26239 +/* make the parent dir on bdst */
26240 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
26245 + a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
26246 + a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
26247 + a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
26248 + a->mvd_h_dst_parent = NULL;
26249 + if (au_dbbot(a->parent) >= a->mvd_bdst)
26250 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26251 + if (!a->mvd_h_dst_parent) {
26252 + err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
26253 + if (unlikely(err)) {
26254 + AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
26257 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26265 +/* lock them all */
26266 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
26269 + struct dentry *h_trap;
26271 + a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
26272 + a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
26273 + err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
26274 + au_opt_udba(a->sb),
26275 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26277 + if (unlikely(err)) {
26278 + AU_MVD_PR(dmsg, "pin_dst failed\n");
26282 + if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
26283 + a->rename_lock = 0;
26284 + au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26285 + AuLsc_DI_PARENT, AuLsc_I_PARENT3,
26286 + au_opt_udba(a->sb),
26287 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26288 + err = au_do_pin(&a->mvd_pin_src);
26290 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26291 + if (unlikely(err)) {
26292 + AU_MVD_PR(dmsg, "pin_src failed\n");
26295 + goto out; /* success */
26298 + a->rename_lock = 1;
26299 + au_pin_hdir_unlock(&a->mvd_pin_dst);
26300 + err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26301 + au_opt_udba(a->sb),
26302 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26304 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26305 + if (unlikely(err)) {
26306 + AU_MVD_PR(dmsg, "pin_src failed\n");
26307 + au_pin_hdir_lock(&a->mvd_pin_dst);
26310 + au_pin_hdir_unlock(&a->mvd_pin_src);
26311 + h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26312 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26314 + err = (h_trap != a->mvd_h_src_parent);
26316 + err = (h_trap != a->mvd_h_dst_parent);
26318 + BUG_ON(err); /* it should never happen */
26319 + if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
26322 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26323 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26324 + au_pin_hdir_lock(&a->mvd_pin_src);
26325 + au_unpin(&a->mvd_pin_src);
26326 + au_pin_hdir_lock(&a->mvd_pin_dst);
26329 + goto out; /* success */
26332 + au_unpin(&a->mvd_pin_dst);
26338 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
26340 + if (!a->rename_lock)
26341 + au_unpin(&a->mvd_pin_src);
26343 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26344 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26345 + au_pin_hdir_lock(&a->mvd_pin_src);
26346 + au_unpin(&a->mvd_pin_src);
26347 + au_pin_hdir_lock(&a->mvd_pin_dst);
26349 + au_unpin(&a->mvd_pin_dst);
26352 +/* copy-down the file */
26353 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
26356 + struct au_cp_generic cpg = {
26357 + .dentry = a->dentry,
26358 + .bdst = a->mvd_bdst,
26359 + .bsrc = a->mvd_bsrc,
26361 + .pin = &a->mvd_pin_dst,
26362 + .flags = AuCpup_DTIME | AuCpup_HOPEN
26365 + AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
26366 + if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26367 + au_fset_cpup(cpg.flags, OVERWRITE);
26368 + if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
26369 + au_fset_cpup(cpg.flags, RWDST);
26370 + err = au_sio_cpdown_simple(&cpg);
26371 + if (unlikely(err))
26372 + AU_MVD_PR(dmsg, "cpdown failed\n");
26379 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
26382 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
26385 + struct path h_path;
26386 + struct au_branch *br;
26387 + struct inode *delegated;
26389 + br = au_sbr(a->sb, a->mvd_bdst);
26390 + h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
26391 + err = PTR_ERR(h_path.dentry);
26392 + if (IS_ERR(h_path.dentry)) {
26393 + AU_MVD_PR(dmsg, "wh_lkup failed\n");
26398 + if (d_is_positive(h_path.dentry)) {
26399 + h_path.mnt = au_br_mnt(br);
26400 + delegated = NULL;
26401 + err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
26402 + &delegated, /*force*/0);
26403 + if (unlikely(err == -EWOULDBLOCK)) {
26404 + pr_warn("cannot retry for NFSv4 delegation"
26405 + " for an internal unlink\n");
26408 + if (unlikely(err))
26409 + AU_MVD_PR(dmsg, "wh_unlink failed\n");
26411 + dput(h_path.dentry);
26419 + * unlink the topmost h_dentry
26421 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
26424 + struct path h_path;
26425 + struct inode *delegated;
26427 + h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
26428 + h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
26429 + delegated = NULL;
26430 + err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
26431 + if (unlikely(err == -EWOULDBLOCK)) {
26432 + pr_warn("cannot retry for NFSv4 delegation"
26433 + " for an internal unlink\n");
26436 + if (unlikely(err))
26437 + AU_MVD_PR(dmsg, "unlink failed\n");
26443 +/* Since mvdown succeeded, we ignore an error of this function */
26444 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
26447 + struct au_branch *br;
26449 + a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
26450 + br = au_sbr(a->sb, a->mvd_bsrc);
26451 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
26453 + br = au_sbr(a->sb, a->mvd_bdst);
26454 + a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
26455 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
26458 + a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
26460 + AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
26464 + * copy-down the file and unlink the bsrc file.
26465 + * - unlink the bdst whout if exist
26466 + * - copy-down the file (with whtmp name and rename)
26467 + * - unlink the bsrc file
26469 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
26473 + err = au_do_mkdir(dmsg, a);
26475 + err = au_do_lock(dmsg, a);
26476 + if (unlikely(err))
26480 + * do not revert the activities we made on bdst since they should be
26481 + * harmless in aufs.
26484 + err = au_do_cpdown(dmsg, a);
26486 + err = au_do_unlink_wh(dmsg, a);
26487 + if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
26488 + err = au_do_unlink(dmsg, a);
26489 + if (unlikely(err))
26492 + AuDbg("%pd2, 0x%x, %d --> %d\n",
26493 + a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
26494 + if (find_lower_writable(a) < 0)
26495 + a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
26497 + if (a->mvdown.flags & AUFS_MVDOWN_STFS)
26498 + au_do_stfs(dmsg, a);
26500 + /* maintain internal array */
26501 + if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
26502 + au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
26503 + au_set_dbtop(a->dentry, a->mvd_bdst);
26504 + au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
26505 + au_set_ibtop(a->inode, a->mvd_bdst);
26507 + /* hide the lower */
26508 + au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
26509 + au_set_dbbot(a->dentry, a->mvd_bsrc);
26510 + au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
26511 + au_set_ibbot(a->inode, a->mvd_bsrc);
26513 + if (au_dbbot(a->dentry) < a->mvd_bdst)
26514 + au_set_dbbot(a->dentry, a->mvd_bdst);
26515 + if (au_ibbot(a->inode) < a->mvd_bdst)
26516 + au_set_ibbot(a->inode, a->mvd_bdst);
26519 + au_do_unlock(dmsg, a);
26525 +/* ---------------------------------------------------------------------- */
26527 +/* make sure the file is idle */
26528 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
26530 + int err, plinked;
26533 + plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
26534 + if (au_dbtop(a->dentry) == a->mvd_bsrc
26535 + && au_dcount(a->dentry) == 1
26536 + && atomic_read(&a->inode->i_count) == 1
26537 + /* && a->mvd_h_src_inode->i_nlink == 1 */
26538 + && (!plinked || !au_plink_test(a->inode))
26539 + && a->inode->i_nlink == 1)
26544 + "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
26545 + a->mvd_bsrc, au_dbtop(a->dentry), au_dcount(a->dentry),
26546 + atomic_read(&a->inode->i_count), a->inode->i_nlink,
26547 + a->mvd_h_src_inode->i_nlink,
26548 + plinked, plinked ? au_plink_test(a->inode) : 0);
26555 +/* make sure the parent dir is fine */
26556 +static int au_mvd_args_parent(const unsigned char dmsg,
26557 + struct au_mvd_args *a)
26560 + aufs_bindex_t bindex;
26563 + if (unlikely(au_alive_dir(a->parent))) {
26565 + AU_MVD_PR(dmsg, "parent dir is dead\n");
26569 + a->bopq = au_dbdiropq(a->parent);
26570 + bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
26571 + AuDbg("b%d\n", bindex);
26572 + if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
26573 + || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
26575 + a->mvd_errno = EAU_MVDOWN_OPAQUE;
26576 + AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
26577 + a->bopq, a->mvd_bdst);
26585 +static int au_mvd_args_intermediate(const unsigned char dmsg,
26586 + struct au_mvd_args *a)
26589 + struct au_dinfo *dinfo, *tmp;
26591 + /* lookup the next lower positive entry */
26593 + tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
26594 + if (unlikely(!tmp))
26599 + dinfo = au_di(a->dentry);
26600 + au_di_cp(tmp, dinfo);
26601 + au_di_swap(tmp, dinfo);
26603 + /* returns the number of positive dentries */
26604 + err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1,
26605 + /* AuLkup_IGNORE_PERM */ 0);
26607 + a->bwh = au_dbwh(a->dentry);
26608 + else if (err > 0)
26609 + a->bfound = au_dbtop(a->dentry);
26611 + au_di_swap(tmp, dinfo);
26612 + au_rw_write_unlock(&tmp->di_rwsem);
26614 + if (unlikely(err < 0))
26615 + AU_MVD_PR(dmsg, "failed look-up lower\n");
26618 + * here, we have these cases.
26620 + * no positive dentry under bsrc. there are more sub-cases.
26622 + * there no whiteout, we can safely move-down.
26625 + * bsrc < bwh && bwh < bdst
26626 + * there is a whiteout on RO branch. cannot proceed.
26628 + * there is a whiteout on the RW target branch. it should
26631 + * there is a whiteout somewhere unrelated branch.
26632 + * -1 < bfound && bfound <= bsrc
26635 + * found, but it is on RO branch between bsrc and bdst. cannot
26638 + * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
26641 + * found, after we create the file on bdst, it will be hidden.
26644 + AuDebugOn(a->bfound == -1
26646 + && a->bwh <= a->mvd_bsrc);
26647 + AuDebugOn(-1 < a->bfound
26648 + && a->bfound <= a->mvd_bsrc);
26651 + if (a->bfound == -1
26652 + && a->mvd_bsrc < a->bwh
26654 + && a->bwh < a->mvd_bdst) {
26655 + a->mvd_errno = EAU_MVDOWN_WHITEOUT;
26656 + AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
26657 + a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
26659 + } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
26660 + a->mvd_errno = EAU_MVDOWN_UPPER;
26661 + AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
26662 + a->mvd_bdst, a->bfound);
26666 + err = 0; /* success */
26673 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
26678 + if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26679 + && a->bfound == a->mvd_bdst)
26685 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
26688 + struct au_branch *br;
26691 + if (unlikely(S_ISDIR(a->inode->i_mode)))
26695 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
26696 + a->mvd_bsrc = au_ibtop(a->inode);
26698 + a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
26699 + if (unlikely(a->mvd_bsrc < 0
26700 + || (a->mvd_bsrc < au_dbtop(a->dentry)
26701 + || au_dbbot(a->dentry) < a->mvd_bsrc
26702 + || !au_h_dptr(a->dentry, a->mvd_bsrc))
26703 + || (a->mvd_bsrc < au_ibtop(a->inode)
26704 + || au_ibbot(a->inode) < a->mvd_bsrc
26705 + || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
26706 + a->mvd_errno = EAU_MVDOWN_NOUPPER;
26707 + AU_MVD_PR(dmsg, "no upper\n");
26711 + if (unlikely(a->mvd_bsrc == au_sbbot(a->sb))) {
26712 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
26713 + AU_MVD_PR(dmsg, "on the bottom\n");
26716 + a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
26717 + br = au_sbr(a->sb, a->mvd_bsrc);
26718 + err = au_br_rdonly(br);
26719 + if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
26720 + if (unlikely(err))
26722 + } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
26723 + || IS_APPEND(a->mvd_h_src_inode))) {
26725 + a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
26731 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
26732 + a->mvd_bdst = find_lower_writable(a);
26733 + if (unlikely(a->mvd_bdst < 0)) {
26734 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
26735 + AU_MVD_PR(dmsg, "no writable lower branch\n");
26739 + a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
26740 + if (unlikely(a->mvd_bdst < 0
26741 + || au_sbbot(a->sb) < a->mvd_bdst)) {
26742 + a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
26743 + AU_MVD_PR(dmsg, "no lower brid\n");
26748 + err = au_mvd_args_busy(dmsg, a);
26750 + err = au_mvd_args_parent(dmsg, a);
26752 + err = au_mvd_args_intermediate(dmsg, a);
26754 + err = au_mvd_args_exist(dmsg, a);
26756 + AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
26763 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
26766 + unsigned char dmsg;
26767 + struct au_mvd_args *args;
26768 + struct inode *inode;
26770 + inode = d_inode(dentry);
26772 + if (unlikely(!capable(CAP_SYS_ADMIN)))
26776 + args = kmalloc(sizeof(*args), GFP_NOFS);
26777 + if (unlikely(!args))
26780 + err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
26782 + /* VERIFY_WRITE */
26783 + err = !access_ok(uarg, sizeof(*uarg));
26784 + if (unlikely(err)) {
26789 + AuDbg("flags 0x%x\n", args->mvdown.flags);
26790 + args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
26791 + args->mvdown.au_errno = 0;
26792 + args->dentry = dentry;
26793 + args->inode = inode;
26794 + args->sb = dentry->d_sb;
26797 + dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
26798 + args->parent = dget_parent(dentry);
26799 + args->dir = d_inode(args->parent);
26800 + inode_lock_nested(args->dir, I_MUTEX_PARENT);
26801 + dput(args->parent);
26802 + if (unlikely(args->parent != dentry->d_parent)) {
26803 + AU_MVD_PR(dmsg, "parent dir is moved\n");
26807 + inode_lock_nested(inode, I_MUTEX_CHILD);
26808 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
26809 + if (unlikely(err))
26812 + di_write_lock_parent(args->parent);
26813 + err = au_mvd_args(dmsg, args);
26814 + if (unlikely(err))
26817 + err = au_do_mvdown(dmsg, args);
26818 + if (unlikely(err))
26821 + au_cpup_attr_timesizes(args->dir);
26822 + au_cpup_attr_timesizes(inode);
26823 + if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
26824 + au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
26825 + /* au_digen_dec(dentry); */
26828 + di_write_unlock(args->parent);
26829 + aufs_read_unlock(dentry, AuLock_DW);
26831 + inode_unlock(inode);
26833 + inode_unlock(args->dir);
26835 + e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
26838 + au_kfree_rcu(args);
26843 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
26844 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
26845 +++ linux/fs/aufs/opts.c 2022-03-21 14:49:05.729966343 +0100
26847 +// SPDX-License-Identifier: GPL-2.0
26849 + * Copyright (C) 2005-2021 Junjiro R. Okajima
26851 + * This program, aufs is free software; you can redistribute it and/or modify
26852 + * it under the terms of the GNU General Public License as published by
26853 + * the Free Software Foundation; either version 2 of the License, or
26854 + * (at your option) any later version.
26856 + * This program is distributed in the hope that it will be useful,
26857 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26858 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26859 + * GNU General Public License for more details.
26861 + * You should have received a copy of the GNU General Public License
26862 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26866 + * mount options/flags
26869 +#include <linux/types.h> /* a distribution requires */
26870 +#include <linux/parser.h>
26873 +/* ---------------------------------------------------------------------- */
26875 +static const char *au_parser_pattern(int val, match_table_t tbl)
26877 + struct match_token *p;
26880 + while (p->pattern) {
26881 + if (p->token == val)
26882 + return p->pattern;
26889 +static const char *au_optstr(int *val, match_table_t tbl)
26891 + struct match_token *p;
26898 + while (p->pattern) {
26900 + && (v & p->token) == p->token) {
26901 + *val &= ~p->token;
26902 + return p->pattern;
26911 +/* ---------------------------------------------------------------------- */
26913 +static match_table_t brperm = {
26914 + {AuBrPerm_RO, AUFS_BRPERM_RO},
26915 + {AuBrPerm_RR, AUFS_BRPERM_RR},
26916 + {AuBrPerm_RW, AUFS_BRPERM_RW},
26920 +static match_table_t brattr = {
26922 + {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
26923 + {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
26924 + /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
26925 + {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
26926 +#ifdef CONFIG_AUFS_FHSM
26927 + {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
26929 +#ifdef CONFIG_AUFS_XATTR
26930 + {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
26931 + {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
26932 + {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
26933 + {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
26934 + {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
26935 + {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
26938 + /* ro/rr branch */
26939 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
26942 + {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
26943 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
26948 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
26955 + p = strchr(str, '+');
26958 + v = match_token(str, table, args);
26960 + if (v & AuBrAttr_CMOO_Mask)
26961 + attr &= ~AuBrAttr_CMOO_Mask;
26966 + pr_warn("ignored branch attribute %s\n", str);
26976 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
26984 + p = au_optstr(&perm, brattr);
26987 + memcpy(q, p, sz + 1);
26993 + p = au_optstr(&perm, brattr);
26997 + memcpy(q, p, sz + 1);
27003 + return q - str->a;
27006 +int au_br_perm_val(char *perm)
27008 + int val, bad, sz;
27010 + substring_t args[MAX_OPT_ARGS];
27011 + au_br_perm_str_t attr;
27013 + p = strchr(perm, '+');
27016 + val = match_token(perm, brperm, args);
27020 + pr_warn("ignored branch permission %s\n", perm);
27021 + val = AuBrPerm_RO;
27027 + val |= br_attr_val(p + 1, brattr, args);
27030 + switch (val & AuBrPerm_Mask) {
27031 + case AuBrPerm_RO:
27032 + case AuBrPerm_RR:
27033 + bad = val & AuBrWAttr_Mask;
27034 + val &= ~AuBrWAttr_Mask;
27036 + case AuBrPerm_RW:
27037 + bad = val & AuBrRAttr_Mask;
27038 + val &= ~AuBrRAttr_Mask;
27043 + * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
27044 + * does not treat it as an error, just warning.
27045 + * this is a tiny guard for the user operation.
27047 + if (val & AuBrAttr_UNPIN) {
27048 + bad |= AuBrAttr_UNPIN;
27049 + val &= ~AuBrAttr_UNPIN;
27052 + if (unlikely(bad)) {
27053 + sz = au_do_optstr_br_attr(&attr, bad);
27055 + pr_warn("ignored branch attribute %s\n", attr.a);
27062 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
27064 + au_br_perm_str_t attr;
27070 + p = au_optstr(&perm, brperm);
27071 + AuDebugOn(!p || !*p);
27073 + memcpy(q, p, sz + 1);
27076 + sz = au_do_optstr_br_attr(&attr, perm);
27079 + memcpy(q, attr.a, sz + 1);
27082 + AuDebugOn(strlen(str->a) >= sizeof(str->a));
27085 +/* ---------------------------------------------------------------------- */
27087 +static match_table_t udbalevel = {
27088 + {AuOpt_UDBA_REVAL, "reval"},
27089 + {AuOpt_UDBA_NONE, "none"},
27090 +#ifdef CONFIG_AUFS_HNOTIFY
27091 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
27092 +#ifdef CONFIG_AUFS_HFSNOTIFY
27093 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
27099 +int au_udba_val(char *str)
27101 + substring_t args[MAX_OPT_ARGS];
27103 + return match_token(str, udbalevel, args);
27106 +const char *au_optstr_udba(int udba)
27108 + return au_parser_pattern(udba, udbalevel);
27111 +/* ---------------------------------------------------------------------- */
27113 +static match_table_t au_wbr_create_policy = {
27114 + {AuWbrCreate_TDP, "tdp"},
27115 + {AuWbrCreate_TDP, "top-down-parent"},
27116 + {AuWbrCreate_RR, "rr"},
27117 + {AuWbrCreate_RR, "round-robin"},
27118 + {AuWbrCreate_MFS, "mfs"},
27119 + {AuWbrCreate_MFS, "most-free-space"},
27120 + {AuWbrCreate_MFSV, "mfs:%d"},
27121 + {AuWbrCreate_MFSV, "most-free-space:%d"},
27123 + /* top-down regardless the parent, and then mfs */
27124 + {AuWbrCreate_TDMFS, "tdmfs:%d"},
27125 + {AuWbrCreate_TDMFSV, "tdmfs:%d:%d"},
27127 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
27128 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
27129 + {AuWbrCreate_PMFS, "pmfs"},
27130 + {AuWbrCreate_PMFSV, "pmfs:%d"},
27131 + {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
27132 + {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
27137 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
27138 + struct au_opt_wbr_create *create)
27141 + unsigned long long ull;
27144 + if (!match_u64(arg, &ull))
27145 + create->mfsrr_watermark = ull;
27147 + pr_err("bad integer in %s\n", str);
27154 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
27155 + struct au_opt_wbr_create *create)
27160 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
27161 + create->mfs_second = n;
27163 + pr_err("bad integer in %s\n", str);
27170 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
27173 + substring_t args[MAX_OPT_ARGS];
27175 + err = match_token(str, au_wbr_create_policy, args);
27176 + create->wbr_create = err;
27178 + case AuWbrCreate_MFSRRV:
27179 + case AuWbrCreate_TDMFSV:
27180 + case AuWbrCreate_PMFSRRV:
27181 + e = au_wbr_mfs_wmark(&args[0], str, create);
27183 + e = au_wbr_mfs_sec(&args[1], str, create);
27187 + case AuWbrCreate_MFSRR:
27188 + case AuWbrCreate_TDMFS:
27189 + case AuWbrCreate_PMFSRR:
27190 + e = au_wbr_mfs_wmark(&args[0], str, create);
27191 + if (unlikely(e)) {
27196 + case AuWbrCreate_MFS:
27197 + case AuWbrCreate_PMFS:
27198 + create->mfs_second = AUFS_MFS_DEF_SEC;
27200 + case AuWbrCreate_MFSV:
27201 + case AuWbrCreate_PMFSV:
27202 + e = au_wbr_mfs_sec(&args[0], str, create);
27211 +const char *au_optstr_wbr_create(int wbr_create)
27213 + return au_parser_pattern(wbr_create, au_wbr_create_policy);
27216 +static match_table_t au_wbr_copyup_policy = {
27217 + {AuWbrCopyup_TDP, "tdp"},
27218 + {AuWbrCopyup_TDP, "top-down-parent"},
27219 + {AuWbrCopyup_BUP, "bup"},
27220 + {AuWbrCopyup_BUP, "bottom-up-parent"},
27221 + {AuWbrCopyup_BU, "bu"},
27222 + {AuWbrCopyup_BU, "bottom-up"},
27226 +int au_wbr_copyup_val(char *str)
27228 + substring_t args[MAX_OPT_ARGS];
27230 + return match_token(str, au_wbr_copyup_policy, args);
27233 +const char *au_optstr_wbr_copyup(int wbr_copyup)
27235 + return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
27238 +/* ---------------------------------------------------------------------- */
27240 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
27241 + aufs_bindex_t bindex)
27244 + struct au_opt_add *add = &opt->add;
27247 + add->bindex = bindex;
27248 + add->perm = AuBrPerm_RO;
27249 + add->pathname = opt_str;
27250 + p = strchr(opt_str, '=');
27254 + add->perm = au_br_perm_val(p);
27257 + err = vfsub_kern_path(add->pathname, AuOpt_LkupDirFlags, &add->path);
27260 + add->perm = AuBrPerm_RO;
27261 + if (au_test_fs_rr(add->path.dentry->d_sb))
27262 + add->perm = AuBrPerm_RR;
27263 + else if (!bindex && !(sb_flags & SB_RDONLY))
27264 + add->perm = AuBrPerm_RW;
27266 + opt->type = Opt_add;
27269 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
27276 +static int au_opt_wbr_create(struct super_block *sb,
27277 + struct au_opt_wbr_create *create)
27280 + struct au_sbinfo *sbinfo;
27282 + SiMustWriteLock(sb);
27284 + err = 1; /* handled */
27285 + sbinfo = au_sbi(sb);
27286 + if (sbinfo->si_wbr_create_ops->fin) {
27287 + err = sbinfo->si_wbr_create_ops->fin(sb);
27292 + sbinfo->si_wbr_create = create->wbr_create;
27293 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
27294 + switch (create->wbr_create) {
27295 + case AuWbrCreate_MFSRRV:
27296 + case AuWbrCreate_MFSRR:
27297 + case AuWbrCreate_TDMFS:
27298 + case AuWbrCreate_TDMFSV:
27299 + case AuWbrCreate_PMFSRR:
27300 + case AuWbrCreate_PMFSRRV:
27301 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
27303 + case AuWbrCreate_MFS:
27304 + case AuWbrCreate_MFSV:
27305 + case AuWbrCreate_PMFS:
27306 + case AuWbrCreate_PMFSV:
27307 + sbinfo->si_wbr_mfs.mfs_expire
27308 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
27312 + if (sbinfo->si_wbr_create_ops->init)
27313 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
27320 + * plus: processed without an error
27321 + * zero: unprocessed
27323 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
27324 + struct au_opts *opts)
27327 + struct au_sbinfo *sbinfo;
27329 + SiMustWriteLock(sb);
27331 + err = 1; /* handled */
27332 + sbinfo = au_sbi(sb);
27333 + switch (opt->type) {
27335 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27336 + sbinfo->si_mntflags |= opt->udba;
27337 + opts->given_udba |= opt->udba;
27342 + au_opt_set(sbinfo->si_mntflags, PLINK);
27344 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27345 + au_plink_put(sb, /*verbose*/1);
27346 + au_opt_clr(sbinfo->si_mntflags, PLINK);
27349 + case Opt_list_plink:
27350 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27351 + au_plink_list(sb);
27356 + au_opt_set(sbinfo->si_mntflags, DIO);
27357 + au_fset_opts(opts->flags, REFRESH_DYAOP);
27359 + au_opt_clr(sbinfo->si_mntflags, DIO);
27360 + au_fset_opts(opts->flags, REFRESH_DYAOP);
27364 + case Opt_fhsm_sec:
27365 + au_fhsm_set(sbinfo, opt->fhsm_second);
27368 + case Opt_diropq_a:
27369 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27371 + case Opt_diropq_w:
27372 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27375 + case Opt_warn_perm:
27377 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
27379 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
27382 + case Opt_verbose:
27384 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
27386 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
27391 + au_opt_set(sbinfo->si_mntflags, SUM);
27393 + au_opt_clr(sbinfo->si_mntflags, SUM);
27394 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
27398 + au_opt_clr(sbinfo->si_mntflags, SUM);
27399 + au_opt_set(sbinfo->si_mntflags, SUM_W);
27402 + case Opt_wbr_create:
27403 + err = au_opt_wbr_create(sb, &opt->wbr_create);
27405 + case Opt_wbr_copyup:
27406 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
27407 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
27411 + sbinfo->si_dirwh = opt->dirwh;
27414 + case Opt_rdcache:
27415 + sbinfo->si_rdcache
27416 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
27419 + sbinfo->si_rdblk = opt->rdblk;
27422 + sbinfo->si_rdhash = opt->rdhash;
27427 + au_opt_set(sbinfo->si_mntflags, SHWH);
27429 + au_opt_clr(sbinfo->si_mntflags, SHWH);
27432 + case Opt_dirperm1:
27434 + au_opt_set(sbinfo->si_mntflags, DIRPERM1);
27436 + au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
27439 + case Opt_trunc_xino:
27441 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
27443 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
27446 + case Opt_trunc_xino_path:
27447 + case Opt_itrunc_xino:
27448 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex,
27454 + case Opt_trunc_xib:
27456 + au_fset_opts(opts->flags, TRUNC_XIB);
27458 + au_fclr_opts(opts->flags, TRUNC_XIB);
27464 + if (!au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27465 + err = au_dr_opt_set(sb);
27470 + au_opt_set(sbinfo->si_mntflags, DIRREN);
27472 + if (au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27473 + err = au_dr_opt_clr(sb, au_ftest_opts(opts->flags,
27479 + au_opt_clr(sbinfo->si_mntflags, DIRREN);
27485 + sb->s_flags |= SB_POSIXACL;
27487 + sb->s_flags &= ~SB_POSIXACL;
27499 + * returns tri-state.
27500 + * plus: processed without an error
27501 + * zero: unprocessed
27504 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
27505 + struct au_opts *opts)
27507 + int err, do_refresh;
27510 + switch (opt->type) {
27512 + opt->add.bindex = au_sbbot(sb) + 1;
27513 + if (opt->add.bindex < 0)
27514 + opt->add.bindex = 0;
27516 + /* Always goto add, not fallthrough */
27517 + case Opt_prepend:
27518 + opt->add.bindex = 0;
27520 + add: /* indented label */
27522 + err = au_br_add(sb, &opt->add,
27523 + au_ftest_opts(opts->flags, REMOUNT));
27526 + au_fset_opts(opts->flags, REFRESH);
27532 + err = au_br_del(sb, &opt->del,
27533 + au_ftest_opts(opts->flags, REMOUNT));
27536 + au_fset_opts(opts->flags, TRUNC_XIB);
27537 + au_fset_opts(opts->flags, REFRESH);
27543 + err = au_br_mod(sb, &opt->mod,
27544 + au_ftest_opts(opts->flags, REMOUNT),
27549 + au_fset_opts(opts->flags, REFRESH);
27556 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
27557 + struct au_opt_xino **opt_xino,
27558 + struct au_opts *opts)
27563 + switch (opt->type) {
27565 + err = au_xino_set(sb, &opt->xino,
27566 + !!au_ftest_opts(opts->flags, REMOUNT));
27568 + *opt_xino = &opt->xino;
27572 + *opt_xino = (void *)-1;
27579 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
27580 + unsigned int pending)
27583 + aufs_bindex_t bindex, bbot;
27584 + unsigned char do_plink, skip, do_free, can_no_dreval;
27585 + struct au_branch *br;
27586 + struct au_wbr *wbr;
27587 + struct dentry *root, *dentry;
27588 + struct inode *dir, *h_dir;
27589 + struct au_sbinfo *sbinfo;
27590 + struct au_hinode *hdir;
27592 + SiMustAnyLock(sb);
27594 + sbinfo = au_sbi(sb);
27595 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
27597 + if (!(sb_flags & SB_RDONLY)) {
27598 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
27599 + pr_warn("first branch should be rw\n");
27600 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
27601 + pr_warn_once("shwh should be used with ro\n");
27604 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
27605 + && !au_opt_test(sbinfo->si_mntflags, XINO))
27606 + pr_warn_once("udba=*notify requires xino\n");
27608 + if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
27609 + pr_warn_once("dirperm1 breaks the protection"
27610 + " by the permission bits on the lower branch\n");
27614 + root = sb->s_root;
27615 + dir = d_inode(root);
27616 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
27617 + can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
27619 + bbot = au_sbbot(sb);
27620 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
27622 + h_dir = au_h_iptr(dir, bindex);
27623 + br = au_sbr(sb, bindex);
27625 + if ((br->br_perm & AuBrAttr_ICEX)
27626 + && !h_dir->i_op->listxattr)
27627 + br->br_perm &= ~AuBrAttr_ICEX;
27628 +#if 0 /* untested */
27629 + if ((br->br_perm & AuBrAttr_ICEX_SEC)
27630 + && (au_br_sb(br)->s_flags & SB_NOSEC))
27631 + br->br_perm &= ~AuBrAttr_ICEX_SEC;
27635 + wbr = br->br_wbr;
27637 + wbr_wh_read_lock(wbr);
27639 + if (!au_br_writable(br->br_perm)) {
27642 + || (!wbr->wbr_whbase
27643 + && !wbr->wbr_plink
27644 + && !wbr->wbr_orph));
27645 + } else if (!au_br_wh_linkable(br->br_perm)) {
27646 + /* skip = (!br->br_whbase && !br->br_orph); */
27647 + skip = (!wbr || !wbr->wbr_whbase);
27648 + if (skip && wbr) {
27650 + skip = !!wbr->wbr_plink;
27652 + skip = !wbr->wbr_plink;
27655 + /* skip = (br->br_whbase && br->br_ohph); */
27656 + skip = (wbr && wbr->wbr_whbase);
27659 + skip = !!wbr->wbr_plink;
27661 + skip = !wbr->wbr_plink;
27665 + wbr_wh_read_unlock(wbr);
27667 + if (can_no_dreval) {
27668 + dentry = br->br_path.dentry;
27669 + spin_lock(&dentry->d_lock);
27670 + if (dentry->d_flags &
27671 + (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
27672 + can_no_dreval = 0;
27673 + spin_unlock(&dentry->d_lock);
27676 + if (au_br_fhsm(br->br_perm)) {
27678 + AuDebugOn(!br->br_fhsm);
27684 + hdir = au_hi(dir, bindex);
27685 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
27687 + wbr_wh_write_lock(wbr);
27688 + err = au_wh_init(br, sb);
27690 + wbr_wh_write_unlock(wbr);
27691 + au_hn_inode_unlock(hdir);
27693 + if (!err && do_free) {
27694 + au_kfree_rcu(wbr);
27695 + br->br_wbr = NULL;
27699 + if (can_no_dreval)
27700 + au_fset_si(sbinfo, NO_DREVAL);
27702 + au_fclr_si(sbinfo, NO_DREVAL);
27705 + au_fset_si(sbinfo, FHSM);
27706 + for (bindex = bbot; bindex >= 0; bindex--) {
27707 + br = au_sbr(sb, bindex);
27708 + if (au_br_fhsm(br->br_perm)) {
27709 + au_fhsm_set_bottom(sb, bindex);
27714 + au_fclr_si(sbinfo, FHSM);
27715 + au_fhsm_set_bottom(sb, -1);
27721 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
27724 + unsigned int tmp;
27725 + aufs_bindex_t bindex, bbot;
27726 + struct au_opt *opt;
27727 + struct au_opt_xino *opt_xino, xino;
27728 + struct au_sbinfo *sbinfo;
27729 + struct au_branch *br;
27730 + struct inode *dir;
27732 + SiMustWriteLock(sb);
27737 + while (err >= 0 && opt->type != Opt_tail)
27738 + err = au_opt_simple(sb, opt++, opts);
27741 + else if (unlikely(err < 0))
27744 + /* disable xino and udba temporary */
27745 + sbinfo = au_sbi(sb);
27746 + tmp = sbinfo->si_mntflags;
27747 + au_opt_clr(sbinfo->si_mntflags, XINO);
27748 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
27751 + while (err >= 0 && opt->type != Opt_tail)
27752 + err = au_opt_br(sb, opt++, opts);
27755 + else if (unlikely(err < 0))
27758 + bbot = au_sbbot(sb);
27759 + if (unlikely(bbot < 0)) {
27761 + pr_err("no branches\n");
27765 + if (au_opt_test(tmp, XINO))
27766 + au_opt_set(sbinfo->si_mntflags, XINO);
27768 + while (!err && opt->type != Opt_tail)
27769 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
27770 + if (unlikely(err))
27773 + err = au_opts_verify(sb, sb->s_flags, tmp);
27774 + if (unlikely(err))
27777 + /* restore xino */
27778 + if (au_opt_test(tmp, XINO) && !opt_xino) {
27779 + xino.file = au_xino_def(sb);
27780 + err = PTR_ERR(xino.file);
27781 + if (IS_ERR(xino.file))
27784 + err = au_xino_set(sb, &xino, /*remount*/0);
27786 + if (unlikely(err))
27790 + /* restore udba */
27791 + tmp &= AuOptMask_UDBA;
27792 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27793 + sbinfo->si_mntflags |= tmp;
27794 + bbot = au_sbbot(sb);
27795 + for (bindex = 0; bindex <= bbot; bindex++) {
27796 + br = au_sbr(sb, bindex);
27797 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
27798 + if (unlikely(err))
27799 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
27801 + /* go on even if err */
27803 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
27804 + dir = d_inode(sb->s_root);
27805 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
27812 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
27815 + unsigned char no_dreval;
27816 + struct inode *dir;
27817 + struct au_opt_xino *opt_xino;
27818 + struct au_opt *opt;
27819 + struct au_sbinfo *sbinfo;
27821 + SiMustWriteLock(sb);
27823 + err = au_dr_opt_flush(sb);
27824 + if (unlikely(err))
27826 + au_fset_opts(opts->flags, DR_FLUSHED);
27828 + dir = d_inode(sb->s_root);
27829 + sbinfo = au_sbi(sb);
27832 + while (err >= 0 && opt->type != Opt_tail) {
27833 + err = au_opt_simple(sb, opt, opts);
27835 + err = au_opt_br(sb, opt, opts);
27837 + err = au_opt_xino(sb, opt, &opt_xino, opts);
27843 + /* go on even err */
27845 + no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
27846 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
27847 + if (unlikely(rerr && !err))
27850 + if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
27851 + au_fset_opts(opts->flags, REFRESH_IDOP);
27853 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
27854 + rerr = au_xib_trunc(sb);
27855 + if (unlikely(rerr && !err))
27859 + /* will be handled by the caller */
27860 + if (!au_ftest_opts(opts->flags, REFRESH)
27861 + && (opts->given_udba
27862 + || au_opt_test(sbinfo->si_mntflags, XINO)
27863 + || au_ftest_opts(opts->flags, REFRESH_IDOP)
27865 + au_fset_opts(opts->flags, REFRESH);
27867 + AuDbg("status 0x%x\n", opts->flags);
27873 +/* ---------------------------------------------------------------------- */
27875 +unsigned int au_opt_udba(struct super_block *sb)
27877 + return au_mntflags(sb) & AuOptMask_UDBA;
27879 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
27880 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
27881 +++ linux/fs/aufs/opts.h 2022-03-21 14:49:05.729966343 +0100
27883 +/* SPDX-License-Identifier: GPL-2.0 */
27885 + * Copyright (C) 2005-2021 Junjiro R. Okajima
27887 + * This program, aufs is free software; you can redistribute it and/or modify
27888 + * it under the terms of the GNU General Public License as published by
27889 + * the Free Software Foundation; either version 2 of the License, or
27890 + * (at your option) any later version.
27892 + * This program is distributed in the hope that it will be useful,
27893 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27894 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27895 + * GNU General Public License for more details.
27897 + * You should have received a copy of the GNU General Public License
27898 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27902 + * mount options/flags
27905 +#ifndef __AUFS_OPTS_H__
27906 +#define __AUFS_OPTS_H__
27910 +#include <linux/fs_parser.h>
27911 +#include <linux/namei.h>
27912 +#include <linux/path.h>
27916 + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
27917 + Opt_idel, Opt_imod,
27918 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
27919 + Opt_xino, Opt_noxino,
27920 + Opt_trunc_xino, Opt_trunc_xino_v,
27921 + Opt_trunc_xino_path, Opt_itrunc_xino,
27924 + Opt_plink, Opt_list_plink,
27927 + Opt_diropq, Opt_diropq_a, Opt_diropq_w,
27929 + Opt_wbr_copyup, Opt_wbr_create,
27931 + Opt_verbose, Opt_noverbose,
27932 + Opt_sum, Opt_wsum,
27936 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
27939 +/* ---------------------------------------------------------------------- */
27942 +#define AuOpt_XINO 1 /* external inode number bitmap
27943 + and translation table */
27944 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
27945 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
27946 +#define AuOpt_UDBA_REVAL (1 << 3)
27947 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
27948 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
27949 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
27950 +#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
27952 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
27953 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
27954 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
27955 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
27956 +#define AuOpt_VERBOSE (1 << 13) /* print the cause of error */
27957 +#define AuOpt_DIO (1 << 14) /* direct io */
27958 +#define AuOpt_DIRREN (1 << 15) /* directory rename */
27960 +#ifndef CONFIG_AUFS_HNOTIFY
27961 +#undef AuOpt_UDBA_HNOTIFY
27962 +#define AuOpt_UDBA_HNOTIFY 0
27964 +#ifndef CONFIG_AUFS_DIRREN
27965 +#undef AuOpt_DIRREN
27966 +#define AuOpt_DIRREN 0
27968 +#ifndef CONFIG_AUFS_SHWH
27970 +#define AuOpt_SHWH 0
27973 +#define AuOpt_Def (AuOpt_XINO \
27974 + | AuOpt_UDBA_REVAL \
27976 + /* | AuOpt_DIRPERM1 */ \
27977 + | AuOpt_WARN_PERM)
27978 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
27979 + | AuOpt_UDBA_REVAL \
27980 + | AuOpt_UDBA_HNOTIFY)
27982 +#define AuOpt_LkupDirFlags (LOOKUP_FOLLOW | LOOKUP_DIRECTORY)
27984 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
27985 +#define au_opt_set(flags, name) do { \
27986 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
27987 + ((flags) |= AuOpt_##name); \
27989 +#define au_opt_set_udba(flags, name) do { \
27990 + (flags) &= ~AuOptMask_UDBA; \
27991 + ((flags) |= AuOpt_##name); \
27993 +#define au_opt_clr(flags, name) do { \
27994 + ((flags) &= ~AuOpt_##name); \
27997 +static inline unsigned int au_opts_plink(unsigned int mntflags)
27999 +#ifdef CONFIG_PROC_FS
28002 + return mntflags & ~AuOpt_PLINK;
28006 +/* ---------------------------------------------------------------------- */
28008 +/* policies to select one among multiple writable branches */
28010 + AuWbrCreate_TDP, /* top down parent */
28011 + AuWbrCreate_RR, /* round robin */
28012 + AuWbrCreate_MFS, /* most free space */
28013 + AuWbrCreate_MFSV, /* mfs with seconds */
28014 + AuWbrCreate_MFSRR, /* mfs then rr */
28015 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
28016 + AuWbrCreate_TDMFS, /* top down regardless parent and mfs */
28017 + AuWbrCreate_TDMFSV, /* top down regardless parent and mfs */
28018 + AuWbrCreate_PMFS, /* parent and mfs */
28019 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
28020 + AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
28021 + AuWbrCreate_PMFSRRV, /* plus seconds */
28023 + AuWbrCreate_Def = AuWbrCreate_TDP
28027 + AuWbrCopyup_TDP, /* top down parent */
28028 + AuWbrCopyup_BUP, /* bottom up parent */
28029 + AuWbrCopyup_BU, /* bottom up */
28031 + AuWbrCopyup_Def = AuWbrCopyup_TDP
28034 +/* ---------------------------------------------------------------------- */
28038 +struct au_opt_add {
28039 + aufs_bindex_t bindex;
28042 + struct path path;
28045 +struct au_opt_del {
28047 + struct path h_path;
28050 +struct au_opt_mod {
28053 + struct dentry *h_root;
28056 +struct au_opt_xino {
28058 + struct file *file;
28061 +struct au_opt_xino_itrunc {
28062 + aufs_bindex_t bindex;
28065 +struct au_opt_wbr_create {
28068 + unsigned long long mfsrr_watermark;
28074 + struct au_opt_xino xino;
28075 + struct au_opt_xino_itrunc xino_itrunc;
28076 + struct au_opt_add add;
28077 + struct au_opt_del del;
28078 + struct au_opt_mod mod;
28081 + unsigned int rdblk;
28082 + unsigned int rdhash;
28084 + struct au_opt_wbr_create wbr_create;
28086 + unsigned int fhsm_second;
28087 + bool tf; /* generic flag, true or false */
28092 +#define AuOpts_REMOUNT 1
28093 +#define AuOpts_REFRESH (1 << 1)
28094 +#define AuOpts_TRUNC_XIB (1 << 2)
28095 +#define AuOpts_REFRESH_DYAOP (1 << 3)
28096 +#define AuOpts_REFRESH_IDOP (1 << 4)
28097 +#define AuOpts_DR_FLUSHED (1 << 5)
28098 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
28099 +#define au_fset_opts(flags, name) \
28100 + do { (flags) |= AuOpts_##name; } while (0)
28101 +#define au_fclr_opts(flags, name) \
28102 + do { (flags) &= ~AuOpts_##name; } while (0)
28104 +#ifndef CONFIG_AUFS_DIRREN
28105 +#undef AuOpts_DR_FLUSHED
28106 +#define AuOpts_DR_FLUSHED 0
28110 + struct au_opt *opt;
28113 + unsigned int given_udba;
28114 + unsigned int flags;
28115 + unsigned long sb_flags;
28118 +/* ---------------------------------------------------------------------- */
28121 +int au_br_perm_val(char *perm);
28122 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
28123 +int au_udba_val(char *str);
28124 +const char *au_optstr_udba(int udba);
28125 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create);
28126 +const char *au_optstr_wbr_create(int wbr_create);
28127 +int au_wbr_copyup_val(char *str);
28128 +const char *au_optstr_wbr_copyup(int wbr_copyup);
28130 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
28131 + aufs_bindex_t bindex);
28132 +struct super_block;
28133 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
28134 + unsigned int pending);
28135 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
28136 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
28138 +unsigned int au_opt_udba(struct super_block *sb);
28141 +int aufs_fsctx_init(struct fs_context *fc);
28142 +extern const struct fs_parameter_spec aufs_fsctx_paramspec[];
28144 +#endif /* __KERNEL__ */
28145 +#endif /* __AUFS_OPTS_H__ */
28146 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
28147 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
28148 +++ linux/fs/aufs/plink.c 2022-03-21 14:49:05.729966343 +0100
28150 +// SPDX-License-Identifier: GPL-2.0
28152 + * Copyright (C) 2005-2021 Junjiro R. Okajima
28154 + * This program, aufs is free software; you can redistribute it and/or modify
28155 + * it under the terms of the GNU General Public License as published by
28156 + * the Free Software Foundation; either version 2 of the License, or
28157 + * (at your option) any later version.
28159 + * This program is distributed in the hope that it will be useful,
28160 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28161 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28162 + * GNU General Public License for more details.
28164 + * You should have received a copy of the GNU General Public License
28165 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28175 + * the pseudo-link maintenance mode.
28176 + * during a user process maintains the pseudo-links,
28177 + * prohibit adding a new plink and branch manipulation.
28181 + * For entry functions which will handle plink, and i_mutex is already held
28183 + * They cannot wait and should return an error at once.
28184 + * Callers has to check the error.
28186 + * For entry functions which will handle plink, but i_mutex is not held
28188 + * They can wait the plink maintenance mode to finish.
28190 + * They behave like F_SETLK and F_SETLKW.
28191 + * If the caller never handle plink, then both flags are unnecessary.
28194 +int au_plink_maint(struct super_block *sb, int flags)
28198 + struct task_struct *parent, *prev;
28199 + struct au_sbinfo *sbi;
28201 + SiMustAnyLock(sb);
28204 + if (!au_opt_test(au_mntflags(sb), PLINK))
28207 + sbi = au_sbi(sb);
28208 + pid = sbi->si_plink_maint_pid;
28209 + if (!pid || pid == current->pid)
28212 + /* todo: it highly depends upon /sbin/mount.aufs */
28214 + parent = current;
28218 + parent = rcu_dereference(parent->real_parent);
28219 + if (parent == prev)
28221 + ppid = task_pid_vnr(parent);
28222 + if (pid == ppid) {
28223 + rcu_read_unlock();
28228 + rcu_read_unlock();
28230 + if (au_ftest_lock(flags, NOPLMW)) {
28231 + /* if there is no i_mutex lock in VFS, we don't need to wait */
28232 + /* AuDebugOn(!lockdep_depth(current)); */
28233 + while (sbi->si_plink_maint_pid) {
28234 + si_read_unlock(sb);
28235 + /* gave up wake_up_bit() */
28236 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
28238 + if (au_ftest_lock(flags, FLUSH))
28239 + au_nwt_flush(&sbi->si_nowait);
28240 + si_noflush_read_lock(sb);
28242 + } else if (au_ftest_lock(flags, NOPLM)) {
28243 + AuDbg("ppid %d, pid %d\n", ppid, pid);
28251 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
28253 + spin_lock(&sbinfo->si_plink_maint_lock);
28254 + sbinfo->si_plink_maint_pid = 0;
28255 + spin_unlock(&sbinfo->si_plink_maint_lock);
28256 + wake_up_all(&sbinfo->si_plink_wq);
28259 +int au_plink_maint_enter(struct super_block *sb)
28262 + struct au_sbinfo *sbinfo;
28265 + sbinfo = au_sbi(sb);
28266 + /* make sure i am the only one in this fs */
28267 + si_write_lock(sb, AuLock_FLUSH);
28268 + if (au_opt_test(au_mntflags(sb), PLINK)) {
28269 + spin_lock(&sbinfo->si_plink_maint_lock);
28270 + if (!sbinfo->si_plink_maint_pid)
28271 + sbinfo->si_plink_maint_pid = current->pid;
28274 + spin_unlock(&sbinfo->si_plink_maint_lock);
28276 + si_write_unlock(sb);
28281 +/* ---------------------------------------------------------------------- */
28283 +#ifdef CONFIG_AUFS_DEBUG
28284 +void au_plink_list(struct super_block *sb)
28287 + struct au_sbinfo *sbinfo;
28288 + struct hlist_bl_head *hbl;
28289 + struct hlist_bl_node *pos;
28290 + struct au_icntnr *icntnr;
28292 + SiMustAnyLock(sb);
28294 + sbinfo = au_sbi(sb);
28295 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28296 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28298 + for (i = 0; i < AuPlink_NHASH; i++) {
28299 + hbl = sbinfo->si_plink + i;
28300 + hlist_bl_lock(hbl);
28301 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28302 + AuDbg("%lu\n", icntnr->vfs_inode.i_ino);
28303 + hlist_bl_unlock(hbl);
28308 +/* is the inode pseudo-linked? */
28309 +int au_plink_test(struct inode *inode)
28312 + struct au_sbinfo *sbinfo;
28313 + struct hlist_bl_head *hbl;
28314 + struct hlist_bl_node *pos;
28315 + struct au_icntnr *icntnr;
28317 + sbinfo = au_sbi(inode->i_sb);
28318 + AuRwMustAnyLock(&sbinfo->si_rwsem);
28319 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
28320 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28323 + i = au_plink_hash(inode->i_ino);
28324 + hbl = sbinfo->si_plink + i;
28325 + hlist_bl_lock(hbl);
28326 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28327 + if (&icntnr->vfs_inode == inode) {
28331 + hlist_bl_unlock(hbl);
28335 +/* ---------------------------------------------------------------------- */
28338 + * generate a name for plink.
28339 + * the file will be stored under AUFS_WH_PLINKDIR.
28341 +/* 20 is max digits length of ulong 64 */
28342 +#define PLINK_NAME_LEN ((20 + 1) * 2)
28344 +static int plink_name(char *name, int len, struct inode *inode,
28345 + aufs_bindex_t bindex)
28348 + struct inode *h_inode;
28350 + h_inode = au_h_iptr(inode, bindex);
28351 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
28355 +struct au_do_plink_lkup_args {
28356 + struct dentry **errp;
28357 + struct qstr *tgtname;
28358 + struct path *h_ppath;
28361 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
28362 + struct path *h_ppath)
28364 + struct dentry *h_dentry;
28365 + struct inode *h_inode;
28367 + h_inode = d_inode(h_ppath->dentry);
28368 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2);
28369 + h_dentry = vfsub_lkup_one(tgtname, h_ppath);
28370 + inode_unlock_shared(h_inode);
28375 +static void au_call_do_plink_lkup(void *args)
28377 + struct au_do_plink_lkup_args *a = args;
28378 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_ppath);
28381 +/* lookup the plink-ed @inode under the branch at @bindex */
28382 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
28384 + struct dentry *h_dentry;
28385 + struct au_branch *br;
28386 + struct path h_ppath;
28388 + char a[PLINK_NAME_LEN];
28389 + struct qstr tgtname = QSTR_INIT(a, 0);
28391 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28393 + br = au_sbr(inode->i_sb, bindex);
28394 + h_ppath.dentry = br->br_wbr->wbr_plink;
28395 + h_ppath.mnt = au_br_mnt(br);
28396 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28398 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28399 + struct au_do_plink_lkup_args args = {
28400 + .errp = &h_dentry,
28401 + .tgtname = &tgtname,
28402 + .h_ppath = &h_ppath
28405 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
28406 + if (unlikely(wkq_err))
28407 + h_dentry = ERR_PTR(wkq_err);
28409 + h_dentry = au_do_plink_lkup(&tgtname, &h_ppath);
28414 +/* create a pseudo-link */
28415 +static int do_whplink(struct qstr *tgt, struct path *h_ppath,
28416 + struct dentry *h_dentry)
28419 + struct path h_path;
28420 + struct inode *h_dir, *delegated;
28422 + h_dir = d_inode(h_ppath->dentry);
28423 + inode_lock_nested(h_dir, AuLsc_I_CHILD2);
28424 + h_path.mnt = h_ppath->mnt;
28426 + h_path.dentry = vfsub_lkup_one(tgt, h_ppath);
28427 + err = PTR_ERR(h_path.dentry);
28428 + if (IS_ERR(h_path.dentry))
28432 + /* wh.plink dir is not monitored */
28433 + /* todo: is it really safe? */
28434 + if (d_is_positive(h_path.dentry)
28435 + && d_inode(h_path.dentry) != d_inode(h_dentry)) {
28436 + delegated = NULL;
28437 + err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
28438 + if (unlikely(err == -EWOULDBLOCK)) {
28439 + pr_warn("cannot retry for NFSv4 delegation"
28440 + " for an internal unlink\n");
28443 + dput(h_path.dentry);
28444 + h_path.dentry = NULL;
28448 + if (!err && d_is_negative(h_path.dentry)) {
28449 + delegated = NULL;
28450 + err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
28451 + if (unlikely(err == -EWOULDBLOCK)) {
28452 + pr_warn("cannot retry for NFSv4 delegation"
28453 + " for an internal link\n");
28457 + dput(h_path.dentry);
28460 + inode_unlock(h_dir);
28464 +struct do_whplink_args {
28466 + struct qstr *tgt;
28467 + struct path *h_ppath;
28468 + struct dentry *h_dentry;
28471 +static void call_do_whplink(void *args)
28473 + struct do_whplink_args *a = args;
28474 + *a->errp = do_whplink(a->tgt, a->h_ppath, a->h_dentry);
28477 +static int whplink(struct dentry *h_dentry, struct inode *inode,
28478 + aufs_bindex_t bindex)
28480 + int err, wkq_err;
28481 + struct au_branch *br;
28482 + struct au_wbr *wbr;
28483 + struct path h_ppath;
28484 + char a[PLINK_NAME_LEN];
28485 + struct qstr tgtname = QSTR_INIT(a, 0);
28487 + br = au_sbr(inode->i_sb, bindex);
28488 + wbr = br->br_wbr;
28489 + h_ppath.dentry = wbr->wbr_plink;
28490 + h_ppath.mnt = au_br_mnt(br);
28491 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28493 + /* always superio. */
28494 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28495 + struct do_whplink_args args = {
28498 + .h_ppath = &h_ppath,
28499 + .h_dentry = h_dentry
28501 + wkq_err = au_wkq_wait(call_do_whplink, &args);
28502 + if (unlikely(wkq_err))
28505 + err = do_whplink(&tgtname, &h_ppath, h_dentry);
28511 + * create a new pseudo-link for @h_dentry on @bindex.
28512 + * the linked inode is held in aufs @inode.
28514 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
28515 + struct dentry *h_dentry)
28517 + struct super_block *sb;
28518 + struct au_sbinfo *sbinfo;
28519 + struct hlist_bl_head *hbl;
28520 + struct hlist_bl_node *pos;
28521 + struct au_icntnr *icntnr;
28522 + int found, err, cnt, i;
28524 + sb = inode->i_sb;
28525 + sbinfo = au_sbi(sb);
28526 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28527 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28529 + found = au_plink_test(inode);
28533 + i = au_plink_hash(inode->i_ino);
28534 + hbl = sbinfo->si_plink + i;
28537 + hlist_bl_lock(hbl);
28538 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink) {
28539 + if (&icntnr->vfs_inode == inode) {
28545 + icntnr = container_of(inode, struct au_icntnr, vfs_inode);
28546 + hlist_bl_add_head(&icntnr->plink, hbl);
28548 + hlist_bl_unlock(hbl);
28550 + cnt = au_hbl_count(hbl);
28551 +#define msg "unexpectedly unbalanced or too many pseudo-links"
28552 + if (cnt > AUFS_PLINK_WARN)
28553 + AuWarn1(msg ", %d\n", cnt);
28555 + err = whplink(h_dentry, inode, bindex);
28556 + if (unlikely(err)) {
28557 + pr_warn("err %d, damaged pseudo link.\n", err);
28558 + au_hbl_del(&icntnr->plink, hbl);
28559 + iput(&icntnr->vfs_inode);
28562 + iput(&icntnr->vfs_inode);
28565 +/* free all plinks */
28566 +void au_plink_put(struct super_block *sb, int verbose)
28569 + struct au_sbinfo *sbinfo;
28570 + struct hlist_bl_head *hbl;
28571 + struct hlist_bl_node *pos, *tmp;
28572 + struct au_icntnr *icntnr;
28574 + SiMustWriteLock(sb);
28576 + sbinfo = au_sbi(sb);
28577 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28578 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28580 + /* no spin_lock since sbinfo is write-locked */
28582 + for (i = 0; i < AuPlink_NHASH; i++) {
28583 + hbl = sbinfo->si_plink + i;
28584 + if (!warned && verbose && !hlist_bl_empty(hbl)) {
28585 + pr_warn("pseudo-link is not flushed");
28588 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink)
28589 + iput(&icntnr->vfs_inode);
28590 + INIT_HLIST_BL_HEAD(hbl);
28594 +void au_plink_clean(struct super_block *sb, int verbose)
28596 + struct dentry *root;
28598 + root = sb->s_root;
28599 + aufs_write_lock(root);
28600 + if (au_opt_test(au_mntflags(sb), PLINK))
28601 + au_plink_put(sb, verbose);
28602 + aufs_write_unlock(root);
28605 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
28608 + aufs_bindex_t btop, bbot, bindex;
28611 + btop = au_ibtop(inode);
28612 + bbot = au_ibbot(inode);
28614 + for (bindex = btop; bindex <= bbot; bindex++) {
28615 + if (!au_h_iptr(inode, bindex)
28616 + || au_ii_br_id(inode, bindex) != br_id)
28618 + au_set_h_iptr(inode, bindex, NULL, 0);
28623 + for (bindex = btop; bindex <= bbot; bindex++)
28624 + if (au_h_iptr(inode, bindex)) {
28634 +/* free the plinks on a branch specified by @br_id */
28635 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
28637 + struct au_sbinfo *sbinfo;
28638 + struct hlist_bl_head *hbl;
28639 + struct hlist_bl_node *pos, *tmp;
28640 + struct au_icntnr *icntnr;
28641 + struct inode *inode;
28644 + SiMustWriteLock(sb);
28646 + sbinfo = au_sbi(sb);
28647 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28648 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28650 + /* no bit_lock since sbinfo is write-locked */
28651 + for (i = 0; i < AuPlink_NHASH; i++) {
28652 + hbl = sbinfo->si_plink + i;
28653 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) {
28654 + inode = au_igrab(&icntnr->vfs_inode);
28655 + ii_write_lock_child(inode);
28656 + do_put = au_plink_do_half_refresh(inode, br_id);
28658 + hlist_bl_del(&icntnr->plink);
28661 + ii_write_unlock(inode);
28666 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
28667 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
28668 +++ linux/fs/aufs/poll.c 2022-03-21 14:49:05.729966343 +0100
28670 +// SPDX-License-Identifier: GPL-2.0
28672 + * Copyright (C) 2005-2021 Junjiro R. Okajima
28674 + * This program, aufs is free software; you can redistribute it and/or modify
28675 + * it under the terms of the GNU General Public License as published by
28676 + * the Free Software Foundation; either version 2 of the License, or
28677 + * (at your option) any later version.
28679 + * This program is distributed in the hope that it will be useful,
28680 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28681 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28682 + * GNU General Public License for more details.
28684 + * You should have received a copy of the GNU General Public License
28685 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28690 + * There is only one filesystem which implements ->poll operation, currently.
28695 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt)
28698 + struct file *h_file;
28699 + struct super_block *sb;
28701 + /* We should pretend an error happened. */
28702 + mask = EPOLLERR /* | EPOLLIN | EPOLLOUT */;
28703 + sb = file->f_path.dentry->d_sb;
28704 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
28706 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
28707 + if (IS_ERR(h_file)) {
28708 + AuDbg("h_file %ld\n", PTR_ERR(h_file));
28712 + mask = vfs_poll(h_file, pt);
28713 + fput(h_file); /* instead of au_read_post() */
28716 + si_read_unlock(sb);
28717 + if (mask & EPOLLERR)
28718 + AuDbg("mask 0x%x\n", mask);
28721 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
28722 --- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
28723 +++ linux/fs/aufs/posix_acl.c 2022-03-21 14:49:05.729966343 +0100
28725 +// SPDX-License-Identifier: GPL-2.0
28727 + * Copyright (C) 2014-2021 Junjiro R. Okajima
28729 + * This program, aufs is free software; you can redistribute it and/or modify
28730 + * it under the terms of the GNU General Public License as published by
28731 + * the Free Software Foundation; either version 2 of the License, or
28732 + * (at your option) any later version.
28734 + * This program is distributed in the hope that it will be useful,
28735 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28736 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28737 + * GNU General Public License for more details.
28739 + * You should have received a copy of the GNU General Public License
28740 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28744 + * posix acl operations
28747 +#include <linux/fs.h>
28750 +struct posix_acl *aufs_get_acl(struct inode *inode, int type, bool rcu)
28752 + struct posix_acl *acl;
28754 + aufs_bindex_t bindex;
28755 + struct inode *h_inode;
28756 + struct super_block *sb;
28758 + acl = ERR_PTR(-ECHILD);
28763 + sb = inode->i_sb;
28764 + si_read_lock(sb, AuLock_FLUSH);
28765 + ii_read_lock_child(inode);
28766 + if (!(sb->s_flags & SB_POSIXACL))
28769 + bindex = au_ibtop(inode);
28770 + h_inode = au_h_iptr(inode, bindex);
28771 + if (unlikely(!h_inode
28772 + || ((h_inode->i_mode & S_IFMT)
28773 + != (inode->i_mode & S_IFMT)))) {
28774 + err = au_busy_or_stale();
28775 + acl = ERR_PTR(err);
28779 + /* always topmost only */
28780 + acl = get_acl(h_inode, type);
28782 + forget_cached_acl(inode, type);
28784 + set_cached_acl(inode, type, acl);
28787 + ii_read_unlock(inode);
28788 + si_read_unlock(sb);
28791 + AuTraceErrPtr(acl);
28795 +int aufs_set_acl(struct user_namespace *userns, struct inode *inode,
28796 + struct posix_acl *acl, int type)
28800 + struct dentry *dentry;
28801 + struct au_sxattr arg = {
28802 + .type = AU_ACL_SET,
28809 + IMustLock(inode);
28811 + if (inode->i_ino == AUFS_ROOT_INO)
28812 + dentry = dget(inode->i_sb->s_root);
28814 + dentry = d_find_alias(inode);
28816 + dentry = d_find_any_alias(inode);
28818 + pr_warn("cannot handle this inode, "
28819 + "please report to aufs-users ML\n");
28825 + ssz = au_sxattr(dentry, inode, &arg);
28826 + /* forget even it if succeeds since the branch might set differently */
28827 + forget_cached_acl(inode, type);
28836 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
28837 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
28838 +++ linux/fs/aufs/procfs.c 2022-03-21 14:49:05.729966343 +0100
28840 +// SPDX-License-Identifier: GPL-2.0
28842 + * Copyright (C) 2010-2021 Junjiro R. Okajima
28844 + * This program, aufs is free software; you can redistribute it and/or modify
28845 + * it under the terms of the GNU General Public License as published by
28846 + * the Free Software Foundation; either version 2 of the License, or
28847 + * (at your option) any later version.
28849 + * This program is distributed in the hope that it will be useful,
28850 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28851 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28852 + * GNU General Public License for more details.
28854 + * You should have received a copy of the GNU General Public License
28855 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28859 + * procfs interfaces
28862 +#include <linux/proc_fs.h>
28865 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
28867 + struct au_sbinfo *sbinfo;
28869 + sbinfo = file->private_data;
28871 + au_plink_maint_leave(sbinfo);
28872 + kobject_put(&sbinfo->si_kobj);
28878 +static void au_procfs_plm_write_clean(struct file *file)
28880 + struct au_sbinfo *sbinfo;
28882 + sbinfo = file->private_data;
28884 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
28887 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
28890 + struct super_block *sb;
28891 + struct au_sbinfo *sbinfo;
28892 + struct hlist_bl_node *pos;
28895 + if (unlikely(file->private_data))
28899 + /* don't use au_sbilist_lock() here */
28900 + hlist_bl_lock(&au_sbilist);
28901 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
28902 + if (id == sysaufs_si_id(sbinfo)) {
28903 + if (kobject_get_unless_zero(&sbinfo->si_kobj))
28904 + sb = sbinfo->si_sb;
28907 + hlist_bl_unlock(&au_sbilist);
28910 + if (unlikely(!sb))
28913 + err = au_plink_maint_enter(sb);
28915 + /* keep kobject_get() */
28916 + file->private_data = sbinfo;
28918 + kobject_put(&sbinfo->si_kobj);
28924 + * Accept a valid "si=xxxx" only.
28925 + * Once it is accepted successfully, accept "clean" too.
28927 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
28928 + size_t count, loff_t *ppos)
28931 + unsigned long id;
28932 + /* last newline is allowed */
28933 + char buf[3 + sizeof(unsigned long) * 2 + 1];
28936 + if (unlikely(!capable(CAP_SYS_ADMIN)))
28940 + if (unlikely(count > sizeof(buf)))
28943 + err = copy_from_user(buf, ubuf, count);
28944 + if (unlikely(err)) {
28951 + if (!strcmp("clean", buf)) {
28952 + au_procfs_plm_write_clean(file);
28953 + goto out_success;
28954 + } else if (unlikely(strncmp("si=", buf, 3)))
28957 + err = kstrtoul(buf + 3, 16, &id);
28958 + if (unlikely(err))
28961 + err = au_procfs_plm_write_si(file, id);
28962 + if (unlikely(err))
28966 + err = count; /* success */
28971 +static const struct proc_ops au_procfs_plm_op = {
28972 + .proc_write = au_procfs_plm_write,
28973 + .proc_release = au_procfs_plm_release
28976 +/* ---------------------------------------------------------------------- */
28978 +static struct proc_dir_entry *au_procfs_dir;
28980 +void au_procfs_fin(void)
28982 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
28983 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
28986 +int __init au_procfs_init(void)
28989 + struct proc_dir_entry *entry;
28992 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
28993 + if (unlikely(!au_procfs_dir))
28996 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | 0200,
28997 + au_procfs_dir, &au_procfs_plm_op);
28998 + if (unlikely(!entry))
29002 + goto out; /* success */
29006 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29010 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
29011 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
29012 +++ linux/fs/aufs/rdu.c 2022-03-21 14:49:05.729966343 +0100
29014 +// SPDX-License-Identifier: GPL-2.0
29016 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29018 + * This program, aufs is free software; you can redistribute it and/or modify
29019 + * it under the terms of the GNU General Public License as published by
29020 + * the Free Software Foundation; either version 2 of the License, or
29021 + * (at your option) any later version.
29023 + * This program is distributed in the hope that it will be useful,
29024 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29025 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29026 + * GNU General Public License for more details.
29028 + * You should have received a copy of the GNU General Public License
29029 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29033 + * readdir in userspace.
29036 +#include <linux/compat.h>
29037 +#include <linux/fs_stack.h>
29038 +#include <linux/security.h>
29041 +/* bits for struct aufs_rdu.flags */
29042 +#define AuRdu_CALLED 1
29043 +#define AuRdu_CONT (1 << 1)
29044 +#define AuRdu_FULL (1 << 2)
29045 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
29046 +#define au_fset_rdu(flags, name) \
29047 + do { (flags) |= AuRdu_##name; } while (0)
29048 +#define au_fclr_rdu(flags, name) \
29049 + do { (flags) &= ~AuRdu_##name; } while (0)
29051 +struct au_rdu_arg {
29052 + struct dir_context ctx;
29053 + struct aufs_rdu *rdu;
29054 + union au_rdu_ent_ul ent;
29055 + unsigned long end;
29057 + struct super_block *sb;
29061 +static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
29062 + loff_t offset, u64 h_ino, unsigned int d_type)
29065 + struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
29066 + struct aufs_rdu *rdu = arg->rdu;
29067 + struct au_rdu_ent ent;
29071 + au_fset_rdu(rdu->cookie.flags, CALLED);
29072 + len = au_rdu_len(nlen);
29073 + if (arg->ent.ul + len < arg->end) {
29075 + ent.bindex = rdu->cookie.bindex;
29076 + ent.type = d_type;
29078 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
29079 + ent.type = DT_UNKNOWN;
29081 + /* unnecessary to support mmap_sem since this is a dir */
29083 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
29085 + if (copy_to_user(arg->ent.e->name, name, nlen))
29087 + /* the terminating NULL */
29088 + if (__put_user(0, arg->ent.e->name + nlen))
29091 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
29092 + arg->ent.ul += len;
29096 + au_fset_rdu(rdu->cookie.flags, FULL);
29098 + rdu->tail = arg->ent;
29102 + /* AuTraceErr(err); */
29106 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
29110 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
29112 + /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
29113 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
29115 + if (unlikely(offset != cookie->h_pos))
29121 + au_fclr_rdu(cookie->flags, CALLED);
29123 + err = vfsub_iterate_dir(h_file, &arg->ctx);
29127 + && au_ftest_rdu(cookie->flags, CALLED)
29128 + && !au_ftest_rdu(cookie->flags, FULL));
29129 + cookie->h_pos = h_file->f_pos;
29136 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
29139 + aufs_bindex_t bbot;
29140 + struct au_rdu_arg arg = {
29142 + .actor = au_rdu_fill
29145 + struct dentry *dentry;
29146 + struct inode *inode;
29147 + struct file *h_file;
29148 + struct au_rdu_cookie *cookie = &rdu->cookie;
29150 + /* VERIFY_WRITE */
29151 + err = !access_ok(rdu->ent.e, rdu->sz);
29152 + if (unlikely(err)) {
29158 + rdu->tail = rdu->ent;
29161 + arg.ent = rdu->ent;
29162 + arg.end = arg.ent.ul;
29163 + arg.end += rdu->sz;
29166 + if (unlikely(!file->f_op->iterate && !file->f_op->iterate_shared))
29169 + err = security_file_permission(file, MAY_READ);
29171 + if (unlikely(err))
29174 + dentry = file->f_path.dentry;
29175 + inode = d_inode(dentry);
29176 + inode_lock_shared(inode);
29178 + arg.sb = inode->i_sb;
29179 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
29180 + if (unlikely(err))
29182 + err = au_alive_dir(dentry);
29183 + if (unlikely(err))
29185 + /* todo: reval? */
29186 + fi_read_lock(file);
29189 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
29190 + && cookie->generation != au_figen(file)))
29195 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
29197 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
29199 + bbot = au_fbtop(file);
29200 + if (cookie->bindex < bbot)
29201 + cookie->bindex = bbot;
29202 + bbot = au_fbbot_dir(file);
29203 + /* AuDbg("b%d, b%d\n", cookie->bindex, bbot); */
29204 + for (; !err && cookie->bindex <= bbot;
29205 + cookie->bindex++, cookie->h_pos = 0) {
29206 + h_file = au_hf_dir(file, cookie->bindex);
29210 + au_fclr_rdu(cookie->flags, FULL);
29211 + err = au_rdu_do(h_file, &arg);
29213 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
29216 + AuDbg("rent %llu\n", rdu->rent);
29218 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
29219 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
29220 + au_fset_rdu(cookie->flags, CONT);
29221 + cookie->generation = au_figen(file);
29224 + ii_read_lock_child(inode);
29225 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibtop(inode)));
29226 + ii_read_unlock(inode);
29229 + fi_read_unlock(file);
29231 + si_read_unlock(arg.sb);
29233 + inode_unlock_shared(inode);
29239 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
29243 + unsigned long long nent;
29244 + union au_rdu_ent_ul *u;
29245 + struct au_rdu_ent ent;
29246 + struct super_block *sb;
29249 + nent = rdu->nent;
29251 + sb = file->f_path.dentry->d_sb;
29252 + si_read_lock(sb, AuLock_FLUSH);
29253 + while (nent-- > 0) {
29254 + /* unnecessary to support mmap_sem since this is a dir */
29255 + err = copy_from_user(&ent, u->e, sizeof(ent));
29257 + /* VERIFY_WRITE */
29258 + err = !access_ok(&u->e->ino, sizeof(ino));
29259 + if (unlikely(err)) {
29265 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
29267 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
29269 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
29271 + if (unlikely(err)) {
29276 + err = __put_user(ino, &u->e->ino);
29277 + if (unlikely(err)) {
29282 + u->ul += au_rdu_len(ent.nlen);
29284 + si_read_unlock(sb);
29289 +/* ---------------------------------------------------------------------- */
29291 +static int au_rdu_verify(struct aufs_rdu *rdu)
29293 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
29294 + "%llu, b%d, 0x%x, g%u}\n",
29295 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
29297 + rdu->rent, rdu->shwh, rdu->full,
29298 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
29299 + rdu->cookie.generation);
29301 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
29305 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
29309 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29312 + struct aufs_rdu rdu;
29313 + void __user *p = (void __user *)arg;
29315 + err = copy_from_user(&rdu, p, sizeof(rdu));
29316 + if (unlikely(err)) {
29321 + err = au_rdu_verify(&rdu);
29322 + if (unlikely(err))
29326 + case AUFS_CTL_RDU:
29327 + err = au_rdu(file, &rdu);
29328 + if (unlikely(err))
29331 + e = copy_to_user(p, &rdu, sizeof(rdu));
29332 + if (unlikely(e)) {
29337 + case AUFS_CTL_RDU_INO:
29338 + err = au_rdu_ino(file, &rdu);
29342 + /* err = -ENOTTY; */
29351 +#ifdef CONFIG_COMPAT
29352 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29355 + struct aufs_rdu rdu;
29356 + void __user *p = compat_ptr(arg);
29358 + /* todo: get_user()? */
29359 + err = copy_from_user(&rdu, p, sizeof(rdu));
29360 + if (unlikely(err)) {
29365 + rdu.ent.e = compat_ptr(rdu.ent.ul);
29366 + err = au_rdu_verify(&rdu);
29367 + if (unlikely(err))
29371 + case AUFS_CTL_RDU:
29372 + err = au_rdu(file, &rdu);
29373 + if (unlikely(err))
29376 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
29377 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
29378 + e = copy_to_user(p, &rdu, sizeof(rdu));
29379 + if (unlikely(e)) {
29384 + case AUFS_CTL_RDU_INO:
29385 + err = au_rdu_ino(file, &rdu);
29389 + /* err = -ENOTTY; */
29398 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
29399 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
29400 +++ linux/fs/aufs/rwsem.h 2022-03-21 14:49:05.729966343 +0100
29402 +/* SPDX-License-Identifier: GPL-2.0 */
29404 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29406 + * This program, aufs is free software; you can redistribute it and/or modify
29407 + * it under the terms of the GNU General Public License as published by
29408 + * the Free Software Foundation; either version 2 of the License, or
29409 + * (at your option) any later version.
29411 + * This program is distributed in the hope that it will be useful,
29412 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29413 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29414 + * GNU General Public License for more details.
29416 + * You should have received a copy of the GNU General Public License
29417 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29421 + * simple read-write semaphore wrappers
29424 +#ifndef __AUFS_RWSEM_H__
29425 +#define __AUFS_RWSEM_H__
29429 +#include "debug.h"
29431 +/* in the future, the name 'au_rwsem' will be totally gone */
29432 +#define au_rwsem rw_semaphore
29434 +/* to debug easier, do not make them inlined functions */
29435 +#define AuRwMustNoWaiters(rw) AuDebugOn(rwsem_is_contended(rw))
29437 +#ifdef CONFIG_LOCKDEP
29438 +/* rwsem_is_locked() is unusable */
29439 +#define AuRwMustReadLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29440 + && !lockdep_recursing(current) \
29442 + && !lockdep_is_held_type(rw, 1))
29443 +#define AuRwMustWriteLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29444 + && !lockdep_recursing(current) \
29446 + && !lockdep_is_held_type(rw, 0))
29447 +#define AuRwMustAnyLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29448 + && !lockdep_recursing(current) \
29450 + && !lockdep_is_held(rw))
29451 +#define AuRwDestroy(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29452 + && !lockdep_recursing(current) \
29454 + && lockdep_is_held(rw))
29456 +#define AuRwMustReadLock(rw) do {} while (0)
29457 +#define AuRwMustWriteLock(rw) do {} while (0)
29458 +#define AuRwMustAnyLock(rw) do {} while (0)
29459 +#define AuRwDestroy(rw) do {} while (0)
29462 +#define au_rw_init(rw) init_rwsem(rw)
29464 +#define au_rw_init_wlock(rw) do { \
29465 + au_rw_init(rw); \
29466 + down_write(rw); \
29469 +#define au_rw_init_wlock_nested(rw, lsc) do { \
29470 + au_rw_init(rw); \
29471 + down_write_nested(rw, lsc); \
29474 +#define au_rw_read_lock(rw) down_read(rw)
29475 +#define au_rw_read_lock_nested(rw, lsc) down_read_nested(rw, lsc)
29476 +#define au_rw_read_unlock(rw) up_read(rw)
29477 +#define au_rw_dgrade_lock(rw) downgrade_write(rw)
29478 +#define au_rw_write_lock(rw) down_write(rw)
29479 +#define au_rw_write_lock_nested(rw, lsc) down_write_nested(rw, lsc)
29480 +#define au_rw_write_unlock(rw) up_write(rw)
29481 +/* why is not _nested version defined? */
29482 +#define au_rw_read_trylock(rw) down_read_trylock(rw)
29483 +#define au_rw_write_trylock(rw) down_write_trylock(rw)
29485 +#endif /* __KERNEL__ */
29486 +#endif /* __AUFS_RWSEM_H__ */
29487 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
29488 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
29489 +++ linux/fs/aufs/sbinfo.c 2022-03-21 14:49:05.729966343 +0100
29491 +// SPDX-License-Identifier: GPL-2.0
29493 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29495 + * This program, aufs is free software; you can redistribute it and/or modify
29496 + * it under the terms of the GNU General Public License as published by
29497 + * the Free Software Foundation; either version 2 of the License, or
29498 + * (at your option) any later version.
29500 + * This program is distributed in the hope that it will be useful,
29501 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29502 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29503 + * GNU General Public License for more details.
29505 + * You should have received a copy of the GNU General Public License
29506 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29510 + * superblock private data
29513 +#include <linux/iversion.h>
29517 + * they are necessary regardless sysfs is disabled.
29519 +void au_si_free(struct kobject *kobj)
29522 + struct au_sbinfo *sbinfo;
29523 + char *locked __maybe_unused; /* debug only */
29525 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
29526 + for (i = 0; i < AuPlink_NHASH; i++)
29527 + AuDebugOn(!hlist_bl_empty(sbinfo->si_plink + i));
29528 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
29530 + AuLCntZero(au_lcnt_read(&sbinfo->si_ninodes, /*do_rev*/0));
29531 + au_lcnt_fin(&sbinfo->si_ninodes, /*do_sync*/0);
29532 + AuLCntZero(au_lcnt_read(&sbinfo->si_nfiles, /*do_rev*/0));
29533 + au_lcnt_fin(&sbinfo->si_nfiles, /*do_sync*/0);
29535 + dbgaufs_si_fin(sbinfo);
29536 + au_rw_write_lock(&sbinfo->si_rwsem);
29537 + au_br_free(sbinfo);
29538 + au_rw_write_unlock(&sbinfo->si_rwsem);
29540 + au_kfree_try_rcu(sbinfo->si_branch);
29541 + mutex_destroy(&sbinfo->si_xib_mtx);
29542 + AuRwDestroy(&sbinfo->si_rwsem);
29544 + au_lcnt_wait_for_fin(&sbinfo->si_ninodes);
29545 + /* si_nfiles is waited too */
29546 + au_kfree_rcu(sbinfo);
29549 +struct au_sbinfo *au_si_alloc(struct super_block *sb)
29551 + struct au_sbinfo *sbinfo;
29555 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
29556 + if (unlikely(!sbinfo))
29559 + /* will be reallocated separately */
29560 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
29561 + if (unlikely(!sbinfo->si_branch))
29564 + err = sysaufs_si_init(sbinfo);
29566 + dbgaufs_si_null(sbinfo);
29567 + err = dbgaufs_si_init(sbinfo);
29568 + if (unlikely(err))
29569 + kobject_put(&sbinfo->si_kobj);
29571 + if (unlikely(err))
29574 + au_nwt_init(&sbinfo->si_nowait);
29575 + au_rw_init_wlock(&sbinfo->si_rwsem);
29577 + au_lcnt_init(&sbinfo->si_ninodes, /*release*/NULL);
29578 + au_lcnt_init(&sbinfo->si_nfiles, /*release*/NULL);
29580 + sbinfo->si_bbot = -1;
29581 + sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
29583 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
29584 + sbinfo->si_wbr_create = AuWbrCreate_Def;
29585 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
29586 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
29588 + au_fhsm_init(sbinfo);
29590 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
29592 + sbinfo->si_xino_jiffy = jiffies;
29593 + sbinfo->si_xino_expire
29594 + = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
29595 + mutex_init(&sbinfo->si_xib_mtx);
29596 + /* leave si_xib_last_pindex and si_xib_next_bit */
29598 + INIT_HLIST_BL_HEAD(&sbinfo->si_aopen);
29600 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
29601 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
29602 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
29603 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
29605 + for (i = 0; i < AuPlink_NHASH; i++)
29606 + INIT_HLIST_BL_HEAD(sbinfo->si_plink + i);
29607 + init_waitqueue_head(&sbinfo->si_plink_wq);
29608 + spin_lock_init(&sbinfo->si_plink_maint_lock);
29610 + INIT_HLIST_BL_HEAD(&sbinfo->si_files);
29612 + /* with getattr by default */
29613 + sbinfo->si_iop_array = aufs_iop;
29615 + /* leave other members for sysaufs and si_mnt. */
29616 + sbinfo->si_sb = sb;
29618 + sb->s_fs_info = sbinfo;
29621 + return sbinfo; /* success */
29624 + au_kfree_try_rcu(sbinfo->si_branch);
29626 + au_kfree_rcu(sbinfo);
29628 + return ERR_PTR(err);
29631 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink)
29634 + struct au_branch **brp;
29636 + AuRwMustWriteLock(&sbinfo->si_rwsem);
29639 + sz = sizeof(*brp) * (sbinfo->si_bbot + 1);
29640 + if (unlikely(!sz))
29641 + sz = sizeof(*brp);
29642 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS,
29645 + sbinfo->si_branch = brp;
29652 +/* ---------------------------------------------------------------------- */
29654 +unsigned int au_sigen_inc(struct super_block *sb)
29656 + unsigned int gen;
29657 + struct inode *inode;
29659 + SiMustWriteLock(sb);
29661 + gen = ++au_sbi(sb)->si_generation;
29662 + au_update_digen(sb->s_root);
29663 + inode = d_inode(sb->s_root);
29664 + au_update_iigen(inode, /*half*/0);
29665 + inode_inc_iversion(inode);
29669 +aufs_bindex_t au_new_br_id(struct super_block *sb)
29671 + aufs_bindex_t br_id;
29673 + struct au_sbinfo *sbinfo;
29675 + SiMustWriteLock(sb);
29677 + sbinfo = au_sbi(sb);
29678 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
29679 + br_id = ++sbinfo->si_last_br_id;
29680 + AuDebugOn(br_id < 0);
29681 + if (br_id && au_br_index(sb, br_id) < 0)
29688 +/* ---------------------------------------------------------------------- */
29690 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
29691 +int si_read_lock(struct super_block *sb, int flags)
29696 + if (au_ftest_lock(flags, FLUSH))
29697 + au_nwt_flush(&au_sbi(sb)->si_nowait);
29699 + si_noflush_read_lock(sb);
29700 + err = au_plink_maint(sb, flags);
29701 + if (unlikely(err))
29702 + si_read_unlock(sb);
29707 +int si_write_lock(struct super_block *sb, int flags)
29711 + if (au_ftest_lock(flags, FLUSH))
29712 + au_nwt_flush(&au_sbi(sb)->si_nowait);
29714 + si_noflush_write_lock(sb);
29715 + err = au_plink_maint(sb, flags);
29716 + if (unlikely(err))
29717 + si_write_unlock(sb);
29722 +/* dentry and super_block lock. call at entry point */
29723 +int aufs_read_lock(struct dentry *dentry, int flags)
29726 + struct super_block *sb;
29728 + sb = dentry->d_sb;
29729 + err = si_read_lock(sb, flags);
29730 + if (unlikely(err))
29733 + if (au_ftest_lock(flags, DW))
29734 + di_write_lock_child(dentry);
29736 + di_read_lock_child(dentry, flags);
29738 + if (au_ftest_lock(flags, GEN)) {
29739 + err = au_digen_test(dentry, au_sigen(sb));
29740 + if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
29741 + AuDebugOn(!err && au_dbrange_test(dentry));
29743 + err = au_dbrange_test(dentry);
29744 + if (unlikely(err))
29745 + aufs_read_unlock(dentry, flags);
29752 +void aufs_read_unlock(struct dentry *dentry, int flags)
29754 + if (au_ftest_lock(flags, DW))
29755 + di_write_unlock(dentry);
29757 + di_read_unlock(dentry, flags);
29758 + si_read_unlock(dentry->d_sb);
29761 +void aufs_write_lock(struct dentry *dentry)
29763 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
29764 + di_write_lock_child(dentry);
29767 +void aufs_write_unlock(struct dentry *dentry)
29769 + di_write_unlock(dentry);
29770 + si_write_unlock(dentry->d_sb);
29773 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
29776 + unsigned int sigen;
29777 + struct super_block *sb;
29780 + err = si_read_lock(sb, flags);
29781 + if (unlikely(err))
29784 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
29786 + if (au_ftest_lock(flags, GEN)) {
29787 + sigen = au_sigen(sb);
29788 + err = au_digen_test(d1, sigen);
29789 + AuDebugOn(!err && au_dbrange_test(d1));
29791 + err = au_digen_test(d2, sigen);
29792 + AuDebugOn(!err && au_dbrange_test(d2));
29794 + if (unlikely(err))
29795 + aufs_read_and_write_unlock2(d1, d2);
29802 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
29804 + di_write_unlock2(d1, d2);
29805 + si_read_unlock(d1->d_sb);
29807 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
29808 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
29809 +++ linux/fs/aufs/super.c 2022-05-23 09:44:37.729839166 +0200
29811 +// SPDX-License-Identifier: GPL-2.0
29813 + * Copyright (C) 2005-2021 Junjiro R. Okajima
29815 + * This program, aufs is free software; you can redistribute it and/or modify
29816 + * it under the terms of the GNU General Public License as published by
29817 + * the Free Software Foundation; either version 2 of the License, or
29818 + * (at your option) any later version.
29820 + * This program is distributed in the hope that it will be useful,
29821 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29822 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29823 + * GNU General Public License for more details.
29825 + * You should have received a copy of the GNU General Public License
29826 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29830 + * mount and super_block operations
29833 +#include <linux/iversion.h>
29834 +#include <linux/mm.h>
29835 +#include <linux/seq_file.h>
29836 +#include <linux/statfs.h>
29837 +#include <linux/vmalloc.h>
29841 + * super_operations
29843 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
29845 + struct au_icntnr *c;
29847 + c = au_cache_alloc_icntnr(sb);
29849 + au_icntnr_init(c);
29850 + inode_set_iversion(&c->vfs_inode, 1); /* sigen(sb); */
29851 + c->iinfo.ii_hinode = NULL;
29852 + return &c->vfs_inode;
29857 +static void aufs_destroy_inode(struct inode *inode)
29859 + if (!au_is_bad_inode(inode))
29860 + au_iinfo_fin(inode);
29863 +static void aufs_free_inode(struct inode *inode)
29865 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
29868 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
29870 + struct inode *inode;
29873 + inode = iget_locked(sb, ino);
29874 + if (unlikely(!inode)) {
29875 + inode = ERR_PTR(-ENOMEM);
29878 + if (!(inode->i_state & I_NEW))
29881 + err = au_xigen_new(inode);
29883 + err = au_iinfo_init(inode);
29885 + inode_inc_iversion(inode);
29887 + iget_failed(inode);
29888 + inode = ERR_PTR(err);
29892 + /* never return NULL */
29893 + AuDebugOn(!inode);
29894 + AuTraceErrPtr(inode);
29898 +/* lock free root dinfo */
29899 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
29902 + aufs_bindex_t bindex, bbot;
29903 + struct path path;
29904 + struct au_hdentry *hdp;
29905 + struct au_branch *br;
29906 + au_br_perm_str_t perm;
29909 + bbot = au_sbbot(sb);
29911 + hdp = au_hdentry(au_di(sb->s_root), bindex);
29912 + for (; !err && bindex <= bbot; bindex++, hdp++) {
29913 + br = au_sbr(sb, bindex);
29914 + path.mnt = au_br_mnt(br);
29915 + path.dentry = hdp->hd_dentry;
29916 + err = au_seq_path(seq, &path);
29918 + au_optstr_br_perm(&perm, br->br_perm);
29919 + seq_printf(seq, "=%s", perm.a);
29920 + if (bindex != bbot)
29921 + seq_putc(seq, ':');
29924 + if (unlikely(err || seq_has_overflowed(seq)))
29930 +static void au_gen_fmt(char *fmt, int len __maybe_unused, const char *pat,
29931 + const char *append)
29936 + while (*pat != ':')
29939 + strcpy(p, append);
29940 + AuDebugOn(strlen(fmt) >= len);
29943 +static void au_show_wbr_create(struct seq_file *m, int v,
29944 + struct au_sbinfo *sbinfo)
29948 + struct au_wbr_mfs *mfs;
29950 + AuRwMustAnyLock(&sbinfo->si_rwsem);
29952 + seq_puts(m, ",create=");
29953 + pat = au_optstr_wbr_create(v);
29954 + mfs = &sbinfo->si_wbr_mfs;
29956 + case AuWbrCreate_TDP:
29957 + case AuWbrCreate_RR:
29958 + case AuWbrCreate_MFS:
29959 + case AuWbrCreate_PMFS:
29960 + seq_puts(m, pat);
29962 + case AuWbrCreate_MFSRR:
29963 + case AuWbrCreate_TDMFS:
29964 + case AuWbrCreate_PMFSRR:
29965 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu");
29966 + seq_printf(m, fmt, mfs->mfsrr_watermark);
29968 + case AuWbrCreate_MFSV:
29969 + case AuWbrCreate_PMFSV:
29970 + au_gen_fmt(fmt, sizeof(fmt), pat, "%lu");
29971 + seq_printf(m, fmt,
29972 + jiffies_to_msecs(mfs->mfs_expire)
29975 + case AuWbrCreate_MFSRRV:
29976 + case AuWbrCreate_TDMFSV:
29977 + case AuWbrCreate_PMFSRRV:
29978 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu:%lu");
29979 + seq_printf(m, fmt, mfs->mfsrr_watermark,
29980 + jiffies_to_msecs(mfs->mfs_expire) / MSEC_PER_SEC);
29987 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
29989 +#ifdef CONFIG_SYSFS
29993 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
29994 + aufs_bindex_t bindex, brid;
29995 + struct qstr *name;
29997 + struct dentry *d, *h_root;
29998 + struct au_branch *br;
30000 + AuRwMustAnyLock(&sbinfo->si_rwsem);
30003 + f = au_sbi(sb)->si_xib;
30007 + /* stop printing the default xino path on the first writable branch */
30009 + bindex = au_xi_root(sb, f->f_path.dentry);
30010 + if (bindex >= 0) {
30011 + br = au_sbr_sb(sb, bindex);
30012 + h_root = au_br_dentry(br);
30015 + d = f->f_path.dentry;
30016 + name = &d->d_name;
30017 + /* safe ->d_parent because the file is unlinked */
30018 + if (d->d_parent == h_root
30019 + && name->len == len
30020 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
30023 + seq_puts(seq, ",xino=");
30024 + err = au_xino_path(seq, f);
30031 +/* seq_file will re-call me in case of too long string */
30032 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
30035 + unsigned int mnt_flags, v;
30036 + struct super_block *sb;
30037 + struct au_sbinfo *sbinfo;
30039 +#define AuBool(name, str) do { \
30040 + v = au_opt_test(mnt_flags, name); \
30041 + if (v != au_opt_test(AuOpt_Def, name)) \
30042 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
30045 +#define AuStr(name, str) do { \
30046 + v = mnt_flags & AuOptMask_##name; \
30047 + if (v != (AuOpt_Def & AuOptMask_##name)) \
30048 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
30051 +#define AuUInt(name, str, val) do { \
30052 + if (val != AUFS_##name##_DEF) \
30053 + seq_printf(m, "," #str "=%u", val); \
30056 + sb = dentry->d_sb;
30057 + if (sb->s_flags & SB_POSIXACL)
30058 + seq_puts(m, ",acl");
30059 +#if 0 /* reserved for future use */
30060 + if (sb->s_flags & SB_I_VERSION)
30061 + seq_puts(m, ",i_version");
30064 + /* lock free root dinfo */
30065 + si_noflush_read_lock(sb);
30066 + sbinfo = au_sbi(sb);
30067 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
30069 + mnt_flags = au_mntflags(sb);
30070 + if (au_opt_test(mnt_flags, XINO)) {
30071 + err = au_show_xino(m, sb);
30072 + if (unlikely(err))
30075 + seq_puts(m, ",noxino");
30077 + AuBool(TRUNC_XINO, trunc_xino);
30078 + AuStr(UDBA, udba);
30079 + AuBool(SHWH, shwh);
30080 + AuBool(PLINK, plink);
30081 + AuBool(DIO, dio);
30082 + AuBool(DIRPERM1, dirperm1);
30084 + v = sbinfo->si_wbr_create;
30085 + if (v != AuWbrCreate_Def)
30086 + au_show_wbr_create(m, v, sbinfo);
30088 + v = sbinfo->si_wbr_copyup;
30089 + if (v != AuWbrCopyup_Def)
30090 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
30092 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
30093 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
30094 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
30096 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
30098 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
30099 + AuUInt(RDCACHE, rdcache, v);
30101 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
30102 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
30104 + au_fhsm_show(m, sbinfo);
30106 + AuBool(DIRREN, dirren);
30107 + AuBool(SUM, sum);
30108 + /* AuBool(SUM_W, wsum); */
30109 + AuBool(WARN_PERM, warn_perm);
30110 + AuBool(VERBOSE, verbose);
30113 + /* be sure to print "br:" last */
30114 + if (!sysaufs_brs) {
30115 + seq_puts(m, ",br:");
30116 + au_show_brs(m, sb);
30118 + si_read_unlock(sb);
30126 +/* ---------------------------------------------------------------------- */
30128 +/* sum mode which returns the summation for statfs(2) */
30130 +static u64 au_add_till_max(u64 a, u64 b)
30138 + return ULLONG_MAX;
30141 +static u64 au_mul_till_max(u64 a, long mul)
30149 + return ULLONG_MAX;
30152 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
30155 + long bsize, factor;
30156 + u64 blocks, bfree, bavail, files, ffree;
30157 + aufs_bindex_t bbot, bindex, i;
30158 + unsigned char shared;
30159 + struct path h_path;
30160 + struct super_block *h_sb;
30163 + bsize = LONG_MAX;
30169 + bbot = au_sbbot(sb);
30170 + for (bindex = 0; bindex <= bbot; bindex++) {
30171 + h_path.mnt = au_sbr_mnt(sb, bindex);
30172 + h_sb = h_path.mnt->mnt_sb;
30174 + for (i = 0; !shared && i < bindex; i++)
30175 + shared = (au_sbr_sb(sb, i) == h_sb);
30179 + /* sb->s_root for NFS is unreliable */
30180 + h_path.dentry = h_path.mnt->mnt_root;
30181 + err = vfs_statfs(&h_path, buf);
30182 + if (unlikely(err))
30185 + if (bsize > buf->f_bsize) {
30187 + * we will reduce bsize, so we have to expand blocks
30188 + * etc. to match them again
30190 + factor = (bsize / buf->f_bsize);
30191 + blocks = au_mul_till_max(blocks, factor);
30192 + bfree = au_mul_till_max(bfree, factor);
30193 + bavail = au_mul_till_max(bavail, factor);
30194 + bsize = buf->f_bsize;
30197 + factor = (buf->f_bsize / bsize);
30198 + blocks = au_add_till_max(blocks,
30199 + au_mul_till_max(buf->f_blocks, factor));
30200 + bfree = au_add_till_max(bfree,
30201 + au_mul_till_max(buf->f_bfree, factor));
30202 + bavail = au_add_till_max(bavail,
30203 + au_mul_till_max(buf->f_bavail, factor));
30204 + files = au_add_till_max(files, buf->f_files);
30205 + ffree = au_add_till_max(ffree, buf->f_ffree);
30208 + buf->f_bsize = bsize;
30209 + buf->f_blocks = blocks;
30210 + buf->f_bfree = bfree;
30211 + buf->f_bavail = bavail;
30212 + buf->f_files = files;
30213 + buf->f_ffree = ffree;
30214 + buf->f_frsize = 0;
30220 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
30223 + struct path h_path;
30224 + struct super_block *sb;
30226 + /* lock free root dinfo */
30227 + sb = dentry->d_sb;
30228 + si_noflush_read_lock(sb);
30229 + if (!au_opt_test(au_mntflags(sb), SUM)) {
30230 + /* sb->s_root for NFS is unreliable */
30231 + h_path.mnt = au_sbr_mnt(sb, 0);
30232 + h_path.dentry = h_path.mnt->mnt_root;
30233 + err = vfs_statfs(&h_path, buf);
30235 + err = au_statfs_sum(sb, buf);
30236 + si_read_unlock(sb);
30239 + buf->f_type = AUFS_SUPER_MAGIC;
30240 + buf->f_namelen = AUFS_MAX_NAMELEN;
30241 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
30243 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
30248 +/* ---------------------------------------------------------------------- */
30250 +static int aufs_sync_fs(struct super_block *sb, int wait)
30253 + aufs_bindex_t bbot, bindex;
30254 + struct au_branch *br;
30255 + struct super_block *h_sb;
30258 + si_noflush_read_lock(sb);
30259 + bbot = au_sbbot(sb);
30260 + for (bindex = 0; bindex <= bbot; bindex++) {
30261 + br = au_sbr(sb, bindex);
30262 + if (!au_br_writable(br->br_perm))
30265 + h_sb = au_sbr_sb(sb, bindex);
30266 + e = vfsub_sync_filesystem(h_sb);
30267 + if (unlikely(e && !err))
30269 + /* go on even if an error happens */
30271 + si_read_unlock(sb);
30276 +/* ---------------------------------------------------------------------- */
30278 +/* final actions when unmounting a file system */
30279 +static void aufs_put_super(struct super_block *sb)
30281 + struct au_sbinfo *sbinfo;
30283 + sbinfo = au_sbi(sb);
30285 + kobject_put(&sbinfo->si_kobj);
30288 +/* ---------------------------------------------------------------------- */
30290 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30291 + struct super_block *sb, void *arg)
30294 + unsigned long long n, sz;
30301 + if (*hint > ULLONG_MAX / sizeof(array)) {
30302 + array = ERR_PTR(-EMFILE);
30303 + pr_err("hint %llu\n", *hint);
30307 + sz = sizeof(array) * *hint;
30308 + array = kzalloc(sz, GFP_NOFS);
30309 + if (unlikely(!array))
30310 + array = vzalloc(sz);
30311 + if (unlikely(!array)) {
30312 + array = ERR_PTR(-ENOMEM);
30316 + n = cb(sb, array, *hint, arg);
30317 + AuDebugOn(n > *hint);
30324 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
30325 + unsigned long long max __maybe_unused,
30328 + unsigned long long n;
30329 + struct inode **p, *inode;
30330 + struct list_head *head;
30335 + spin_lock(&sb->s_inode_list_lock);
30336 + list_for_each_entry(inode, head, i_sb_list) {
30337 + if (!au_is_bad_inode(inode)
30338 + && au_ii(inode)->ii_btop >= 0) {
30339 + spin_lock(&inode->i_lock);
30340 + if (atomic_read(&inode->i_count)) {
30344 + AuDebugOn(n > max);
30346 + spin_unlock(&inode->i_lock);
30349 + spin_unlock(&sb->s_inode_list_lock);
30354 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
30356 + struct au_sbinfo *sbi;
30358 + sbi = au_sbi(sb);
30359 + *max = au_lcnt_read(&sbi->si_ninodes, /*do_rev*/1);
30360 + return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
30363 +void au_iarray_free(struct inode **a, unsigned long long max)
30365 + unsigned long long ull;
30367 + for (ull = 0; ull < max; ull++)
30372 +/* ---------------------------------------------------------------------- */
30375 + * refresh dentry and inode at remount time.
30377 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
30378 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
30379 + struct dentry *parent)
30383 + di_write_lock_child(dentry);
30384 + di_read_lock_parent(parent, AuLock_IR);
30385 + err = au_refresh_dentry(dentry, parent);
30386 + if (!err && dir_flags)
30387 + au_hn_reset(d_inode(dentry), dir_flags);
30388 + di_read_unlock(parent, AuLock_IR);
30389 + di_write_unlock(dentry);
30394 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
30395 + struct au_sbinfo *sbinfo,
30396 + const unsigned int dir_flags, unsigned int do_idop)
30399 + struct dentry *parent;
30402 + parent = dget_parent(dentry);
30403 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
30404 + if (d_really_is_positive(dentry)) {
30405 + if (!d_is_dir(dentry))
30406 + err = au_do_refresh(dentry, /*dir_flags*/0,
30409 + err = au_do_refresh(dentry, dir_flags, parent);
30410 + if (unlikely(err))
30411 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
30414 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
30415 + AuDbgDentry(dentry);
30421 + au_refresh_dop(dentry, /*force_reval*/0);
30423 + au_refresh_dop(dentry, /*force_reval*/1);
30429 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
30431 + int err, i, j, ndentry, e;
30432 + unsigned int sigen;
30433 + struct au_dcsub_pages dpages;
30434 + struct au_dpage *dpage;
30435 + struct dentry **dentries, *d;
30436 + struct au_sbinfo *sbinfo;
30437 + struct dentry *root = sb->s_root;
30438 + const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
30441 + au_refresh_dop(root, /*force_reval*/0);
30443 + err = au_dpages_init(&dpages, GFP_NOFS);
30444 + if (unlikely(err))
30446 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
30447 + if (unlikely(err))
30450 + sigen = au_sigen(sb);
30451 + sbinfo = au_sbi(sb);
30452 + for (i = 0; i < dpages.ndpage; i++) {
30453 + dpage = dpages.dpages + i;
30454 + dentries = dpage->dentries;
30455 + ndentry = dpage->ndentry;
30456 + for (j = 0; j < ndentry; j++) {
30458 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
30460 + if (unlikely(e && !err))
30462 + /* go on even err */
30467 + au_dpages_free(&dpages);
30472 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
30475 + unsigned int sigen;
30476 + unsigned long long max, ull;
30477 + struct inode *inode, **array;
30479 + array = au_iarray_alloc(sb, &max);
30480 + err = PTR_ERR(array);
30481 + if (IS_ERR(array))
30485 + sigen = au_sigen(sb);
30486 + for (ull = 0; ull < max; ull++) {
30487 + inode = array[ull];
30488 + if (unlikely(!inode))
30492 + ii_write_lock_child(inode);
30493 + if (au_iigen(inode, NULL) != sigen) {
30494 + e = au_refresh_hinode_self(inode);
30495 + if (unlikely(e)) {
30496 + au_refresh_iop(inode, /*force_getattr*/1);
30497 + pr_err("error %d, i%lu\n", e, inode->i_ino);
30500 + /* go on even if err */
30503 + if (!e && do_idop)
30504 + au_refresh_iop(inode, /*force_getattr*/0);
30505 + ii_write_unlock(inode);
30508 + au_iarray_free(array, max);
30514 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
30517 + unsigned int udba;
30518 + aufs_bindex_t bindex, bbot;
30519 + struct dentry *root;
30520 + struct inode *inode;
30521 + struct au_branch *br;
30522 + struct au_sbinfo *sbi;
30524 + au_sigen_inc(sb);
30525 + sbi = au_sbi(sb);
30526 + au_fclr_si(sbi, FAILED_REFRESH_DIR);
30528 + root = sb->s_root;
30529 + DiMustNoWaiters(root);
30530 + inode = d_inode(root);
30531 + IiMustNoWaiters(inode);
30533 + udba = au_opt_udba(sb);
30534 + bbot = au_sbbot(sb);
30535 + for (bindex = 0; bindex <= bbot; bindex++) {
30536 + br = au_sbr(sb, bindex);
30537 + err = au_hnotify_reset_br(udba, br, br->br_perm);
30538 + if (unlikely(err))
30539 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
30541 + /* go on even if err */
30543 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
30546 + if (au_ftest_si(sbi, NO_DREVAL)) {
30547 + AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
30548 + sb->s_d_op = &aufs_dop_noreval;
30549 + AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
30550 + sbi->si_iop_array = aufs_iop_nogetattr;
30552 + AuDebugOn(sb->s_d_op == &aufs_dop);
30553 + sb->s_d_op = &aufs_dop;
30554 + AuDebugOn(sbi->si_iop_array == aufs_iop);
30555 + sbi->si_iop_array = aufs_iop;
30557 + pr_info("reset to %ps and %ps\n",
30558 + sb->s_d_op, sbi->si_iop_array);
30561 + di_write_unlock(root);
30562 + err = au_refresh_d(sb, do_idop);
30563 + e = au_refresh_i(sb, do_idop);
30564 + if (unlikely(e && !err))
30566 + /* aufs_write_lock() calls ..._child() */
30567 + di_write_lock_child(root);
30569 + au_cpup_attr_all(inode, /*force*/1);
30571 + if (unlikely(err))
30572 + AuIOErr("refresh failed, ignored, %d\n", err);
30575 +const struct super_operations aufs_sop = {
30576 + .alloc_inode = aufs_alloc_inode,
30577 + .destroy_inode = aufs_destroy_inode,
30578 + .free_inode = aufs_free_inode,
30579 + /* always deleting, no clearing */
30580 + .drop_inode = generic_delete_inode,
30581 + .show_options = aufs_show_options,
30582 + .statfs = aufs_statfs,
30583 + .put_super = aufs_put_super,
30584 + .sync_fs = aufs_sync_fs
30587 +/* ---------------------------------------------------------------------- */
30589 +int au_alloc_root(struct super_block *sb)
30592 + struct inode *inode;
30593 + struct dentry *root;
30596 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
30597 + err = PTR_ERR(inode);
30598 + if (IS_ERR(inode))
30601 + inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
30602 + inode->i_fop = &aufs_dir_fop;
30603 + inode->i_mode = S_IFDIR;
30604 + set_nlink(inode, 2);
30605 + unlock_new_inode(inode);
30607 + root = d_make_root(inode);
30608 + if (unlikely(!root))
30610 + err = PTR_ERR(root);
30611 + if (IS_ERR(root))
30614 + err = au_di_init(root);
30616 + sb->s_root = root;
30617 + return 0; /* success */
30625 +/* ---------------------------------------------------------------------- */
30627 +static void aufs_kill_sb(struct super_block *sb)
30629 + struct au_sbinfo *sbinfo;
30630 + struct dentry *root;
30632 + sbinfo = au_sbi(sb);
30636 + au_sbilist_del(sb);
30638 + root = sb->s_root;
30640 + aufs_write_lock(root);
30642 + __si_write_lock(sb);
30645 + if (sbinfo->si_wbr_create_ops->fin)
30646 + sbinfo->si_wbr_create_ops->fin(sb);
30647 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
30648 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
30649 + au_remount_refresh(sb, /*do_idop*/0);
30651 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
30652 + au_plink_put(sb, /*verbose*/1);
30655 + au_dr_opt_flush(sb);
30658 + aufs_write_unlock(root);
30660 + __si_write_unlock(sb);
30662 + sbinfo->si_sb = NULL;
30663 + au_nwt_flush(&sbinfo->si_nowait);
30666 + kill_anon_super(sb);
30669 +struct file_system_type aufs_fs_type = {
30670 + .name = AUFS_FSTYPE,
30671 + /* a race between rename and others */
30672 + .fs_flags = FS_RENAME_DOES_D_MOVE
30674 + /*| FS_ALLOW_IDMAP*/
30676 + .init_fs_context = aufs_fsctx_init,
30677 + .parameters = aufs_fsctx_paramspec,
30678 + .kill_sb = aufs_kill_sb,
30679 + /* no need to __module_get() and module_put(). */
30680 + .owner = THIS_MODULE,
30682 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
30683 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
30684 +++ linux/fs/aufs/super.h 2022-03-21 14:49:05.729966343 +0100
30686 +/* SPDX-License-Identifier: GPL-2.0 */
30688 + * Copyright (C) 2005-2021 Junjiro R. Okajima
30690 + * This program, aufs is free software; you can redistribute it and/or modify
30691 + * it under the terms of the GNU General Public License as published by
30692 + * the Free Software Foundation; either version 2 of the License, or
30693 + * (at your option) any later version.
30695 + * This program is distributed in the hope that it will be useful,
30696 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30697 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30698 + * GNU General Public License for more details.
30700 + * You should have received a copy of the GNU General Public License
30701 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30705 + * super_block operations
30708 +#ifndef __AUFS_SUPER_H__
30709 +#define __AUFS_SUPER_H__
30713 +#include <linux/fs.h>
30714 +#include <linux/kobject.h>
30717 +#include "rwsem.h"
30720 +/* policies to select one among multiple writable branches */
30721 +struct au_wbr_copyup_operations {
30722 + int (*copyup)(struct dentry *dentry);
30725 +#define AuWbr_DIR 1 /* target is a dir */
30726 +#define AuWbr_PARENT (1 << 1) /* always require a parent */
30728 +#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
30729 +#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
30730 +#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
30732 +struct au_wbr_create_operations {
30733 + int (*create)(struct dentry *dentry, unsigned int flags);
30734 + int (*init)(struct super_block *sb);
30735 + int (*fin)(struct super_block *sb);
30738 +struct au_wbr_mfs {
30739 + struct mutex mfs_lock; /* protect this structure */
30740 + unsigned long mfs_jiffy;
30741 + unsigned long mfs_expire;
30742 + aufs_bindex_t mfs_bindex;
30744 + unsigned long long mfsrr_bytes;
30745 + unsigned long long mfsrr_watermark;
30748 +#define AuPlink_NHASH 100
30749 +static inline int au_plink_hash(ino_t ino)
30751 + return ino % AuPlink_NHASH;
30754 +/* File-based Hierarchical Storage Management */
30756 +#ifdef CONFIG_AUFS_FHSM
30757 + /* allow only one process who can receive the notification */
30758 + spinlock_t fhsm_spin;
30760 + wait_queue_head_t fhsm_wqh;
30761 + atomic_t fhsm_readable;
30763 + /* these are protected by si_rwsem */
30764 + unsigned long fhsm_expire;
30765 + aufs_bindex_t fhsm_bottom;
30770 +struct au_sbinfo {
30771 + /* nowait tasks in the system-wide workqueue */
30772 + struct au_nowait_tasks si_nowait;
30775 + * tried sb->s_umount, but failed due to the dependency between i_mutex.
30776 + * rwsem for au_sbinfo is necessary.
30778 + struct au_rwsem si_rwsem;
30781 + * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
30784 + au_lcnt_t si_ninodes, si_nfiles;
30786 + /* branch management */
30787 + unsigned int si_generation;
30789 + /* see AuSi_ flags */
30790 + unsigned char au_si_status;
30792 + aufs_bindex_t si_bbot;
30794 + /* dirty trick to keep br_id plus */
30795 + unsigned int si_last_br_id :
30796 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
30797 + struct au_branch **si_branch;
30799 + /* policy to select a writable branch */
30800 + unsigned char si_wbr_copyup;
30801 + unsigned char si_wbr_create;
30802 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
30803 + struct au_wbr_create_operations *si_wbr_create_ops;
30805 + /* round robin */
30806 + atomic_t si_wbr_rr_next;
30808 + /* most free space */
30809 + struct au_wbr_mfs si_wbr_mfs;
30811 + /* File-based Hierarchical Storage Management */
30812 + struct au_fhsm si_fhsm;
30814 + /* mount flags */
30815 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
30816 + unsigned int si_mntflags;
30818 + /* external inode number (bitmap and translation table) */
30819 + loff_t si_ximaxent; /* max entries in a xino */
30821 + struct file *si_xib;
30822 + struct mutex si_xib_mtx; /* protect xib members */
30823 + unsigned long *si_xib_buf;
30824 + unsigned long si_xib_last_pindex;
30825 + int si_xib_next_bit;
30827 + unsigned long si_xino_jiffy;
30828 + unsigned long si_xino_expire;
30829 + /* reserved for future use */
30830 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
30832 +#ifdef CONFIG_AUFS_EXPORT
30833 + /* i_generation */
30834 + /* todo: make xigen file an array to support many inode numbers */
30835 + struct file *si_xigen;
30836 + atomic_t si_xigen_next;
30839 + /* dirty trick to support atomic_open */
30840 + struct hlist_bl_head si_aopen;
30842 + /* vdir parameters */
30843 + unsigned long si_rdcache; /* max cache time in jiffies */
30844 + unsigned int si_rdblk; /* deblk size */
30845 + unsigned int si_rdhash; /* hash size */
30848 + * If the number of whiteouts are larger than si_dirwh, leave all of
30849 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
30850 + * future fsck.aufs or kernel thread will remove them later.
30851 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
30853 + unsigned int si_dirwh;
30855 + /* pseudo_link list */
30856 + struct hlist_bl_head si_plink[AuPlink_NHASH];
30857 + wait_queue_head_t si_plink_wq;
30858 + spinlock_t si_plink_maint_lock;
30859 + pid_t si_plink_maint_pid;
30862 + struct hlist_bl_head si_files;
30864 + /* with/without getattr, brother of sb->s_d_op */
30865 + const struct inode_operations *si_iop_array;
30868 + * sysfs and lifetime management.
30869 + * this is not a small structure and it may be a waste of memory in case
30870 + * of sysfs is disabled, particularly when many aufs-es are mounted.
30871 + * but using sysfs is majority.
30873 + struct kobject si_kobj;
30874 +#ifdef CONFIG_DEBUG_FS
30875 + struct dentry *si_dbgaufs;
30876 + struct dentry *si_dbgaufs_plink;
30877 + struct dentry *si_dbgaufs_xib;
30878 +#ifdef CONFIG_AUFS_EXPORT
30879 + struct dentry *si_dbgaufs_xigen;
30883 +#ifdef CONFIG_AUFS_SBILIST
30884 + struct hlist_bl_node si_list;
30887 + /* dirty, necessary for unmounting, sysfs and sysrq */
30888 + struct super_block *si_sb;
30891 +/* sbinfo status flags */
30893 + * set true when refresh_dirs() failed at remount time.
30894 + * then try refreshing dirs at access time again.
30895 + * if it is false, refreshing dirs at access time is unnecessary
30897 +#define AuSi_FAILED_REFRESH_DIR 1
30898 +#define AuSi_FHSM (1 << 1) /* fhsm is active now */
30899 +#define AuSi_NO_DREVAL (1 << 2) /* disable all d_revalidate */
30901 +#ifndef CONFIG_AUFS_FHSM
30903 +#define AuSi_FHSM 0
30906 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
30907 + unsigned int flag)
30909 + AuRwMustAnyLock(&sbi->si_rwsem);
30910 + return sbi->au_si_status & flag;
30912 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
30913 +#define au_fset_si(sbinfo, name) do { \
30914 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30915 + (sbinfo)->au_si_status |= AuSi_##name; \
30917 +#define au_fclr_si(sbinfo, name) do { \
30918 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30919 + (sbinfo)->au_si_status &= ~AuSi_##name; \
30922 +/* ---------------------------------------------------------------------- */
30924 +/* policy to select one among writable branches */
30925 +#define AuWbrCopyup(sbinfo, ...) \
30926 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
30927 +#define AuWbrCreate(sbinfo, ...) \
30928 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
30930 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
30931 +#define AuLock_DW 1 /* write-lock dentry */
30932 +#define AuLock_IR (1 << 1) /* read-lock inode */
30933 +#define AuLock_IW (1 << 2) /* write-lock inode */
30934 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
30935 +#define AuLock_DIRS (1 << 4) /* target is a pair of dirs */
30936 + /* except RENAME_EXCHANGE */
30937 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
30938 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
30939 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
30940 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
30941 +#define au_fset_lock(flags, name) \
30942 + do { (flags) |= AuLock_##name; } while (0)
30943 +#define au_fclr_lock(flags, name) \
30944 + do { (flags) &= ~AuLock_##name; } while (0)
30946 +/* ---------------------------------------------------------------------- */
30949 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
30951 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
30952 + unsigned long long max, void *arg);
30953 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30954 + struct super_block *sb, void *arg);
30955 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
30956 +void au_iarray_free(struct inode **a, unsigned long long max);
30958 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop);
30959 +extern const struct super_operations aufs_sop;
30960 +int au_alloc_root(struct super_block *sb);
30961 +extern struct file_system_type aufs_fs_type;
30964 +void au_si_free(struct kobject *kobj);
30965 +struct au_sbinfo *au_si_alloc(struct super_block *sb);
30966 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink);
30968 +unsigned int au_sigen_inc(struct super_block *sb);
30969 +aufs_bindex_t au_new_br_id(struct super_block *sb);
30971 +int si_read_lock(struct super_block *sb, int flags);
30972 +int si_write_lock(struct super_block *sb, int flags);
30973 +int aufs_read_lock(struct dentry *dentry, int flags);
30974 +void aufs_read_unlock(struct dentry *dentry, int flags);
30975 +void aufs_write_lock(struct dentry *dentry);
30976 +void aufs_write_unlock(struct dentry *dentry);
30977 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
30978 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
30980 +/* wbr_policy.c */
30981 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
30982 +extern struct au_wbr_create_operations au_wbr_create_ops[];
30983 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
30984 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
30985 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop);
30988 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
30990 +#ifdef CONFIG_AUFS_FHSM
30993 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
30997 + spin_lock(&fhsm->fhsm_spin);
30998 + pid = fhsm->fhsm_pid;
30999 + spin_unlock(&fhsm->fhsm_spin);
31004 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
31005 +void au_fhsm_wrote_all(struct super_block *sb, int force);
31006 +int au_fhsm_fd(struct super_block *sb, int oflags);
31007 +int au_fhsm_br_alloc(struct au_branch *br);
31008 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
31009 +void au_fhsm_fin(struct super_block *sb);
31010 +void au_fhsm_init(struct au_sbinfo *sbinfo);
31011 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
31012 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
31014 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
31016 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
31017 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
31018 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
31019 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
31020 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
31021 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
31022 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
31023 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
31024 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
31027 +/* ---------------------------------------------------------------------- */
31029 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
31031 + return sb->s_fs_info;
31034 +/* ---------------------------------------------------------------------- */
31036 +#ifdef CONFIG_AUFS_EXPORT
31037 +int au_test_nfsd(void);
31038 +void au_export_init(struct super_block *sb);
31039 +void au_xigen_inc(struct inode *inode);
31040 +int au_xigen_new(struct inode *inode);
31041 +int au_xigen_set(struct super_block *sb, struct path *path);
31042 +void au_xigen_clr(struct super_block *sb);
31044 +static inline int au_busy_or_stale(void)
31046 + if (!au_test_nfsd())
31051 +AuStubInt0(au_test_nfsd, void)
31052 +AuStubVoid(au_export_init, struct super_block *sb)
31053 +AuStubVoid(au_xigen_inc, struct inode *inode)
31054 +AuStubInt0(au_xigen_new, struct inode *inode)
31055 +AuStubInt0(au_xigen_set, struct super_block *sb, struct path *path)
31056 +AuStubVoid(au_xigen_clr, struct super_block *sb)
31057 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
31058 +#endif /* CONFIG_AUFS_EXPORT */
31060 +/* ---------------------------------------------------------------------- */
31062 +#ifdef CONFIG_AUFS_SBILIST
31064 +extern struct hlist_bl_head au_sbilist;
31066 +static inline void au_sbilist_init(void)
31068 + INIT_HLIST_BL_HEAD(&au_sbilist);
31071 +static inline void au_sbilist_add(struct super_block *sb)
31073 + au_hbl_add(&au_sbi(sb)->si_list, &au_sbilist);
31076 +static inline void au_sbilist_del(struct super_block *sb)
31078 + au_hbl_del(&au_sbi(sb)->si_list, &au_sbilist);
31081 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
31082 +static inline void au_sbilist_lock(void)
31084 + hlist_bl_lock(&au_sbilist);
31087 +static inline void au_sbilist_unlock(void)
31089 + hlist_bl_unlock(&au_sbilist);
31091 +#define AuGFP_SBILIST GFP_ATOMIC
31093 +AuStubVoid(au_sbilist_lock, void)
31094 +AuStubVoid(au_sbilist_unlock, void)
31095 +#define AuGFP_SBILIST GFP_NOFS
31096 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
31098 +AuStubVoid(au_sbilist_init, void)
31099 +AuStubVoid(au_sbilist_add, struct super_block *sb)
31100 +AuStubVoid(au_sbilist_del, struct super_block *sb)
31101 +AuStubVoid(au_sbilist_lock, void)
31102 +AuStubVoid(au_sbilist_unlock, void)
31103 +#define AuGFP_SBILIST GFP_NOFS
31106 +/* ---------------------------------------------------------------------- */
31108 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
31111 + * This function is a dynamic '__init' function actually,
31112 + * so the tiny check for si_rwsem is unnecessary.
31114 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
31115 +#ifdef CONFIG_DEBUG_FS
31116 + sbinfo->si_dbgaufs = NULL;
31117 + sbinfo->si_dbgaufs_plink = NULL;
31118 + sbinfo->si_dbgaufs_xib = NULL;
31119 +#ifdef CONFIG_AUFS_EXPORT
31120 + sbinfo->si_dbgaufs_xigen = NULL;
31125 +/* ---------------------------------------------------------------------- */
31127 +/* current->atomic_flags */
31128 +/* this value should never corrupt the ones defined in linux/sched.h */
31129 +#define PFA_AUFS 0x10
31131 +TASK_PFA_TEST(AUFS, test_aufs) /* task_test_aufs */
31132 +TASK_PFA_SET(AUFS, aufs) /* task_set_aufs */
31133 +TASK_PFA_CLEAR(AUFS, aufs) /* task_clear_aufs */
31135 +static inline int si_pid_test(struct super_block *sb)
31137 + return !!task_test_aufs(current);
31140 +static inline void si_pid_clr(struct super_block *sb)
31142 + AuDebugOn(!task_test_aufs(current));
31143 + task_clear_aufs(current);
31146 +static inline void si_pid_set(struct super_block *sb)
31148 + AuDebugOn(task_test_aufs(current));
31149 + task_set_aufs(current);
31152 +/* ---------------------------------------------------------------------- */
31154 +/* lock superblock. mainly for entry point functions */
31155 +#define __si_read_lock(sb) au_rw_read_lock(&au_sbi(sb)->si_rwsem)
31156 +#define __si_write_lock(sb) au_rw_write_lock(&au_sbi(sb)->si_rwsem)
31157 +#define __si_read_trylock(sb) au_rw_read_trylock(&au_sbi(sb)->si_rwsem)
31158 +#define __si_write_trylock(sb) au_rw_write_trylock(&au_sbi(sb)->si_rwsem)
31160 +#define __si_read_trylock_nested(sb) \
31161 + au_rw_read_trylock_nested(&au_sbi(sb)->si_rwsem)
31162 +#define __si_write_trylock_nested(sb) \
31163 + au_rw_write_trylock_nested(&au_sbi(sb)->si_rwsem)
31166 +#define __si_read_unlock(sb) au_rw_read_unlock(&au_sbi(sb)->si_rwsem)
31167 +#define __si_write_unlock(sb) au_rw_write_unlock(&au_sbi(sb)->si_rwsem)
31168 +#define __si_downgrade_lock(sb) au_rw_dgrade_lock(&au_sbi(sb)->si_rwsem)
31170 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
31171 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
31172 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
31174 +static inline void si_noflush_read_lock(struct super_block *sb)
31176 + __si_read_lock(sb);
31180 +static inline int si_noflush_read_trylock(struct super_block *sb)
31184 + locked = __si_read_trylock(sb);
31190 +static inline void si_noflush_write_lock(struct super_block *sb)
31192 + __si_write_lock(sb);
31196 +static inline int si_noflush_write_trylock(struct super_block *sb)
31200 + locked = __si_write_trylock(sb);
31206 +#if 0 /* reserved */
31207 +static inline int si_read_trylock(struct super_block *sb, int flags)
31209 + if (au_ftest_lock(flags, FLUSH))
31210 + au_nwt_flush(&au_sbi(sb)->si_nowait);
31211 + return si_noflush_read_trylock(sb);
31215 +static inline void si_read_unlock(struct super_block *sb)
31218 + __si_read_unlock(sb);
31221 +#if 0 /* reserved */
31222 +static inline int si_write_trylock(struct super_block *sb, int flags)
31224 + if (au_ftest_lock(flags, FLUSH))
31225 + au_nwt_flush(&au_sbi(sb)->si_nowait);
31226 + return si_noflush_write_trylock(sb);
31230 +static inline void si_write_unlock(struct super_block *sb)
31233 + __si_write_unlock(sb);
31236 +#if 0 /* reserved */
31237 +static inline void si_downgrade_lock(struct super_block *sb)
31239 + __si_downgrade_lock(sb);
31243 +/* ---------------------------------------------------------------------- */
31245 +static inline aufs_bindex_t au_sbbot(struct super_block *sb)
31247 + SiMustAnyLock(sb);
31248 + return au_sbi(sb)->si_bbot;
31251 +static inline unsigned int au_mntflags(struct super_block *sb)
31253 + SiMustAnyLock(sb);
31254 + return au_sbi(sb)->si_mntflags;
31257 +static inline unsigned int au_sigen(struct super_block *sb)
31259 + SiMustAnyLock(sb);
31260 + return au_sbi(sb)->si_generation;
31263 +static inline struct au_branch *au_sbr(struct super_block *sb,
31264 + aufs_bindex_t bindex)
31266 + SiMustAnyLock(sb);
31267 + return au_sbi(sb)->si_branch[0 + bindex];
31270 +static inline loff_t au_xi_maxent(struct super_block *sb)
31272 + SiMustAnyLock(sb);
31273 + return au_sbi(sb)->si_ximaxent;
31276 +#endif /* __KERNEL__ */
31277 +#endif /* __AUFS_SUPER_H__ */
31278 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
31279 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
31280 +++ linux/fs/aufs/sysaufs.c 2022-05-23 09:44:37.729839166 +0200
31282 +// SPDX-License-Identifier: GPL-2.0
31284 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31286 + * This program, aufs is free software; you can redistribute it and/or modify
31287 + * it under the terms of the GNU General Public License as published by
31288 + * the Free Software Foundation; either version 2 of the License, or
31289 + * (at your option) any later version.
31291 + * This program is distributed in the hope that it will be useful,
31292 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31293 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31294 + * GNU General Public License for more details.
31296 + * You should have received a copy of the GNU General Public License
31297 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31301 + * sysfs interface and lifetime management
31302 + * they are necessary regardless sysfs is disabled.
31305 +#include <linux/random.h>
31308 +unsigned long sysaufs_si_mask;
31309 +struct kset *sysaufs_kset;
31311 +#define AuSiAttr(_name) { \
31312 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
31313 + .show = sysaufs_si_##_name, \
31316 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
31317 +struct attribute *sysaufs_si_attrs[] = {
31318 + &sysaufs_si_attr_xi_path.attr,
31321 +ATTRIBUTE_GROUPS(sysaufs_si);
31323 +static const struct sysfs_ops au_sbi_ops = {
31324 + .show = sysaufs_si_show
31327 +static struct kobj_type au_sbi_ktype = {
31328 + .release = au_si_free,
31329 + .sysfs_ops = &au_sbi_ops,
31330 + .default_groups = sysaufs_si_groups
31333 +/* ---------------------------------------------------------------------- */
31335 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
31339 + sbinfo->si_kobj.kset = sysaufs_kset;
31340 + /* cf. sysaufs_name() */
31341 + err = kobject_init_and_add
31342 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
31343 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
31348 +void sysaufs_fin(void)
31350 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31351 + kset_unregister(sysaufs_kset);
31354 +int __init sysaufs_init(void)
31359 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
31360 + } while (!sysaufs_si_mask);
31363 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
31364 + if (unlikely(!sysaufs_kset))
31366 + err = PTR_ERR(sysaufs_kset);
31367 + if (IS_ERR(sysaufs_kset))
31369 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31370 + if (unlikely(err))
31371 + kset_unregister(sysaufs_kset);
31376 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
31377 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
31378 +++ linux/fs/aufs/sysaufs.h 2022-03-21 14:49:05.729966343 +0100
31380 +/* SPDX-License-Identifier: GPL-2.0 */
31382 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31384 + * This program, aufs is free software; you can redistribute it and/or modify
31385 + * it under the terms of the GNU General Public License as published by
31386 + * the Free Software Foundation; either version 2 of the License, or
31387 + * (at your option) any later version.
31389 + * This program is distributed in the hope that it will be useful,
31390 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31391 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31392 + * GNU General Public License for more details.
31394 + * You should have received a copy of the GNU General Public License
31395 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31399 + * sysfs interface and mount lifetime management
31402 +#ifndef __SYSAUFS_H__
31403 +#define __SYSAUFS_H__
31407 +#include <linux/sysfs.h>
31408 +#include "module.h"
31410 +struct super_block;
31413 +struct sysaufs_si_attr {
31414 + struct attribute attr;
31415 + int (*show)(struct seq_file *seq, struct super_block *sb);
31418 +/* ---------------------------------------------------------------------- */
31421 +extern unsigned long sysaufs_si_mask;
31422 +extern struct kset *sysaufs_kset;
31423 +extern struct attribute *sysaufs_si_attrs[];
31424 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
31425 +int __init sysaufs_init(void);
31426 +void sysaufs_fin(void);
31428 +/* ---------------------------------------------------------------------- */
31430 +/* some people doesn't like to show a pointer in kernel */
31431 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
31433 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
31436 +#define SysaufsSiNamePrefix "si_"
31437 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
31438 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
31440 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
31441 + sysaufs_si_id(sbinfo));
31445 +#ifdef CONFIG_SYSFS
31447 +extern struct attribute_group *sysaufs_attr_group;
31449 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
31450 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31452 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
31453 +#ifdef CONFIG_COMPAT
31454 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
31457 +void sysaufs_br_init(struct au_branch *br);
31458 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
31459 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
31461 +#define sysaufs_brs_init() do {} while (0)
31464 +#define sysaufs_attr_group NULL
31466 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
31467 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
31468 + struct attribute *attr, char *buf)
31469 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
31470 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
31471 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
31473 +static inline void sysaufs_brs_init(void)
31478 +#endif /* CONFIG_SYSFS */
31480 +#endif /* __KERNEL__ */
31481 +#endif /* __SYSAUFS_H__ */
31482 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
31483 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
31484 +++ linux/fs/aufs/sysfs.c 2022-03-21 14:49:05.729966343 +0100
31486 +// SPDX-License-Identifier: GPL-2.0
31488 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31490 + * This program, aufs is free software; you can redistribute it and/or modify
31491 + * it under the terms of the GNU General Public License as published by
31492 + * the Free Software Foundation; either version 2 of the License, or
31493 + * (at your option) any later version.
31495 + * This program is distributed in the hope that it will be useful,
31496 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31497 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31498 + * GNU General Public License for more details.
31500 + * You should have received a copy of the GNU General Public License
31501 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31505 + * sysfs interface
31508 +#include <linux/compat.h>
31509 +#include <linux/seq_file.h>
31512 +#ifdef CONFIG_AUFS_FS_MODULE
31513 +/* this entry violates the "one line per file" policy of sysfs */
31514 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
31518 + static char *conf =
31519 +/* this file is generated at compiling */
31520 +#include "conf.str"
31523 + err = snprintf(buf, PAGE_SIZE, conf);
31524 + if (unlikely(err >= PAGE_SIZE))
31529 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
31532 +static struct attribute *au_attr[] = {
31533 +#ifdef CONFIG_AUFS_FS_MODULE
31534 + &au_config_attr.attr,
31536 + NULL, /* need to NULL terminate the list of attributes */
31539 +static struct attribute_group sysaufs_attr_group_body = {
31543 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
31545 +/* ---------------------------------------------------------------------- */
31547 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
31551 + SiMustAnyLock(sb);
31554 + if (au_opt_test(au_mntflags(sb), XINO)) {
31555 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
31556 + seq_putc(seq, '\n');
31562 + * the lifetime of branch is independent from the entry under sysfs.
31563 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
31566 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
31567 + aufs_bindex_t bindex, int idx)
31570 + struct path path;
31571 + struct dentry *root;
31572 + struct au_branch *br;
31573 + au_br_perm_str_t perm;
31575 + AuDbg("b%d\n", bindex);
31578 + root = sb->s_root;
31579 + di_read_lock_parent(root, !AuLock_IR);
31580 + br = au_sbr(sb, bindex);
31583 + case AuBrSysfs_BR:
31584 + path.mnt = au_br_mnt(br);
31585 + path.dentry = au_h_dptr(root, bindex);
31586 + err = au_seq_path(seq, &path);
31588 + au_optstr_br_perm(&perm, br->br_perm);
31589 + seq_printf(seq, "=%s\n", perm.a);
31592 + case AuBrSysfs_BRID:
31593 + seq_printf(seq, "%d\n", br->br_id);
31596 + di_read_unlock(root, !AuLock_IR);
31597 + if (unlikely(err || seq_has_overflowed(seq)))
31603 +/* ---------------------------------------------------------------------- */
31605 +static struct seq_file *au_seq(char *p, ssize_t len)
31607 + struct seq_file *seq;
31609 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
31611 + /* mutex_init(&seq.lock); */
31614 + return seq; /* success */
31617 + seq = ERR_PTR(-ENOMEM);
31621 +#define SysaufsBr_PREFIX "br"
31622 +#define SysaufsBrid_PREFIX "brid"
31624 +/* todo: file size may exceed PAGE_SIZE */
31625 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31631 + aufs_bindex_t bbot;
31632 + struct au_sbinfo *sbinfo;
31633 + struct super_block *sb;
31634 + struct seq_file *seq;
31636 + struct attribute **cattr;
31638 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
31639 + sb = sbinfo->si_sb;
31642 + * prevent a race condition between sysfs and aufs.
31643 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
31644 + * prohibits maintaining the sysfs entries.
31645 + * hew we acquire read lock after sysfs_get_active_two().
31646 + * on the other hand, the remount process may maintain the sysfs/aufs
31647 + * entries after acquiring write lock.
31648 + * it can cause a deadlock.
31649 + * simply we gave up processing read here.
31652 + if (unlikely(!si_noflush_read_trylock(sb)))
31655 + seq = au_seq(buf, PAGE_SIZE);
31656 + err = PTR_ERR(seq);
31660 + name = (void *)attr->name;
31661 + cattr = sysaufs_si_attrs;
31663 + if (!strcmp(name, (*cattr)->name)) {
31664 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
31671 + if (!strncmp(name, SysaufsBrid_PREFIX,
31672 + sizeof(SysaufsBrid_PREFIX) - 1)) {
31673 + idx = AuBrSysfs_BRID;
31674 + name += sizeof(SysaufsBrid_PREFIX) - 1;
31675 + } else if (!strncmp(name, SysaufsBr_PREFIX,
31676 + sizeof(SysaufsBr_PREFIX) - 1)) {
31677 + idx = AuBrSysfs_BR;
31678 + name += sizeof(SysaufsBr_PREFIX) - 1;
31682 + err = kstrtol(name, 10, &l);
31684 + bbot = au_sbbot(sb);
31686 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
31693 + err = seq->count;
31694 + /* sysfs limit */
31695 + if (unlikely(err == PAGE_SIZE))
31698 + au_kfree_rcu(seq);
31700 + si_read_unlock(sb);
31705 +/* ---------------------------------------------------------------------- */
31707 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
31711 + aufs_bindex_t bindex, bbot;
31714 + struct seq_file *seq;
31715 + struct au_branch *br;
31717 + si_read_lock(sb, AuLock_FLUSH);
31718 + bbot = au_sbbot(sb);
31724 + buf = (void *)__get_free_page(GFP_NOFS);
31725 + if (unlikely(!buf))
31728 + seq = au_seq(buf, PAGE_SIZE);
31729 + err = PTR_ERR(seq);
31733 + sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
31734 + for (bindex = 0; bindex <= bbot; bindex++, arg++) {
31735 + /* VERIFY_WRITE */
31736 + err = !access_ok(arg, sizeof(*arg));
31737 + if (unlikely(err))
31740 + br = au_sbr(sb, bindex);
31741 + brid = br->br_id;
31742 + BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
31743 + err = __put_user(brid, &arg->id);
31744 + if (unlikely(err))
31747 + BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
31748 + err = __put_user(br->br_perm, &arg->perm);
31749 + if (unlikely(err))
31752 + err = au_seq_path(seq, &br->br_path);
31753 + if (unlikely(err))
31755 + seq_putc(seq, '\0');
31756 + if (!seq_has_overflowed(seq)) {
31757 + err = copy_to_user(arg->path, seq->buf, seq->count);
31759 + if (unlikely(err))
31766 + if (unlikely(err))
31770 + au_kfree_rcu(seq);
31772 + free_page((unsigned long)buf);
31774 + si_read_unlock(sb);
31778 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
31780 + return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
31783 +#ifdef CONFIG_COMPAT
31784 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
31786 + return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
31790 +/* ---------------------------------------------------------------------- */
31792 +void sysaufs_br_init(struct au_branch *br)
31795 + struct au_brsysfs *br_sysfs;
31796 + struct attribute *attr;
31798 + br_sysfs = br->br_sysfs;
31799 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31800 + attr = &br_sysfs->attr;
31801 + sysfs_attr_init(attr);
31802 + attr->name = br_sysfs->name;
31803 + attr->mode = 0444;
31808 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
31810 + struct au_branch *br;
31811 + struct kobject *kobj;
31812 + struct au_brsysfs *br_sysfs;
31814 + aufs_bindex_t bbot;
31816 + if (!sysaufs_brs)
31819 + kobj = &au_sbi(sb)->si_kobj;
31820 + bbot = au_sbbot(sb);
31821 + for (; bindex <= bbot; bindex++) {
31822 + br = au_sbr(sb, bindex);
31823 + br_sysfs = br->br_sysfs;
31824 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31825 + sysfs_remove_file(kobj, &br_sysfs->attr);
31831 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
31834 + aufs_bindex_t bbot;
31835 + struct kobject *kobj;
31836 + struct au_branch *br;
31837 + struct au_brsysfs *br_sysfs;
31839 + if (!sysaufs_brs)
31842 + kobj = &au_sbi(sb)->si_kobj;
31843 + bbot = au_sbbot(sb);
31844 + for (; bindex <= bbot; bindex++) {
31845 + br = au_sbr(sb, bindex);
31846 + br_sysfs = br->br_sysfs;
31847 + snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
31848 + SysaufsBr_PREFIX "%d", bindex);
31849 + snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
31850 + SysaufsBrid_PREFIX "%d", bindex);
31851 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31852 + err = sysfs_create_file(kobj, &br_sysfs->attr);
31853 + if (unlikely(err))
31854 + pr_warn("failed %s under sysfs(%d)\n",
31855 + br_sysfs->name, err);
31860 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
31861 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
31862 +++ linux/fs/aufs/sysrq.c 2022-03-21 14:49:05.729966343 +0100
31864 +// SPDX-License-Identifier: GPL-2.0
31866 + * Copyright (C) 2005-2021 Junjiro R. Okajima
31868 + * This program, aufs is free software; you can redistribute it and/or modify
31869 + * it under the terms of the GNU General Public License as published by
31870 + * the Free Software Foundation; either version 2 of the License, or
31871 + * (at your option) any later version.
31873 + * This program is distributed in the hope that it will be useful,
31874 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31875 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31876 + * GNU General Public License for more details.
31878 + * You should have received a copy of the GNU General Public License
31879 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31883 + * magic sysrq handler
31886 +/* #include <linux/sysrq.h> */
31887 +#include <linux/writeback.h>
31890 +/* ---------------------------------------------------------------------- */
31892 +static void sysrq_sb(struct super_block *sb)
31895 + struct au_sbinfo *sbinfo;
31896 + struct file *file;
31897 + struct hlist_bl_head *files;
31898 + struct hlist_bl_node *pos;
31899 + struct au_finfo *finfo;
31902 + plevel = au_plevel;
31903 + au_plevel = KERN_WARNING;
31905 + /* since we define pr_fmt, call printk directly */
31906 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
31908 + sbinfo = au_sbi(sb);
31909 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
31910 + pr("superblock\n");
31913 +#if 0 /* reserved */
31915 + int err, i, j, ndentry;
31916 + struct au_dcsub_pages dpages;
31917 + struct au_dpage *dpage;
31919 + err = au_dpages_init(&dpages, GFP_ATOMIC);
31920 + if (unlikely(err))
31922 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
31924 + for (i = 0; i < dpages.ndpage; i++) {
31925 + dpage = dpages.dpages + i;
31926 + ndentry = dpage->ndentry;
31927 + for (j = 0; j < ndentry; j++)
31928 + au_dpri_dentry(dpage->dentries[j]);
31930 + au_dpages_free(&dpages);
31934 + pr("isolated inode\n");
31935 + spin_lock(&sb->s_inode_list_lock);
31936 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
31937 + spin_lock(&i->i_lock);
31938 + if (hlist_empty(&i->i_dentry))
31939 + au_dpri_inode(i);
31940 + spin_unlock(&i->i_lock);
31942 + spin_unlock(&sb->s_inode_list_lock);
31945 + files = &au_sbi(sb)->si_files;
31946 + hlist_bl_lock(files);
31947 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
31950 + file = finfo->fi_file;
31951 + mode = file_inode(file)->i_mode;
31952 + if (!special_file(mode))
31953 + au_dpri_file(file);
31955 + hlist_bl_unlock(files);
31959 + au_plevel = plevel;
31962 +/* ---------------------------------------------------------------------- */
31964 +/* module parameter */
31965 +static char *aufs_sysrq_key = "a";
31966 +module_param_named(sysrq, aufs_sysrq_key, charp, 0444);
31967 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
31969 +static void au_sysrq(int key __maybe_unused)
31971 + struct au_sbinfo *sbinfo;
31972 + struct hlist_bl_node *pos;
31975 + au_sbilist_lock();
31976 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
31977 + sysrq_sb(sbinfo->si_sb);
31978 + au_sbilist_unlock();
31982 +static struct sysrq_key_op au_sysrq_op = {
31983 + .handler = au_sysrq,
31984 + .help_msg = "Aufs",
31985 + .action_msg = "Aufs",
31986 + .enable_mask = SYSRQ_ENABLE_DUMP
31989 +/* ---------------------------------------------------------------------- */
31991 +int __init au_sysrq_init(void)
31997 + key = *aufs_sysrq_key;
31998 + if ('a' <= key && key <= 'z')
31999 + err = register_sysrq_key(key, &au_sysrq_op);
32000 + if (unlikely(err))
32001 + pr_err("err %d, sysrq=%c\n", err, key);
32005 +void au_sysrq_fin(void)
32009 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
32010 + if (unlikely(err))
32011 + pr_err("err %d (ignored)\n", err);
32013 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
32014 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
32015 +++ linux/fs/aufs/vdir.c 2022-03-21 14:49:05.733299677 +0100
32017 +// SPDX-License-Identifier: GPL-2.0
32019 + * Copyright (C) 2005-2021 Junjiro R. Okajima
32021 + * This program, aufs is free software; you can redistribute it and/or modify
32022 + * it under the terms of the GNU General Public License as published by
32023 + * the Free Software Foundation; either version 2 of the License, or
32024 + * (at your option) any later version.
32026 + * This program is distributed in the hope that it will be useful,
32027 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32028 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32029 + * GNU General Public License for more details.
32031 + * You should have received a copy of the GNU General Public License
32032 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32036 + * virtual or vertical directory
32039 +#include <linux/iversion.h>
32042 +static unsigned int calc_size(int nlen)
32044 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
32047 +static int set_deblk_end(union au_vdir_deblk_p *p,
32048 + union au_vdir_deblk_p *deblk_end)
32050 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
32051 + p->de->de_str.len = 0;
32055 + return -1; /* error */
32058 +/* returns true or false */
32059 +static int is_deblk_end(union au_vdir_deblk_p *p,
32060 + union au_vdir_deblk_p *deblk_end)
32062 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
32063 + return !p->de->de_str.len;
32067 +static unsigned char *last_deblk(struct au_vdir *vdir)
32069 + return vdir->vd_deblk[vdir->vd_nblk - 1];
32072 +/* ---------------------------------------------------------------------- */
32074 +/* estimate the appropriate size for name hash table */
32075 +unsigned int au_rdhash_est(loff_t sz)
32083 + if (sz < AUFS_RDHASH_DEF)
32084 + n = AUFS_RDHASH_DEF;
32085 + /* pr_info("n %u\n", n); */
32090 + * the allocated memory has to be freed by
32091 + * au_nhash_wh_free() or au_nhash_de_free().
32093 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
32095 + struct hlist_head *head;
32099 + sz = sizeof(*nhash->nh_head) * num_hash;
32100 + head = kmalloc(sz, gfp);
32102 + nhash->nh_num = num_hash;
32103 + nhash->nh_head = head;
32104 + for (u = 0; u < num_hash; u++)
32105 + INIT_HLIST_HEAD(head++);
32106 + return 0; /* success */
32112 +static void nhash_count(struct hlist_head *head)
32114 +#if 0 /* debugging */
32116 + struct hlist_node *pos;
32119 + hlist_for_each(pos, head)
32121 + pr_info("%lu\n", n);
32125 +static void au_nhash_wh_do_free(struct hlist_head *head)
32127 + struct au_vdir_wh *pos;
32128 + struct hlist_node *node;
32130 + hlist_for_each_entry_safe(pos, node, head, wh_hash)
32131 + au_kfree_rcu(pos);
32134 +static void au_nhash_de_do_free(struct hlist_head *head)
32136 + struct au_vdir_dehstr *pos;
32137 + struct hlist_node *node;
32139 + hlist_for_each_entry_safe(pos, node, head, hash)
32140 + au_cache_free_vdir_dehstr(pos);
32143 +static void au_nhash_do_free(struct au_nhash *nhash,
32144 + void (*free)(struct hlist_head *head))
32147 + struct hlist_head *head;
32149 + n = nhash->nh_num;
32153 + head = nhash->nh_head;
32154 + while (n-- > 0) {
32155 + nhash_count(head);
32158 + au_kfree_try_rcu(nhash->nh_head);
32161 +void au_nhash_wh_free(struct au_nhash *whlist)
32163 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
32166 +static void au_nhash_de_free(struct au_nhash *delist)
32168 + au_nhash_do_free(delist, au_nhash_de_do_free);
32171 +/* ---------------------------------------------------------------------- */
32173 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
32177 + unsigned int u, n;
32178 + struct hlist_head *head;
32179 + struct au_vdir_wh *pos;
32182 + n = whlist->nh_num;
32183 + head = whlist->nh_head;
32184 + for (u = 0; u < n; u++, head++)
32185 + hlist_for_each_entry(pos, head, wh_hash)
32186 + if (pos->wh_bindex == btgt && ++num > limit)
32191 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
32192 + unsigned char *name,
32193 + unsigned int len)
32196 + /* const unsigned int magic_bit = 12; */
32198 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
32205 + /* v = hash_long(v, magic_bit); */
32206 + v %= nhash->nh_num;
32207 + return nhash->nh_head + v;
32210 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
32213 + return str->len == nlen && !memcmp(str->name, name, nlen);
32216 +/* returns found or not */
32217 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
32219 + struct hlist_head *head;
32220 + struct au_vdir_wh *pos;
32221 + struct au_vdir_destr *str;
32223 + head = au_name_hash(whlist, name, nlen);
32224 + hlist_for_each_entry(pos, head, wh_hash) {
32225 + str = &pos->wh_str;
32226 + AuDbg("%.*s\n", str->len, str->name);
32227 + if (au_nhash_test_name(str, name, nlen))
32233 +/* returns found(true) or not */
32234 +static int test_known(struct au_nhash *delist, char *name, int nlen)
32236 + struct hlist_head *head;
32237 + struct au_vdir_dehstr *pos;
32238 + struct au_vdir_destr *str;
32240 + head = au_name_hash(delist, name, nlen);
32241 + hlist_for_each_entry(pos, head, hash) {
32243 + AuDbg("%.*s\n", str->len, str->name);
32244 + if (au_nhash_test_name(str, name, nlen))
32250 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
32251 + unsigned char d_type)
32253 +#ifdef CONFIG_AUFS_SHWH
32254 + wh->wh_ino = ino;
32255 + wh->wh_type = d_type;
32259 +/* ---------------------------------------------------------------------- */
32261 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
32262 + unsigned int d_type, aufs_bindex_t bindex,
32263 + unsigned char shwh)
32266 + struct au_vdir_destr *str;
32267 + struct au_vdir_wh *wh;
32269 + AuDbg("%.*s\n", nlen, name);
32270 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
32273 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
32274 + if (unlikely(!wh))
32278 + wh->wh_bindex = bindex;
32280 + au_shwh_init_wh(wh, ino, d_type);
32281 + str = &wh->wh_str;
32283 + memcpy(str->name, name, nlen);
32284 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
32291 +static int append_deblk(struct au_vdir *vdir)
32294 + unsigned long ul;
32295 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
32296 + union au_vdir_deblk_p p, deblk_end;
32297 + unsigned char **o;
32300 + o = au_krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
32301 + GFP_NOFS, /*may_shrink*/0);
32302 + if (unlikely(!o))
32305 + vdir->vd_deblk = o;
32306 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
32308 + ul = vdir->vd_nblk++;
32309 + vdir->vd_deblk[ul] = p.deblk;
32310 + vdir->vd_last.ul = ul;
32311 + vdir->vd_last.p.deblk = p.deblk;
32312 + deblk_end.deblk = p.deblk + deblk_sz;
32313 + err = set_deblk_end(&p, &deblk_end);
32320 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
32321 + unsigned int d_type, struct au_nhash *delist)
32325 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
32326 + union au_vdir_deblk_p p, *room, deblk_end;
32327 + struct au_vdir_dehstr *dehstr;
32329 + p.deblk = last_deblk(vdir);
32330 + deblk_end.deblk = p.deblk + deblk_sz;
32331 + room = &vdir->vd_last.p;
32332 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
32333 + || !is_deblk_end(room, &deblk_end));
32335 + sz = calc_size(nlen);
32336 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
32337 + err = append_deblk(vdir);
32338 + if (unlikely(err))
32341 + p.deblk = last_deblk(vdir);
32342 + deblk_end.deblk = p.deblk + deblk_sz;
32344 + AuDebugOn(room->deblk != p.deblk);
32348 + dehstr = au_cache_alloc_vdir_dehstr();
32349 + if (unlikely(!dehstr))
32352 + dehstr->str = &room->de->de_str;
32353 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
32354 + room->de->de_ino = ino;
32355 + room->de->de_type = d_type;
32356 + room->de->de_str.len = nlen;
32357 + memcpy(room->de->de_str.name, name, nlen);
32360 + room->deblk += sz;
32361 + if (unlikely(set_deblk_end(room, &deblk_end)))
32362 + err = append_deblk(vdir);
32369 +/* ---------------------------------------------------------------------- */
32371 +void au_vdir_free(struct au_vdir *vdir)
32373 + unsigned char **deblk;
32375 + deblk = vdir->vd_deblk;
32376 + while (vdir->vd_nblk--)
32377 + au_kfree_try_rcu(*deblk++);
32378 + au_kfree_try_rcu(vdir->vd_deblk);
32379 + au_cache_free_vdir(vdir);
32382 +static struct au_vdir *alloc_vdir(struct file *file)
32384 + struct au_vdir *vdir;
32385 + struct super_block *sb;
32388 + sb = file->f_path.dentry->d_sb;
32389 + SiMustAnyLock(sb);
32392 + vdir = au_cache_alloc_vdir();
32393 + if (unlikely(!vdir))
32396 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
32397 + if (unlikely(!vdir->vd_deblk))
32400 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
32401 + if (!vdir->vd_deblk_sz) {
32402 + /* estimate the appropriate size for deblk */
32403 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
32404 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
32406 + vdir->vd_nblk = 0;
32407 + vdir->vd_version = 0;
32408 + vdir->vd_jiffy = 0;
32409 + err = append_deblk(vdir);
32411 + return vdir; /* success */
32413 + au_kfree_try_rcu(vdir->vd_deblk);
32416 + au_cache_free_vdir(vdir);
32418 + vdir = ERR_PTR(err);
32422 +static int reinit_vdir(struct au_vdir *vdir)
32425 + union au_vdir_deblk_p p, deblk_end;
32427 + while (vdir->vd_nblk > 1) {
32428 + au_kfree_try_rcu(vdir->vd_deblk[vdir->vd_nblk - 1]);
32429 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
32432 + p.deblk = vdir->vd_deblk[0];
32433 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
32434 + err = set_deblk_end(&p, &deblk_end);
32435 + /* keep vd_dblk_sz */
32436 + vdir->vd_last.ul = 0;
32437 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32438 + vdir->vd_version = 0;
32439 + vdir->vd_jiffy = 0;
32444 +/* ---------------------------------------------------------------------- */
32446 +#define AuFillVdir_CALLED 1
32447 +#define AuFillVdir_WHABLE (1 << 1)
32448 +#define AuFillVdir_SHWH (1 << 2)
32449 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
32450 +#define au_fset_fillvdir(flags, name) \
32451 + do { (flags) |= AuFillVdir_##name; } while (0)
32452 +#define au_fclr_fillvdir(flags, name) \
32453 + do { (flags) &= ~AuFillVdir_##name; } while (0)
32455 +#ifndef CONFIG_AUFS_SHWH
32456 +#undef AuFillVdir_SHWH
32457 +#define AuFillVdir_SHWH 0
32460 +struct fillvdir_arg {
32461 + struct dir_context ctx;
32462 + struct file *file;
32463 + struct au_vdir *vdir;
32464 + struct au_nhash delist;
32465 + struct au_nhash whlist;
32466 + aufs_bindex_t bindex;
32467 + unsigned int flags;
32471 +static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
32472 + loff_t offset __maybe_unused, u64 h_ino,
32473 + unsigned int d_type)
32475 + struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
32476 + char *name = (void *)__name;
32477 + struct super_block *sb;
32479 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
32482 + sb = arg->file->f_path.dentry->d_sb;
32483 + au_fset_fillvdir(arg->flags, CALLED);
32485 + if (nlen <= AUFS_WH_PFX_LEN
32486 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
32487 + if (test_known(&arg->delist, name, nlen)
32488 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
32489 + goto out; /* already exists or whiteouted */
32491 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
32493 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
32494 + d_type = DT_UNKNOWN;
32495 + arg->err = append_de(arg->vdir, name, nlen, ino,
32496 + d_type, &arg->delist);
32498 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
32499 + name += AUFS_WH_PFX_LEN;
32500 + nlen -= AUFS_WH_PFX_LEN;
32501 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
32502 + goto out; /* already whiteouted */
32504 + ino = 0; /* just to suppress a warning */
32506 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
32509 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
32510 + d_type = DT_UNKNOWN;
32511 + arg->err = au_nhash_append_wh
32512 + (&arg->whlist, name, nlen, ino, d_type,
32513 + arg->bindex, shwh);
32519 + arg->vdir->vd_jiffy = jiffies;
32521 + AuTraceErr(arg->err);
32525 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
32526 + struct au_nhash *whlist, struct au_nhash *delist)
32528 +#ifdef CONFIG_AUFS_SHWH
32530 + unsigned int nh, u;
32531 + struct hlist_head *head;
32532 + struct au_vdir_wh *pos;
32533 + struct hlist_node *n;
32535 + struct au_vdir_destr *destr;
32537 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
32540 + o = p = (void *)__get_free_page(GFP_NOFS);
32541 + if (unlikely(!p))
32545 + nh = whlist->nh_num;
32546 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32547 + p += AUFS_WH_PFX_LEN;
32548 + for (u = 0; u < nh; u++) {
32549 + head = whlist->nh_head + u;
32550 + hlist_for_each_entry_safe(pos, n, head, wh_hash) {
32551 + destr = &pos->wh_str;
32552 + memcpy(p, destr->name, destr->len);
32553 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
32554 + pos->wh_ino, pos->wh_type, delist);
32555 + if (unlikely(err))
32560 + free_page((unsigned long)o);
32570 +static int au_do_read_vdir(struct fillvdir_arg *arg)
32573 + unsigned int rdhash;
32575 + aufs_bindex_t bbot, bindex, btop;
32576 + unsigned char shwh;
32577 + struct file *hf, *file;
32578 + struct super_block *sb;
32580 + file = arg->file;
32581 + sb = file->f_path.dentry->d_sb;
32582 + SiMustAnyLock(sb);
32584 + rdhash = au_sbi(sb)->si_rdhash;
32586 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
32587 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
32588 + if (unlikely(err))
32590 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
32591 + if (unlikely(err))
32597 + if (au_opt_test(au_mntflags(sb), SHWH)) {
32599 + au_fset_fillvdir(arg->flags, SHWH);
32601 + btop = au_fbtop(file);
32602 + bbot = au_fbbot_dir(file);
32603 + for (bindex = btop; !err && bindex <= bbot; bindex++) {
32604 + hf = au_hf_dir(file, bindex);
32608 + offset = vfsub_llseek(hf, 0, SEEK_SET);
32610 + if (unlikely(offset))
32613 + arg->bindex = bindex;
32614 + au_fclr_fillvdir(arg->flags, WHABLE);
32616 + || (bindex != bbot
32617 + && au_br_whable(au_sbr_perm(sb, bindex))))
32618 + au_fset_fillvdir(arg->flags, WHABLE);
32621 + au_fclr_fillvdir(arg->flags, CALLED);
32623 + err = vfsub_iterate_dir(hf, &arg->ctx);
32626 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
32629 + * dir_relax() may be good for concurrency, but aufs should not
32630 + * use it since it will cause a lockdep problem.
32634 + if (!err && shwh)
32635 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
32637 + au_nhash_wh_free(&arg->whlist);
32640 + au_nhash_de_free(&arg->delist);
32645 +static int read_vdir(struct file *file, int may_read)
32648 + unsigned long expire;
32649 + unsigned char do_read;
32650 + struct fillvdir_arg arg = {
32652 + .actor = fillvdir
32655 + struct inode *inode;
32656 + struct au_vdir *vdir, *allocated;
32659 + inode = file_inode(file);
32660 + IMustLock(inode);
32661 + IiMustWriteLock(inode);
32662 + SiMustAnyLock(inode->i_sb);
32664 + allocated = NULL;
32666 + expire = au_sbi(inode->i_sb)->si_rdcache;
32667 + vdir = au_ivdir(inode);
32670 + vdir = alloc_vdir(file);
32671 + err = PTR_ERR(vdir);
32672 + if (IS_ERR(vdir))
32675 + allocated = vdir;
32676 + } else if (may_read
32677 + && (!inode_eq_iversion(inode, vdir->vd_version)
32678 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
32680 + err = reinit_vdir(vdir);
32681 + if (unlikely(err))
32686 + return 0; /* success */
32690 + err = au_do_read_vdir(&arg);
32692 + /* file->f_pos = 0; */ /* todo: ctx->pos? */
32693 + vdir->vd_version = inode_query_iversion(inode);
32694 + vdir->vd_last.ul = 0;
32695 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32697 + au_set_ivdir(inode, allocated);
32698 + } else if (allocated)
32699 + au_vdir_free(allocated);
32705 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
32708 + unsigned long ul, n;
32709 + const unsigned int deblk_sz = src->vd_deblk_sz;
32711 + AuDebugOn(tgt->vd_nblk != 1);
32714 + if (tgt->vd_nblk < src->vd_nblk) {
32715 + unsigned char **p;
32717 + p = au_krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
32718 + GFP_NOFS, /*may_shrink*/0);
32719 + if (unlikely(!p))
32721 + tgt->vd_deblk = p;
32724 + if (tgt->vd_deblk_sz != deblk_sz) {
32725 + unsigned char *p;
32727 + tgt->vd_deblk_sz = deblk_sz;
32728 + p = au_krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS,
32729 + /*may_shrink*/1);
32730 + if (unlikely(!p))
32732 + tgt->vd_deblk[0] = p;
32734 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
32735 + tgt->vd_version = src->vd_version;
32736 + tgt->vd_jiffy = src->vd_jiffy;
32738 + n = src->vd_nblk;
32739 + for (ul = 1; ul < n; ul++) {
32740 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
32742 + if (unlikely(!tgt->vd_deblk[ul]))
32746 + tgt->vd_nblk = n;
32747 + tgt->vd_last.ul = tgt->vd_last.ul;
32748 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
32749 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
32750 + - src->vd_deblk[src->vd_last.ul];
32752 + return 0; /* success */
32755 + rerr = reinit_vdir(tgt);
32760 +int au_vdir_init(struct file *file)
32763 + struct inode *inode;
32764 + struct au_vdir *vdir_cache, *allocated;
32766 + /* test file->f_pos here instead of ctx->pos */
32767 + err = read_vdir(file, !file->f_pos);
32768 + if (unlikely(err))
32771 + allocated = NULL;
32772 + vdir_cache = au_fvdir_cache(file);
32773 + if (!vdir_cache) {
32774 + vdir_cache = alloc_vdir(file);
32775 + err = PTR_ERR(vdir_cache);
32776 + if (IS_ERR(vdir_cache))
32778 + allocated = vdir_cache;
32779 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
32780 + /* test file->f_pos here instead of ctx->pos */
32781 + err = reinit_vdir(vdir_cache);
32782 + if (unlikely(err))
32785 + return 0; /* success */
32787 + inode = file_inode(file);
32788 + err = copy_vdir(vdir_cache, au_ivdir(inode));
32790 + file->f_version = inode_query_iversion(inode);
32792 + au_set_fvdir_cache(file, allocated);
32793 + } else if (allocated)
32794 + au_vdir_free(allocated);
32800 +static loff_t calc_offset(struct au_vdir *vdir)
32803 + union au_vdir_deblk_p p;
32805 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
32806 + offset = vdir->vd_last.p.deblk - p.deblk;
32807 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
32811 +/* returns true or false */
32812 +static int seek_vdir(struct file *file, struct dir_context *ctx)
32815 + unsigned int deblk_sz;
32816 + unsigned long ul, n;
32818 + union au_vdir_deblk_p p, deblk_end;
32819 + struct au_vdir *vdir_cache;
32822 + vdir_cache = au_fvdir_cache(file);
32823 + offset = calc_offset(vdir_cache);
32824 + AuDbg("offset %lld\n", offset);
32825 + if (ctx->pos == offset)
32828 + vdir_cache->vd_last.ul = 0;
32829 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
32834 + deblk_sz = vdir_cache->vd_deblk_sz;
32835 + ul = div64_u64(ctx->pos, deblk_sz);
32836 + AuDbg("ul %lu\n", ul);
32837 + if (ul >= vdir_cache->vd_nblk)
32840 + n = vdir_cache->vd_nblk;
32841 + for (; ul < n; ul++) {
32842 + p.deblk = vdir_cache->vd_deblk[ul];
32843 + deblk_end.deblk = p.deblk + deblk_sz;
32845 + offset *= deblk_sz;
32846 + while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
32849 + l = calc_size(p.de->de_str.len);
32853 + if (!is_deblk_end(&p, &deblk_end)) {
32855 + vdir_cache->vd_last.ul = ul;
32856 + vdir_cache->vd_last.p = p;
32864 + AuDbg("valid %d\n", !valid);
32868 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
32870 + unsigned int l, deblk_sz;
32871 + union au_vdir_deblk_p deblk_end;
32872 + struct au_vdir *vdir_cache;
32873 + struct au_vdir_de *de;
32875 + if (!seek_vdir(file, ctx))
32878 + vdir_cache = au_fvdir_cache(file);
32879 + deblk_sz = vdir_cache->vd_deblk_sz;
32881 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32882 + deblk_end.deblk += deblk_sz;
32883 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
32884 + de = vdir_cache->vd_last.p.de;
32885 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
32886 + de->de_str.len, de->de_str.name, ctx->pos,
32887 + (unsigned long)de->de_ino, de->de_type);
32888 + if (unlikely(!dir_emit(ctx, de->de_str.name,
32889 + de->de_str.len, de->de_ino,
32891 + /* todo: ignore the error caused by udba? */
32892 + /* return err; */
32896 + l = calc_size(de->de_str.len);
32897 + vdir_cache->vd_last.p.deblk += l;
32900 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
32901 + vdir_cache->vd_last.ul++;
32902 + vdir_cache->vd_last.p.deblk
32903 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32904 + ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
32913 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
32914 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
32915 +++ linux/fs/aufs/vfsub.c 2022-03-21 14:49:05.733299677 +0100
32917 +// SPDX-License-Identifier: GPL-2.0
32919 + * Copyright (C) 2005-2021 Junjiro R. Okajima
32921 + * This program, aufs is free software; you can redistribute it and/or modify
32922 + * it under the terms of the GNU General Public License as published by
32923 + * the Free Software Foundation; either version 2 of the License, or
32924 + * (at your option) any later version.
32926 + * This program is distributed in the hope that it will be useful,
32927 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32928 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32929 + * GNU General Public License for more details.
32931 + * You should have received a copy of the GNU General Public License
32932 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32936 + * sub-routines for VFS
32939 +#include <linux/mnt_namespace.h>
32940 +#include <linux/nsproxy.h>
32941 +#include <linux/security.h>
32942 +#include <linux/splice.h>
32945 +#ifdef CONFIG_AUFS_BR_FUSE
32946 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb)
32948 + if (!au_test_fuse(h_sb) || !au_userns)
32951 + return is_current_mnt_ns(mnt) ? 0 : -EACCES;
32955 +int vfsub_sync_filesystem(struct super_block *h_sb)
32960 + down_read(&h_sb->s_umount);
32961 + err = sync_filesystem(h_sb);
32962 + up_read(&h_sb->s_umount);
32968 +/* ---------------------------------------------------------------------- */
32970 +int vfsub_update_h_iattr(struct path *h_path, int *did)
32974 + struct super_block *h_sb;
32977 + * Always needs h_path->mnt for LSM or FUSE branch.
32979 + AuDebugOn(!h_path->mnt);
32981 + /* for remote fs, leave work for its getattr or d_revalidate */
32982 + /* for bad i_attr fs, handle them in aufs_getattr() */
32983 + /* still some fs may acquire i_mutex. we need to skip them */
32987 + h_sb = h_path->dentry->d_sb;
32988 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
32990 + err = vfsub_getattr(h_path, &st);
32995 +/* ---------------------------------------------------------------------- */
32997 +struct file *vfsub_dentry_open(struct path *path, int flags)
32999 + return dentry_open(path, flags /* | __FMODE_NONOTIFY */,
33003 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
33005 + struct file *file;
33008 + file = filp_open(path,
33009 + oflags /* | __FMODE_NONOTIFY */,
33012 + if (IS_ERR(file))
33014 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33021 + * Ideally this function should call VFS:do_last() in order to keep all its
33022 + * checkings. But it is very hard for aufs to regenerate several VFS internal
33023 + * structure such as nameidata. This is a second (or third) best approach.
33024 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
33026 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33027 + struct vfsub_aopen_args *args)
33030 + struct au_branch *br = args->br;
33031 + struct file *file = args->file;
33032 + /* copied from linux/fs/namei.c:atomic_open() */
33033 + struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
33036 + AuDebugOn(!dir->i_op->atomic_open);
33038 + err = au_br_test_oflag(args->open_flag, br);
33039 + if (unlikely(err))
33042 + au_lcnt_inc(&br->br_nfiles);
33043 + file->f_path.dentry = DENTRY_NOT_SET;
33044 + file->f_path.mnt = au_br_mnt(br);
33045 + AuDbg("%ps\n", dir->i_op->atomic_open);
33046 + err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
33047 + args->create_mode);
33048 + if (unlikely(err < 0)) {
33049 + au_lcnt_dec(&br->br_nfiles);
33053 + /* temporary workaround for nfsv4 branch */
33054 + if (au_test_nfs(dir->i_sb))
33055 + nfs_mark_for_revalidate(dir);
33057 + if (file->f_mode & FMODE_CREATED)
33058 + fsnotify_create(dir, dentry);
33059 + if (!(file->f_mode & FMODE_OPENED)) {
33060 + au_lcnt_dec(&br->br_nfiles);
33064 + /* todo: call VFS:may_open() here */
33065 + /* todo: ima_file_check() too? */
33066 + if (!err && (args->open_flag & __FMODE_EXEC))
33067 + err = deny_write_access(file);
33069 + fsnotify_open(file);
33071 + au_lcnt_dec(&br->br_nfiles);
33072 + /* note that the file is created and still opened */
33078 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
33082 + err = kern_path(name, flags, path);
33083 + if (!err && d_is_positive(path->dentry))
33084 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33088 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33089 + struct path *ppath, int len)
33091 + struct path path;
33093 + path.dentry = lookup_one_len_unlocked(name, ppath->dentry, len);
33094 + if (IS_ERR(path.dentry))
33096 + if (d_is_positive(path.dentry)) {
33097 + path.mnt = ppath->mnt;
33098 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33102 + AuTraceErrPtr(path.dentry);
33103 + return path.dentry;
33106 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33109 + struct path path;
33111 + /* VFS checks it too, but by WARN_ON_ONCE() */
33112 + IMustLock(d_inode(ppath->dentry));
33114 + path.dentry = lookup_one_len(name, ppath->dentry, len);
33115 + if (IS_ERR(path.dentry))
33117 + if (d_is_positive(path.dentry)) {
33118 + path.mnt = ppath->mnt;
33119 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33123 + AuTraceErrPtr(path.dentry);
33124 + return path.dentry;
33127 +void vfsub_call_lkup_one(void *args)
33129 + struct vfsub_lkup_one_args *a = args;
33130 + *a->errp = vfsub_lkup_one(a->name, a->ppath);
33133 +/* ---------------------------------------------------------------------- */
33135 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33136 + struct dentry *d2, struct au_hinode *hdir2)
33138 + struct dentry *d;
33141 + d = lock_rename(d1, d2);
33143 + au_hn_suspend(hdir1);
33144 + if (hdir1 != hdir2)
33145 + au_hn_suspend(hdir2);
33150 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33151 + struct dentry *d2, struct au_hinode *hdir2)
33153 + au_hn_resume(hdir1);
33154 + if (hdir1 != hdir2)
33155 + au_hn_resume(hdir2);
33157 + unlock_rename(d1, d2);
33161 +/* ---------------------------------------------------------------------- */
33163 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
33166 + struct dentry *d;
33167 + struct user_namespace *userns;
33171 + d = path->dentry;
33172 + path->dentry = d->d_parent;
33173 + err = security_path_mknod(path, d, mode, 0);
33174 + path->dentry = d;
33175 + if (unlikely(err))
33177 + userns = mnt_user_ns(path->mnt);
33180 + err = vfs_create(userns, dir, path->dentry, mode, want_excl);
33183 + struct path tmp = *path;
33186 + vfsub_update_h_iattr(&tmp, &did);
33188 + tmp.dentry = path->dentry->d_parent;
33189 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33198 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
33201 + struct dentry *d;
33202 + struct user_namespace *userns;
33206 + d = path->dentry;
33207 + path->dentry = d->d_parent;
33208 + err = security_path_symlink(path, d, symname);
33209 + path->dentry = d;
33210 + if (unlikely(err))
33212 + userns = mnt_user_ns(path->mnt);
33215 + err = vfs_symlink(userns, dir, path->dentry, symname);
33218 + struct path tmp = *path;
33221 + vfsub_update_h_iattr(&tmp, &did);
33223 + tmp.dentry = path->dentry->d_parent;
33224 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33233 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
33236 + struct dentry *d;
33237 + struct user_namespace *userns;
33241 + d = path->dentry;
33242 + path->dentry = d->d_parent;
33243 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
33244 + path->dentry = d;
33245 + if (unlikely(err))
33247 + userns = mnt_user_ns(path->mnt);
33250 + err = vfs_mknod(userns, dir, path->dentry, mode, dev);
33253 + struct path tmp = *path;
33256 + vfsub_update_h_iattr(&tmp, &did);
33258 + tmp.dentry = path->dentry->d_parent;
33259 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33268 +static int au_test_nlink(struct inode *inode)
33270 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
33272 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
33273 + || inode->i_nlink < link_max)
33278 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
33279 + struct inode **delegated_inode)
33282 + struct dentry *d;
33283 + struct user_namespace *userns;
33287 + err = au_test_nlink(d_inode(src_dentry));
33288 + if (unlikely(err))
33291 + /* we don't call may_linkat() */
33292 + d = path->dentry;
33293 + path->dentry = d->d_parent;
33294 + err = security_path_link(src_dentry, path, d);
33295 + path->dentry = d;
33296 + if (unlikely(err))
33298 + userns = mnt_user_ns(path->mnt);
33301 + err = vfs_link(src_dentry, userns, dir, path->dentry, delegated_inode);
33304 + struct path tmp = *path;
33307 + /* fuse has different memory inode for the same inumber */
33308 + vfsub_update_h_iattr(&tmp, &did);
33310 + tmp.dentry = path->dentry->d_parent;
33311 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33312 + tmp.dentry = src_dentry;
33313 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33322 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
33323 + struct inode *dir, struct path *path,
33324 + struct inode **delegated_inode, unsigned int flags)
33327 + struct renamedata rd;
33328 + struct path tmp = {
33331 + struct dentry *d;
33334 + IMustLock(src_dir);
33336 + d = path->dentry;
33337 + path->dentry = d->d_parent;
33338 + tmp.dentry = src_dentry->d_parent;
33339 + err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
33340 + path->dentry = d;
33341 + if (unlikely(err))
33344 + rd.old_mnt_userns = mnt_user_ns(path->mnt);
33345 + rd.old_dir = src_dir;
33346 + rd.old_dentry = src_dentry;
33347 + rd.new_mnt_userns = rd.old_mnt_userns;
33348 + rd.new_dir = dir;
33349 + rd.new_dentry = path->dentry;
33350 + rd.delegated_inode = delegated_inode;
33351 + rd.flags = flags;
33353 + err = vfs_rename(&rd);
33358 + tmp.dentry = d->d_parent;
33359 + vfsub_update_h_iattr(&tmp, &did);
33361 + tmp.dentry = src_dentry;
33362 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33363 + tmp.dentry = src_dentry->d_parent;
33364 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33373 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
33376 + struct dentry *d;
33377 + struct user_namespace *userns;
33381 + d = path->dentry;
33382 + path->dentry = d->d_parent;
33383 + err = security_path_mkdir(path, d, mode);
33384 + path->dentry = d;
33385 + if (unlikely(err))
33387 + userns = mnt_user_ns(path->mnt);
33390 + err = vfs_mkdir(userns, dir, path->dentry, mode);
33393 + struct path tmp = *path;
33396 + vfsub_update_h_iattr(&tmp, &did);
33398 + tmp.dentry = path->dentry->d_parent;
33399 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33408 +int vfsub_rmdir(struct inode *dir, struct path *path)
33411 + struct dentry *d;
33412 + struct user_namespace *userns;
33416 + d = path->dentry;
33417 + path->dentry = d->d_parent;
33418 + err = security_path_rmdir(path, d);
33419 + path->dentry = d;
33420 + if (unlikely(err))
33422 + userns = mnt_user_ns(path->mnt);
33425 + err = vfs_rmdir(userns, dir, path->dentry);
33428 + struct path tmp = {
33429 + .dentry = path->dentry->d_parent,
33433 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33440 +/* ---------------------------------------------------------------------- */
33442 +/* todo: support mmap_sem? */
33443 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
33449 + err = vfs_read(file, ubuf, count, ppos);
33452 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33456 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
33462 + err = kernel_read(file, kbuf, count, ppos);
33466 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33470 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
33476 + err = vfs_write(file, ubuf, count, ppos);
33479 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33483 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
33488 + err = kernel_write(file, kbuf, count, ppos);
33491 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33495 +int vfsub_flush(struct file *file, fl_owner_t id)
33500 + if (file->f_op->flush) {
33501 + if (!au_test_nfs(file->f_path.dentry->d_sb))
33502 + err = file->f_op->flush(file, id);
33505 + err = file->f_op->flush(file, id);
33509 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
33515 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
33519 + AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
33522 + err = iterate_dir(file, ctx);
33525 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33530 +long vfsub_splice_to(struct file *in, loff_t *ppos,
33531 + struct pipe_inode_info *pipe, size_t len,
33532 + unsigned int flags)
33537 + err = do_splice_to(in, ppos, pipe, len, flags);
33539 + file_accessed(in);
33541 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
33545 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
33546 + loff_t *ppos, size_t len, unsigned int flags)
33551 + err = do_splice_from(pipe, out, ppos, len, flags);
33554 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
33558 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
33562 + /* file can be NULL */
33564 + err = vfs_fsync(file, datasync);
33568 + AuDebugOn(!file);
33569 + path = &file->f_path;
33571 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33576 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
33577 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33578 + struct file *h_file)
33581 + struct inode *h_inode;
33582 + struct super_block *h_sb;
33583 + struct user_namespace *h_userns;
33586 + err = vfsub_truncate(h_path, length);
33590 + h_inode = d_inode(h_path->dentry);
33591 + h_sb = h_inode->i_sb;
33593 + sb_start_write(h_sb);
33595 + err = security_path_truncate(h_path);
33597 + h_userns = mnt_user_ns(h_path->mnt);
33599 + err = do_truncate(h_userns, h_path->dentry, length, attr,
33604 + sb_end_write(h_sb);
33611 +/* ---------------------------------------------------------------------- */
33613 +struct au_vfsub_mkdir_args {
33615 + struct inode *dir;
33616 + struct path *path;
33620 +static void au_call_vfsub_mkdir(void *args)
33622 + struct au_vfsub_mkdir_args *a = args;
33623 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
33626 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
33628 + int err, do_sio, wkq_err;
33629 + struct user_namespace *userns;
33631 + userns = mnt_user_ns(path->mnt);
33632 + do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33635 + err = vfsub_mkdir(dir, path, mode);
33638 + struct au_vfsub_mkdir_args args = {
33644 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
33645 + if (unlikely(wkq_err))
33652 +struct au_vfsub_rmdir_args {
33654 + struct inode *dir;
33655 + struct path *path;
33658 +static void au_call_vfsub_rmdir(void *args)
33660 + struct au_vfsub_rmdir_args *a = args;
33661 + *a->errp = vfsub_rmdir(a->dir, a->path);
33664 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
33666 + int err, do_sio, wkq_err;
33667 + struct user_namespace *userns;
33669 + userns = mnt_user_ns(path->mnt);
33670 + do_sio = au_test_h_perm_sio(userns, dir, MAY_EXEC | MAY_WRITE);
33673 + err = vfsub_rmdir(dir, path);
33676 + struct au_vfsub_rmdir_args args = {
33681 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
33682 + if (unlikely(wkq_err))
33689 +/* ---------------------------------------------------------------------- */
33691 +struct notify_change_args {
33693 + struct path *path;
33694 + struct iattr *ia;
33695 + struct inode **delegated_inode;
33698 +static void call_notify_change(void *args)
33700 + struct notify_change_args *a = args;
33701 + struct inode *h_inode;
33702 + struct user_namespace *userns;
33704 + h_inode = d_inode(a->path->dentry);
33705 + IMustLock(h_inode);
33707 + *a->errp = -EPERM;
33708 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
33709 + userns = mnt_user_ns(a->path->mnt);
33711 + *a->errp = notify_change(userns, a->path->dentry, a->ia,
33712 + a->delegated_inode);
33715 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
33717 + AuTraceErr(*a->errp);
33720 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33721 + struct inode **delegated_inode)
33724 + struct notify_change_args args = {
33728 + .delegated_inode = delegated_inode
33731 + call_notify_change(&args);
33736 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33737 + struct inode **delegated_inode)
33739 + int err, wkq_err;
33740 + struct notify_change_args args = {
33744 + .delegated_inode = delegated_inode
33747 + wkq_err = au_wkq_wait(call_notify_change, &args);
33748 + if (unlikely(wkq_err))
33754 +/* ---------------------------------------------------------------------- */
33756 +struct unlink_args {
33758 + struct inode *dir;
33759 + struct path *path;
33760 + struct inode **delegated_inode;
33763 +static void call_unlink(void *args)
33765 + struct unlink_args *a = args;
33766 + struct dentry *d = a->path->dentry;
33767 + struct inode *h_inode;
33768 + struct user_namespace *userns;
33769 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
33770 + && au_dcount(d) == 1);
33772 + IMustLock(a->dir);
33774 + a->path->dentry = d->d_parent;
33775 + *a->errp = security_path_unlink(a->path, d);
33776 + a->path->dentry = d;
33777 + if (unlikely(*a->errp))
33780 + if (!stop_sillyrename)
33783 + if (d_is_positive(d)) {
33784 + h_inode = d_inode(d);
33788 + userns = mnt_user_ns(a->path->mnt);
33790 + *a->errp = vfs_unlink(userns, a->dir, d, a->delegated_inode);
33793 + struct path tmp = {
33794 + .dentry = d->d_parent,
33795 + .mnt = a->path->mnt
33797 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33800 + if (!stop_sillyrename)
33805 + AuTraceErr(*a->errp);
33809 + * @dir: must be locked.
33810 + * @dentry: target dentry.
33812 +int vfsub_unlink(struct inode *dir, struct path *path,
33813 + struct inode **delegated_inode, int force)
33816 + struct unlink_args args = {
33820 + .delegated_inode = delegated_inode
33824 + call_unlink(&args);
33828 + wkq_err = au_wkq_wait(call_unlink, &args);
33829 + if (unlikely(wkq_err))
33835 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
33836 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
33837 +++ linux/fs/aufs/vfsub.h 2022-03-21 14:49:05.733299677 +0100
33839 +/* SPDX-License-Identifier: GPL-2.0 */
33841 + * Copyright (C) 2005-2021 Junjiro R. Okajima
33843 + * This program, aufs is free software; you can redistribute it and/or modify
33844 + * it under the terms of the GNU General Public License as published by
33845 + * the Free Software Foundation; either version 2 of the License, or
33846 + * (at your option) any later version.
33848 + * This program is distributed in the hope that it will be useful,
33849 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33850 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33851 + * GNU General Public License for more details.
33853 + * You should have received a copy of the GNU General Public License
33854 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33858 + * sub-routines for VFS
33861 +#ifndef __AUFS_VFSUB_H__
33862 +#define __AUFS_VFSUB_H__
33866 +#include <linux/fs.h>
33867 +#include <linux/mount.h>
33868 +#include <linux/posix_acl.h>
33869 +#include <linux/xattr.h>
33870 +#include "debug.h"
33872 +/* copied from linux/fs/internal.h */
33873 +/* todo: BAD approach!! */
33874 +extern void __mnt_drop_write(struct vfsmount *);
33875 +extern struct file *alloc_empty_file(int, const struct cred *);
33877 +/* ---------------------------------------------------------------------- */
33879 +/* lock subclass for lower inode */
33880 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
33881 +/* reduce? gave up. */
33883 + AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
33884 + AuLsc_I_PARENT, /* lower inode, parent first */
33885 + AuLsc_I_PARENT2, /* copyup dirs */
33886 + AuLsc_I_PARENT3, /* copyup wh */
33892 +/* to debug easier, do not make them inlined functions */
33893 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
33894 +#define IMustLock(i) AuDebugOn(!inode_is_locked(i))
33896 +/* ---------------------------------------------------------------------- */
33898 +static inline void vfsub_drop_nlink(struct inode *inode)
33900 + AuDebugOn(!inode->i_nlink);
33901 + drop_nlink(inode);
33904 +static inline void vfsub_dead_dir(struct inode *inode)
33906 + AuDebugOn(!S_ISDIR(inode->i_mode));
33907 + inode->i_flags |= S_DEAD;
33908 + clear_nlink(inode);
33911 +static inline int vfsub_native_ro(struct inode *inode)
33913 + return sb_rdonly(inode->i_sb)
33914 + || IS_RDONLY(inode)
33915 + /* || IS_APPEND(inode) */
33916 + || IS_IMMUTABLE(inode);
33919 +#ifdef CONFIG_AUFS_BR_FUSE
33920 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb);
33922 +AuStubInt0(vfsub_test_mntns, struct vfsmount *mnt, struct super_block *h_sb);
33925 +int vfsub_sync_filesystem(struct super_block *h_sb);
33927 +/* ---------------------------------------------------------------------- */
33929 +int vfsub_update_h_iattr(struct path *h_path, int *did);
33930 +struct file *vfsub_dentry_open(struct path *path, int flags);
33931 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
33933 +struct vfsub_aopen_args {
33934 + struct file *file;
33935 + unsigned int open_flag;
33936 + umode_t create_mode;
33937 + struct au_branch *br;
33939 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33940 + struct vfsub_aopen_args *args);
33941 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
33943 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33944 + struct path *ppath, int len);
33945 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33948 +struct vfsub_lkup_one_args {
33949 + struct dentry **errp;
33950 + struct qstr *name;
33951 + struct path *ppath;
33954 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
33955 + struct path *ppath)
33957 + return vfsub_lookup_one_len(name->name, ppath, name->len);
33960 +void vfsub_call_lkup_one(void *args);
33962 +/* ---------------------------------------------------------------------- */
33964 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
33969 + err = mnt_want_write(mnt);
33974 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
33977 + mnt_drop_write(mnt);
33981 +#if 0 /* reserved */
33982 +static inline void vfsub_mnt_drop_write_file(struct file *file)
33985 + mnt_drop_write_file(file);
33990 +/* ---------------------------------------------------------------------- */
33993 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33994 + struct dentry *d2, struct au_hinode *hdir2);
33995 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33996 + struct dentry *d2, struct au_hinode *hdir2);
33998 +int vfsub_create(struct inode *dir, struct path *path, int mode,
34000 +int vfsub_symlink(struct inode *dir, struct path *path,
34001 + const char *symname);
34002 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
34003 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
34004 + struct path *path, struct inode **delegated_inode);
34005 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
34006 + struct inode *hdir, struct path *path,
34007 + struct inode **delegated_inode, unsigned int flags);
34008 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
34009 +int vfsub_rmdir(struct inode *dir, struct path *path);
34011 +/* ---------------------------------------------------------------------- */
34013 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
34015 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
34017 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
34019 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
34021 +int vfsub_flush(struct file *file, fl_owner_t id);
34022 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
34024 +static inline loff_t vfsub_f_size_read(struct file *file)
34026 + return i_size_read(file_inode(file));
34029 +static inline unsigned int vfsub_file_flags(struct file *file)
34031 + unsigned int flags;
34033 + spin_lock(&file->f_lock);
34034 + flags = file->f_flags;
34035 + spin_unlock(&file->f_lock);
34040 +static inline int vfsub_file_execed(struct file *file)
34042 + /* todo: direct access f_flags */
34043 + return !!(vfsub_file_flags(file) & __FMODE_EXEC);
34046 +#if 0 /* reserved */
34047 +static inline void vfsub_file_accessed(struct file *h_file)
34049 + file_accessed(h_file);
34050 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
34054 +#if 0 /* reserved */
34055 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
34056 + struct dentry *h_dentry)
34058 + struct path h_path = {
34059 + .dentry = h_dentry,
34062 + touch_atime(&h_path);
34063 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
34067 +static inline int vfsub_update_time(struct inode *h_inode,
34068 + struct timespec64 *ts, int flags)
34070 + return inode_update_time(h_inode, ts, flags);
34071 + /* no vfsub_update_h_iattr() since we don't have struct path */
34074 +#ifdef CONFIG_FS_POSIX_ACL
34075 +static inline int vfsub_acl_chmod(struct user_namespace *h_userns,
34076 + struct inode *h_inode, umode_t h_mode)
34080 + err = posix_acl_chmod(h_userns, h_inode, h_mode);
34081 + if (err == -EOPNOTSUPP)
34086 +AuStubInt0(vfsub_acl_chmod, struct user_namespace *h_userns,
34087 + struct inode *h_inode, umode_t h_mode);
34090 +long vfsub_splice_to(struct file *in, loff_t *ppos,
34091 + struct pipe_inode_info *pipe, size_t len,
34092 + unsigned int flags);
34093 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
34094 + loff_t *ppos, size_t len, unsigned int flags);
34096 +static inline long vfsub_truncate(struct path *path, loff_t length)
34101 + err = vfs_truncate(path, length);
34106 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
34107 + struct file *h_file);
34108 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
34111 + * re-use branch fs's ioctl(FICLONE) while aufs itself doesn't support such
34114 +static inline loff_t vfsub_clone_file_range(struct file *src, struct file *dst,
34120 + err = vfs_clone_file_range(src, 0, dst, 0, len, /*remap_flags*/0);
34126 +/* copy_file_range(2) is a systemcall */
34127 +static inline ssize_t vfsub_copy_file_range(struct file *src, loff_t src_pos,
34128 + struct file *dst, loff_t dst_pos,
34129 + size_t len, unsigned int flags)
34134 + ssz = vfs_copy_file_range(src, src_pos, dst, dst_pos, len, flags);
34140 +/* ---------------------------------------------------------------------- */
34142 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
34147 + err = vfs_llseek(file, offset, origin);
34152 +/* ---------------------------------------------------------------------- */
34154 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
34155 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
34156 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
34157 + struct inode **delegated_inode);
34158 +int vfsub_notify_change(struct path *path, struct iattr *ia,
34159 + struct inode **delegated_inode);
34160 +int vfsub_unlink(struct inode *dir, struct path *path,
34161 + struct inode **delegated_inode, int force);
34163 +static inline int vfsub_getattr(const struct path *path, struct kstat *st)
34165 + return vfs_getattr(path, st, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
34168 +/* ---------------------------------------------------------------------- */
34170 +static inline int vfsub_setxattr(struct user_namespace *userns,
34171 + struct dentry *dentry, const char *name,
34172 + const void *value, size_t size, int flags)
34177 + err = vfs_setxattr(userns, dentry, name, value, size, flags);
34183 +static inline int vfsub_removexattr(struct user_namespace *userns,
34184 + struct dentry *dentry, const char *name)
34189 + err = vfs_removexattr(userns, dentry, name);
34195 +#endif /* __KERNEL__ */
34196 +#endif /* __AUFS_VFSUB_H__ */
34197 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
34198 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
34199 +++ linux/fs/aufs/wbr_policy.c 2022-03-21 14:49:05.733299677 +0100
34201 +// SPDX-License-Identifier: GPL-2.0
34203 + * Copyright (C) 2005-2021 Junjiro R. Okajima
34205 + * This program, aufs is free software; you can redistribute it and/or modify
34206 + * it under the terms of the GNU General Public License as published by
34207 + * the Free Software Foundation; either version 2 of the License, or
34208 + * (at your option) any later version.
34210 + * This program is distributed in the hope that it will be useful,
34211 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
34212 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
34213 + * GNU General Public License for more details.
34215 + * You should have received a copy of the GNU General Public License
34216 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
34220 + * policies for selecting one among multiple writable branches
34223 +#include <linux/statfs.h>
34226 +/* subset of cpup_attr() */
34227 +static noinline_for_stack
34228 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
34232 + struct inode *h_isrc;
34234 + h_isrc = d_inode(h_src);
34235 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
34236 + ia.ia_mode = h_isrc->i_mode;
34237 + ia.ia_uid = h_isrc->i_uid;
34238 + ia.ia_gid = h_isrc->i_gid;
34239 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
34240 + au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
34241 + /* no delegation since it is just created */
34242 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34244 + /* is this nfs only? */
34245 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
34246 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
34247 + ia.ia_mode = h_isrc->i_mode;
34248 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34254 +#define AuCpdown_PARENT_OPQ 1
34255 +#define AuCpdown_WHED (1 << 1)
34256 +#define AuCpdown_MADE_DIR (1 << 2)
34257 +#define AuCpdown_DIROPQ (1 << 3)
34258 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
34259 +#define au_fset_cpdown(flags, name) \
34260 + do { (flags) |= AuCpdown_##name; } while (0)
34261 +#define au_fclr_cpdown(flags, name) \
34262 + do { (flags) &= ~AuCpdown_##name; } while (0)
34264 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
34265 + unsigned int *flags)
34268 + struct dentry *opq_dentry;
34270 + opq_dentry = au_diropq_create(dentry, bdst);
34271 + err = PTR_ERR(opq_dentry);
34272 + if (IS_ERR(opq_dentry))
34274 + dput(opq_dentry);
34275 + au_fset_cpdown(*flags, DIROPQ);
34281 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
34282 + struct inode *dir, aufs_bindex_t bdst)
34285 + struct path h_path;
34286 + struct au_branch *br;
34288 + br = au_sbr(dentry->d_sb, bdst);
34289 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
34290 + err = PTR_ERR(h_path.dentry);
34291 + if (IS_ERR(h_path.dentry))
34295 + if (d_is_positive(h_path.dentry)) {
34296 + h_path.mnt = au_br_mnt(br);
34297 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
34300 + dput(h_path.dentry);
34306 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
34307 + struct au_pin *pin,
34308 + struct dentry *h_parent, void *arg)
34311 + aufs_bindex_t bopq, btop;
34312 + struct path h_path;
34313 + struct dentry *parent;
34314 + struct inode *h_dir, *h_inode, *inode, *dir;
34315 + unsigned int *flags = arg;
34317 + btop = au_dbtop(dentry);
34318 + /* dentry is di-locked */
34319 + parent = dget_parent(dentry);
34320 + dir = d_inode(parent);
34321 + h_dir = d_inode(h_parent);
34322 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
34323 + IMustLock(h_dir);
34325 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
34326 + if (unlikely(err < 0))
34328 + h_path.dentry = au_h_dptr(dentry, bdst);
34329 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
34330 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path, 0755);
34331 + if (unlikely(err))
34333 + au_fset_cpdown(*flags, MADE_DIR);
34335 + bopq = au_dbdiropq(dentry);
34336 + au_fclr_cpdown(*flags, WHED);
34337 + au_fclr_cpdown(*flags, DIROPQ);
34338 + if (au_dbwh(dentry) == bdst)
34339 + au_fset_cpdown(*flags, WHED);
34340 + if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
34341 + au_fset_cpdown(*flags, PARENT_OPQ);
34342 + h_inode = d_inode(h_path.dentry);
34343 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
34344 + if (au_ftest_cpdown(*flags, WHED)) {
34345 + err = au_cpdown_dir_opq(dentry, bdst, flags);
34346 + if (unlikely(err)) {
34347 + inode_unlock(h_inode);
34352 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, btop));
34353 + inode_unlock(h_inode);
34354 + if (unlikely(err))
34357 + if (au_ftest_cpdown(*flags, WHED)) {
34358 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
34359 + if (unlikely(err))
34363 + inode = d_inode(dentry);
34364 + if (au_ibbot(inode) < bdst)
34365 + au_set_ibbot(inode, bdst);
34366 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
34367 + au_hi_flags(inode, /*isdir*/1));
34368 + au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
34369 + goto out; /* success */
34373 + if (au_ftest_cpdown(*flags, DIROPQ)) {
34374 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
34375 + rerr = au_diropq_remove(dentry, bdst);
34376 + inode_unlock(h_inode);
34377 + if (unlikely(rerr)) {
34378 + AuIOErr("failed removing diropq for %pd b%d (%d)\n",
34379 + dentry, bdst, rerr);
34385 + if (au_ftest_cpdown(*flags, MADE_DIR)) {
34386 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
34387 + if (unlikely(rerr)) {
34388 + AuIOErr("failed removing %pd b%d (%d)\n",
34389 + dentry, bdst, rerr);
34394 + au_set_h_dptr(dentry, bdst, NULL);
34395 + if (au_dbbot(dentry) == bdst)
34396 + au_update_dbbot(dentry);
34402 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
34405 + unsigned int flags;
34408 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
34413 +/* ---------------------------------------------------------------------- */
34415 +/* policies for create */
34417 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
34419 + int err, i, j, ndentry;
34420 + aufs_bindex_t bopq;
34421 + struct au_dcsub_pages dpages;
34422 + struct au_dpage *dpage;
34423 + struct dentry **dentries, *parent, *d;
34425 + err = au_dpages_init(&dpages, GFP_NOFS);
34426 + if (unlikely(err))
34428 + parent = dget_parent(dentry);
34429 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
34430 + if (unlikely(err))
34434 + for (i = 0; i < dpages.ndpage; i++) {
34435 + dpage = dpages.dpages + i;
34436 + dentries = dpage->dentries;
34437 + ndentry = dpage->ndentry;
34438 + for (j = 0; j < ndentry; j++) {
34440 + di_read_lock_parent2(d, !AuLock_IR);
34441 + bopq = au_dbdiropq(d);
34442 + di_read_unlock(d, !AuLock_IR);
34443 + if (bopq >= 0 && bopq < err)
34450 + au_dpages_free(&dpages);
34455 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
34457 + for (; bindex >= 0; bindex--)
34458 + if (!au_br_rdonly(au_sbr(sb, bindex)))
34463 +/* top down parent */
34464 +static int au_wbr_create_tdp(struct dentry *dentry,
34465 + unsigned int flags __maybe_unused)
34468 + aufs_bindex_t btop, bindex;
34469 + struct super_block *sb;
34470 + struct dentry *parent, *h_parent;
34472 + sb = dentry->d_sb;
34473 + btop = au_dbtop(dentry);
34475 + if (!au_br_rdonly(au_sbr(sb, btop)))
34479 + parent = dget_parent(dentry);
34480 + for (bindex = au_dbtop(parent); bindex < btop; bindex++) {
34481 + h_parent = au_h_dptr(parent, bindex);
34482 + if (!h_parent || d_is_negative(h_parent))
34485 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
34492 + /* bottom up here */
34493 + if (unlikely(err < 0)) {
34494 + err = au_wbr_bu(sb, btop - 1);
34496 + err = au_wbr_nonopq(dentry, err);
34500 + AuDbg("b%d\n", err);
34504 +/* ---------------------------------------------------------------------- */
34506 +/* an exception for the policy other than tdp */
34507 +static int au_wbr_create_exp(struct dentry *dentry)
34510 + aufs_bindex_t bwh, bdiropq;
34511 + struct dentry *parent;
34514 + bwh = au_dbwh(dentry);
34515 + parent = dget_parent(dentry);
34516 + bdiropq = au_dbdiropq(parent);
34518 + if (bdiropq >= 0)
34519 + err = min(bdiropq, bwh);
34522 + AuDbg("%d\n", err);
34523 + } else if (bdiropq >= 0) {
34525 + AuDbg("%d\n", err);
34530 + err = au_wbr_nonopq(dentry, err);
34532 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
34535 + AuDbg("%d\n", err);
34539 +/* ---------------------------------------------------------------------- */
34542 +static int au_wbr_create_init_rr(struct super_block *sb)
34546 + err = au_wbr_bu(sb, au_sbbot(sb));
34547 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
34550 + AuDbg("b%d\n", err);
34554 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
34558 + aufs_bindex_t bindex, bbot;
34559 + struct super_block *sb;
34562 + err = au_wbr_create_exp(dentry);
34566 + sb = dentry->d_sb;
34567 + next = &au_sbi(sb)->si_wbr_rr_next;
34568 + bbot = au_sbbot(sb);
34570 + for (bindex = 0; bindex <= bbot; bindex++) {
34571 + if (!au_ftest_wbr(flags, DIR)) {
34572 + err = atomic_dec_return(next) + 1;
34573 + /* modulo for 0 is meaningless */
34574 + if (unlikely(!err))
34575 + err = atomic_dec_return(next) + 1;
34577 + err = atomic_read(next);
34578 + AuDbg("%d\n", err);
34581 + AuDbg("%d\n", err);
34582 + if (!au_br_rdonly(au_sbr(sb, err)))
34588 + err = au_wbr_nonopq(dentry, err);
34591 + AuDbg("%d\n", err);
34595 +/* ---------------------------------------------------------------------- */
34597 +/* most free space */
34598 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
34600 + struct super_block *sb;
34601 + struct au_branch *br;
34602 + struct au_wbr_mfs *mfs;
34603 + struct dentry *h_parent;
34604 + aufs_bindex_t bindex, bbot;
34606 + unsigned long long b, bavail;
34607 + struct path h_path;
34608 + /* reduce the stack usage */
34609 + struct kstatfs *st;
34611 + st = kmalloc(sizeof(*st), GFP_NOFS);
34612 + if (unlikely(!st)) {
34613 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
34618 + sb = dentry->d_sb;
34619 + mfs = &au_sbi(sb)->si_wbr_mfs;
34620 + MtxMustLock(&mfs->mfs_lock);
34621 + mfs->mfs_bindex = -EROFS;
34622 + mfs->mfsrr_bytes = 0;
34625 + bbot = au_sbbot(sb);
34627 + bindex = au_dbtop(parent);
34628 + bbot = au_dbtaildir(parent);
34631 + for (; bindex <= bbot; bindex++) {
34633 + h_parent = au_h_dptr(parent, bindex);
34634 + if (!h_parent || d_is_negative(h_parent))
34637 + br = au_sbr(sb, bindex);
34638 + if (au_br_rdonly(br))
34641 + /* sb->s_root for NFS is unreliable */
34642 + h_path.mnt = au_br_mnt(br);
34643 + h_path.dentry = h_path.mnt->mnt_root;
34644 + err = vfs_statfs(&h_path, st);
34645 + if (unlikely(err)) {
34646 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
34650 + /* when the available size is equal, select the lower one */
34651 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
34652 + || sizeof(b) < sizeof(st->f_bsize));
34653 + b = st->f_bavail * st->f_bsize;
34654 + br->br_wbr->wbr_bytes = b;
34655 + if (b >= bavail) {
34657 + mfs->mfs_bindex = bindex;
34658 + mfs->mfs_jiffy = jiffies;
34662 + mfs->mfsrr_bytes = bavail;
34663 + AuDbg("b%d\n", mfs->mfs_bindex);
34664 + au_kfree_rcu(st);
34667 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
34670 + struct dentry *parent;
34671 + struct super_block *sb;
34672 + struct au_wbr_mfs *mfs;
34674 + err = au_wbr_create_exp(dentry);
34678 + sb = dentry->d_sb;
34680 + if (au_ftest_wbr(flags, PARENT))
34681 + parent = dget_parent(dentry);
34682 + mfs = &au_sbi(sb)->si_wbr_mfs;
34683 + mutex_lock(&mfs->mfs_lock);
34684 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34685 + || mfs->mfs_bindex < 0
34686 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
34687 + au_mfs(dentry, parent);
34688 + mutex_unlock(&mfs->mfs_lock);
34689 + err = mfs->mfs_bindex;
34693 + err = au_wbr_nonopq(dentry, err);
34696 + AuDbg("b%d\n", err);
34700 +static int au_wbr_create_init_mfs(struct super_block *sb)
34702 + struct au_wbr_mfs *mfs;
34704 + mfs = &au_sbi(sb)->si_wbr_mfs;
34705 + mutex_init(&mfs->mfs_lock);
34706 + mfs->mfs_jiffy = 0;
34707 + mfs->mfs_bindex = -EROFS;
34712 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
34714 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
34718 +/* ---------------------------------------------------------------------- */
34720 +/* top down regardless parent, and then mfs */
34721 +static int au_wbr_create_tdmfs(struct dentry *dentry,
34722 + unsigned int flags __maybe_unused)
34725 + aufs_bindex_t bwh, btail, bindex, bfound, bmfs;
34726 + unsigned long long watermark;
34727 + struct super_block *sb;
34728 + struct au_wbr_mfs *mfs;
34729 + struct au_branch *br;
34730 + struct dentry *parent;
34732 + sb = dentry->d_sb;
34733 + mfs = &au_sbi(sb)->si_wbr_mfs;
34734 + mutex_lock(&mfs->mfs_lock);
34735 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34736 + || mfs->mfs_bindex < 0)
34737 + au_mfs(dentry, /*parent*/NULL);
34738 + watermark = mfs->mfsrr_watermark;
34739 + bmfs = mfs->mfs_bindex;
34740 + mutex_unlock(&mfs->mfs_lock);
34742 + /* another style of au_wbr_create_exp() */
34743 + bwh = au_dbwh(dentry);
34744 + parent = dget_parent(dentry);
34745 + btail = au_dbtaildir(parent);
34746 + if (bwh >= 0 && bwh < btail)
34749 + err = au_wbr_nonopq(dentry, btail);
34750 + if (unlikely(err < 0))
34754 + for (bindex = 0; bindex <= btail; bindex++) {
34755 + br = au_sbr(sb, bindex);
34756 + if (au_br_rdonly(br))
34758 + if (br->br_wbr->wbr_bytes > watermark) {
34769 + AuDbg("b%d\n", err);
34773 +/* ---------------------------------------------------------------------- */
34775 +/* most free space and then round robin */
34776 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
34779 + struct au_wbr_mfs *mfs;
34781 + err = au_wbr_create_mfs(dentry, flags);
34783 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
34784 + mutex_lock(&mfs->mfs_lock);
34785 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
34786 + err = au_wbr_create_rr(dentry, flags);
34787 + mutex_unlock(&mfs->mfs_lock);
34790 + AuDbg("b%d\n", err);
34794 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
34798 + au_wbr_create_init_mfs(sb); /* ignore */
34799 + err = au_wbr_create_init_rr(sb);
34804 +/* ---------------------------------------------------------------------- */
34806 +/* top down parent and most free space */
34807 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
34810 + unsigned long long b;
34811 + aufs_bindex_t bindex, btop, bbot;
34812 + struct super_block *sb;
34813 + struct dentry *parent, *h_parent;
34814 + struct au_branch *br;
34816 + err = au_wbr_create_tdp(dentry, flags);
34817 + if (unlikely(err < 0))
34819 + parent = dget_parent(dentry);
34820 + btop = au_dbtop(parent);
34821 + bbot = au_dbtaildir(parent);
34822 + if (btop == bbot)
34823 + goto out_parent; /* success */
34825 + e2 = au_wbr_create_mfs(dentry, flags);
34827 + goto out_parent; /* success */
34829 + /* when the available size is equal, select upper one */
34830 + sb = dentry->d_sb;
34831 + br = au_sbr(sb, err);
34832 + b = br->br_wbr->wbr_bytes;
34833 + AuDbg("b%d, %llu\n", err, b);
34835 + for (bindex = btop; bindex <= bbot; bindex++) {
34836 + h_parent = au_h_dptr(parent, bindex);
34837 + if (!h_parent || d_is_negative(h_parent))
34840 + br = au_sbr(sb, bindex);
34841 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
34842 + b = br->br_wbr->wbr_bytes;
34844 + AuDbg("b%d, %llu\n", err, b);
34849 + err = au_wbr_nonopq(dentry, err);
34854 + AuDbg("b%d\n", err);
34858 +/* ---------------------------------------------------------------------- */
34861 + * - top down parent
34862 + * - most free space with parent
34863 + * - most free space round-robin regardless parent
34865 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
34868 + unsigned long long watermark;
34869 + struct super_block *sb;
34870 + struct au_branch *br;
34871 + struct au_wbr_mfs *mfs;
34873 + err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
34874 + if (unlikely(err < 0))
34877 + sb = dentry->d_sb;
34878 + br = au_sbr(sb, err);
34879 + mfs = &au_sbi(sb)->si_wbr_mfs;
34880 + mutex_lock(&mfs->mfs_lock);
34881 + watermark = mfs->mfsrr_watermark;
34882 + mutex_unlock(&mfs->mfs_lock);
34883 + if (br->br_wbr->wbr_bytes < watermark)
34884 + /* regardless the parent dir */
34885 + err = au_wbr_create_mfsrr(dentry, flags);
34888 + AuDbg("b%d\n", err);
34892 +/* ---------------------------------------------------------------------- */
34894 +/* policies for copyup */
34896 +/* top down parent */
34897 +static int au_wbr_copyup_tdp(struct dentry *dentry)
34899 + return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
34902 +/* bottom up parent */
34903 +static int au_wbr_copyup_bup(struct dentry *dentry)
34906 + aufs_bindex_t bindex, btop;
34907 + struct dentry *parent, *h_parent;
34908 + struct super_block *sb;
34911 + sb = dentry->d_sb;
34912 + parent = dget_parent(dentry);
34913 + btop = au_dbtop(parent);
34914 + for (bindex = au_dbtop(dentry); bindex >= btop; bindex--) {
34915 + h_parent = au_h_dptr(parent, bindex);
34916 + if (!h_parent || d_is_negative(h_parent))
34919 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
34926 + /* bottom up here */
34927 + if (unlikely(err < 0))
34928 + err = au_wbr_bu(sb, btop - 1);
34930 + AuDbg("b%d\n", err);
34935 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop)
34939 + err = au_wbr_bu(dentry->d_sb, btop);
34940 + AuDbg("b%d\n", err);
34942 + err = au_wbr_nonopq(dentry, err);
34944 + AuDbg("b%d\n", err);
34948 +static int au_wbr_copyup_bu(struct dentry *dentry)
34951 + aufs_bindex_t btop;
34953 + btop = au_dbtop(dentry);
34954 + err = au_wbr_do_copyup_bu(dentry, btop);
34958 +/* ---------------------------------------------------------------------- */
34960 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
34961 + [AuWbrCopyup_TDP] = {
34962 + .copyup = au_wbr_copyup_tdp
34964 + [AuWbrCopyup_BUP] = {
34965 + .copyup = au_wbr_copyup_bup
34967 + [AuWbrCopyup_BU] = {
34968 + .copyup = au_wbr_copyup_bu
34972 +struct au_wbr_create_operations au_wbr_create_ops[] = {
34973 + [AuWbrCreate_TDP] = {
34974 + .create = au_wbr_create_tdp
34976 + [AuWbrCreate_RR] = {
34977 + .create = au_wbr_create_rr,
34978 + .init = au_wbr_create_init_rr
34980 + [AuWbrCreate_MFS] = {
34981 + .create = au_wbr_create_mfs,
34982 + .init = au_wbr_create_init_mfs,
34983 + .fin = au_wbr_create_fin_mfs
34985 + [AuWbrCreate_MFSV] = {
34986 + .create = au_wbr_create_mfs,
34987 + .init = au_wbr_create_init_mfs,
34988 + .fin = au_wbr_create_fin_mfs
34990 + [AuWbrCreate_MFSRR] = {
34991 + .create = au_wbr_create_mfsrr,
34992 + .init = au_wbr_create_init_mfsrr,
34993 + .fin = au_wbr_create_fin_mfs
34995 + [AuWbrCreate_MFSRRV] = {
34996 + .create = au_wbr_create_mfsrr,
34997 + .init = au_wbr_create_init_mfsrr,
34998 + .fin = au_wbr_create_fin_mfs
35000 + [AuWbrCreate_TDMFS] = {
35001 + .create = au_wbr_create_tdmfs,
35002 + .init = au_wbr_create_init_mfs,
35003 + .fin = au_wbr_create_fin_mfs
35005 + [AuWbrCreate_TDMFSV] = {
35006 + .create = au_wbr_create_tdmfs,
35007 + .init = au_wbr_create_init_mfs,
35008 + .fin = au_wbr_create_fin_mfs
35010 + [AuWbrCreate_PMFS] = {
35011 + .create = au_wbr_create_pmfs,
35012 + .init = au_wbr_create_init_mfs,
35013 + .fin = au_wbr_create_fin_mfs
35015 + [AuWbrCreate_PMFSV] = {
35016 + .create = au_wbr_create_pmfs,
35017 + .init = au_wbr_create_init_mfs,
35018 + .fin = au_wbr_create_fin_mfs
35020 + [AuWbrCreate_PMFSRR] = {
35021 + .create = au_wbr_create_pmfsrr,
35022 + .init = au_wbr_create_init_mfsrr,
35023 + .fin = au_wbr_create_fin_mfs
35025 + [AuWbrCreate_PMFSRRV] = {
35026 + .create = au_wbr_create_pmfsrr,
35027 + .init = au_wbr_create_init_mfsrr,
35028 + .fin = au_wbr_create_fin_mfs
35031 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
35032 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
35033 +++ linux/fs/aufs/whout.c 2022-03-21 14:49:05.733299677 +0100
35035 +// SPDX-License-Identifier: GPL-2.0
35037 + * Copyright (C) 2005-2021 Junjiro R. Okajima
35039 + * This program, aufs is free software; you can redistribute it and/or modify
35040 + * it under the terms of the GNU General Public License as published by
35041 + * the Free Software Foundation; either version 2 of the License, or
35042 + * (at your option) any later version.
35044 + * This program is distributed in the hope that it will be useful,
35045 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35046 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35047 + * GNU General Public License for more details.
35049 + * You should have received a copy of the GNU General Public License
35050 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35054 + * whiteout for logical deletion and opaque directory
35059 +#define WH_MASK 0444
35062 + * If a directory contains this file, then it is opaque. We start with the
35063 + * .wh. flag so that it is blocked by lookup.
35065 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
35066 + sizeof(AUFS_WH_DIROPQ) - 1);
35069 + * generate whiteout name, which is NOT terminated by NULL.
35070 + * @name: original d_name.name
35071 + * @len: original d_name.len
35072 + * @wh: whiteout qstr
35073 + * returns zero when succeeds, otherwise error.
35074 + * succeeded value as wh->name should be freed by kfree().
35076 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
35080 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
35081 + return -ENAMETOOLONG;
35083 + wh->len = name->len + AUFS_WH_PFX_LEN;
35084 + p = kmalloc(wh->len, GFP_NOFS);
35087 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35088 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
35095 +/* ---------------------------------------------------------------------- */
35098 + * test if the @wh_name exists under @h_ppath.
35099 + * @try_sio specifies the necessary of super-io.
35101 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
35102 + struct qstr *wh_name, int try_sio)
35105 + struct dentry *wh_dentry;
35108 + wh_dentry = vfsub_lkup_one(wh_name, h_ppath);
35110 + wh_dentry = au_sio_lkup_one(h_userns, wh_name, h_ppath);
35111 + err = PTR_ERR(wh_dentry);
35112 + if (IS_ERR(wh_dentry)) {
35113 + if (err == -ENAMETOOLONG)
35119 + if (d_is_negative(wh_dentry))
35120 + goto out_wh; /* success */
35123 + if (d_is_reg(wh_dentry))
35124 + goto out_wh; /* success */
35127 + AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
35128 + wh_dentry, d_inode(wh_dentry)->i_mode);
35137 + * test if the @h_path->dentry sets opaque or not.
35139 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path)
35142 + struct inode *h_dir;
35144 + h_dir = d_inode(h_path->dentry);
35145 + err = au_wh_test(h_userns, h_path, &diropq_name,
35146 + au_test_h_perm_sio(h_userns, h_dir, MAY_EXEC));
35151 + * returns a negative dentry whose name is unique and temporary.
35153 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
35154 + struct qstr *prefix)
35156 + struct dentry *dentry;
35158 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
35160 + /* strict atomic_t is unnecessary here */
35161 + static unsigned short cnt;
35163 + struct path h_ppath;
35164 + struct user_namespace *h_userns;
35166 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
35169 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
35170 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
35171 + dentry = ERR_PTR(-ENAMETOOLONG);
35172 + if (unlikely(qs.len > NAME_MAX))
35174 + dentry = ERR_PTR(-ENOMEM);
35175 + name = kmalloc(qs.len + 1, GFP_NOFS);
35176 + if (unlikely(!name))
35180 + /* doubly whiteout-ed */
35181 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
35182 + p = name + AUFS_WH_PFX_LEN * 2;
35183 + memcpy(p, prefix->name, prefix->len);
35184 + p += prefix->len;
35186 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
35188 + h_ppath.dentry = h_parent;
35189 + h_ppath.mnt = au_br_mnt(br);
35190 + h_userns = au_br_userns(br);
35192 + for (i = 0; i < 3; i++) {
35193 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
35194 + dentry = au_sio_lkup_one(h_userns, &qs, &h_ppath);
35195 + if (IS_ERR(dentry) || d_is_negative(dentry))
35199 + /* pr_warn("could not get random name\n"); */
35200 + dentry = ERR_PTR(-EEXIST);
35201 + AuDbg("%.*s\n", AuLNPair(&qs));
35205 + if (name != defname)
35206 + au_kfree_try_rcu(name);
35208 + AuTraceErrPtr(dentry);
35213 + * rename the @h_dentry on @br to the whiteouted temporary name.
35215 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
35218 + struct path h_path = {
35219 + .mnt = au_br_mnt(br)
35221 + struct inode *h_dir, *delegated;
35222 + struct dentry *h_parent;
35224 + h_parent = h_dentry->d_parent; /* dir inode is locked */
35225 + h_dir = d_inode(h_parent);
35226 + IMustLock(h_dir);
35228 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
35229 + err = PTR_ERR(h_path.dentry);
35230 + if (IS_ERR(h_path.dentry))
35233 + /* under the same dir, no need to lock_rename() */
35234 + delegated = NULL;
35235 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated,
35238 + if (unlikely(err == -EWOULDBLOCK)) {
35239 + pr_warn("cannot retry for NFSv4 delegation"
35240 + " for an internal rename\n");
35243 + dput(h_path.dentry);
35250 +/* ---------------------------------------------------------------------- */
35252 + * functions for removing a whiteout
35255 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
35258 + struct inode *delegated;
35261 + * forces superio when the dir has a sticky bit.
35262 + * this may be a violation of unix fs semantics.
35264 + force = (h_dir->i_mode & S_ISVTX)
35265 + && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
35266 + delegated = NULL;
35267 + err = vfsub_unlink(h_dir, h_path, &delegated, force);
35268 + if (unlikely(err == -EWOULDBLOCK)) {
35269 + pr_warn("cannot retry for NFSv4 delegation"
35270 + " for an internal unlink\n");
35276 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
35277 + struct dentry *dentry)
35281 + err = do_unlink_wh(h_dir, h_path);
35282 + if (!err && dentry)
35283 + au_set_dbwh(dentry, -1);
35288 +static int unlink_wh_name(struct path *h_ppath, struct qstr *wh)
35291 + struct path h_path;
35294 + h_path.dentry = vfsub_lkup_one(wh, h_ppath);
35295 + if (IS_ERR(h_path.dentry))
35296 + err = PTR_ERR(h_path.dentry);
35298 + if (d_is_reg(h_path.dentry)) {
35299 + h_path.mnt = h_ppath->mnt;
35300 + err = do_unlink_wh(d_inode(h_ppath->dentry), &h_path);
35302 + dput(h_path.dentry);
35308 +/* ---------------------------------------------------------------------- */
35310 + * initialize/clean whiteout for a branch
35313 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
35317 + struct inode *delegated;
35319 + if (d_is_negative(whpath->dentry))
35323 + err = vfsub_rmdir(h_dir, whpath);
35325 + delegated = NULL;
35326 + err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
35327 + if (unlikely(err == -EWOULDBLOCK)) {
35328 + pr_warn("cannot retry for NFSv4 delegation"
35329 + " for an internal unlink\n");
35333 + if (unlikely(err))
35334 + pr_warn("failed removing %pd (%d), ignored.\n",
35335 + whpath->dentry, err);
35338 +static int test_linkable(struct dentry *h_root)
35340 + struct inode *h_dir = d_inode(h_root);
35342 + if (h_dir->i_op->link)
35345 + pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
35346 + h_root, au_sbtype(h_root->d_sb));
35347 + return -ENOSYS; /* the branch doesn't have its ->link() */
35350 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
35351 +static int au_whdir(struct inode *h_dir, struct path *path)
35356 + if (d_is_negative(path->dentry)) {
35359 + if (au_test_nfs(path->dentry->d_sb))
35361 + err = vfsub_mkdir(h_dir, path, mode);
35362 + } else if (d_is_dir(path->dentry))
35365 + pr_err("unknown %pd exists\n", path->dentry);
35370 +struct au_wh_base {
35371 + const struct qstr *name;
35372 + struct dentry *dentry;
35375 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
35376 + struct path *h_path)
35378 + h_path->dentry = base[AuBrWh_BASE].dentry;
35379 + au_wh_clean(h_dir, h_path, /*isdir*/0);
35380 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35381 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35382 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35383 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35387 + * returns tri-state,
35388 + * minus: error, caller should print the message
35390 + * plus: error, caller should NOT print the message
35392 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
35393 + int do_plink, struct au_wh_base base[],
35394 + struct path *h_path)
35397 + struct inode *h_dir;
35399 + h_dir = d_inode(h_root);
35400 + h_path->dentry = base[AuBrWh_BASE].dentry;
35401 + au_wh_clean(h_dir, h_path, /*isdir*/0);
35402 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35404 + err = test_linkable(h_root);
35405 + if (unlikely(err)) {
35410 + err = au_whdir(h_dir, h_path);
35411 + if (unlikely(err))
35413 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35415 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35416 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35417 + err = au_whdir(h_dir, h_path);
35418 + if (unlikely(err))
35420 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35427 + * for the moment, aufs supports the branch filesystem which does not support
35428 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
35429 + * copyup failed. finally, such filesystem will not be used as the writable
35432 + * returns tri-state, see above.
35434 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
35435 + int do_plink, struct au_wh_base base[],
35436 + struct path *h_path)
35439 + struct inode *h_dir;
35441 + WbrWhMustWriteLock(wbr);
35443 + err = test_linkable(h_root);
35444 + if (unlikely(err)) {
35450 + * todo: should this create be done in /sbin/mount.aufs helper?
35453 + h_dir = d_inode(h_root);
35454 + if (d_is_negative(base[AuBrWh_BASE].dentry)) {
35455 + h_path->dentry = base[AuBrWh_BASE].dentry;
35456 + err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
35457 + } else if (d_is_reg(base[AuBrWh_BASE].dentry))
35460 + pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
35461 + if (unlikely(err))
35464 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35466 + err = au_whdir(h_dir, h_path);
35467 + if (unlikely(err))
35469 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35471 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35472 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
35474 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35475 + err = au_whdir(h_dir, h_path);
35476 + if (unlikely(err))
35478 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35485 + * initialize the whiteout base file/dir for @br.
35487 +int au_wh_init(struct au_branch *br, struct super_block *sb)
35490 + const unsigned char do_plink
35491 + = !!au_opt_test(au_mntflags(sb), PLINK);
35492 + struct inode *h_dir;
35493 + struct path path = br->br_path;
35494 + struct dentry *h_root = path.dentry;
35495 + struct au_wbr *wbr = br->br_wbr;
35496 + static const struct qstr base_name[] = {
35497 + [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
35498 + sizeof(AUFS_BASE_NAME) - 1),
35499 + [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
35500 + sizeof(AUFS_PLINKDIR_NAME) - 1),
35501 + [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
35502 + sizeof(AUFS_ORPHDIR_NAME) - 1)
35504 + struct au_wh_base base[] = {
35505 + [AuBrWh_BASE] = {
35506 + .name = base_name + AuBrWh_BASE,
35509 + [AuBrWh_PLINK] = {
35510 + .name = base_name + AuBrWh_PLINK,
35513 + [AuBrWh_ORPH] = {
35514 + .name = base_name + AuBrWh_ORPH,
35520 + WbrWhMustWriteLock(wbr);
35522 + for (i = 0; i < AuBrWh_Last; i++) {
35523 + /* doubly whiteouted */
35524 + struct dentry *d;
35526 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
35527 + err = PTR_ERR(d);
35531 + base[i].dentry = d;
35533 + && wbr->wbr_wh[i]
35534 + && wbr->wbr_wh[i] != base[i].dentry);
35538 + for (i = 0; i < AuBrWh_Last; i++) {
35539 + dput(wbr->wbr_wh[i]);
35540 + wbr->wbr_wh[i] = NULL;
35544 + if (!au_br_writable(br->br_perm)) {
35545 + h_dir = d_inode(h_root);
35546 + au_wh_init_ro(h_dir, base, &path);
35547 + } else if (!au_br_wh_linkable(br->br_perm)) {
35548 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
35554 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
35560 + goto out; /* success */
35563 + pr_err("an error(%d) on the writable branch %pd(%s)\n",
35564 + err, h_root, au_sbtype(h_root->d_sb));
35566 + for (i = 0; i < AuBrWh_Last; i++)
35567 + dput(base[i].dentry);
35571 +/* ---------------------------------------------------------------------- */
35573 + * whiteouts are all hard-linked usually.
35574 + * when its link count reaches a ceiling, we create a new whiteout base
35575 + * asynchronously.
35578 +struct reinit_br_wh {
35579 + struct super_block *sb;
35580 + struct au_branch *br;
35583 +static void reinit_br_wh(void *arg)
35586 + aufs_bindex_t bindex;
35587 + struct path h_path;
35588 + struct reinit_br_wh *a = arg;
35589 + struct au_wbr *wbr;
35590 + struct inode *dir, *delegated;
35591 + struct dentry *h_root;
35592 + struct au_hinode *hdir;
35595 + wbr = a->br->br_wbr;
35596 + /* big aufs lock */
35597 + si_noflush_write_lock(a->sb);
35598 + if (!au_br_writable(a->br->br_perm))
35600 + bindex = au_br_index(a->sb, a->br->br_id);
35601 + if (unlikely(bindex < 0))
35604 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
35605 + dir = d_inode(a->sb->s_root);
35606 + hdir = au_hi(dir, bindex);
35607 + h_root = au_h_dptr(a->sb->s_root, bindex);
35608 + AuDebugOn(h_root != au_br_dentry(a->br));
35610 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35611 + wbr_wh_write_lock(wbr);
35612 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
35615 + h_path.dentry = wbr->wbr_whbase;
35616 + h_path.mnt = au_br_mnt(a->br);
35617 + delegated = NULL;
35618 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
35620 + if (unlikely(err == -EWOULDBLOCK)) {
35621 + pr_warn("cannot retry for NFSv4 delegation"
35622 + " for an internal unlink\n");
35626 + pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
35629 + dput(wbr->wbr_whbase);
35630 + wbr->wbr_whbase = NULL;
35632 + err = au_wh_init(a->br, a->sb);
35633 + wbr_wh_write_unlock(wbr);
35634 + au_hn_inode_unlock(hdir);
35635 + di_read_unlock(a->sb->s_root, AuLock_IR);
35637 + au_fhsm_wrote(a->sb, bindex, /*force*/0);
35641 + atomic_dec(&wbr->wbr_wh_running);
35642 + au_lcnt_dec(&a->br->br_count);
35643 + si_write_unlock(a->sb);
35644 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
35646 + if (unlikely(err))
35647 + AuIOErr("err %d\n", err);
35650 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
35652 + int do_dec, wkq_err;
35653 + struct reinit_br_wh *arg;
35656 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
35659 + /* ignore ENOMEM */
35660 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
35663 + * dec(wh_running), kfree(arg) and dec(br_count)
35664 + * in reinit function
35668 + au_lcnt_inc(&br->br_count);
35669 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
35670 + if (unlikely(wkq_err)) {
35671 + atomic_dec(&br->br_wbr->wbr_wh_running);
35672 + au_lcnt_dec(&br->br_count);
35673 + au_kfree_rcu(arg);
35680 + atomic_dec(&br->br_wbr->wbr_wh_running);
35683 +/* ---------------------------------------------------------------------- */
35686 + * create the whiteout @wh.
35688 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
35689 + struct dentry *wh)
35692 + struct path h_path = {
35695 + struct au_branch *br;
35696 + struct au_wbr *wbr;
35697 + struct dentry *h_parent;
35698 + struct inode *h_dir, *delegated;
35700 + h_parent = wh->d_parent; /* dir inode is locked */
35701 + h_dir = d_inode(h_parent);
35702 + IMustLock(h_dir);
35704 + br = au_sbr(sb, bindex);
35705 + h_path.mnt = au_br_mnt(br);
35706 + wbr = br->br_wbr;
35707 + wbr_wh_read_lock(wbr);
35708 + if (wbr->wbr_whbase) {
35709 + delegated = NULL;
35710 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
35711 + if (unlikely(err == -EWOULDBLOCK)) {
35712 + pr_warn("cannot retry for NFSv4 delegation"
35713 + " for an internal link\n");
35716 + if (!err || err != -EMLINK)
35719 + /* link count full. re-initialize br_whbase. */
35720 + kick_reinit_br_wh(sb, br);
35723 + /* return this error in this context */
35724 + err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
35726 + au_fhsm_wrote(sb, bindex, /*force*/0);
35729 + wbr_wh_read_unlock(wbr);
35733 +/* ---------------------------------------------------------------------- */
35736 + * create or remove the diropq.
35738 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
35739 + unsigned int flags)
35741 + struct dentry *opq_dentry;
35742 + struct super_block *sb;
35743 + struct au_branch *br;
35744 + struct path h_path;
35747 + sb = dentry->d_sb;
35748 + br = au_sbr(sb, bindex);
35749 + h_path.dentry = au_h_dptr(dentry, bindex);
35750 + h_path.mnt = au_br_mnt(br);
35751 + opq_dentry = vfsub_lkup_one(&diropq_name, &h_path);
35752 + if (IS_ERR(opq_dentry))
35755 + if (au_ftest_diropq(flags, CREATE)) {
35756 + err = link_or_create_wh(sb, bindex, opq_dentry);
35758 + au_set_dbdiropq(dentry, bindex);
35759 + goto out; /* success */
35762 + h_path.dentry = opq_dentry;
35763 + err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &h_path);
35765 + au_set_dbdiropq(dentry, -1);
35767 + dput(opq_dentry);
35768 + opq_dentry = ERR_PTR(err);
35771 + return opq_dentry;
35774 +struct do_diropq_args {
35775 + struct dentry **errp;
35776 + struct dentry *dentry;
35777 + aufs_bindex_t bindex;
35778 + unsigned int flags;
35781 +static void call_do_diropq(void *args)
35783 + struct do_diropq_args *a = args;
35784 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
35787 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35788 + unsigned int flags)
35790 + struct dentry *diropq, *h_dentry;
35791 + struct user_namespace *h_userns;
35793 + h_userns = au_sbr_userns(dentry->d_sb, bindex);
35794 + h_dentry = au_h_dptr(dentry, bindex);
35795 + if (!au_test_h_perm_sio(h_userns, d_inode(h_dentry),
35796 + MAY_EXEC | MAY_WRITE))
35797 + diropq = do_diropq(dentry, bindex, flags);
35800 + struct do_diropq_args args = {
35802 + .dentry = dentry,
35803 + .bindex = bindex,
35807 + wkq_err = au_wkq_wait(call_do_diropq, &args);
35808 + if (unlikely(wkq_err))
35809 + diropq = ERR_PTR(wkq_err);
35815 +/* ---------------------------------------------------------------------- */
35818 + * lookup whiteout dentry.
35819 + * @h_parent: lower parent dentry which must exist and be locked
35820 + * @base_name: name of dentry which will be whiteouted
35821 + * returns dentry for whiteout.
35823 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35824 + struct au_branch *br)
35827 + struct qstr wh_name;
35828 + struct dentry *wh_dentry;
35829 + struct path h_path;
35831 + err = au_wh_name_alloc(&wh_name, base_name);
35832 + wh_dentry = ERR_PTR(err);
35834 + h_path.dentry = h_parent;
35835 + h_path.mnt = au_br_mnt(br);
35836 + wh_dentry = vfsub_lkup_one(&wh_name, &h_path);
35837 + au_kfree_try_rcu(wh_name.name);
35839 + return wh_dentry;
35843 + * link/create a whiteout for @dentry on @bindex.
35845 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35846 + struct dentry *h_parent)
35848 + struct dentry *wh_dentry;
35849 + struct super_block *sb;
35852 + sb = dentry->d_sb;
35853 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
35854 + if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
35855 + err = link_or_create_wh(sb, bindex, wh_dentry);
35857 + au_set_dbwh(dentry, bindex);
35858 + au_fhsm_wrote(sb, bindex, /*force*/0);
35861 + wh_dentry = ERR_PTR(err);
35865 + return wh_dentry;
35868 +/* ---------------------------------------------------------------------- */
35870 +/* Delete all whiteouts in this directory on branch bindex. */
35871 +static int del_wh_children(struct path *h_path, struct au_nhash *whlist,
35872 + aufs_bindex_t bindex)
35875 + unsigned long ul, n;
35876 + struct qstr wh_name;
35878 + struct hlist_head *head;
35879 + struct au_vdir_wh *pos;
35880 + struct au_vdir_destr *str;
35883 + p = (void *)__get_free_page(GFP_NOFS);
35884 + wh_name.name = p;
35885 + if (unlikely(!wh_name.name))
35889 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35890 + p += AUFS_WH_PFX_LEN;
35891 + n = whlist->nh_num;
35892 + head = whlist->nh_head;
35893 + for (ul = 0; !err && ul < n; ul++, head++) {
35894 + hlist_for_each_entry(pos, head, wh_hash) {
35895 + if (pos->wh_bindex != bindex)
35898 + str = &pos->wh_str;
35899 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
35900 + memcpy(p, str->name, str->len);
35901 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
35902 + err = unlink_wh_name(h_path, &wh_name);
35907 + AuIOErr("whiteout name too long %.*s\n",
35908 + str->len, str->name);
35913 + free_page((unsigned long)wh_name.name);
35919 +struct del_wh_children_args {
35921 + struct path *h_path;
35922 + struct au_nhash *whlist;
35923 + aufs_bindex_t bindex;
35926 +static void call_del_wh_children(void *args)
35928 + struct del_wh_children_args *a = args;
35929 + *a->errp = del_wh_children(a->h_path, a->whlist, a->bindex);
35932 +/* ---------------------------------------------------------------------- */
35934 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
35936 + struct au_whtmp_rmdir *whtmp;
35938 + unsigned int rdhash;
35940 + SiMustAnyLock(sb);
35942 + whtmp = kzalloc(sizeof(*whtmp), gfp);
35943 + if (unlikely(!whtmp)) {
35944 + whtmp = ERR_PTR(-ENOMEM);
35948 + /* no estimation for dir size */
35949 + rdhash = au_sbi(sb)->si_rdhash;
35951 + rdhash = AUFS_RDHASH_DEF;
35952 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
35953 + if (unlikely(err)) {
35954 + au_kfree_rcu(whtmp);
35955 + whtmp = ERR_PTR(err);
35962 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
35965 + au_lcnt_dec(&whtmp->br->br_count);
35966 + dput(whtmp->wh_dentry);
35967 + iput(whtmp->dir);
35968 + au_nhash_wh_free(&whtmp->whlist);
35969 + au_kfree_rcu(whtmp);
35973 + * rmdir the whiteouted temporary named dir @h_dentry.
35974 + * @whlist: whiteouted children.
35976 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
35977 + struct dentry *wh_dentry, struct au_nhash *whlist)
35980 + unsigned int h_nlink;
35981 + struct path wh_path;
35982 + struct inode *wh_inode, *h_dir;
35983 + struct au_branch *br;
35984 + struct user_namespace *h_userns;
35986 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
35987 + IMustLock(h_dir);
35989 + br = au_sbr(dir->i_sb, bindex);
35990 + wh_path.dentry = wh_dentry;
35991 + wh_path.mnt = au_br_mnt(br);
35992 + h_userns = au_br_userns(br);
35993 + wh_inode = d_inode(wh_dentry);
35994 + inode_lock_nested(wh_inode, AuLsc_I_CHILD);
35997 + * someone else might change some whiteouts while we were sleeping.
35998 + * it means this whlist may have an obsoleted entry.
36000 + if (!au_test_h_perm_sio(h_userns, wh_inode, MAY_EXEC | MAY_WRITE))
36001 + err = del_wh_children(&wh_path, whlist, bindex);
36004 + struct del_wh_children_args args = {
36006 + .h_path = &wh_path,
36007 + .whlist = whlist,
36011 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
36012 + if (unlikely(wkq_err))
36015 + inode_unlock(wh_inode);
36018 + h_nlink = h_dir->i_nlink;
36019 + err = vfsub_rmdir(h_dir, &wh_path);
36020 + /* some fs doesn't change the parent nlink in some cases */
36021 + h_nlink -= h_dir->i_nlink;
36025 + if (au_ibtop(dir) == bindex) {
36026 + /* todo: dir->i_mutex is necessary */
36027 + au_cpup_attr_timesizes(dir);
36029 + vfsub_drop_nlink(dir);
36031 + return 0; /* success */
36034 + pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
36038 +static void call_rmdir_whtmp(void *args)
36041 + aufs_bindex_t bindex;
36042 + struct au_whtmp_rmdir *a = args;
36043 + struct super_block *sb;
36044 + struct dentry *h_parent;
36045 + struct inode *h_dir;
36046 + struct au_hinode *hdir;
36048 + /* rmdir by nfsd may cause deadlock with this i_mutex */
36049 + /* inode_lock(a->dir); */
36051 + sb = a->dir->i_sb;
36052 + si_read_lock(sb, !AuLock_FLUSH);
36053 + if (!au_br_writable(a->br->br_perm))
36055 + bindex = au_br_index(sb, a->br->br_id);
36056 + if (unlikely(bindex < 0))
36060 + ii_write_lock_parent(a->dir);
36061 + h_parent = dget_parent(a->wh_dentry);
36062 + h_dir = d_inode(h_parent);
36063 + hdir = au_hi(a->dir, bindex);
36064 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
36065 + if (unlikely(err))
36067 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
36068 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
36071 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
36072 + au_hn_inode_unlock(hdir);
36073 + vfsub_mnt_drop_write(au_br_mnt(a->br));
36077 + ii_write_unlock(a->dir);
36079 + /* inode_unlock(a->dir); */
36080 + au_whtmp_rmdir_free(a);
36081 + si_read_unlock(sb);
36082 + au_nwt_done(&au_sbi(sb)->si_nowait);
36083 + if (unlikely(err))
36084 + AuIOErr("err %d\n", err);
36087 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36088 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
36091 + struct super_block *sb;
36095 + /* all post-process will be done in do_rmdir_whtmp(). */
36097 + args->dir = au_igrab(dir);
36098 + args->br = au_sbr(sb, bindex);
36099 + au_lcnt_inc(&args->br->br_count);
36100 + args->wh_dentry = dget(wh_dentry);
36101 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
36102 + if (unlikely(wkq_err)) {
36103 + pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
36104 + au_whtmp_rmdir_free(args);
36107 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
36108 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
36109 +++ linux/fs/aufs/whout.h 2022-03-21 14:49:05.733299677 +0100
36111 +/* SPDX-License-Identifier: GPL-2.0 */
36113 + * Copyright (C) 2005-2021 Junjiro R. Okajima
36115 + * This program, aufs is free software; you can redistribute it and/or modify
36116 + * it under the terms of the GNU General Public License as published by
36117 + * the Free Software Foundation; either version 2 of the License, or
36118 + * (at your option) any later version.
36120 + * This program is distributed in the hope that it will be useful,
36121 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36122 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36123 + * GNU General Public License for more details.
36125 + * You should have received a copy of the GNU General Public License
36126 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36130 + * whiteout for logical deletion and opaque directory
36133 +#ifndef __AUFS_WHOUT_H__
36134 +#define __AUFS_WHOUT_H__
36141 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
36142 +int au_wh_test(struct user_namespace *h_userns, struct path *h_ppath,
36143 + struct qstr *wh_name, int try_sio);
36144 +int au_diropq_test(struct user_namespace *h_userns, struct path *h_path);
36146 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
36147 + struct qstr *prefix);
36148 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
36149 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
36150 + struct dentry *dentry);
36151 +int au_wh_init(struct au_branch *br, struct super_block *sb);
36153 +/* diropq flags */
36154 +#define AuDiropq_CREATE 1
36155 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
36156 +#define au_fset_diropq(flags, name) \
36157 + do { (flags) |= AuDiropq_##name; } while (0)
36158 +#define au_fclr_diropq(flags, name) \
36159 + do { (flags) &= ~AuDiropq_##name; } while (0)
36161 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
36162 + unsigned int flags);
36163 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
36164 + struct au_branch *br);
36165 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
36166 + struct dentry *h_parent);
36168 +/* real rmdir for the whiteout-ed dir */
36169 +struct au_whtmp_rmdir {
36170 + struct inode *dir;
36171 + struct au_branch *br;
36172 + struct dentry *wh_dentry;
36173 + struct au_nhash whlist;
36176 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
36177 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
36178 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36179 + struct dentry *wh_dentry, struct au_nhash *whlist);
36180 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36181 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
36183 +/* ---------------------------------------------------------------------- */
36185 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
36186 + aufs_bindex_t bindex)
36188 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
36191 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
36193 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
36196 +#endif /* __KERNEL__ */
36197 +#endif /* __AUFS_WHOUT_H__ */
36198 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
36199 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
36200 +++ linux/fs/aufs/wkq.c 2022-03-21 14:49:05.733299677 +0100
36202 +// SPDX-License-Identifier: GPL-2.0
36204 + * Copyright (C) 2005-2021 Junjiro R. Okajima
36206 + * This program, aufs is free software; you can redistribute it and/or modify
36207 + * it under the terms of the GNU General Public License as published by
36208 + * the Free Software Foundation; either version 2 of the License, or
36209 + * (at your option) any later version.
36211 + * This program is distributed in the hope that it will be useful,
36212 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36213 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36214 + * GNU General Public License for more details.
36216 + * You should have received a copy of the GNU General Public License
36217 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36221 + * workqueue for asynchronous/super-io operations
36222 + * todo: try new credential scheme
36225 +#include <linux/module.h>
36228 +/* internal workqueue named AUFS_WKQ_NAME */
36230 +static struct workqueue_struct *au_wkq;
36232 +struct au_wkinfo {
36233 + struct work_struct wk;
36234 + struct kobject *kobj;
36236 + unsigned int flags; /* see wkq.h */
36238 + au_wkq_func_t func;
36241 +#ifdef CONFIG_LOCKDEP
36243 + struct held_lock **hlock;
36246 + struct completion *comp;
36249 +/* ---------------------------------------------------------------------- */
36251 + * Aufs passes some operations to the workqueue such as the internal copyup.
36252 + * This scheme looks rather unnatural for LOCKDEP debugging feature, since the
36253 + * job run by workqueue depends upon the locks acquired in the other task.
36254 + * Delegating a small operation to the workqueue, aufs passes its lockdep
36255 + * information too. And the job in the workqueue restores the info in order to
36256 + * pretend as if it acquired those locks. This is just to make LOCKDEP work
36257 + * correctly and expectedly.
36260 +#ifndef CONFIG_LOCKDEP
36261 +AuStubInt0(au_wkq_lockdep_alloc, struct au_wkinfo *wkinfo);
36262 +AuStubVoid(au_wkq_lockdep_free, struct au_wkinfo *wkinfo);
36263 +AuStubVoid(au_wkq_lockdep_pre, struct au_wkinfo *wkinfo);
36264 +AuStubVoid(au_wkq_lockdep_post, struct au_wkinfo *wkinfo);
36265 +AuStubVoid(au_wkq_lockdep_init, struct au_wkinfo *wkinfo);
36267 +static void au_wkq_lockdep_init(struct au_wkinfo *wkinfo)
36269 + wkinfo->hlock = NULL;
36270 + wkinfo->dont_check = 0;
36277 +static int au_wkq_lockdep_test(struct lock_class_key *key, const char *name)
36279 + static DEFINE_SPINLOCK(spin);
36282 + struct lock_class_key *key;
36284 + { .name = "&sbinfo->si_rwsem" },
36285 + { .name = "&finfo->fi_rwsem" },
36286 + { .name = "&dinfo->di_rwsem" },
36287 + { .name = "&iinfo->ii_rwsem" }
36292 + /* lockless read from 'set.' see below */
36293 + if (set == ARRAY_SIZE(a)) {
36294 + for (i = 0; i < ARRAY_SIZE(a); i++)
36295 + if (a[i].key == key)
36300 + spin_lock(&spin);
36302 + for (i = 0; i < ARRAY_SIZE(a); i++)
36303 + if (a[i].key == key) {
36304 + spin_unlock(&spin);
36307 + for (i = 0; i < ARRAY_SIZE(a); i++) {
36309 + if (unlikely(a[i].key == key)) { /* rare but possible */
36310 + spin_unlock(&spin);
36315 + if (strstr(a[i].name, name)) {
36317 + * the order of these three lines is important for the
36318 + * lockless read above.
36321 + spin_unlock(&spin);
36323 + /* AuDbg("%d, %s\n", set, name); */
36327 + spin_unlock(&spin);
36336 +static int au_wkq_lockdep_alloc(struct au_wkinfo *wkinfo)
36339 + struct task_struct *curr;
36340 + struct held_lock **hl, *held_locks, *p;
36344 + wkinfo->dont_check = lockdep_recursing(curr);
36345 + if (wkinfo->dont_check)
36347 + n = curr->lockdep_depth;
36352 + wkinfo->hlock = kmalloc_array(n + 1, sizeof(*wkinfo->hlock), GFP_NOFS);
36353 + if (unlikely(!wkinfo->hlock))
36357 +#if 0 /* left for debugging */
36358 + if (0 && au_debug_test())
36359 + lockdep_print_held_locks(curr);
36361 + held_locks = curr->held_locks;
36362 + hl = wkinfo->hlock;
36364 + p = held_locks++;
36365 + if (au_wkq_lockdep_test(p->instance->key, p->instance->name))
36374 +static void au_wkq_lockdep_free(struct au_wkinfo *wkinfo)
36376 + au_kfree_try_rcu(wkinfo->hlock);
36379 +static void au_wkq_lockdep_pre(struct au_wkinfo *wkinfo)
36381 + struct held_lock *p, **hl = wkinfo->hlock;
36384 + if (wkinfo->dont_check)
36388 + while ((p = *hl++)) { /* assignment */
36389 + subclass = lockdep_hlock_class(p)->subclass;
36390 + /* AuDbg("%s, %d\n", p->instance->name, subclass); */
36392 + rwsem_acquire_read(p->instance, subclass, 0,
36393 + /*p->acquire_ip*/_RET_IP_);
36395 + rwsem_acquire(p->instance, subclass, 0,
36396 + /*p->acquire_ip*/_RET_IP_);
36400 +static void au_wkq_lockdep_post(struct au_wkinfo *wkinfo)
36402 + struct held_lock *p, **hl = wkinfo->hlock;
36404 + if (wkinfo->dont_check)
36408 + while ((p = *hl++)) /* assignment */
36409 + rwsem_release(p->instance, /*p->acquire_ip*/_RET_IP_);
36413 +static void wkq_func(struct work_struct *wk)
36415 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
36417 + AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
36418 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
36420 + au_wkq_lockdep_pre(wkinfo);
36421 + wkinfo->func(wkinfo->args);
36422 + au_wkq_lockdep_post(wkinfo);
36423 + if (au_ftest_wkq(wkinfo->flags, WAIT))
36424 + complete(wkinfo->comp);
36426 + kobject_put(wkinfo->kobj);
36427 + module_put(THIS_MODULE); /* todo: ?? */
36428 + au_kfree_rcu(wkinfo);
36433 + * Since struct completion is large, try allocating it dynamically.
36435 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
36437 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
36439 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
36441 + init_completion(*comp);
36442 + wkinfo->comp = *comp;
36448 +static void au_wkq_comp_free(struct completion *comp)
36450 + au_kfree_rcu(comp);
36453 +static void au_wkq_run(struct au_wkinfo *wkinfo)
36455 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
36456 + if (au_wkq_test()) {
36457 + AuWarn1("wkq from wkq, unless silly-rename on NFS,"
36458 + " due to a dead dir by UDBA,"
36459 + " or async xino write?\n");
36460 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
36463 + au_dbg_verify_kthread();
36465 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
36466 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
36467 + queue_work(au_wkq, &wkinfo->wk);
36469 + INIT_WORK(&wkinfo->wk, wkq_func);
36470 + schedule_work(&wkinfo->wk);
36475 + * Be careful. It is easy to make deadlock happen.
36476 + * processA: lock, wkq and wait
36477 + * processB: wkq and wait, lock in wkq
36480 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
36483 + AuWkqCompDeclare(comp);
36484 + struct au_wkinfo wkinfo = {
36490 + err = au_wkq_comp_alloc(&wkinfo, &comp);
36491 + if (unlikely(err))
36493 + err = au_wkq_lockdep_alloc(&wkinfo);
36494 + if (unlikely(err))
36497 + au_wkq_run(&wkinfo);
36498 + /* no timeout, no interrupt */
36499 + wait_for_completion(wkinfo.comp);
36501 + au_wkq_lockdep_free(&wkinfo);
36504 + au_wkq_comp_free(comp);
36506 + destroy_work_on_stack(&wkinfo.wk);
36511 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
36512 + * problem in a concurrent umounting.
36514 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36515 + unsigned int flags)
36518 + struct au_wkinfo *wkinfo;
36520 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
36523 + * wkq_func() must free this wkinfo.
36524 + * it highly depends upon the implementation of workqueue.
36527 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
36529 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
36530 + wkinfo->flags = flags & ~AuWkq_WAIT;
36531 + wkinfo->func = func;
36532 + wkinfo->args = args;
36533 + wkinfo->comp = NULL;
36534 + au_wkq_lockdep_init(wkinfo);
36535 + kobject_get(wkinfo->kobj);
36536 + __module_get(THIS_MODULE); /* todo: ?? */
36538 + au_wkq_run(wkinfo);
36541 + au_nwt_done(&au_sbi(sb)->si_nowait);
36547 +/* ---------------------------------------------------------------------- */
36549 +void au_nwt_init(struct au_nowait_tasks *nwt)
36551 + atomic_set(&nwt->nw_len, 0);
36552 + /* smp_mb(); */ /* atomic_set */
36553 + init_waitqueue_head(&nwt->nw_wq);
36556 +void au_wkq_fin(void)
36558 + destroy_workqueue(au_wkq);
36561 +int __init au_wkq_init(void)
36566 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
36567 + if (IS_ERR(au_wkq))
36568 + err = PTR_ERR(au_wkq);
36569 + else if (!au_wkq)
36574 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
36575 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
36576 +++ linux/fs/aufs/wkq.h 2022-03-21 14:49:05.733299677 +0100
36578 +/* SPDX-License-Identifier: GPL-2.0 */
36580 + * Copyright (C) 2005-2021 Junjiro R. Okajima
36582 + * This program, aufs is free software; you can redistribute it and/or modify
36583 + * it under the terms of the GNU General Public License as published by
36584 + * the Free Software Foundation; either version 2 of the License, or
36585 + * (at your option) any later version.
36587 + * This program is distributed in the hope that it will be useful,
36588 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36589 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36590 + * GNU General Public License for more details.
36592 + * You should have received a copy of the GNU General Public License
36593 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36597 + * workqueue for asynchronous/super-io operations
36598 + * todo: try new credentials management scheme
36601 +#ifndef __AUFS_WKQ_H__
36602 +#define __AUFS_WKQ_H__
36606 +#include <linux/wait.h>
36608 +struct super_block;
36610 +/* ---------------------------------------------------------------------- */
36613 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
36615 +struct au_nowait_tasks {
36617 + wait_queue_head_t nw_wq;
36620 +/* ---------------------------------------------------------------------- */
36622 +typedef void (*au_wkq_func_t)(void *args);
36625 +#define AuWkq_WAIT 1
36626 +#define AuWkq_NEST (1 << 1)
36627 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
36628 +#define au_fset_wkq(flags, name) \
36629 + do { (flags) |= AuWkq_##name; } while (0)
36630 +#define au_fclr_wkq(flags, name) \
36631 + do { (flags) &= ~AuWkq_##name; } while (0)
36634 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
36635 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36636 + unsigned int flags);
36637 +void au_nwt_init(struct au_nowait_tasks *nwt);
36638 +int __init au_wkq_init(void);
36639 +void au_wkq_fin(void);
36641 +/* ---------------------------------------------------------------------- */
36643 +static inline int au_wkq_test(void)
36645 + return current->flags & PF_WQ_WORKER;
36648 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
36650 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
36653 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
36655 + if (atomic_dec_and_test(&nwt->nw_len))
36656 + wake_up_all(&nwt->nw_wq);
36659 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
36661 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
36665 +#endif /* __KERNEL__ */
36666 +#endif /* __AUFS_WKQ_H__ */
36667 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
36668 --- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
36669 +++ linux/fs/aufs/xattr.c 2022-03-21 14:49:05.733299677 +0100
36671 +// SPDX-License-Identifier: GPL-2.0
36673 + * Copyright (C) 2014-2021 Junjiro R. Okajima
36675 + * This program, aufs is free software; you can redistribute it and/or modify
36676 + * it under the terms of the GNU General Public License as published by
36677 + * the Free Software Foundation; either version 2 of the License, or
36678 + * (at your option) any later version.
36680 + * This program is distributed in the hope that it will be useful,
36681 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36682 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36683 + * GNU General Public License for more details.
36685 + * You should have received a copy of the GNU General Public License
36686 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36690 + * handling xattr functions
36693 +#include <linux/fs.h>
36694 +#include <linux/posix_acl_xattr.h>
36695 +#include <linux/xattr.h>
36698 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
36700 + if (!ignore_flags)
36708 + if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
36713 +#define cmp(brattr, prefix) do { \
36714 + if (!strncmp(name, XATTR_##prefix##_PREFIX, \
36715 + XATTR_##prefix##_PREFIX_LEN)) { \
36716 + if (ignore_flags & AuBrAttr_ICEX_##brattr) \
36722 + cmp(SEC, SECURITY);
36723 + cmp(SYS, SYSTEM);
36724 + cmp(TR, TRUSTED);
36728 + if (ignore_flags & AuBrAttr_ICEX_OTH)
36735 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
36737 +static int au_do_cpup_xattr(struct path *h_dst, struct path *h_src,
36738 + char *name, char **buf, unsigned int ignore_flags,
36739 + unsigned int verbose)
36743 + struct inode *h_idst;
36744 + struct dentry *h_dst_dentry, *h_src_dentry;
36745 + struct user_namespace *h_dst_userns, *h_src_userns;
36747 + h_src_userns = mnt_user_ns(h_src->mnt);
36748 + h_src_dentry = h_src->dentry;
36749 + ssz = vfs_getxattr_alloc(h_src_userns, h_src_dentry, name, buf, 0,
36752 + if (unlikely(err <= 0)) {
36753 + if (err == -ENODATA
36754 + || (err == -EOPNOTSUPP
36755 + && ((ignore_flags & au_xattr_out_of_list)
36756 + || (au_test_nfs_noacl(d_inode(h_src_dentry))
36757 + && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
36759 + XATTR_NAME_POSIX_ACL_DEFAULT))))
36762 + if (err && (verbose || au_debug_test()))
36763 + pr_err("%s, err %d\n", name, err);
36767 + /* unlock it temporary */
36768 + h_dst_userns = mnt_user_ns(h_dst->mnt);
36769 + h_dst_dentry = h_dst->dentry;
36770 + h_idst = d_inode(h_dst_dentry);
36771 + inode_unlock(h_idst);
36772 + err = vfsub_setxattr(h_dst_userns, h_dst_dentry, name, *buf, ssz,
36774 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36775 + if (unlikely(err)) {
36776 + if (verbose || au_debug_test())
36777 + pr_err("%s, err %d\n", name, err);
36778 + err = au_xattr_ignore(err, name, ignore_flags);
36785 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
36786 + unsigned int verbose)
36788 + int err, unlocked, acl_access, acl_default;
36790 + struct dentry *h_dst_dentry, *h_src_dentry;
36791 + struct inode *h_isrc, *h_idst;
36792 + char *value, *p, *o, *e;
36794 + /* try stopping to update the source inode while we are referencing */
36795 + /* there should not be the parent-child relationship between them */
36796 + h_dst_dentry = h_dst->dentry;
36797 + h_idst = d_inode(h_dst_dentry);
36798 + h_src_dentry = h_src->dentry;
36799 + h_isrc = d_inode(h_src_dentry);
36800 + inode_unlock(h_idst);
36801 + inode_lock_shared_nested(h_isrc, AuLsc_I_CHILD);
36802 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36805 + /* some filesystems don't list POSIX ACL, for example tmpfs */
36806 + ssz = vfs_listxattr(h_src_dentry, NULL, 0);
36808 + if (unlikely(err < 0)) {
36810 + if (err == -ENODATA
36811 + || err == -EOPNOTSUPP)
36812 + err = 0; /* ignore */
36821 + p = kmalloc(ssz, GFP_NOFS);
36823 + if (unlikely(!p))
36825 + err = vfs_listxattr(h_src_dentry, p, ssz);
36827 + inode_unlock_shared(h_isrc);
36829 + AuDbg("err %d, ssz %zd\n", err, ssz);
36830 + if (unlikely(err < 0))
36838 + while (!err && p < e) {
36839 + acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
36840 + sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
36841 + acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
36842 + sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
36844 + err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
36846 + p += strlen(p) + 1;
36849 + ignore_flags |= au_xattr_out_of_list;
36850 + if (!err && !acl_access) {
36851 + err = au_do_cpup_xattr(h_dst, h_src,
36852 + XATTR_NAME_POSIX_ACL_ACCESS, &value,
36853 + ignore_flags, verbose);
36856 + if (!err && !acl_default) {
36857 + err = au_do_cpup_xattr(h_dst, h_src,
36858 + XATTR_NAME_POSIX_ACL_DEFAULT, &value,
36859 + ignore_flags, verbose);
36863 + au_kfree_try_rcu(value);
36866 + au_kfree_try_rcu(o);
36869 + inode_unlock_shared(h_isrc);
36874 +/* ---------------------------------------------------------------------- */
36876 +static int au_smack_reentering(struct super_block *sb)
36878 +#if IS_ENABLED(CONFIG_SECURITY_SMACK) || IS_ENABLED(CONFIG_SECURITY_SELINUX)
36880 + * as a part of lookup, smack_d_instantiate() is called, and it calls
36881 + * i_op->getxattr(). ouch.
36883 + return si_pid_test(sb);
36894 +struct au_lgxattr {
36902 + const char *name;
36909 +static ssize_t au_lgxattr(struct dentry *dentry, struct inode *inode,
36910 + struct au_lgxattr *arg)
36914 + struct path h_path;
36915 + struct super_block *sb;
36917 + sb = dentry->d_sb;
36918 + reenter = au_smack_reentering(sb);
36920 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
36921 + if (unlikely(err))
36924 + err = au_h_path_getattr(dentry, inode, /*force*/1, &h_path, reenter);
36925 + if (unlikely(err))
36927 + if (unlikely(!h_path.dentry))
36928 + /* illegally overlapped or something */
36929 + goto out_di; /* pretending success */
36931 + /* always topmost entry only */
36932 + switch (arg->type) {
36933 + case AU_XATTR_LIST:
36934 + err = vfs_listxattr(h_path.dentry,
36935 + arg->u.list.list, arg->u.list.size);
36937 + case AU_XATTR_GET:
36938 + AuDebugOn(d_is_negative(h_path.dentry));
36939 + err = vfs_getxattr(mnt_user_ns(h_path.mnt), h_path.dentry,
36940 + arg->u.get.name, arg->u.get.value,
36941 + arg->u.get.size);
36947 + di_read_unlock(dentry, AuLock_IR);
36950 + si_read_unlock(sb);
36956 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
36958 + struct au_lgxattr arg = {
36959 + .type = AU_XATTR_LIST,
36966 + return au_lgxattr(dentry, /*inode*/NULL, &arg);
36969 +static ssize_t au_getxattr(struct dentry *dentry, struct inode *inode,
36970 + const char *name, void *value, size_t size)
36972 + struct au_lgxattr arg = {
36973 + .type = AU_XATTR_GET,
36981 + return au_lgxattr(dentry, inode, &arg);
36984 +static int au_setxattr(struct dentry *dentry, struct inode *inode,
36985 + const char *name, const void *value, size_t size,
36988 + struct au_sxattr arg = {
36989 + .type = AU_XATTR_SET,
36998 + return au_sxattr(dentry, inode, &arg);
37001 +/* ---------------------------------------------------------------------- */
37003 +static int au_xattr_get(const struct xattr_handler *handler,
37004 + struct dentry *dentry, struct inode *inode,
37005 + const char *name, void *buffer, size_t size)
37007 + return au_getxattr(dentry, inode, name, buffer, size);
37010 +static int au_xattr_set(const struct xattr_handler *handler,
37011 + struct user_namespace *userns,
37012 + struct dentry *dentry, struct inode *inode,
37013 + const char *name, const void *value, size_t size,
37016 + return au_setxattr(dentry, inode, name, value, size, flags);
37019 +static const struct xattr_handler au_xattr_handler = {
37022 + .get = au_xattr_get,
37023 + .set = au_xattr_set
37026 +static const struct xattr_handler *au_xattr_handlers[] = {
37027 +#ifdef CONFIG_FS_POSIX_ACL
37028 + &posix_acl_access_xattr_handler,
37029 + &posix_acl_default_xattr_handler,
37031 + &au_xattr_handler, /* must be last */
37035 +void au_xattr_init(struct super_block *sb)
37037 + sb->s_xattr = au_xattr_handlers;
37039 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
37040 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
37041 +++ linux/fs/aufs/xino.c 2022-03-21 14:49:05.733299677 +0100
37043 +// SPDX-License-Identifier: GPL-2.0
37045 + * Copyright (C) 2005-2021 Junjiro R. Okajima
37047 + * This program, aufs is free software; you can redistribute it and/or modify
37048 + * it under the terms of the GNU General Public License as published by
37049 + * the Free Software Foundation; either version 2 of the License, or
37050 + * (at your option) any later version.
37052 + * This program is distributed in the hope that it will be useful,
37053 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
37054 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
37055 + * GNU General Public License for more details.
37057 + * You should have received a copy of the GNU General Public License
37058 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
37062 + * external inode number translation table and bitmap
37064 + * things to consider
37066 + * + au_xino object
37067 + * + XINO files (xino, xib, xigen)
37068 + * + dynamic debugfs entries (xiN)
37069 + * + static debugfs entries (xib, xigen)
37070 + * + static sysfs entry (xi_path)
37071 + * - several entry points to handle them.
37072 + * + mount(2) without xino option (default)
37073 + * + mount(2) with xino option
37074 + * + mount(2) with noxino option
37076 + * + remount with add/del branches
37077 + * + remount with xino/noxino options
37080 +#include <linux/seq_file.h>
37081 +#include <linux/statfs.h>
37084 +static aufs_bindex_t sbr_find_shared(struct super_block *sb, aufs_bindex_t btop,
37085 + aufs_bindex_t bbot,
37086 + struct super_block *h_sb)
37088 + /* todo: try binary-search if the branches are many */
37089 + for (; btop <= bbot; btop++)
37090 + if (h_sb == au_sbr_sb(sb, btop))
37096 + * find another branch who is on the same filesystem of the specified
37097 + * branch{@btgt}. search until @bbot.
37099 +static aufs_bindex_t is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
37100 + aufs_bindex_t bbot)
37102 + aufs_bindex_t bindex;
37103 + struct super_block *tgt_sb;
37105 + tgt_sb = au_sbr_sb(sb, btgt);
37106 + bindex = sbr_find_shared(sb, /*btop*/0, btgt - 1, tgt_sb);
37108 + bindex = sbr_find_shared(sb, btgt + 1, bbot, tgt_sb);
37113 +/* ---------------------------------------------------------------------- */
37116 + * stop unnecessary notify events at creating xino files
37119 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry)
37121 + aufs_bindex_t bfound, bindex, bbot;
37122 + struct dentry *parent;
37123 + struct au_branch *br;
37126 + parent = dentry->d_parent; /* safe d_parent access */
37127 + bbot = au_sbbot(sb);
37128 + for (bindex = 0; bindex <= bbot; bindex++) {
37129 + br = au_sbr(sb, bindex);
37130 + if (au_br_dentry(br) == parent) {
37136 + AuDbg("bfound b%d\n", bfound);
37140 +struct au_xino_lock_dir {
37141 + struct au_hinode *hdir;
37142 + struct dentry *parent;
37143 + struct inode *dir;
37146 +static struct dentry *au_dget_parent_lock(struct dentry *dentry,
37147 + unsigned int lsc)
37149 + struct dentry *parent;
37150 + struct inode *dir;
37152 + parent = dget_parent(dentry);
37153 + dir = d_inode(parent);
37154 + inode_lock_nested(dir, lsc);
37155 +#if 0 /* it should not happen */
37156 + spin_lock(&dentry->d_lock);
37157 + if (unlikely(dentry->d_parent != parent)) {
37158 + spin_unlock(&dentry->d_lock);
37159 + inode_unlock(dir);
37164 + spin_unlock(&dentry->d_lock);
37171 +static void au_xino_lock_dir(struct super_block *sb, struct path *xipath,
37172 + struct au_xino_lock_dir *ldir)
37174 + aufs_bindex_t bindex;
37176 + ldir->hdir = NULL;
37177 + bindex = au_xi_root(sb, xipath->dentry);
37178 + if (bindex >= 0) {
37179 + /* rw branch root */
37180 + ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
37181 + au_hn_inode_lock_nested(ldir->hdir, AuLsc_I_PARENT);
37184 + ldir->parent = au_dget_parent_lock(xipath->dentry,
37186 + ldir->dir = d_inode(ldir->parent);
37190 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
37193 + au_hn_inode_unlock(ldir->hdir);
37195 + inode_unlock(ldir->dir);
37196 + dput(ldir->parent);
37200 +/* ---------------------------------------------------------------------- */
37203 + * create and set a new xino file
37205 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
37208 + struct file *file;
37209 + struct dentry *h_parent, *d;
37210 + struct inode *h_dir, *inode;
37212 + static DEFINE_MUTEX(mtx);
37215 + * at mount-time, and the xino file is the default path,
37216 + * hnotify is disabled so we have no notify events to ignore.
37217 + * when a user specified the xino, we cannot get au_hdir to be ignored.
37220 + mutex_lock(&mtx);
37221 + file = vfsub_filp_open(fpath, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37222 + /* | __FMODE_NONOTIFY */,
37224 + if (IS_ERR(file)) {
37226 + mutex_unlock(&mtx);
37228 + pr_err("open %s(%ld)\n", fpath, PTR_ERR(file));
37232 + /* keep file count */
37234 + d = file->f_path.dentry;
37235 + h_parent = au_dget_parent_lock(d, AuLsc_I_PARENT);
37237 + mutex_unlock(&mtx);
37238 + /* mnt_want_write() is unnecessary here */
37239 + h_dir = d_inode(h_parent);
37240 + inode = file_inode(file);
37241 + /* no delegation since it is just created */
37242 + if (inode->i_nlink)
37243 + err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
37245 + inode_unlock(h_dir);
37247 + if (unlikely(err)) {
37249 + pr_err("unlink %s(%d)\n", fpath, err);
37254 + if (unlikely(sb && sb == d->d_sb)) {
37256 + pr_err("%s must be outside\n", fpath);
37259 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
37261 + pr_err("xino doesn't support %s(%s)\n",
37262 + fpath, au_sbtype(d->d_sb));
37265 + return file; /* success */
37269 + file = ERR_PTR(err);
37274 + * create a new xinofile at the same place/path as @base.
37276 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
37277 + struct file *copy_src)
37279 + struct file *file;
37280 + struct dentry *dentry;
37281 + struct inode *dir, *delegated;
37282 + struct qstr *name;
37283 + struct path ppath, path;
37284 + int err, do_unlock;
37285 + struct au_xino_lock_dir ldir;
37288 + au_xino_lock_dir(sb, base, &ldir);
37289 + dentry = base->dentry;
37290 + ppath.dentry = dentry->d_parent; /* dir inode is locked */
37291 + ppath.mnt = base->mnt;
37292 + dir = d_inode(ppath.dentry);
37295 + name = &dentry->d_name;
37296 + path.dentry = vfsub_lookup_one_len(name->name, &ppath, name->len);
37297 + if (IS_ERR(path.dentry)) {
37298 + file = (void *)path.dentry;
37299 + pr_err("%pd lookup err %ld\n", dentry, PTR_ERR(path.dentry));
37303 + /* no need to mnt_want_write() since we call dentry_open() later */
37304 + err = vfs_create(mnt_user_ns(base->mnt), dir, path.dentry, 0666, NULL);
37305 + if (unlikely(err)) {
37306 + file = ERR_PTR(err);
37307 + pr_err("%pd create err %d\n", dentry, err);
37311 + path.mnt = base->mnt;
37312 + file = vfsub_dentry_open(&path,
37313 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37314 + /* | __FMODE_NONOTIFY */);
37315 + if (IS_ERR(file)) {
37316 + pr_err("%pd open err %ld\n", dentry, PTR_ERR(file));
37320 + delegated = NULL;
37321 + err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
37322 + au_xino_unlock_dir(&ldir);
37324 + if (unlikely(err == -EWOULDBLOCK)) {
37325 + pr_warn("cannot retry for NFSv4 delegation"
37326 + " for an internal unlink\n");
37329 + if (unlikely(err)) {
37330 + pr_err("%pd unlink err %d\n", dentry, err);
37335 + /* no one can touch copy_src xino */
37336 + err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
37337 + if (unlikely(err)) {
37338 + pr_err("%pd copy err %d\n", dentry, err);
37342 + goto out_dput; /* success */
37346 + file = ERR_PTR(err);
37348 + dput(path.dentry);
37351 + au_xino_unlock_dir(&ldir);
37355 +struct file *au_xino_file1(struct au_xino *xi)
37357 + struct file *file;
37358 + unsigned int u, nfile;
37361 + nfile = xi->xi_nfile;
37362 + for (u = 0; u < nfile; u++) {
37363 + file = xi->xi_file[u];
37371 +static int au_xino_file_set(struct au_xino *xi, int idx, struct file *file)
37382 + if (idx < xi->xi_nfile) {
37383 + f = xi->xi_file[idx];
37387 + p = au_kzrealloc(xi->xi_file,
37388 + sizeof(*xi->xi_file) * xi->xi_nfile,
37389 + sizeof(*xi->xi_file) * (idx + 1),
37390 + GFP_NOFS, /*may_shrink*/0);
37392 + MtxMustLock(&xi->xi_mtx);
37394 + xi->xi_nfile = idx + 1;
37402 + xi->xi_file[idx] = file;
37409 + * if @xinew->xi is not set, then create new xigen file.
37411 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew)
37413 + struct file *file;
37416 + SiMustAnyLock(sb);
37418 + file = au_xino_create2(sb, xinew->base, xinew->copy_src);
37419 + if (IS_ERR(file)) {
37420 + err = PTR_ERR(file);
37421 + pr_err("%s[%d], err %d\n",
37422 + xinew->xi ? "xino" : "xigen",
37423 + xinew->idx, err);
37428 + err = au_xino_file_set(xinew->xi, xinew->idx, file);
37431 + /* todo: make xigen file an array */
37432 + /* err = au_xigen_file_set(sb, xinew->idx, file); */
37435 + if (unlikely(err))
37436 + file = ERR_PTR(err);
37442 +/* ---------------------------------------------------------------------- */
37445 + * truncate xino files
37447 +static int au_xino_do_trunc(struct super_block *sb, aufs_bindex_t bindex,
37448 + int idx, struct kstatfs *st)
37452 + struct file *file, *new_xino;
37453 + struct au_xi_new xinew = {
37458 + xinew.xi = au_sbr(sb, bindex)->br_xino;
37459 + file = au_xino_file(xinew.xi, idx);
37463 + xinew.base = &file->f_path;
37464 + err = vfs_statfs(xinew.base, st);
37465 + if (unlikely(err)) {
37466 + AuErr1("statfs err %d, ignored\n", err);
37471 + blocks = file_inode(file)->i_blocks;
37472 + pr_info("begin truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37473 + bindex, idx, (u64)blocks, st->f_bfree, st->f_blocks);
37475 + xinew.copy_src = file;
37476 + new_xino = au_xi_new(sb, &xinew);
37477 + if (IS_ERR(new_xino)) {
37478 + err = PTR_ERR(new_xino);
37479 + pr_err("xino(b%d-%d), err %d, ignored\n", bindex, idx, err);
37483 + err = vfs_statfs(&new_xino->f_path, st);
37485 + pr_info("end truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37486 + bindex, idx, (u64)file_inode(new_xino)->i_blocks,
37487 + st->f_bfree, st->f_blocks);
37489 + AuErr1("statfs err %d, ignored\n", err);
37497 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin)
37500 + unsigned long jiffy;
37501 + aufs_bindex_t bbot;
37502 + struct kstatfs *st;
37503 + struct au_branch *br;
37504 + struct au_xino *xi;
37507 + st = kmalloc(sizeof(*st), GFP_NOFS);
37508 + if (unlikely(!st))
37512 + bbot = au_sbbot(sb);
37513 + if (unlikely(bindex < 0 || bbot < bindex))
37518 + br = au_sbr(sb, bindex);
37519 + xi = br->br_xino;
37520 + for (i = idx_begin; !err && i < xi->xi_nfile; i++)
37521 + err = au_xino_do_trunc(sb, bindex, i, st);
37523 + au_sbi(sb)->si_xino_jiffy = jiffy;
37526 + au_kfree_rcu(st);
37531 +struct xino_do_trunc_args {
37532 + struct super_block *sb;
37533 + struct au_branch *br;
37537 +static void xino_do_trunc(void *_args)
37539 + struct xino_do_trunc_args *args = _args;
37540 + struct super_block *sb;
37541 + struct au_branch *br;
37542 + struct inode *dir;
37544 + aufs_bindex_t bindex;
37548 + dir = d_inode(sb->s_root);
37552 + si_noflush_write_lock(sb);
37553 + ii_read_lock_parent(dir);
37554 + bindex = au_br_index(sb, br->br_id);
37555 + err = au_xino_trunc(sb, bindex, idx);
37556 + ii_read_unlock(dir);
37557 + if (unlikely(err))
37558 + pr_warn("err b%d, (%d)\n", bindex, err);
37559 + atomic_dec(&br->br_xino->xi_truncating);
37560 + au_lcnt_dec(&br->br_count);
37561 + si_write_unlock(sb);
37562 + au_nwt_done(&au_sbi(sb)->si_nowait);
37563 + au_kfree_rcu(args);
37567 + * returns the index in the xi_file array whose corresponding file is necessary
37568 + * to truncate, or -1 which means no need to truncate.
37570 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
37574 + struct kstatfs st;
37575 + struct au_sbinfo *sbinfo;
37576 + struct au_xino *xi;
37577 + struct file *file;
37579 + /* todo: si_xino_expire and the ratio should be customizable */
37580 + sbinfo = au_sbi(sb);
37581 + if (time_before(jiffies,
37582 + sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
37585 + /* truncation border */
37586 + xi = br->br_xino;
37587 + for (u = 0; u < xi->xi_nfile; u++) {
37588 + file = au_xino_file(xi, u);
37592 + err = vfs_statfs(&file->f_path, &st);
37593 + if (unlikely(err)) {
37594 + AuErr1("statfs err %d, ignored\n", err);
37597 + if (div64_u64(st.f_bfree * 100, st.f_blocks)
37598 + >= AUFS_XINO_DEF_TRUNC)
37605 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
37608 + struct xino_do_trunc_args *args;
37611 + idx = xino_trunc_test(sb, br);
37615 + if (atomic_inc_return(&br->br_xino->xi_truncating) > 1)
37618 + /* lock and kfree() will be called in trunc_xino() */
37619 + args = kmalloc(sizeof(*args), GFP_NOFS);
37620 + if (unlikely(!args)) {
37621 + AuErr1("no memory\n");
37625 + au_lcnt_inc(&br->br_count);
37629 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
37631 + return; /* success */
37633 + pr_err("wkq %d\n", wkq_err);
37634 + au_lcnt_dec(&br->br_count);
37635 + au_kfree_rcu(args);
37638 + atomic_dec(&br->br_xino->xi_truncating);
37641 +/* ---------------------------------------------------------------------- */
37643 +struct au_xi_calc {
37648 +static void au_xi_calc(struct super_block *sb, ino_t h_ino,
37649 + struct au_xi_calc *calc)
37653 + maxent = au_xi_maxent(sb);
37654 + calc->idx = div64_u64_rem(h_ino, maxent, &calc->pos);
37655 + calc->pos *= sizeof(ino_t);
37658 +static int au_xino_do_new_async(struct super_block *sb, struct au_branch *br,
37659 + struct au_xi_calc *calc)
37662 + struct file *file;
37663 + struct au_xino *xi = br->br_xino;
37664 + struct au_xi_new xinew = {
37668 + SiMustAnyLock(sb);
37674 + mutex_lock(&xi->xi_mtx);
37675 + file = au_xino_file(xi, calc->idx);
37679 + file = au_xino_file(xi, /*idx*/-1);
37680 + AuDebugOn(!file);
37681 + xinew.idx = calc->idx;
37682 + xinew.base = &file->f_path;
37683 + /* xinew.copy_src = NULL; */
37684 + file = au_xi_new(sb, &xinew);
37685 + if (IS_ERR(file))
37686 + err = PTR_ERR(file);
37689 + mutex_unlock(&xi->xi_mtx);
37694 +struct au_xino_do_new_async_args {
37695 + struct super_block *sb;
37696 + struct au_branch *br;
37697 + struct au_xi_calc calc;
37701 +struct au_xi_writing {
37702 + struct hlist_bl_node node;
37703 + ino_t h_ino, ino;
37706 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37709 +static void au_xino_call_do_new_async(void *args)
37711 + struct au_xino_do_new_async_args *a = args;
37712 + struct au_branch *br;
37713 + struct super_block *sb;
37714 + struct au_sbinfo *sbi;
37715 + struct inode *root;
37716 + struct file *file;
37717 + struct au_xi_writing *del, *p;
37718 + struct hlist_bl_head *hbl;
37719 + struct hlist_bl_node *pos;
37724 + sbi = au_sbi(sb);
37725 + si_noflush_read_lock(sb);
37726 + root = d_inode(sb->s_root);
37727 + ii_read_lock_child(root);
37728 + err = au_xino_do_new_async(sb, br, &a->calc);
37729 + if (unlikely(err)) {
37730 + AuIOErr("err %d\n", err);
37734 + file = au_xino_file(br->br_xino, a->calc.idx);
37735 + AuDebugOn(!file);
37736 + err = au_xino_do_write(file, &a->calc, a->ino);
37737 + if (unlikely(err)) {
37738 + AuIOErr("err %d\n", err);
37743 + hbl = &br->br_xino->xi_writing;
37744 + hlist_bl_lock(hbl);
37745 + au_hbl_for_each(pos, hbl) {
37746 + p = container_of(pos, struct au_xi_writing, node);
37747 + if (p->ino == a->ino) {
37749 + hlist_bl_del(&p->node);
37753 + hlist_bl_unlock(hbl);
37754 + au_kfree_rcu(del);
37757 + au_lcnt_dec(&br->br_count);
37758 + ii_read_unlock(root);
37759 + si_read_unlock(sb);
37760 + au_nwt_done(&sbi->si_nowait);
37765 + * create a new xino file asynchronously
37767 +static int au_xino_new_async(struct super_block *sb, struct au_branch *br,
37768 + struct au_xi_calc *calc, ino_t ino)
37771 + struct au_xino_do_new_async_args *arg;
37774 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
37775 + if (unlikely(!arg))
37780 + arg->calc = *calc;
37782 + au_lcnt_inc(&br->br_count);
37783 + err = au_wkq_nowait(au_xino_call_do_new_async, arg, sb, AuWkq_NEST);
37784 + if (unlikely(err)) {
37785 + pr_err("wkq %d\n", err);
37786 + au_lcnt_dec(&br->br_count);
37787 + au_kfree_rcu(arg);
37795 + * read @ino from xinofile for the specified branch{@sb, @bindex}
37796 + * at the position of @h_ino.
37798 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37803 + struct au_xi_calc calc;
37804 + struct au_sbinfo *sbinfo;
37805 + struct file *file;
37806 + struct au_xino *xi;
37807 + struct hlist_bl_head *hbl;
37808 + struct hlist_bl_node *pos;
37809 + struct au_xi_writing *p;
37812 + if (!au_opt_test(au_mntflags(sb), XINO))
37813 + return 0; /* no xino */
37816 + au_xi_calc(sb, h_ino, &calc);
37817 + xi = au_sbr(sb, bindex)->br_xino;
37818 + file = au_xino_file(xi, calc.idx);
37820 + hbl = &xi->xi_writing;
37821 + hlist_bl_lock(hbl);
37822 + au_hbl_for_each(pos, hbl) {
37823 + p = container_of(pos, struct au_xi_writing, node);
37824 + if (p->h_ino == h_ino) {
37825 + AuDbg("hi%llu, i%llu, found\n",
37826 + (u64)p->h_ino, (u64)p->ino);
37831 + hlist_bl_unlock(hbl);
37833 + } else if (vfsub_f_size_read(file) < calc.pos + sizeof(*ino))
37834 + return 0; /* no xino */
37836 + sbinfo = au_sbi(sb);
37837 + sz = xino_fread(file, ino, sizeof(*ino), &calc.pos);
37838 + if (sz == sizeof(*ino))
37839 + return 0; /* success */
37842 + if (unlikely(sz >= 0)) {
37844 + AuIOErr("xino read error (%zd)\n", sz);
37849 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37854 + sz = xino_fwrite(file, &ino, sizeof(ino), &calc->pos);
37855 + if (sz == sizeof(ino))
37856 + return 0; /* success */
37858 + AuIOErr("write failed (%zd)\n", sz);
37863 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
37864 + * at the position of @h_ino.
37865 + * even if @ino is zero, it is written to the xinofile and means no entry.
37866 + * if the size of the xino file on a specific filesystem exceeds the watermark,
37867 + * try truncating it.
37869 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37873 + unsigned int mnt_flags;
37874 + struct au_xi_calc calc;
37875 + struct file *file;
37876 + struct au_branch *br;
37877 + struct au_xino *xi;
37878 + struct au_xi_writing *p;
37880 + SiMustAnyLock(sb);
37882 + mnt_flags = au_mntflags(sb);
37883 + if (!au_opt_test(mnt_flags, XINO))
37886 + au_xi_calc(sb, h_ino, &calc);
37887 + br = au_sbr(sb, bindex);
37888 + xi = br->br_xino;
37889 + file = au_xino_file(xi, calc.idx);
37891 + /* store the inum pair into the list */
37892 + p = kmalloc(sizeof(*p), GFP_NOFS | __GFP_NOFAIL);
37893 + p->h_ino = h_ino;
37895 + au_hbl_add(&p->node, &xi->xi_writing);
37897 + /* create and write a new xino file asynchronously */
37898 + err = au_xino_new_async(sb, br, &calc, ino);
37900 + return 0; /* success */
37904 + err = au_xino_do_write(file, &calc, ino);
37906 + br = au_sbr(sb, bindex);
37907 + if (au_opt_test(mnt_flags, TRUNC_XINO)
37908 + && au_test_fs_trunc_xino(au_br_sb(br)))
37909 + xino_try_trunc(sb, br);
37910 + return 0; /* success */
37914 + AuIOErr("write failed (%d)\n", err);
37918 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37921 +/* todo: unnecessary to support mmap_sem since kernel-space? */
37922 +ssize_t xino_fread(struct file *file, void *kbuf, size_t size, loff_t *pos)
37926 + const int prevent_endless = 10;
37930 + err = vfsub_read_k(file, kbuf, size, pos);
37931 + if (err == -EINTR
37932 + && !au_wkq_test()
37933 + && fatal_signal_pending(current)) {
37934 + err = xino_fread_wkq(file, kbuf, size, pos);
37935 + BUG_ON(err == -EINTR);
37937 + } while (i++ < prevent_endless
37938 + && (err == -EAGAIN || err == -EINTR));
37940 +#if 0 /* reserved for future use */
37942 + fsnotify_access(file->f_path.dentry);
37948 +struct xino_fread_args {
37950 + struct file *file;
37956 +static void call_xino_fread(void *args)
37958 + struct xino_fread_args *a = args;
37959 + *a->errp = xino_fread(a->file, a->buf, a->size, a->pos);
37962 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37967 + struct xino_fread_args args = {
37975 + wkq_err = au_wkq_wait(call_xino_fread, &args);
37976 + if (unlikely(wkq_err))
37982 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
37985 +static ssize_t do_xino_fwrite(struct file *file, void *kbuf, size_t size,
37990 + const int prevent_endless = 10;
37994 + err = vfsub_write_k(file, kbuf, size, pos);
37995 + if (err == -EINTR
37996 + && !au_wkq_test()
37997 + && fatal_signal_pending(current)) {
37998 + err = xino_fwrite_wkq(file, kbuf, size, pos);
37999 + BUG_ON(err == -EINTR);
38001 + } while (i++ < prevent_endless
38002 + && (err == -EAGAIN || err == -EINTR));
38004 +#if 0 /* reserved for future use */
38006 + fsnotify_modify(file->f_path.dentry);
38012 +struct do_xino_fwrite_args {
38014 + struct file *file;
38020 +static void call_do_xino_fwrite(void *args)
38022 + struct do_xino_fwrite_args *a = args;
38023 + *a->errp = do_xino_fwrite(a->file, a->buf, a->size, a->pos);
38026 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38031 + struct do_xino_fwrite_args args = {
38040 + * it breaks RLIMIT_FSIZE and normal user's limit,
38041 + * users should care about quota and real 'filesystem full.'
38043 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
38044 + if (unlikely(wkq_err))
38050 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos)
38054 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
38056 + err = do_xino_fwrite(file, buf, size, pos);
38060 + err = xino_fwrite_wkq(file, buf, size, pos);
38067 +/* ---------------------------------------------------------------------- */
38070 + * inode number bitmap
38072 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
38073 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
38077 + AuDebugOn(bit < 0 || page_bits <= bit);
38078 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
38082 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
38084 + AuDebugOn(ino < AUFS_FIRST_INO);
38085 + ino -= AUFS_FIRST_INO;
38086 + *pindex = ino / page_bits;
38087 + *bit = ino % page_bits;
38090 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
38095 + struct au_sbinfo *sbinfo;
38096 + struct file *xib;
38097 + unsigned long *p;
38099 + sbinfo = au_sbi(sb);
38100 + MtxMustLock(&sbinfo->si_xib_mtx);
38101 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
38102 + || !au_opt_test(sbinfo->si_mntflags, XINO));
38104 + if (pindex == sbinfo->si_xib_last_pindex)
38107 + xib = sbinfo->si_xib;
38108 + p = sbinfo->si_xib_buf;
38109 + pos = sbinfo->si_xib_last_pindex;
38110 + pos *= PAGE_SIZE;
38111 + sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38112 + if (unlikely(sz != PAGE_SIZE))
38116 + pos *= PAGE_SIZE;
38117 + if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
38118 + sz = xino_fread(xib, p, PAGE_SIZE, &pos);
38120 + memset(p, 0, PAGE_SIZE);
38121 + sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38123 + if (sz == PAGE_SIZE) {
38124 + sbinfo->si_xib_last_pindex = pindex;
38125 + return 0; /* success */
38129 + AuIOErr1("write failed (%zd)\n", sz);
38136 +static void au_xib_clear_bit(struct inode *inode)
38139 + unsigned long pindex;
38140 + struct super_block *sb;
38141 + struct au_sbinfo *sbinfo;
38143 + AuDebugOn(inode->i_nlink);
38145 + sb = inode->i_sb;
38146 + xib_calc_bit(inode->i_ino, &pindex, &bit);
38147 + AuDebugOn(page_bits <= bit);
38148 + sbinfo = au_sbi(sb);
38149 + mutex_lock(&sbinfo->si_xib_mtx);
38150 + err = xib_pindex(sb, pindex);
38152 + clear_bit(bit, sbinfo->si_xib_buf);
38153 + sbinfo->si_xib_next_bit = bit;
38155 + mutex_unlock(&sbinfo->si_xib_mtx);
38158 +/* ---------------------------------------------------------------------- */
38161 + * truncate a xino bitmap file
38165 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
38169 + unsigned long pindex;
38170 + loff_t pos, pend;
38171 + struct au_sbinfo *sbinfo;
38173 + unsigned long *p;
38176 + sbinfo = au_sbi(sb);
38177 + MtxMustLock(&sbinfo->si_xib_mtx);
38178 + p = sbinfo->si_xib_buf;
38179 + pend = vfsub_f_size_read(file);
38181 + while (pos < pend) {
38182 + sz = xino_fread(file, page, PAGE_SIZE, &pos);
38184 + if (unlikely(sz <= 0))
38188 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
38189 + if (unlikely(*ino < AUFS_FIRST_INO))
38192 + xib_calc_bit(*ino, &pindex, &bit);
38193 + AuDebugOn(page_bits <= bit);
38194 + err = xib_pindex(sb, pindex);
38206 +static int xib_restore(struct super_block *sb)
38209 + unsigned int nfile;
38210 + aufs_bindex_t bindex, bbot;
38212 + struct au_branch *br;
38213 + struct au_xino *xi;
38214 + struct file *file;
38217 + page = (void *)__get_free_page(GFP_NOFS);
38218 + if (unlikely(!page))
38222 + bbot = au_sbbot(sb);
38223 + for (bindex = 0; !err && bindex <= bbot; bindex++)
38224 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0) {
38225 + br = au_sbr(sb, bindex);
38226 + xi = br->br_xino;
38227 + nfile = xi->xi_nfile;
38228 + for (i = 0; i < nfile; i++) {
38229 + file = au_xino_file(xi, i);
38231 + err = do_xib_restore(sb, file, page);
38234 + AuDbg("skip shared b%d\n", bindex);
38235 + free_page((unsigned long)page);
38241 +int au_xib_trunc(struct super_block *sb)
38246 + struct au_sbinfo *sbinfo;
38247 + unsigned long *p;
38248 + struct file *file;
38250 + SiMustWriteLock(sb);
38253 + sbinfo = au_sbi(sb);
38254 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
38257 + file = sbinfo->si_xib;
38258 + if (vfsub_f_size_read(file) <= PAGE_SIZE)
38261 + file = au_xino_create2(sb, &sbinfo->si_xib->f_path, NULL);
38262 + err = PTR_ERR(file);
38263 + if (IS_ERR(file))
38265 + fput(sbinfo->si_xib);
38266 + sbinfo->si_xib = file;
38268 + p = sbinfo->si_xib_buf;
38269 + memset(p, 0, PAGE_SIZE);
38271 + sz = xino_fwrite(sbinfo->si_xib, p, PAGE_SIZE, &pos);
38272 + if (unlikely(sz != PAGE_SIZE)) {
38274 + AuIOErr("err %d\n", err);
38280 + mutex_lock(&sbinfo->si_xib_mtx);
38281 + /* mnt_want_write() is unnecessary here */
38282 + err = xib_restore(sb);
38283 + mutex_unlock(&sbinfo->si_xib_mtx);
38289 +/* ---------------------------------------------------------------------- */
38291 +struct au_xino *au_xino_alloc(unsigned int nfile)
38293 + struct au_xino *xi;
38295 + xi = kzalloc(sizeof(*xi), GFP_NOFS);
38296 + if (unlikely(!xi))
38298 + xi->xi_nfile = nfile;
38299 + xi->xi_file = kcalloc(nfile, sizeof(*xi->xi_file), GFP_NOFS);
38300 + if (unlikely(!xi->xi_file))
38303 + xi->xi_nondir.total = 8; /* initial size */
38304 + xi->xi_nondir.array = kcalloc(xi->xi_nondir.total, sizeof(ino_t),
38306 + if (unlikely(!xi->xi_nondir.array))
38309 + spin_lock_init(&xi->xi_nondir.spin);
38310 + init_waitqueue_head(&xi->xi_nondir.wqh);
38311 + mutex_init(&xi->xi_mtx);
38312 + INIT_HLIST_BL_HEAD(&xi->xi_writing);
38313 + atomic_set(&xi->xi_truncating, 0);
38314 + kref_init(&xi->xi_kref);
38315 + goto out; /* success */
38318 + au_kfree_try_rcu(xi->xi_file);
38320 + au_kfree_rcu(xi);
38326 +static int au_xino_init(struct au_branch *br, int idx, struct file *file)
38329 + struct au_xino *xi;
38332 + xi = au_xino_alloc(idx + 1);
38333 + if (unlikely(!xi)) {
38340 + xi->xi_file[idx] = file;
38341 + AuDebugOn(br->br_xino);
38342 + br->br_xino = xi;
38348 +static void au_xino_release(struct kref *kref)
38350 + struct au_xino *xi;
38352 + unsigned long ul;
38353 + struct hlist_bl_head *hbl;
38354 + struct hlist_bl_node *pos, *n;
38355 + struct au_xi_writing *p;
38357 + xi = container_of(kref, struct au_xino, xi_kref);
38358 + for (i = 0; i < xi->xi_nfile; i++)
38359 + if (xi->xi_file[i])
38360 + fput(xi->xi_file[i]);
38361 + for (i = xi->xi_nondir.total - 1; i >= 0; i--)
38362 + AuDebugOn(xi->xi_nondir.array[i]);
38363 + mutex_destroy(&xi->xi_mtx);
38364 + hbl = &xi->xi_writing;
38365 + ul = au_hbl_count(hbl);
38366 + if (unlikely(ul)) {
38367 + pr_warn("xi_writing %lu\n", ul);
38368 + hlist_bl_lock(hbl);
38369 + hlist_bl_for_each_entry_safe(p, pos, n, hbl, node) {
38370 + hlist_bl_del(&p->node);
38371 + /* kmemleak reported au_kfree_rcu() doesn't free it */
38374 + hlist_bl_unlock(hbl);
38376 + au_kfree_try_rcu(xi->xi_file);
38377 + au_kfree_try_rcu(xi->xi_nondir.array);
38378 + au_kfree_rcu(xi);
38381 +int au_xino_put(struct au_branch *br)
38384 + struct au_xino *xi;
38387 + xi = br->br_xino;
38389 + br->br_xino = NULL;
38390 + ret = kref_put(&xi->xi_kref, au_xino_release);
38396 +/* ---------------------------------------------------------------------- */
38399 + * xino mount option handlers
38403 +static void xino_clear_xib(struct super_block *sb)
38405 + struct au_sbinfo *sbinfo;
38407 + SiMustWriteLock(sb);
38409 + sbinfo = au_sbi(sb);
38410 + if (sbinfo->si_xib)
38411 + fput(sbinfo->si_xib);
38412 + sbinfo->si_xib = NULL;
38413 + if (sbinfo->si_xib_buf)
38414 + free_page((unsigned long)sbinfo->si_xib_buf);
38415 + sbinfo->si_xib_buf = NULL;
38418 +static int au_xino_set_xib(struct super_block *sb, struct path *path)
38422 + struct au_sbinfo *sbinfo;
38423 + struct file *file;
38424 + struct super_block *xi_sb;
38426 + SiMustWriteLock(sb);
38428 + sbinfo = au_sbi(sb);
38429 + file = au_xino_create2(sb, path, sbinfo->si_xib);
38430 + err = PTR_ERR(file);
38431 + if (IS_ERR(file))
38433 + if (sbinfo->si_xib)
38434 + fput(sbinfo->si_xib);
38435 + sbinfo->si_xib = file;
38436 + xi_sb = file_inode(file)->i_sb;
38437 + sbinfo->si_ximaxent = xi_sb->s_maxbytes;
38438 + if (unlikely(sbinfo->si_ximaxent < PAGE_SIZE)) {
38440 + pr_err("s_maxbytes(%llu) on %s is too small\n",
38441 + (u64)sbinfo->si_ximaxent, au_sbtype(xi_sb));
38444 + sbinfo->si_ximaxent /= sizeof(ino_t);
38447 + if (!sbinfo->si_xib_buf)
38448 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
38449 + if (unlikely(!sbinfo->si_xib_buf))
38452 + sbinfo->si_xib_last_pindex = 0;
38453 + sbinfo->si_xib_next_bit = 0;
38454 + if (vfsub_f_size_read(file) < PAGE_SIZE) {
38456 + err = xino_fwrite(file, sbinfo->si_xib_buf, PAGE_SIZE, &pos);
38457 + if (unlikely(err != PAGE_SIZE))
38461 + goto out; /* success */
38464 + if (sbinfo->si_xib_buf)
38465 + free_page((unsigned long)sbinfo->si_xib_buf);
38466 + sbinfo->si_xib_buf = NULL;
38470 + fput(sbinfo->si_xib);
38471 + sbinfo->si_xib = NULL;
38477 +/* xino for each branch */
38478 +static void xino_clear_br(struct super_block *sb)
38480 + aufs_bindex_t bindex, bbot;
38481 + struct au_branch *br;
38483 + bbot = au_sbbot(sb);
38484 + for (bindex = 0; bindex <= bbot; bindex++) {
38485 + br = au_sbr(sb, bindex);
38491 +static void au_xino_set_br_shared(struct super_block *sb, struct au_branch *br,
38492 + aufs_bindex_t bshared)
38494 + struct au_branch *brshared;
38496 + brshared = au_sbr(sb, bshared);
38497 + AuDebugOn(!brshared->br_xino);
38498 + AuDebugOn(!brshared->br_xino->xi_file);
38499 + if (br->br_xino != brshared->br_xino) {
38500 + au_xino_get(brshared);
38502 + br->br_xino = brshared->br_xino;
38506 +struct au_xino_do_set_br {
38507 + struct au_branch *br;
38509 + aufs_bindex_t bshared;
38512 +static int au_xino_do_set_br(struct super_block *sb, struct path *path,
38513 + struct au_xino_do_set_br *args)
38516 + struct au_xi_calc calc;
38517 + struct file *file;
38518 + struct au_branch *br;
38519 + struct au_xi_new xinew = {
38524 + xinew.xi = br->br_xino;
38525 + au_xi_calc(sb, args->h_ino, &calc);
38526 + xinew.copy_src = au_xino_file(xinew.xi, calc.idx);
38527 + if (args->bshared >= 0)
38528 + /* shared xino */
38529 + au_xino_set_br_shared(sb, br, args->bshared);
38530 + else if (!xinew.xi) {
38532 + err = au_xino_init(br, calc.idx, xinew.copy_src);
38533 + if (unlikely(err))
38537 + /* force re-creating */
38538 + xinew.xi = br->br_xino;
38539 + xinew.idx = calc.idx;
38540 + mutex_lock(&xinew.xi->xi_mtx);
38541 + file = au_xi_new(sb, &xinew);
38542 + mutex_unlock(&xinew.xi->xi_mtx);
38543 + err = PTR_ERR(file);
38544 + if (IS_ERR(file))
38546 + AuDebugOn(!file);
38548 + err = au_xino_do_write(file, &calc, AUFS_ROOT_INO);
38549 + if (unlikely(err))
38557 +static int au_xino_set_br(struct super_block *sb, struct path *path)
38560 + aufs_bindex_t bindex, bbot;
38561 + struct au_xino_do_set_br args;
38562 + struct inode *inode;
38564 + SiMustWriteLock(sb);
38566 + bbot = au_sbbot(sb);
38567 + inode = d_inode(sb->s_root);
38568 + for (bindex = 0; bindex <= bbot; bindex++) {
38569 + args.h_ino = au_h_iptr(inode, bindex)->i_ino;
38570 + args.br = au_sbr(sb, bindex);
38571 + args.bshared = is_sb_shared(sb, bindex, bindex - 1);
38572 + err = au_xino_do_set_br(sb, path, &args);
38573 + if (unlikely(err))
38581 +void au_xino_clr(struct super_block *sb)
38583 + struct au_sbinfo *sbinfo;
38585 + au_xigen_clr(sb);
38586 + xino_clear_xib(sb);
38587 + xino_clear_br(sb);
38588 + dbgaufs_brs_del(sb, 0);
38589 + sbinfo = au_sbi(sb);
38590 + /* lvalue, do not call au_mntflags() */
38591 + au_opt_clr(sbinfo->si_mntflags, XINO);
38594 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount)
38597 + struct dentry *dentry, *parent, *cur_dentry, *cur_parent;
38598 + struct qstr *dname, *cur_name;
38599 + struct file *cur_xino;
38600 + struct au_sbinfo *sbinfo;
38601 + struct path *path, *cur_path;
38603 + SiMustWriteLock(sb);
38606 + sbinfo = au_sbi(sb);
38607 + path = &xiopt->file->f_path;
38608 + dentry = path->dentry;
38609 + parent = dget_parent(dentry);
38612 + cur_xino = sbinfo->si_xib;
38614 + cur_path = &cur_xino->f_path;
38615 + cur_dentry = cur_path->dentry;
38616 + cur_parent = dget_parent(cur_dentry);
38617 + cur_name = &cur_dentry->d_name;
38618 + dname = &dentry->d_name;
38619 + skip = (cur_parent == parent
38620 + && au_qstreq(dname, cur_name));
38621 + dput(cur_parent);
38627 + au_opt_set(sbinfo->si_mntflags, XINO);
38628 + err = au_xino_set_xib(sb, path);
38629 + /* si_x{read,write} are set */
38631 + err = au_xigen_set(sb, path);
38633 + err = au_xino_set_br(sb, path);
38635 + dbgaufs_brs_add(sb, 0, /*topdown*/1);
38636 + goto out; /* success */
38640 + AuIOErr("failed setting xino(%d).\n", err);
38649 + * create a xinofile at the default place/path.
38651 +struct file *au_xino_def(struct super_block *sb)
38653 + struct file *file;
38655 + struct au_branch *br;
38656 + struct super_block *h_sb;
38657 + struct path path;
38658 + aufs_bindex_t bbot, bindex, bwr;
38661 + bbot = au_sbbot(sb);
38663 + for (bindex = 0; bindex <= bbot; bindex++) {
38664 + br = au_sbr(sb, bindex);
38665 + if (au_br_writable(br->br_perm)
38666 + && !au_test_fs_bad_xino(au_br_sb(br))) {
38673 + file = ERR_PTR(-ENOMEM);
38674 + page = (void *)__get_free_page(GFP_NOFS);
38675 + if (unlikely(!page))
38677 + path.mnt = au_br_mnt(br);
38678 + path.dentry = au_h_dptr(sb->s_root, bwr);
38679 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
38680 + file = (void *)p;
38681 + if (!IS_ERR(p)) {
38682 + strcat(p, "/" AUFS_XINO_FNAME);
38683 + AuDbg("%s\n", p);
38684 + file = au_xino_create(sb, p, /*silent*/0, /*wbrtop*/1);
38686 + free_page((unsigned long)page);
38688 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0,
38690 + if (IS_ERR(file))
38692 + h_sb = file->f_path.dentry->d_sb;
38693 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
38694 + pr_err("xino doesn't support %s(%s)\n",
38695 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
38697 + file = ERR_PTR(-EINVAL);
38705 +/* ---------------------------------------------------------------------- */
38708 + * initialize the xinofile for the specified branch @br
38709 + * at the place/path where @base_file indicates.
38710 + * test whether another branch is on the same filesystem or not,
38711 + * if found then share the xinofile with another branch.
38713 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
38714 + struct path *base)
38717 + struct au_xino_do_set_br args = {
38722 + args.bshared = sbr_find_shared(sb, /*btop*/0, au_sbbot(sb),
38724 + err = au_xino_do_set_br(sb, base, &args);
38725 + if (unlikely(err))
38731 +/* ---------------------------------------------------------------------- */
38734 + * get an unused inode number from bitmap
38736 +ino_t au_xino_new_ino(struct super_block *sb)
38739 + unsigned long *p, pindex, ul, pend;
38740 + struct au_sbinfo *sbinfo;
38741 + struct file *file;
38742 + int free_bit, err;
38744 + if (!au_opt_test(au_mntflags(sb), XINO))
38745 + return iunique(sb, AUFS_FIRST_INO);
38747 + sbinfo = au_sbi(sb);
38748 + mutex_lock(&sbinfo->si_xib_mtx);
38749 + p = sbinfo->si_xib_buf;
38750 + free_bit = sbinfo->si_xib_next_bit;
38751 + if (free_bit < page_bits && !test_bit(free_bit, p))
38752 + goto out; /* success */
38753 + free_bit = find_first_zero_bit(p, page_bits);
38754 + if (free_bit < page_bits)
38755 + goto out; /* success */
38757 + pindex = sbinfo->si_xib_last_pindex;
38758 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
38759 + err = xib_pindex(sb, ul);
38760 + if (unlikely(err))
38762 + free_bit = find_first_zero_bit(p, page_bits);
38763 + if (free_bit < page_bits)
38764 + goto out; /* success */
38767 + file = sbinfo->si_xib;
38768 + pend = vfsub_f_size_read(file) / PAGE_SIZE;
38769 + for (ul = pindex + 1; ul <= pend; ul++) {
38770 + err = xib_pindex(sb, ul);
38771 + if (unlikely(err))
38773 + free_bit = find_first_zero_bit(p, page_bits);
38774 + if (free_bit < page_bits)
38775 + goto out; /* success */
38780 + set_bit(free_bit, p);
38781 + sbinfo->si_xib_next_bit = free_bit + 1;
38782 + pindex = sbinfo->si_xib_last_pindex;
38783 + mutex_unlock(&sbinfo->si_xib_mtx);
38784 + ino = xib_calc_ino(pindex, free_bit);
38785 + AuDbg("i%lu\n", (unsigned long)ino);
38788 + mutex_unlock(&sbinfo->si_xib_mtx);
38793 +/* for s_op->delete_inode() */
38794 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
38797 + unsigned int mnt_flags;
38798 + aufs_bindex_t bindex, bbot, bi;
38799 + unsigned char try_trunc;
38800 + struct au_iinfo *iinfo;
38801 + struct super_block *sb;
38802 + struct au_hinode *hi;
38803 + struct inode *h_inode;
38804 + struct au_branch *br;
38805 + struct au_xi_calc calc;
38806 + struct file *file;
38808 + AuDebugOn(au_is_bad_inode(inode));
38810 + sb = inode->i_sb;
38811 + mnt_flags = au_mntflags(sb);
38812 + if (!au_opt_test(mnt_flags, XINO)
38813 + || inode->i_ino == AUFS_ROOT_INO)
38817 + au_xigen_inc(inode);
38818 + au_xib_clear_bit(inode);
38821 + iinfo = au_ii(inode);
38822 + bindex = iinfo->ii_btop;
38826 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
38827 + hi = au_hinode(iinfo, bindex);
38828 + bbot = iinfo->ii_bbot;
38829 + for (; bindex <= bbot; bindex++, hi++) {
38830 + h_inode = hi->hi_inode;
38832 + || (!unlinked && h_inode->i_nlink))
38835 + /* inode may not be revalidated */
38836 + bi = au_br_index(sb, hi->hi_id);
38840 + br = au_sbr(sb, bi);
38841 + au_xi_calc(sb, h_inode->i_ino, &calc);
38842 + file = au_xino_file(br->br_xino, calc.idx);
38843 + if (IS_ERR_OR_NULL(file))
38846 + err = au_xino_do_write(file, &calc, /*ino*/0);
38847 + if (!err && try_trunc
38848 + && au_test_fs_trunc_xino(au_br_sb(br)))
38849 + xino_try_trunc(sb, br);
38853 +/* ---------------------------------------------------------------------- */
38855 +static int au_xinondir_find(struct au_xino *xi, ino_t h_ino)
38857 + int found, total, i;
38860 + total = xi->xi_nondir.total;
38861 + for (i = 0; i < total; i++) {
38862 + if (xi->xi_nondir.array[i] != h_ino)
38871 +static int au_xinondir_expand(struct au_xino *xi)
38876 + BUILD_BUG_ON(KMALLOC_MAX_SIZE > INT_MAX);
38879 + sz = xi->xi_nondir.total * sizeof(ino_t);
38880 + if (unlikely(sz > KMALLOC_MAX_SIZE / 2))
38882 + p = au_kzrealloc(xi->xi_nondir.array, sz, sz << 1, GFP_ATOMIC,
38883 + /*may_shrink*/0);
38885 + xi->xi_nondir.array = p;
38886 + xi->xi_nondir.total <<= 1;
38887 + AuDbg("xi_nondir.total %d\n", xi->xi_nondir.total);
38895 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
38896 + ino_t h_ino, int idx)
38898 + struct au_xino *xi;
38900 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
38901 + xi = au_sbr(sb, bindex)->br_xino;
38902 + AuDebugOn(idx < 0 || xi->xi_nondir.total <= idx);
38904 + spin_lock(&xi->xi_nondir.spin);
38905 + AuDebugOn(xi->xi_nondir.array[idx] != h_ino);
38906 + xi->xi_nondir.array[idx] = 0;
38907 + spin_unlock(&xi->xi_nondir.spin);
38908 + wake_up_all(&xi->xi_nondir.wqh);
38911 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
38914 + int err, found, empty;
38915 + struct au_xino *xi;
38919 + if (!au_opt_test(au_mntflags(sb), XINO))
38920 + goto out; /* no xino */
38922 + xi = au_sbr(sb, bindex)->br_xino;
38925 + spin_lock(&xi->xi_nondir.spin);
38926 + found = au_xinondir_find(xi, h_ino);
38927 + if (found == -1) {
38928 + empty = au_xinondir_find(xi, /*h_ino*/0);
38929 + if (empty == -1) {
38930 + empty = xi->xi_nondir.total;
38931 + err = au_xinondir_expand(xi);
38932 + if (unlikely(err))
38935 + xi->xi_nondir.array[empty] = h_ino;
38938 + spin_unlock(&xi->xi_nondir.spin);
38939 + wait_event(xi->xi_nondir.wqh,
38940 + xi->xi_nondir.array[found] != h_ino);
38945 + spin_unlock(&xi->xi_nondir.spin);
38950 +/* ---------------------------------------------------------------------- */
38952 +int au_xino_path(struct seq_file *seq, struct file *file)
38956 + err = au_seq_path(seq, &file->f_path);
38957 + if (unlikely(err))
38960 +#define Deleted "\\040(deleted)"
38961 + seq->count -= sizeof(Deleted) - 1;
38962 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
38963 + sizeof(Deleted) - 1));
38969 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
38970 --- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
38971 +++ linux/include/uapi/linux/aufs_type.h 2022-05-23 09:44:37.729839166 +0200
38973 +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
38975 + * Copyright (C) 2005-2021 Junjiro R. Okajima
38977 + * This program, aufs is free software; you can redistribute it and/or modify
38978 + * it under the terms of the GNU General Public License as published by
38979 + * the Free Software Foundation; either version 2 of the License, or
38980 + * (at your option) any later version.
38982 + * This program is distributed in the hope that it will be useful,
38983 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
38984 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
38985 + * GNU General Public License for more details.
38987 + * You should have received a copy of the GNU General Public License
38988 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
38991 +#ifndef __AUFS_TYPE_H__
38992 +#define __AUFS_TYPE_H__
38994 +#define AUFS_NAME "aufs"
38998 + * define it before including all other headers.
38999 + * sched.h may use pr_* macros before defining "current", so define the
39000 + * no-current version first, and re-define later.
39002 +#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
39003 +#include <linux/sched.h>
39005 +#define pr_fmt(fmt) \
39006 + AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
39007 + (int)sizeof(current->comm), current->comm, current->pid
39008 +#include <linux/limits.h>
39010 +#include <stdint.h>
39011 +#include <sys/types.h>
39012 +#include <limits.h>
39013 +#endif /* __KERNEL__ */
39015 +#define AUFS_VERSION "5.x-rcN-20220418"
39017 +/* todo? move this to linux-2.6.19/include/magic.h */
39018 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
39020 +/* ---------------------------------------------------------------------- */
39023 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
39024 +typedef int8_t aufs_bindex_t;
39025 +#define AUFS_BRANCH_MAX 127
39027 +typedef int16_t aufs_bindex_t;
39028 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
39029 +#define AUFS_BRANCH_MAX 511
39030 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
39031 +#define AUFS_BRANCH_MAX 1023
39032 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
39033 +#define AUFS_BRANCH_MAX 32767
39037 +#ifndef AUFS_BRANCH_MAX
39038 +#error unknown CONFIG_AUFS_BRANCH_MAX value
39040 +#endif /* __KERNEL__ */
39042 +/* ---------------------------------------------------------------------- */
39044 +#define AUFS_FSTYPE AUFS_NAME
39046 +#define AUFS_ROOT_INO 2
39047 +#define AUFS_FIRST_INO 11
39049 +#define AUFS_WH_PFX ".wh."
39050 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
39051 +#define AUFS_WH_TMP_LEN 4
39052 +/* a limit for rmdir/rename a dir and copyup */
39053 +#define AUFS_MAX_NAMELEN (NAME_MAX \
39054 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
39056 + - AUFS_WH_TMP_LEN) /* hex */
39057 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
39058 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
39059 +#define AUFS_XINO_DEF_SEC 30 /* seconds */
39060 +#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
39061 +#define AUFS_DIRWH_DEF 3
39062 +#define AUFS_RDCACHE_DEF 10 /* seconds */
39063 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
39064 +#define AUFS_RDBLK_DEF 512 /* bytes */
39065 +#define AUFS_RDHASH_DEF 32
39066 +#define AUFS_WKQ_NAME AUFS_NAME "d"
39067 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
39068 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
39069 +#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
39070 +#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
39072 +/* pseudo-link maintenace under /proc */
39073 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
39074 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
39075 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
39077 +/* dirren, renamed dir */
39078 +#define AUFS_DR_INFO_PFX AUFS_WH_PFX ".dr."
39079 +#define AUFS_DR_BRHINO_NAME AUFS_WH_PFX "hino"
39080 +/* whiteouted doubly */
39081 +#define AUFS_WH_DR_INFO_PFX AUFS_WH_PFX AUFS_DR_INFO_PFX
39082 +#define AUFS_WH_DR_BRHINO AUFS_WH_PFX AUFS_DR_BRHINO_NAME
39084 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
39085 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
39087 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
39088 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
39089 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
39091 +/* doubly whiteouted */
39092 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
39093 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
39094 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
39096 +/* branch permissions and attributes */
39097 +#define AUFS_BRPERM_RW "rw"
39098 +#define AUFS_BRPERM_RO "ro"
39099 +#define AUFS_BRPERM_RR "rr"
39100 +#define AUFS_BRATTR_COO_REG "coo_reg"
39101 +#define AUFS_BRATTR_COO_ALL "coo_all"
39102 +#define AUFS_BRATTR_FHSM "fhsm"
39103 +#define AUFS_BRATTR_UNPIN "unpin"
39104 +#define AUFS_BRATTR_ICEX "icex"
39105 +#define AUFS_BRATTR_ICEX_SEC "icexsec"
39106 +#define AUFS_BRATTR_ICEX_SYS "icexsys"
39107 +#define AUFS_BRATTR_ICEX_TR "icextr"
39108 +#define AUFS_BRATTR_ICEX_USR "icexusr"
39109 +#define AUFS_BRATTR_ICEX_OTH "icexoth"
39110 +#define AUFS_BRRATTR_WH "wh"
39111 +#define AUFS_BRWATTR_NLWH "nolwh"
39112 +#define AUFS_BRWATTR_MOO "moo"
39114 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
39115 +#define AuBrPerm_RO (1 << 1) /* readonly */
39116 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
39117 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
39119 +#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
39120 +#define AuBrAttr_COO_ALL (1 << 4)
39121 +#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
39123 +#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
39124 +#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
39125 + branch. meaningless since
39126 + linux-3.18-rc1 */
39128 +/* ignore error in copying XATTR */
39129 +#define AuBrAttr_ICEX_SEC (1 << 7)
39130 +#define AuBrAttr_ICEX_SYS (1 << 8)
39131 +#define AuBrAttr_ICEX_TR (1 << 9)
39132 +#define AuBrAttr_ICEX_USR (1 << 10)
39133 +#define AuBrAttr_ICEX_OTH (1 << 11)
39134 +#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
39135 + | AuBrAttr_ICEX_SYS \
39136 + | AuBrAttr_ICEX_TR \
39137 + | AuBrAttr_ICEX_USR \
39138 + | AuBrAttr_ICEX_OTH)
39140 +#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
39141 +#define AuBrRAttr_Mask AuBrRAttr_WH
39143 +#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
39144 +#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
39145 +#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
39147 +#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
39149 +/* #warning test userspace */
39151 +#ifndef CONFIG_AUFS_FHSM
39152 +#undef AuBrAttr_FHSM
39153 +#define AuBrAttr_FHSM 0
39155 +#ifndef CONFIG_AUFS_XATTR
39156 +#undef AuBrAttr_ICEX
39157 +#define AuBrAttr_ICEX 0
39158 +#undef AuBrAttr_ICEX_SEC
39159 +#define AuBrAttr_ICEX_SEC 0
39160 +#undef AuBrAttr_ICEX_SYS
39161 +#define AuBrAttr_ICEX_SYS 0
39162 +#undef AuBrAttr_ICEX_TR
39163 +#define AuBrAttr_ICEX_TR 0
39164 +#undef AuBrAttr_ICEX_USR
39165 +#define AuBrAttr_ICEX_USR 0
39166 +#undef AuBrAttr_ICEX_OTH
39167 +#define AuBrAttr_ICEX_OTH 0
39171 +/* the longest combination */
39172 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
39173 +#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
39174 + "+" AUFS_BRATTR_COO_REG \
39175 + "+" AUFS_BRATTR_FHSM \
39176 + "+" AUFS_BRATTR_UNPIN \
39177 + "+" AUFS_BRATTR_ICEX_SEC \
39178 + "+" AUFS_BRATTR_ICEX_SYS \
39179 + "+" AUFS_BRATTR_ICEX_USR \
39180 + "+" AUFS_BRATTR_ICEX_OTH \
39181 + "+" AUFS_BRWATTR_NLWH)
39184 + char a[AuBrPermStrSz];
39185 +} au_br_perm_str_t;
39187 +static inline int au_br_writable(int brperm)
39189 + return brperm & AuBrPerm_RW;
39192 +static inline int au_br_whable(int brperm)
39194 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
39197 +static inline int au_br_wh_linkable(int brperm)
39199 + return !(brperm & AuBrWAttr_NoLinkWH);
39202 +static inline int au_br_cmoo(int brperm)
39204 + return brperm & AuBrAttr_CMOO_Mask;
39207 +static inline int au_br_fhsm(int brperm)
39209 + return brperm & AuBrAttr_FHSM;
39212 +/* ---------------------------------------------------------------------- */
39216 + /* readdir in userspace */
39220 + AuCtl_WBR_FD, /* pathconf wrapper */
39221 + AuCtl_IBUSY, /* busy inode */
39222 + AuCtl_MVDOWN, /* move-down */
39223 + AuCtl_BR, /* info about branches */
39224 + AuCtl_FHSM_FD /* connection for fhsm */
39227 +/* borrowed from linux/include/linux/kernel.h */
39229 +#ifdef _GNU_SOURCE
39230 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
39232 +#define ALIGN(x, a) (((x) + (a) - 1) & ~((a) - 1))
39234 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
39237 +/* borrowed from linux/include/linux/compiler-gcc3.h */
39239 +#define __aligned(x) __attribute__((aligned(x)))
39244 +#define __packed __attribute__((packed))
39248 +struct au_rdu_cookie {
39253 + uint32_t generation;
39256 +struct au_rdu_ent {
39265 +static inline int au_rdu_len(int nlen)
39267 + /* include the terminating NULL */
39268 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
39269 + sizeof(uint64_t));
39272 +union au_rdu_ent_ul {
39273 + struct au_rdu_ent __user *e;
39285 + uint64_t sz; /* AuCtl_RDU */
39286 + uint64_t nent; /* AuCtl_RDU_INO */
39288 + union au_rdu_ent_ul ent;
39289 + uint16_t verify[AufsCtlRduV_End];
39291 + /* input/output */
39295 + union au_rdu_ent_ul tail;
39296 + /* number of entries which were added in a single call */
39301 + struct au_rdu_cookie cookie;
39304 +/* ---------------------------------------------------------------------- */
39306 +/* dirren. the branch is identified by the filename who contains this */
39307 +struct au_drinfo {
39310 + uint8_t oldnamelen;
39311 + uint64_t _padding;
39313 + uint8_t oldname[];
39316 +struct au_drinfo_fdata {
39318 + struct au_drinfo drinfo;
39321 +#define AUFS_DRINFO_MAGIC_V1 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x01)
39323 +#define AUFS_DRINFO_MAGIC_V2 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x02)
39325 +/* ---------------------------------------------------------------------- */
39327 +struct aufs_wbr_fd {
39332 +/* ---------------------------------------------------------------------- */
39334 +struct aufs_ibusy {
39335 + uint64_t ino, h_ino;
39339 +/* ---------------------------------------------------------------------- */
39341 +/* error code for move-down */
39342 +/* the actual message strings are implemented in aufs-util.git */
39344 + EAU_MVDOWN_OPAQUE = 1,
39345 + EAU_MVDOWN_WHITEOUT,
39346 + EAU_MVDOWN_UPPER,
39347 + EAU_MVDOWN_BOTTOM,
39348 + EAU_MVDOWN_NOUPPER,
39349 + EAU_MVDOWN_NOLOWERBR,
39353 +/* flags for move-down */
39354 +#define AUFS_MVDOWN_DMSG 1
39355 +#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
39356 +#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
39357 +#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
39358 +#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
39359 +#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
39360 +#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
39361 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
39362 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
39363 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
39364 +#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
39365 +#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
39366 +#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
39368 +/* index for move-down */
39370 + AUFS_MVDOWN_UPPER,
39371 + AUFS_MVDOWN_LOWER,
39372 + AUFS_MVDOWN_NARRAY
39376 + * additional info of move-down
39377 + * number of free blocks and inodes.
39378 + * subset of struct kstatfs, but smaller and always 64bit.
39380 +struct aufs_stfs {
39381 + uint64_t f_blocks;
39382 + uint64_t f_bavail;
39383 + uint64_t f_files;
39384 + uint64_t f_ffree;
39387 +struct aufs_stbr {
39388 + int16_t brid; /* optional input */
39389 + int16_t bindex; /* output */
39390 + struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
39393 +struct aufs_mvdown {
39394 + uint32_t flags; /* input/output */
39395 + struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
39396 + int8_t au_errno; /* output */
39399 +/* ---------------------------------------------------------------------- */
39401 +union aufs_brinfo {
39402 + /* PATH_MAX may differ between kernel-space and user-space */
39403 + char _spacer[4096];
39411 +/* ---------------------------------------------------------------------- */
39413 +#define AuCtlType 'A'
39414 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
39415 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
39416 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
39417 + struct aufs_wbr_fd)
39418 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
39419 +#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
39420 + struct aufs_mvdown)
39421 +#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
39422 +#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
39424 +#endif /* __AUFS_TYPE_H__ */
39425 SPDX-License-Identifier: GPL-2.0
39426 aufs5.x-rcN loopback patch
39428 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
39429 index ba35d0d6847a..63e256b40842 100644
39430 --- a/drivers/block/loop.c
39431 +++ b/drivers/block/loop.c
39432 @@ -509,6 +509,15 @@ static inline void loop_update_dio(struct loop_device *lo)
39436 +static struct file *loop_real_file(struct file *file)
39438 + struct file *f = NULL;
39440 + if (file->f_path.dentry->d_sb->s_op->real_loop)
39441 + f = file->f_path.dentry->d_sb->s_op->real_loop(file);
39445 static void loop_reread_partitions(struct loop_device *lo)
39448 @@ -566,6 +575,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39450 struct file *file = fget(arg);
39451 struct file *old_file;
39452 + struct file *f, *virt_file = NULL, *old_virt_file;
39456 @@ -585,11 +595,19 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39457 if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
39460 + f = loop_real_file(file);
39462 + virt_file = file;
39467 error = loop_validate_file(file, bdev);
39471 old_file = lo->lo_backing_file;
39472 + old_virt_file = lo->lo_backing_virt_file;
39476 @@ -602,6 +620,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39477 blk_mq_freeze_queue(lo->lo_queue);
39478 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
39479 lo->lo_backing_file = file;
39480 + lo->lo_backing_virt_file = virt_file;
39481 lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
39482 mapping_set_gfp_mask(file->f_mapping,
39483 lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39484 @@ -624,6 +643,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39488 + if (old_virt_file)
39489 + fput(old_virt_file);
39491 loop_reread_partitions(lo);
39493 @@ -632,6 +653,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39494 loop_global_unlock(lo, is_loop);
39502 @@ -970,6 +993,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39503 const struct loop_config *config)
39505 struct file *file = fget(config->fd);
39506 + struct file *f, *virt_file = NULL;
39507 struct inode *inode;
39508 struct address_space *mapping;
39510 @@ -985,6 +1009,13 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39511 /* This is safe, since we have a reference from open(). */
39512 __module_get(THIS_MODULE);
39514 + f = loop_real_file(file);
39516 + virt_file = file;
39522 * If we don't hold exclusive handle for the device, upgrade to it
39523 * here to avoid changing device under exclusive owner.
39524 @@ -1050,6 +1081,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39525 lo->use_dio = lo->lo_flags & LO_FLAGS_DIRECT_IO;
39526 lo->lo_device = bdev;
39527 lo->lo_backing_file = file;
39528 + lo->lo_backing_virt_file = virt_file;
39529 lo->old_gfp_mask = mapping_gfp_mask(mapping);
39530 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39532 @@ -1100,6 +1132,8 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39533 bd_abort_claiming(bdev, loop_configure);
39538 /* This is safe: open() is still holding a reference. */
39539 module_put(THIS_MODULE);
39541 @@ -1108,6 +1142,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39542 static void __loop_clr_fd(struct loop_device *lo, bool release)
39545 + struct file *virt_filp = lo->lo_backing_virt_file;
39546 gfp_t gfp = lo->old_gfp_mask;
39547 struct loop_worker *pos, *worker;
39549 @@ -1153,6 +1188,7 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39550 spin_lock_irq(&lo->lo_lock);
39551 filp = lo->lo_backing_file;
39552 lo->lo_backing_file = NULL;
39553 + lo->lo_backing_virt_file = NULL;
39554 spin_unlock_irq(&lo->lo_lock);
39556 lo->lo_device = NULL;
39557 @@ -1214,6 +1250,8 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39558 * fput can take open_mutex which is usually taken before lo_mutex.
39565 static int loop_clr_fd(struct loop_device *lo)
39566 diff --git a/drivers/block/loop.h b/drivers/block/loop.h
39567 index 082d4b6bfc6a..61feca64688d 100644
39568 --- a/drivers/block/loop.h
39569 +++ b/drivers/block/loop.h
39570 @@ -34,7 +34,7 @@ struct loop_device {
39572 char lo_file_name[LO_NAME_SIZE];
39574 - struct file * lo_backing_file;
39575 + struct file *lo_backing_file, *lo_backing_virt_file;
39576 struct block_device *lo_device;
39578 gfp_t old_gfp_mask;
39579 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
39580 index 99ceca144044..e49dfe855032 100644
39581 --- a/fs/aufs/f_op.c
39582 +++ b/fs/aufs/f_op.c
39583 @@ -304,7 +304,7 @@ static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
39584 if (IS_ERR(h_file))
39587 - if (au_test_loopback_kthread()) {
39588 + if (0 && au_test_loopback_kthread()) {
39589 au_warn_loopback(h_file->f_path.dentry->d_sb);
39590 if (file->f_mapping != h_file->f_mapping) {
39591 file->f_mapping = h_file->f_mapping;
39592 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
39593 index 74347bd75b38..5ef888a1d53f 100644
39594 --- a/fs/aufs/loop.c
39595 +++ b/fs/aufs/loop.c
39596 @@ -133,3 +133,19 @@ void au_loopback_fin(void)
39597 symbol_put(loop_backing_file);
39598 au_kfree_try_rcu(au_warn_loopback_array);
39601 +/* ---------------------------------------------------------------------- */
39603 +/* support the loopback block device insude aufs */
39605 +struct file *aufs_real_loop(struct file *file)
39609 + BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
39610 + fi_read_lock(file);
39611 + f = au_hf_top(file);
39612 + fi_read_unlock(file);
39616 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
39617 index 7293bee427f9..3345c098d0d4 100644
39618 --- a/fs/aufs/loop.h
39619 +++ b/fs/aufs/loop.h
39620 @@ -26,6 +26,8 @@ void au_warn_loopback(struct super_block *h_sb);
39622 int au_loopback_init(void);
39623 void au_loopback_fin(void);
39625 +struct file *aufs_real_loop(struct file *file);
39627 AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
39629 @@ -36,6 +38,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
39631 AuStubInt0(au_loopback_init, void)
39632 AuStubVoid(au_loopback_fin, void)
39634 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
39635 #endif /* BLK_DEV_LOOP */
39637 #endif /* __KERNEL__ */
39638 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
39639 index 7d9bcc074c87..0eefe350c01c 100644
39640 --- a/fs/aufs/super.c
39641 +++ b/fs/aufs/super.c
39642 @@ -758,7 +758,10 @@ const struct super_operations aufs_sop = {
39643 .show_options = aufs_show_options,
39644 .statfs = aufs_statfs,
39645 .put_super = aufs_put_super,
39646 - .sync_fs = aufs_sync_fs
39647 + .sync_fs = aufs_sync_fs,
39648 +#ifdef CONFIG_AUFS_BDEV_LOOP
39649 + .real_loop = aufs_real_loop
39653 /* ---------------------------------------------------------------------- */
39654 diff --git a/include/linux/fs.h b/include/linux/fs.h
39655 index 06c9242932a2..9190ce754010 100644
39656 --- a/include/linux/fs.h
39657 +++ b/include/linux/fs.h
39658 @@ -2113,6 +2113,10 @@ struct super_operations {
39659 struct shrink_control *);
39660 long (*free_cached_objects)(struct super_block *,
39661 struct shrink_control *);
39662 +#if IS_ENABLED(CONFIG_BLK_DEV_LOOP) || IS_ENABLED(CONFIG_BLK_DEV_LOOP_MODULE)
39664 + struct file *(*real_loop)(struct file *);