1 SPDX-License-Identifier: GPL-2.0
2 aufs6.x-rcN kbuild patch
4 diff --git a/fs/Kconfig b/fs/Kconfig
5 index cc07a0cd3172..5ee4f1be48ff 100644
8 @@ -316,6 +316,7 @@ 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 834f1c3dba46..b5c97dd026cd 100644
20 @@ -137,3 +137,4 @@ obj-$(CONFIG_EFIVAR_FS) += efivarfs/
21 obj-$(CONFIG_EROFS_FS) += erofs/
22 obj-$(CONFIG_VBOXSF_FS) += vboxsf/
23 obj-$(CONFIG_ZONEFS_FS) += zonefs/
24 +obj-$(CONFIG_AUFS_FS) += aufs/
25 SPDX-License-Identifier: GPL-2.0
26 aufs6.x-rcN base patch
28 diff --git a/MAINTAINERS b/MAINTAINERS
29 index 7e0b87d5aa2e..ce2984ca2f06 100644
32 @@ -3387,6 +3387,19 @@ 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 bc31bb7072a2..131294601819 100644
54 --- a/drivers/block/loop.c
55 +++ b/drivers/block/loop.c
56 @@ -645,6 +645,24 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
62 + * no get/put for file.
64 +struct file *loop_backing_file(struct super_block *sb)
67 + struct loop_device *l;
70 + if (MAJOR(sb->s_dev) == LOOP_MAJOR) {
71 + l = sb->s_bdev->bd_disk->private_data;
72 + ret = l->lo_backing_file;
76 +EXPORT_SYMBOL_GPL(loop_backing_file);
78 /* loop sysfs attributes */
80 static ssize_t loop_attr_show(struct device *dev, char *page,
81 diff --git a/fs/dcache.c b/fs/dcache.c
82 index 52e6d5fdab6b..519321f32f95 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 b622be119706..9ea58b7bb580 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 54847db5b819..ebb7b4b057ea 100644
120 @@ -849,6 +849,12 @@ 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 3e06611d19ae..97a66df75d4c 100644
137 @@ -865,8 +865,8 @@ 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 @@ -876,9 +876,9 @@ 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 21a981680856..1e3054e5367d 100644
163 --- a/include/linux/fs.h
164 +++ b/include/linux/fs.h
165 @@ -1066,6 +1066,7 @@ 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 @@ -1793,6 +1794,7 @@ struct file_operations {
174 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
175 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
176 int (*check_flags)(int);
177 + int (*setfl)(struct file *, unsigned long);
178 int (*flock) (struct file *, int, struct file_lock *);
179 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
180 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
181 diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
182 index b32256e9e944..16bd7230f44b 100644
183 --- a/include/linux/lockdep.h
184 +++ b/include/linux/lockdep.h
185 @@ -249,6 +249,8 @@ static inline int lockdep_match_key(struct lockdep_map *lock,
186 return lock->key == key;
189 +struct lock_class *lockdep_hlock_class(struct held_lock *hlock);
194 diff --git a/include/linux/mnt_namespace.h b/include/linux/mnt_namespace.h
195 index 8f882f5881e8..6b9808f09843 100644
196 --- a/include/linux/mnt_namespace.h
197 +++ b/include/linux/mnt_namespace.h
198 @@ -7,12 +7,15 @@ struct mnt_namespace;
200 struct user_namespace;
204 extern struct mnt_namespace *copy_mnt_ns(unsigned long, struct mnt_namespace *,
205 struct user_namespace *, struct fs_struct *);
206 extern void put_mnt_ns(struct mnt_namespace *ns);
207 extern struct ns_common *from_mnt_ns(struct mnt_namespace *);
209 +extern int is_current_mnt_ns(struct vfsmount *mnt);
211 extern const struct file_operations proc_mounts_operations;
212 extern const struct file_operations proc_mountinfo_operations;
213 extern const struct file_operations proc_mountstats_operations;
214 diff --git a/include/linux/splice.h b/include/linux/splice.h
215 index a55179fd60fc..8e21c53cf883 100644
216 --- a/include/linux/splice.h
217 +++ b/include/linux/splice.h
218 @@ -93,4 +93,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
220 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
221 extern const struct pipe_buf_operations default_pipe_buf_ops;
223 +extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
224 + loff_t *ppos, size_t len, unsigned int flags);
225 +extern long do_splice_to(struct file *in, loff_t *ppos,
226 + struct pipe_inode_info *pipe, size_t len,
227 + unsigned int flags);
229 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
230 index dcd1d5bfc1e0..85d1f7e0bc07 100644
231 --- a/kernel/locking/lockdep.c
232 +++ b/kernel/locking/lockdep.c
233 @@ -218,7 +218,7 @@ unsigned long max_lock_class_idx;
234 struct lock_class lock_classes[MAX_LOCKDEP_KEYS];
235 DECLARE_BITMAP(lock_classes_in_use, MAX_LOCKDEP_KEYS);
237 -static inline struct lock_class *hlock_class(struct held_lock *hlock)
238 +inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
240 unsigned int class_idx = hlock->class_idx;
242 @@ -239,6 +239,7 @@ static inline struct lock_class *hlock_class(struct held_lock *hlock)
244 return lock_classes + class_idx;
246 +#define hlock_class(hlock) lockdep_hlock_class(hlock)
248 #ifdef CONFIG_LOCK_STAT
249 static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
250 SPDX-License-Identifier: GPL-2.0
251 aufs6.x-rcN mmap patch
253 diff --git a/fs/proc/base.c b/fs/proc/base.c
254 index 05452c3b9872..6b92a8430130 100644
257 @@ -2218,7 +2218,7 @@ static int map_files_get_link(struct dentry *dentry, struct path *path)
259 vma = find_exact_vma(mm, vm_start, vm_end);
260 if (vma && vma->vm_file) {
261 - *path = vma->vm_file->f_path;
262 + *path = vma_pr_or_file(vma)->f_path;
266 diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
267 index 4d3493579458..42edd9a42c78 100644
268 --- a/fs/proc/nommu.c
269 +++ b/fs/proc/nommu.c
270 @@ -39,7 +39,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
271 file = region->vm_file;
274 - struct inode *inode = file_inode(region->vm_file);
275 + struct inode *inode;
277 + file = vmr_pr_or_file(region);
278 + inode = file_inode(file);
279 dev = inode->i_sb->s_dev;
282 diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
283 index 420510f6a545..edfc0941806a 100644
284 --- a/fs/proc/task_mmu.c
285 +++ b/fs/proc/task_mmu.c
286 @@ -285,7 +285,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma)
287 const char *name = NULL;
290 - struct inode *inode = file_inode(vma->vm_file);
291 + struct inode *inode;
293 + file = vma_pr_or_file(vma);
294 + inode = file_inode(file);
295 dev = inode->i_sb->s_dev;
297 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
298 @@ -1943,7 +1946,7 @@ static int show_numa_map(struct seq_file *m, void *v)
299 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
300 struct vm_area_struct *vma = v;
301 struct numa_maps *md = &numa_priv->md;
302 - struct file *file = vma->vm_file;
303 + struct file *file = vma_pr_or_file(vma);
304 struct mm_struct *mm = vma->vm_mm;
305 struct mempolicy *pol;
307 diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
308 index 0ec35072a8e5..f0f0a03eb2b5 100644
309 --- a/fs/proc/task_nommu.c
310 +++ b/fs/proc/task_nommu.c
311 @@ -150,7 +150,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
315 - struct inode *inode = file_inode(vma->vm_file);
316 + struct inode *inode;
318 + file = vma_pr_or_file(vma);
319 + inode = file_inode(file);
320 dev = inode->i_sb->s_dev;
322 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
323 diff --git a/include/linux/mm.h b/include/linux/mm.h
324 index 27ce77080c79..5f07b972242d 100644
325 --- a/include/linux/mm.h
326 +++ b/include/linux/mm.h
327 @@ -2353,6 +2353,43 @@ static inline void unmap_shared_mapping_range(struct address_space *mapping,
328 unmap_mapping_range(mapping, holebegin, holelen, 0);
331 +#if 1 /* IS_ENABLED(CONFIG_AUFS_FS) */
332 +extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
333 +extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
335 +extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
336 +extern void vma_do_fput(struct vm_area_struct *, const char[], int);
338 +#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
340 +#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
342 +#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
343 +#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
346 +extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
347 +extern void vmr_do_fput(struct vm_region *, const char[], int);
349 +#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
351 +#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
352 +#endif /* !CONFIG_MMU */
356 +#define vma_file_update_time(vma) file_update_time((vma)->vm_file)
357 +#define vma_pr_or_file(vma) (vma)->vm_file
358 +#define vma_get_file(vma) get_file((vma)->vm_file)
359 +#define vma_fput(vma) fput((vma)->vm_file)
362 +#define vmr_pr_or_file(region) (region)->vm_file
363 +#define vmr_fput(region) fput((region)->vm_file)
364 +#endif /* !CONFIG_MMU */
366 +#endif /* CONFIG_AUFS_FS */
368 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
369 void *buf, int len, unsigned int gup_flags);
370 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
371 diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
372 index 306a3d1a0fa6..56374f0eb825 100644
373 --- a/include/linux/mm_types.h
374 +++ b/include/linux/mm_types.h
375 @@ -449,6 +449,9 @@ struct vm_region {
376 unsigned long vm_top; /* region allocated to here */
377 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
378 struct file *vm_file; /* the backing file or NULL */
379 +#if 1 /* IS_ENABLED(CONFIG_AUFS_FS) */
380 + struct file *vm_prfile; /* the virtual backing file or NULL */
383 int vm_usage; /* region usage count (access under nommu_region_sem) */
384 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
385 @@ -548,6 +551,9 @@ struct vm_area_struct {
386 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
388 struct file * vm_file; /* File we map to (can be NULL). */
389 +#if 1 /* IS_ENABLED(CONFIG_AUFS_FS) */
390 + struct file *vm_prfile; /* shadow of vm_file */
392 void * vm_private_data; /* was vm_pte (shared mem) */
394 #ifdef CONFIG_ANON_VMA_NAME
395 diff --git a/kernel/fork.c b/kernel/fork.c
396 index ed4e01daccaa..e945a1639283 100644
399 @@ -733,7 +733,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
401 struct address_space *mapping = file->f_mapping;
405 i_mmap_lock_write(mapping);
406 if (tmp->vm_flags & VM_SHARED)
407 mapping_allow_writable(mapping);
408 diff --git a/mm/Makefile b/mm/Makefile
409 index e29afc890cde..c467601299d4 100644
412 @@ -137,3 +137,4 @@ obj-$(CONFIG_IO_MAPPING) += io-mapping.o
413 obj-$(CONFIG_HAVE_BOOTMEM_INFO_NODE) += bootmem_info.o
414 obj-$(CONFIG_GENERIC_IOREMAP) += ioremap.o
415 obj-$(CONFIG_SHRINKER_DEBUG) += shrinker_debug.o
417 diff --git a/mm/filemap.c b/mm/filemap.c
418 index b4c9bd368b7e..ec4f4e5c1a92 100644
421 @@ -3564,7 +3564,7 @@ vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf)
422 vm_fault_t ret = VM_FAULT_LOCKED;
424 sb_start_pagefault(mapping->host->i_sb);
425 - file_update_time(vmf->vma->vm_file);
426 + vma_file_update_time(vmf->vma);
428 if (folio->mapping != mapping) {
430 diff --git a/mm/mmap.c b/mm/mmap.c
431 index 13678edaa22c..4de00f38822c 100644
434 @@ -140,7 +140,7 @@ static void remove_vma(struct vm_area_struct *vma, bool unreachable)
435 if (vma->vm_ops && vma->vm_ops->close)
436 vma->vm_ops->close(vma);
438 - fput(vma->vm_file);
440 mpol_put(vma_policy(vma));
443 @@ -2269,7 +2269,7 @@ int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma,
447 - get_file(new->vm_file);
450 if (new->vm_ops && new->vm_ops->open)
451 new->vm_ops->open(new);
452 @@ -2774,7 +2774,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
454 if (file || vma->vm_file) {
456 - fput(vma->vm_file);
460 /* Undo any partial mapping done by a device driver. */
461 @@ -2843,6 +2843,9 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
462 unsigned long populate = 0;
463 unsigned long ret = -EINVAL;
465 +#if 1 /* IS_ENABLED(CONFIG_AUFS_FS) */
466 + struct file *prfile;
469 pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/mm/remap_file_pages.rst.\n",
470 current->comm, current->pid);
471 @@ -2901,10 +2904,34 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
472 if (vma->vm_flags & VM_LOCKED)
475 +#if 1 /* IS_ENABLED(CONFIG_AUFS_FS) */
477 + file = vma->vm_file;
478 + prfile = vma->vm_prfile;
479 + ret = do_mmap(vma->vm_file, start, size,
480 + prot, flags, pgoff, &populate, NULL);
481 + if (!IS_ERR_VALUE(ret) && file && prfile) {
482 + struct vm_area_struct *new_vma;
484 + new_vma = find_vma(mm, ret);
485 + if (!new_vma->vm_prfile)
486 + new_vma->vm_prfile = prfile;
487 + if (new_vma != vma)
491 + * two fput()s instead of vma_fput(vma),
492 + * coz vma may not be available anymore.
498 file = get_file(vma->vm_file);
499 ret = do_mmap(vma->vm_file, start, size,
500 prot, flags, pgoff, &populate, NULL);
502 +#endif /* CONFIG_AUFS_FS */
504 mmap_write_unlock(mm);
506 @@ -3250,7 +3277,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
507 if (anon_vma_clone(new_vma, vma))
508 goto out_free_mempol;
509 if (new_vma->vm_file)
510 - get_file(new_vma->vm_file);
511 + vma_get_file(new_vma);
512 if (new_vma->vm_ops && new_vma->vm_ops->open)
513 new_vma->vm_ops->open(new_vma);
514 vma_start_write(new_vma);
515 @@ -3266,7 +3293,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
516 new_vma->vm_ops->close(new_vma);
518 if (new_vma->vm_file)
519 - fput(new_vma->vm_file);
522 unlink_anon_vmas(new_vma);
524 diff --git a/mm/nommu.c b/mm/nommu.c
525 index f670d9979a26..94d41bc45915 100644
528 @@ -523,7 +523,7 @@ static void __put_nommu_region(struct vm_region *region)
529 up_write(&nommu_region_sem);
532 - fput(region->vm_file);
535 /* IO memory and memory shared directly out of the pagecache
536 * from ramfs/tmpfs mustn't be released here */
537 @@ -602,7 +602,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
538 if (vma->vm_ops && vma->vm_ops->close)
539 vma->vm_ops->close(vma);
541 - fput(vma->vm_file);
543 put_nommu_region(vma->vm_region);
546 @@ -1124,7 +1124,7 @@ unsigned long do_mmap(struct file *file,
547 goto error_just_free;
550 - fput(region->vm_file);
552 kmem_cache_free(vm_region_jar, region);
555 @@ -1206,10 +1206,10 @@ unsigned long do_mmap(struct file *file,
559 - fput(region->vm_file);
561 kmem_cache_free(vm_region_jar, region);
563 - fput(vma->vm_file);
568 diff --git a/mm/prfile.c b/mm/prfile.c
570 index 000000000000..8f820a235364
574 +// SPDX-License-Identifier: GPL-2.0
576 + * Mainly for aufs which mmap(2) different file and wants to print different
577 + * path in /proc/PID/maps.
578 + * Call these functions via macros defined in linux/mm.h.
580 + * See Documentation/filesystems/aufs/design/06mmap.txt
582 + * Copyright (c) 2014-2022 Junjro R. Okajima
583 + * Copyright (c) 2014 Ian Campbell
586 +#include <linux/mm.h>
587 +#include <linux/file.h>
588 +#include <linux/fs.h>
590 +/* #define PRFILE_TRACE */
591 +static inline void prfile_trace(struct file *f, struct file *pr,
592 + const char func[], int line, const char func2[])
596 + pr_info("%s:%d: %s, %pD2\n", func, line, func2, f);
600 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
603 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
605 + prfile_trace(f, pr, func, line, __func__);
606 + file_update_time(f);
608 + file_update_time(pr);
611 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
614 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
616 + prfile_trace(f, pr, func, line, __func__);
617 + return (f && pr) ? pr : f;
620 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
622 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
624 + prfile_trace(f, pr, func, line, __func__);
630 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
632 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
634 + prfile_trace(f, pr, func, line, __func__);
641 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
644 + struct file *f = region->vm_file, *pr = region->vm_prfile;
646 + prfile_trace(f, pr, func, line, __func__);
647 + return (f && pr) ? pr : f;
650 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
652 + struct file *f = region->vm_file, *pr = region->vm_prfile;
654 + prfile_trace(f, pr, func, line, __func__);
659 +#endif /* !CONFIG_MMU */
660 SPDX-License-Identifier: GPL-2.0
661 aufs6.x-rcN standalone patch
663 diff --git a/fs/dcache.c b/fs/dcache.c
664 index 519321f32f95..267e0c65914a 100644
667 @@ -1450,6 +1450,7 @@ void d_walk(struct dentry *parent, void *data,
671 +EXPORT_SYMBOL_GPL(d_walk);
674 struct vfsmount *mnt;
675 @@ -3052,6 +3053,7 @@ void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
677 write_sequnlock(&rename_lock);
679 +EXPORT_SYMBOL_GPL(d_exchange);
682 * d_ancestor - search for an ancestor
683 diff --git a/fs/exec.c b/fs/exec.c
684 index a466e797c8e2..c60353b5214f 100644
687 @@ -112,6 +112,7 @@ bool path_noexec(const struct path *path)
688 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
689 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
691 +EXPORT_SYMBOL_GPL(path_noexec);
695 diff --git a/fs/fcntl.c b/fs/fcntl.c
696 index 9ea58b7bb580..99fef189bcd6 100644
699 @@ -87,6 +87,7 @@ int setfl(int fd, struct file *filp, unsigned long arg)
703 +EXPORT_SYMBOL_GPL(setfl);
705 static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
707 diff --git a/fs/file_table.c b/fs/file_table.c
708 index 372653b92617..69128fe39b6d 100644
709 --- a/fs/file_table.c
710 +++ b/fs/file_table.c
711 @@ -199,6 +199,7 @@ struct file *alloc_empty_file(int flags, const struct cred *cred)
713 return ERR_PTR(-ENFILE);
715 +EXPORT_SYMBOL_GPL(alloc_empty_file);
718 * Variant of alloc_empty_file() that doesn't check and modify nr_files.
719 diff --git a/fs/namespace.c b/fs/namespace.c
720 index ebb7b4b057ea..6ee47677de2f 100644
723 @@ -457,6 +457,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
724 mnt_dec_writers(real_mount(mnt));
727 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
730 * mnt_drop_write - give up write access to a mount
731 @@ -854,6 +855,7 @@ int is_current_mnt_ns(struct vfsmount *mnt)
733 return check_mnt(real_mount(mnt));
735 +EXPORT_SYMBOL_GPL(is_current_mnt_ns);
738 * vfsmount lock must be held for write
739 @@ -2062,6 +2064,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
743 +EXPORT_SYMBOL_GPL(iterate_mounts);
745 static void lock_mnt_tree(struct mount *mnt)
747 diff --git a/fs/notify/group.c b/fs/notify/group.c
748 index 1de6631a3925..3008eb37a18d 100644
749 --- a/fs/notify/group.c
750 +++ b/fs/notify/group.c
751 @@ -100,6 +100,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
753 refcount_inc(&group->refcnt);
755 +EXPORT_SYMBOL_GPL(fsnotify_get_group);
758 * Drop a reference to a group. Free it if it's through.
759 diff --git a/fs/open.c b/fs/open.c
760 index 4478adcc4f3a..5a984ff90580 100644
763 @@ -67,6 +67,7 @@ int do_truncate(struct mnt_idmap *idmap, struct dentry *dentry,
764 inode_unlock(dentry->d_inode);
767 +EXPORT_SYMBOL_GPL(do_truncate);
769 long vfs_truncate(const struct path *path, loff_t length)
771 diff --git a/fs/read_write.c b/fs/read_write.c
772 index a21ba3be7dbe..ec40061176b3 100644
773 --- a/fs/read_write.c
774 +++ b/fs/read_write.c
775 @@ -477,6 +477,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
779 +EXPORT_SYMBOL_GPL(vfs_read);
781 static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
783 @@ -592,6 +593,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
784 file_end_write(file);
787 +EXPORT_SYMBOL_GPL(vfs_write);
789 /* file_ppos returns &file->f_pos or NULL if file is stream */
790 static inline loff_t *file_ppos(struct file *file)
791 diff --git a/fs/splice.c b/fs/splice.c
792 index 97a66df75d4c..19c8d35f5ebd 100644
795 @@ -872,6 +872,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
796 return warn_unsupported(out, "write");
797 return out->f_op->splice_write(pipe, out, ppos, len, flags);
799 +EXPORT_SYMBOL_GPL(do_splice_from);
802 * Attempt to initiate a splice from a file to a pipe.
803 @@ -901,6 +902,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
804 return warn_unsupported(in, "read");
805 return in->f_op->splice_read(in, ppos, pipe, len, flags);
807 +EXPORT_SYMBOL_GPL(do_splice_to);
810 * splice_direct_to_actor - splices data directly between two non-pipes
811 diff --git a/fs/xattr.c b/fs/xattr.c
812 index fcf67d80d7f9..174dfb9285fb 100644
815 @@ -406,6 +406,7 @@ vfs_getxattr_alloc(struct mnt_idmap *idmap, struct dentry *dentry,
816 *xattr_value = value;
819 +EXPORT_SYMBOL_GPL(vfs_getxattr_alloc);
822 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
823 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
824 index 85d1f7e0bc07..5e20504eed24 100644
825 --- a/kernel/locking/lockdep.c
826 +++ b/kernel/locking/lockdep.c
827 @@ -239,6 +239,7 @@ inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
829 return lock_classes + class_idx;
831 +EXPORT_SYMBOL_GPL(lockdep_hlock_class);
832 #define hlock_class(hlock) lockdep_hlock_class(hlock)
834 #ifdef CONFIG_LOCK_STAT
835 diff --git a/kernel/task_work.c b/kernel/task_work.c
836 index 065e1ef8fc8d..c623c6f0c645 100644
837 --- a/kernel/task_work.c
838 +++ b/kernel/task_work.c
839 @@ -182,3 +182,4 @@ void task_work_run(void)
843 +EXPORT_SYMBOL_GPL(task_work_run);
844 diff --git a/security/security.c b/security/security.c
845 index d5ff7ff45b77..5cca1f5774ef 100644
846 --- a/security/security.c
847 +++ b/security/security.c
848 @@ -1711,6 +1711,7 @@ int security_path_rmdir(const struct path *dir, struct dentry *dentry)
850 return call_int_hook(path_rmdir, 0, dir, dentry);
852 +EXPORT_SYMBOL_GPL(security_path_rmdir);
855 * security_path_unlink() - Check if removing a hard link is allowed
856 @@ -1746,6 +1747,7 @@ int security_path_symlink(const struct path *dir, struct dentry *dentry,
858 return call_int_hook(path_symlink, 0, dir, dentry, old_name);
860 +EXPORT_SYMBOL_GPL(security_path_symlink);
863 * security_path_link - Check if creating a hard link is allowed
864 @@ -1764,6 +1766,7 @@ int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
866 return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
868 +EXPORT_SYMBOL_GPL(security_path_link);
871 * security_path_rename() - Check if renaming a file is allowed
872 @@ -1825,6 +1828,7 @@ int security_path_chmod(const struct path *path, umode_t mode)
874 return call_int_hook(path_chmod, 0, path, mode);
876 +EXPORT_SYMBOL_GPL(security_path_chmod);
879 * security_path_chown() - Check if changing the file's owner/group is allowed
880 @@ -1842,6 +1846,7 @@ int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
882 return call_int_hook(path_chown, 0, path, uid, gid);
884 +EXPORT_SYMBOL_GPL(security_path_chown);
887 * security_path_chroot() - Check if changing the root directory is allowed
888 @@ -2071,6 +2076,7 @@ int security_inode_permission(struct inode *inode, int mask)
890 return call_int_hook(inode_permission, 0, inode, mask);
892 +EXPORT_SYMBOL_GPL(security_inode_permission);
895 * security_inode_setattr() - Check if setting file attributes is allowed
896 @@ -2549,6 +2555,7 @@ int security_file_permission(struct file *file, int mask)
898 return fsnotify_perm(file, mask);
900 +EXPORT_SYMBOL_GPL(security_file_permission);
903 * security_file_alloc() - Allocate and init a file's LSM blob
904 @@ -2815,6 +2822,7 @@ int security_file_truncate(struct file *file)
906 return call_int_hook(file_truncate, 0, file);
908 +EXPORT_SYMBOL_GPL(security_file_truncate);
911 * security_task_alloc() - Allocate a task's LSM blob
912 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
913 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
914 +++ linux/Documentation/ABI/testing/debugfs-aufs 2022-11-05 23:02:18.955889283 +0100
916 +What: /debug/aufs/si_<id>/
918 +Contact: J. R. Okajima <hooanon05g@gmail.com>
920 + Under /debug/aufs, a directory named si_<id> is created
921 + per aufs mount, where <id> is a unique id generated
924 +What: /debug/aufs/si_<id>/plink
926 +Contact: J. R. Okajima <hooanon05g@gmail.com>
928 + It has three lines and shows the information about the
929 + pseudo-link. The first line is a single number
930 + representing a number of buckets. The second line is a
931 + number of pseudo-links per buckets (separated by a
932 + blank). The last line is a single number representing a
933 + total number of psedo-links.
934 + When the aufs mount option 'noplink' is specified, it
935 + will show "1\n0\n0\n".
937 +What: /debug/aufs/si_<id>/xib
939 +Contact: J. R. Okajima <hooanon05g@gmail.com>
941 + It shows the consumed blocks by xib (External Inode Number
942 + Bitmap), its block size and file size.
943 + When the aufs mount option 'noxino' is specified, it
944 + will be empty. About XINO files, see the aufs manual.
946 +What: /debug/aufs/si_<id>/xi<branch-index>
948 +Contact: J. R. Okajima <hooanon05g@gmail.com>
950 + It shows the consumed blocks by xino (External Inode Number
951 + Translation Table), its link count, block size and file
953 + Due to the file size limit, there may exist multiple
954 + xino files per branch. In this case, "-N" is added to
955 + the filename and it corresponds to the index of the
956 + internal xino array. "-0" is omitted.
957 + When the aufs mount option 'noxino' is specified, Those
958 + entries won't exist. About XINO files, see the aufs
961 +What: /debug/aufs/si_<id>/xigen
963 +Contact: J. R. Okajima <hooanon05g@gmail.com>
965 + It shows the consumed blocks by xigen (External Inode
966 + Generation Table), its block size and file size.
967 + If CONFIG_AUFS_EXPORT is disabled, this entry will not
969 + When the aufs mount option 'noxino' is specified, it
970 + will be empty. About XINO files, see the aufs manual.
971 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
972 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
973 +++ linux/Documentation/ABI/testing/sysfs-aufs 2022-11-05 23:02:18.955889283 +0100
975 +What: /sys/fs/aufs/si_<id>/
977 +Contact: J. R. Okajima <hooanon05g@gmail.com>
979 + Under /sys/fs/aufs, a directory named si_<id> is created
980 + per aufs mount, where <id> is a unique id generated
983 +What: /sys/fs/aufs/si_<id>/br<idx>
985 +Contact: J. R. Okajima <hooanon05g@gmail.com>
987 + It shows the abolute path of a member directory (which
988 + is called branch) in aufs, and its permission.
990 +What: /sys/fs/aufs/si_<id>/brid<idx>
992 +Contact: J. R. Okajima <hooanon05g@gmail.com>
994 + It shows the id of a member directory (which is called
997 +What: /sys/fs/aufs/si_<id>/xi_path
999 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1001 + It shows the abolute path of XINO (External Inode Number
1002 + Bitmap, Translation Table and Generation Table) file
1003 + even if it is the default path.
1004 + When the aufs mount option 'noxino' is specified, it
1005 + will be empty. About XINO files, see the aufs manual.
1006 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1007 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
1008 +++ linux/Documentation/filesystems/aufs/design/01intro.txt 2022-11-05 23:02:18.955889283 +0100
1011 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1013 +# This program is free software; you can redistribute it and/or modify
1014 +# it under the terms of the GNU General Public License as published by
1015 +# the Free Software Foundation; either version 2 of the License, or
1016 +# (at your option) any later version.
1018 +# This program is distributed in the hope that it will be useful,
1019 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1020 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1021 +# GNU General Public License for more details.
1023 +# You should have received a copy of the GNU General Public License
1024 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1027 +----------------------------------------
1029 +aufs [ei ju: ef es] | /ey-yoo-ef-es/ | [a u f s]
1030 +1. abbrev. for "advanced multi-layered unification filesystem".
1031 +2. abbrev. for "another unionfs".
1032 +3. abbrev. for "auf das" in German which means "on the" in English.
1033 + Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1034 + But "Filesystem aufs Filesystem" is hard to understand.
1035 +4. abbrev. for "African Urban Fashion Show".
1037 +AUFS is a filesystem with features:
1038 +- multi layered stackable unification filesystem, the member directory
1039 + is called as a branch.
1040 +- branch permission and attribute, 'readonly', 'real-readonly',
1041 + 'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1043 +- internal "file copy-on-write".
1044 +- logical deletion, whiteout.
1045 +- dynamic branch manipulation, adding, deleting and changing permission.
1046 +- allow bypassing aufs, user's direct branch access.
1047 +- external inode number translation table and bitmap which maintains the
1048 + persistent aufs inode number.
1049 +- seekable directory, including NFS readdir.
1050 +- file mapping, mmap and sharing pages.
1051 +- pseudo-link, hardlink over branches.
1052 +- loopback mounted filesystem as a branch.
1053 +- several policies to select one among multiple writable branches.
1054 +- revert a single systemcall when an error occurs in aufs.
1058 +Multi Layered Stackable Unification Filesystem
1059 +----------------------------------------------------------------------
1060 +Most people already knows what it is.
1061 +It is a filesystem which unifies several directories and provides a
1062 +merged single directory. When users access a file, the access will be
1063 +passed/re-directed/converted (sorry, I am not sure which English word is
1064 +correct) to the real file on the member filesystem. The member
1065 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1066 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1067 +readonly branch is handled by creating 'whiteout' on the upper writable
1070 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1071 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1072 +different approaches to implement the merged-view.
1073 +The former tries putting it into VFS, and the latter implements as a
1074 +separate filesystem.
1075 +(If I misunderstand about these implementations, please let me know and
1076 +I shall correct it. Because it is a long time ago when I read their
1077 +source files last time).
1079 +UnionMount's approach will be able to small, but may be hard to share
1080 +branches between several UnionMount since the whiteout in it is
1081 +implemented in the inode on branch filesystem and always
1082 +shared. According to Bharata's post, readdir does not seems to be
1084 +There are several missing features known in this implementations such as
1085 +- for users, the inode number may change silently. eg. copy-up.
1086 +- link(2) may break by copy-up.
1087 +- read(2) may get an obsoleted filedata (fstat(2) too).
1088 +- fcntl(F_SETLK) may be broken by copy-up.
1089 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1092 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1093 +merged into mainline. This is another implementation of UnionMount as a
1094 +separated filesystem. All the limitations and known problems which
1095 +UnionMount are equally inherited to "overlay" filesystem.
1097 +Unionfs has a longer history. When I started implementing a stackable
1098 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1099 +inode, dentry and file objects and they have an array pointing lower
1100 +same kind objects. After contributing many patches for Unionfs, I
1101 +re-started my project AUFS (Jun 2006).
1103 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1104 +implemented my own ideas, approaches and enhancements and it became
1105 +totally different one.
1107 +Comparing DM snapshot and fs based implementation
1108 +- the number of bytes to be copied between devices is much smaller.
1109 +- the type of filesystem must be one and only.
1110 +- the fs must be writable, no readonly fs, even for the lower original
1111 + device. so the compression fs will not be usable. but if we use
1112 + loopback mount, we may address this issue.
1114 + mount /cdrom/squashfs.img /sq
1115 + losetup /sq/ext2.img
1116 + losetup /somewhere/cow
1117 + dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1118 +- it will be difficult (or needs more operations) to extract the
1119 + difference between the original device and COW.
1120 +- DM snapshot-merge may help a lot when users try merging. in the
1121 + fs-layer union, users will use rsync(1).
1123 +You may want to read my old paper "Filesystems in LiveCD"
1124 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1127 +Several characters/aspects/persona of aufs
1128 +----------------------------------------------------------------------
1130 +Aufs has several characters, aspects or persona.
1131 +1. a filesystem, callee of VFS helper
1132 +2. sub-VFS, caller of VFS helper for branches
1133 +3. a virtual filesystem which maintains persistent inode number
1134 +4. reader/writer of files on branches such like an application
1136 +1. Callee of VFS Helper
1137 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1138 +unlink(2) from an application reaches sys_unlink() kernel function and
1139 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1140 +calls filesystem specific unlink operation. Actually aufs implements the
1141 +unlink operation but it behaves like a redirector.
1143 +2. Caller of VFS Helper for Branches
1144 +aufs_unlink() passes the unlink request to the branch filesystem as if
1145 +it were called from VFS. So the called unlink operation of the branch
1146 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1147 +every necessary pre/post operation for the branch filesystem.
1148 +- acquire the lock for the parent dir on a branch
1149 +- lookup in a branch
1150 +- revalidate dentry on a branch
1151 +- mnt_want_write() for a branch
1152 +- vfs_unlink() for a branch
1153 +- mnt_drop_write() for a branch
1154 +- release the lock on a branch
1156 +3. Persistent Inode Number
1157 +One of the most important issue for a filesystem is to maintain inode
1158 +numbers. This is particularly important to support exporting a
1159 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1160 +backend block device for its own. But some storage is necessary to
1161 +keep and maintain the inode numbers. It may be a large space and may not
1162 +suit to keep in memory. Aufs rents some space from its first writable
1163 +branch filesystem (by default) and creates file(s) on it. These files
1164 +are created by aufs internally and removed soon (currently) keeping
1166 +Note: Because these files are removed, they are totally gone after
1167 + unmounting aufs. It means the inode numbers are not persistent
1168 + across unmount or reboot. I have a plan to make them really
1169 + persistent which will be important for aufs on NFS server.
1171 +4. Read/Write Files Internally (copy-on-write)
1172 +Because a branch can be readonly, when you write a file on it, aufs will
1173 +"copy-up" it to the upper writable branch internally. And then write the
1174 +originally requested thing to the file. Generally kernel doesn't
1175 +open/read/write file actively. In aufs, even a single write may cause a
1176 +internal "file copy". This behaviour is very similar to cp(1) command.
1178 +Some people may think it is better to pass such work to user space
1179 +helper, instead of doing in kernel space. Actually I am still thinking
1180 +about it. But currently I have implemented it in kernel space.
1181 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1182 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1183 +++ linux/Documentation/filesystems/aufs/design/02struct.txt 2022-11-05 23:02:18.955889283 +0100
1186 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1188 +# This program is free software; you can redistribute it and/or modify
1189 +# it under the terms of the GNU General Public License as published by
1190 +# the Free Software Foundation; either version 2 of the License, or
1191 +# (at your option) any later version.
1193 +# This program is distributed in the hope that it will be useful,
1194 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1195 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1196 +# GNU General Public License for more details.
1198 +# You should have received a copy of the GNU General Public License
1199 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1201 +Basic Aufs Internal Structure
1203 +Superblock/Inode/Dentry/File Objects
1204 +----------------------------------------------------------------------
1205 +As like an ordinary filesystem, aufs has its own
1206 +superblock/inode/dentry/file objects. All these objects have a
1207 +dynamically allocated array and store the same kind of pointers to the
1208 +lower filesystem, branch.
1209 +For example, when you build a union with one readwrite branch and one
1210 +readonly, mounted /au, /rw and /ro respectively.
1212 +- /ro/fileA exists but /rw/fileA
1214 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1215 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1216 +- [0] = NULL (because /rw/fileA doesn't exist)
1219 +This style of an array is essentially same to the aufs
1220 +superblock/inode/dentry/file objects.
1222 +Because aufs supports manipulating branches, ie. add/delete/change
1223 +branches dynamically, these objects has its own generation. When
1224 +branches are changed, the generation in aufs superblock is
1225 +incremented. And a generation in other object are compared when it is
1226 +accessed. When a generation in other objects are obsoleted, aufs
1227 +refreshes the internal array.
1231 +----------------------------------------------------------------------
1232 +Additionally aufs superblock has some data for policies to select one
1233 +among multiple writable branches, XIB files, pseudo-links and kobject.
1234 +See below in detail.
1235 +About the policies which supports copy-down a directory, see
1236 +wbr_policy.txt too.
1239 +Branch and XINO(External Inode Number Translation Table)
1240 +----------------------------------------------------------------------
1241 +Every branch has its own xino (external inode number translation table)
1242 +file. The xino file is created and unlinked by aufs internally. When two
1243 +members of a union exist on the same filesystem, they share the single
1245 +The struct of a xino file is simple, just a sequence of aufs inode
1246 +numbers which is indexed by the lower inode number.
1247 +In the above sample, assume the inode number of /ro/fileA is i111 and
1248 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1249 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1251 +When the inode numbers are not contiguous, the xino file will be sparse
1252 +which has a hole in it and doesn't consume as much disk space as it
1253 +might appear. If your branch filesystem consumes disk space for such
1254 +holes, then you should specify 'xino=' option at mounting aufs.
1256 +Aufs has a mount option to free the disk blocks for such holes in XINO
1257 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1258 +meet a problem of disk shortage due to XINO files, then you should try
1259 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1260 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1261 +the holes in XINO files.
1263 +Also a writable branch has three kinds of "whiteout bases". All these
1264 +are existed when the branch is joined to aufs, and their names are
1265 +whiteout-ed doubly, so that users will never see their names in aufs
1267 +1. a regular file which will be hardlinked to all whiteouts.
1268 +2. a directory to store a pseudo-link.
1269 +3. a directory to store an "orphan"-ed file temporary.
1272 + When you remove a file on a readonly branch, aufs handles it as a
1273 + logical deletion and creates a whiteout on the upper writable branch
1274 + as a hardlink of this file in order not to consume inode on the
1277 + See below, Pseudo-link.
1279 + When "fileC" exists on the lower readonly branch only and it is
1280 + opened and removed with its parent dir, and then user writes
1281 + something into it, then aufs copies-up fileC to this
1282 + directory. Because there is no other dir to store fileC. After
1283 + creating a file under this dir, the file is unlinked.
1285 +Because aufs supports manipulating branches, ie. add/delete/change
1286 +dynamically, a branch has its own id. When the branch order changes,
1287 +aufs finds the new index by searching the branch id.
1291 +----------------------------------------------------------------------
1292 +Assume "fileA" exists on the lower readonly branch only and it is
1293 +hardlinked to "fileB" on the branch. When you write something to fileA,
1294 +aufs copies-up it to the upper writable branch. Additionally aufs
1295 +creates a hardlink under the Pseudo-link Directory of the writable
1296 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1297 +simple list. If fileB is read after unlinking fileA, aufs returns
1298 +filedata from the pseudo-link instead of the lower readonly
1299 +branch. Because the pseudo-link is based upon the inode, to keep the
1300 +inode number by xino (see above) is essentially necessary.
1302 +All the hardlinks under the Pseudo-link Directory of the writable branch
1303 +should be restored in a proper location later. Aufs provides a utility
1304 +to do this. The userspace helpers executed at remounting and unmounting
1306 +During this utility is running, it puts aufs into the pseudo-link
1307 +maintenance mode. In this mode, only the process which began the
1308 +maintenance mode (and its child processes) is allowed to operate in
1309 +aufs. Some other processes which are not related to the pseudo-link will
1310 +be allowed to run too, but the rest have to return an error or wait
1311 +until the maintenance mode ends. If a process already acquires an inode
1312 +mutex (in VFS), it has to return an error.
1315 +XIB(external inode number bitmap)
1316 +----------------------------------------------------------------------
1317 +Addition to the xino file per a branch, aufs has an external inode number
1318 +bitmap in a superblock object. It is also an internal file such like a
1320 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1322 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1324 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1325 +reduce the number of consumed disk blocks for these files.
1328 +Virtual or Vertical Dir, and Readdir in Userspace
1329 +----------------------------------------------------------------------
1330 +In order to support multiple layers (branches), aufs readdir operation
1331 +constructs a virtual dir block on memory. For readdir, aufs calls
1332 +vfs_readdir() internally for each dir on branches, merges their entries
1333 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1334 +object. So the file object has its entry list until it is closed. The
1335 +entry list will be updated when the file position is zero and becomes
1336 +obsoleted. This decision is made in aufs automatically.
1338 +The dynamically allocated memory block for the name of entries has a
1339 +unit of 512 bytes (by default) and stores the names contiguously (no
1340 +padding). Another block for each entry is handled by kmem_cache too.
1341 +During building dir blocks, aufs creates hash list and judging whether
1342 +the entry is whiteouted by its upper branch or already listed.
1343 +The merged result is cached in the corresponding inode object and
1344 +maintained by a customizable life-time option.
1346 +Some people may call it can be a security hole or invite DoS attack
1347 +since the opened and once readdir-ed dir (file object) holds its entry
1348 +list and becomes a pressure for system memory. But I'd say it is similar
1349 +to files under /proc or /sys. The virtual files in them also holds a
1350 +memory page (generally) while they are opened. When an idea to reduce
1351 +memory for them is introduced, it will be applied to aufs too.
1352 +For those who really hate this situation, I've developed readdir(3)
1353 +library which operates this merging in userspace. You just need to set
1354 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1355 +kernel space for readdir(3).
1359 +----------------------------------------------------------------------
1360 +Aufs sometimes requires privilege access to a branch. For instance,
1361 +in copy-up/down operation. When a user process is going to make changes
1362 +to a file which exists in the lower readonly branch only, and the mode
1363 +of one of ancestor directories may not be writable by a user
1364 +process. Here aufs copy-up the file with its ancestors and they may
1365 +require privilege to set its owner/group/mode/etc.
1366 +This is a typical case of a application character of aufs (see
1369 +Aufs uses workqueue synchronously for this case. It creates its own
1370 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1371 +passes the request to call mkdir or write (for example), and wait for
1372 +its completion. This approach solves a problem of a signal handler
1374 +If aufs didn't adopt the workqueue and changed the privilege of the
1375 +process, then the process may receive the unexpected SIGXFSZ or other
1378 +Also aufs uses the system global workqueue ("events" kernel thread) too
1379 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1380 +whiteout base and etc. This is unrelated to a privilege.
1381 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1382 +superblock at the beginning, at the same time waits for the completion
1383 +of all queued asynchronous tasks.
1387 +----------------------------------------------------------------------
1388 +The whiteout in aufs is very similar to Unionfs's. That is represented
1389 +by its filename. UnionMount takes an approach of a file mode, but I am
1390 +afraid several utilities (find(1) or something) will have to support it.
1392 +Basically the whiteout represents "logical deletion" which stops aufs to
1393 +lookup further, but also it represents "dir is opaque" which also stop
1396 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1397 +In order to make several functions in a single systemcall to be
1398 +revertible, aufs adopts an approach to rename a directory to a temporary
1399 +unique whiteouted name.
1400 +For example, in rename(2) dir where the target dir already existed, aufs
1401 +renames the target dir to a temporary unique whiteouted name before the
1402 +actual rename on a branch, and then handles other actions (make it opaque,
1403 +update the attributes, etc). If an error happens in these actions, aufs
1404 +simply renames the whiteouted name back and returns an error. If all are
1405 +succeeded, aufs registers a function to remove the whiteouted unique
1406 +temporary name completely and asynchronously to the system global
1411 +----------------------------------------------------------------------
1412 +It is a well-known feature or concept.
1413 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1414 +internally and makes change to the new file on the upper writable branch.
1415 +When the trigger systemcall does not update the timestamps of the parent
1416 +dir, aufs reverts it after copy-up.
1419 +Move-down (aufs3.9 and later)
1420 +----------------------------------------------------------------------
1421 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1422 +the lower readonly branch to the upper writable branch when a user
1423 +changes something about the file.
1424 +"Move-down" is an opposite action of copy-up. Basically this action is
1425 +ran manually instead of automatically and internally.
1426 +For desgin and implementation, aufs has to consider these issues.
1427 +- whiteout for the file may exist on the lower branch.
1428 +- ancestor directories may not exist on the lower branch.
1429 +- diropq for the ancestor directories may exist on the upper branch.
1430 +- free space on the lower branch will reduce.
1431 +- another access to the file may happen during moving-down, including
1432 + UDBA (see "Revalidate Dentry and UDBA").
1433 +- the file should not be hard-linked nor pseudo-linked. they should be
1434 + handled by auplink utility later.
1436 +Sometimes users want to move-down a file from the upper writable branch
1437 +to the lower readonly or writable branch. For instance,
1438 +- the free space of the upper writable branch is going to run out.
1439 +- create a new intermediate branch between the upper and lower branch.
1442 +For this purpose, use "aumvdown" command in aufs-util.git.
1443 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1444 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt 1970-01-01 01:00:00.000000000 +0100
1445 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt 2022-11-05 23:02:18.955889283 +0100
1448 +# Copyright (C) 2015-2022 Junjiro R. Okajima
1450 +# This program is free software; you can redistribute it and/or modify
1451 +# it under the terms of the GNU General Public License as published by
1452 +# the Free Software Foundation; either version 2 of the License, or
1453 +# (at your option) any later version.
1455 +# This program is distributed in the hope that it will be useful,
1456 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1457 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1458 +# GNU General Public License for more details.
1460 +# You should have received a copy of the GNU General Public License
1461 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1463 +Support for a branch who has its ->atomic_open()
1464 +----------------------------------------------------------------------
1465 +The filesystems who implement its ->atomic_open() are not majority. For
1466 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1467 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1468 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1469 +sure whether all filesystems who have ->atomic_open() behave like this,
1470 +but NFSv4 surely returns the error.
1472 +In order to support ->atomic_open() for aufs, there are a few
1475 +A. Introduce aufs_atomic_open()
1476 + - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1478 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1479 + an aufs user Pip Cet's approach
1480 + - calls aufs_create(), VFS finish_open() and notify_change().
1481 + - pass fake-mode to finish_open(), and then correct the mode by
1483 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1484 + - no aufs_atomic_open().
1485 + - aufs_lookup() registers the TID to an aufs internal object.
1486 + - aufs_create() does nothing when the matching TID is registered, but
1487 + registers the mode.
1488 + - aufs_open() calls branch fs's ->atomic_open() when the matching
1489 + TID is registered.
1490 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1492 + - no aufs_atomic_open().
1493 + - aufs_create() registers the TID to an internal object. this info
1494 + represents "this process created this file just now."
1495 + - when aufs gets EACCES from branch fs's ->open(), then confirm the
1496 + registered TID and re-try open() with superuser's credential.
1498 +Pros and cons for each approach.
1501 + - straightforward but highly depends upon VFS internal.
1502 + - the atomic behavaiour is kept.
1503 + - some of parameters such as nameidata are hard to reproduce for
1507 + - easy to implement.
1508 + - the atomic behavaiour is lost.
1510 + - the atomic behavaiour is kept.
1511 + - dirty and tricky.
1512 + - VFS checks whether the file is created correctly after calling
1513 + ->create(), which means this approach doesn't work.
1515 + - easy to implement.
1516 + - the atomic behavaiour is lost.
1517 + - to open a file with superuser's credential and give it to a user
1518 + process is a bad idea, since the file object keeps the credential
1519 + in it. It may affect LSM or something. This approach doesn't work
1522 +The approach A is ideal, but it hard to implement. So here is a
1523 +variation of A, which is to be implemented.
1525 +A-1. Introduce aufs_atomic_open()
1526 + - calls branch fs ->atomic_open() if exists. otherwise calls
1527 + vfs_create() and finish_open().
1528 + - the demerit is that the several checks after branch fs
1529 + ->atomic_open() are lost. in the ordinary case, the checks are
1530 + done by VFS:do_last(), lookup_open() and atomic_open(). some can
1531 + be implemented in aufs, but not all I am afraid.
1532 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1533 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1534 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2022-11-05 23:02:18.959222617 +0100
1537 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1539 +# This program is free software; you can redistribute it and/or modify
1540 +# it under the terms of the GNU General Public License as published by
1541 +# the Free Software Foundation; either version 2 of the License, or
1542 +# (at your option) any later version.
1544 +# This program is distributed in the hope that it will be useful,
1545 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1546 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1547 +# GNU General Public License for more details.
1549 +# You should have received a copy of the GNU General Public License
1550 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1553 +----------------------------------------------------------------------
1554 +Since aufs has a character of sub-VFS (see Introduction), it operates
1555 +lookup for branches as VFS does. It may be a heavy work. But almost all
1556 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1557 +directly connected to its parent. Digging down the directory hierarchy
1558 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1561 +When a branch is a remote filesystem, aufs basically relies upon its
1562 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1564 +For d_revalidate, aufs implements three levels of revalidate tests. See
1565 +"Revalidate Dentry and UDBA" in detail.
1568 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1569 +----------------------------------------------------------------------
1570 +Let's try case study.
1571 +- aufs has two branches, upper readwrite and lower readonly.
1573 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1574 +- user invoked "chmod a+rx /au/dirA"
1575 +- the internal copy-up is activated and "/rw/dirA" is created and its
1576 + permission bits are set to world readable.
1577 +- then "/au/dirA" becomes world readable?
1579 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1580 +or it may be a natively readonly filesystem. If aufs respects the lower
1581 +branch, it should not respond readdir request from other users. But user
1582 +allowed it by chmod. Should really aufs rejects showing the entries
1585 +To be honest, I don't have a good solution for this case. So aufs
1586 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1588 +When dirperm1 is specified, aufs checks only the highest one for the
1589 +directory permission, and shows the entries. Otherwise, as usual, checks
1590 +every dir existing on all branches and rejects the request.
1592 +As a side effect, dirperm1 option improves the performance of aufs
1593 +because the number of permission check is reduced when the number of
1597 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1598 +----------------------------------------------------------------------
1599 +Generally VFS helpers re-validate a dentry as a part of lookup.
1600 +0. digging down the directory hierarchy.
1601 +1. lock the parent dir by its i_mutex.
1602 +2. lookup the final (child) entry.
1604 +4. call the actual operation (create, unlink, etc.)
1605 +5. unlock the parent dir
1607 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1608 +called. Actually aufs implements it and checks the dentry on a branch is
1610 +But it is not enough. Because aufs has to release the lock for the
1611 +parent dir on a branch at the end of ->lookup() (step 2) and
1612 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1614 +If the file on a branch is changed directly, eg. bypassing aufs, after
1615 +aufs released the lock, then the subsequent operation may cause
1616 +something unpleasant result.
1618 +This situation is a result of VFS architecture, ->lookup() and
1619 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1620 +design from VFS's point of view. It is just not suitable for sub-VFS
1623 +Aufs supports such case by three level of revalidation which is
1624 +selectable by user.
1625 +1. Simple Revalidate
1626 + Addition to the native flow in VFS's, confirm the child-parent
1627 + relationship on the branch just after locking the parent dir on the
1628 + branch in the "actual operation" (step 4). When this validation
1629 + fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1630 + checks the validation of the dentry on branches.
1631 +2. Monitor Changes Internally by Inotify/Fsnotify
1632 + Addition to above, in the "actual operation" (step 4) aufs re-lookup
1633 + the dentry on the branch, and returns EBUSY if it finds different
1635 + Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1636 + during it is in cache. When the event is notified, aufs registers a
1637 + function to kernel 'events' thread by schedule_work(). And the
1638 + function sets some special status to the cached aufs dentry and inode
1639 + private data. If they are not cached, then aufs has nothing to
1640 + do. When the same file is accessed through aufs (step 0-3) later,
1641 + aufs will detect the status and refresh all necessary data.
1642 + In this mode, aufs has to ignore the event which is fired by aufs
1644 +3. No Extra Validation
1645 + This is the simplest test and doesn't add any additional revalidation
1646 + test, and skip the revalidation in step 4. It is useful and improves
1647 + aufs performance when system surely hide the aufs branches from user,
1648 + by over-mounting something (or another method).
1649 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1650 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1651 +++ linux/Documentation/filesystems/aufs/design/04branch.txt 2022-11-05 23:02:18.959222617 +0100
1654 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1656 +# This program is free software; you can redistribute it and/or modify
1657 +# it under the terms of the GNU General Public License as published by
1658 +# the Free Software Foundation; either version 2 of the License, or
1659 +# (at your option) any later version.
1661 +# This program is distributed in the hope that it will be useful,
1662 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1663 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1664 +# GNU General Public License for more details.
1666 +# You should have received a copy of the GNU General Public License
1667 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1669 +Branch Manipulation
1671 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1672 +and changing its permission/attribute, there are a lot of works to do.
1676 +----------------------------------------------------------------------
1677 +o Confirm the adding dir exists outside of aufs, including loopback
1678 + mount, and its various attributes.
1679 +o Initialize the xino file and whiteout bases if necessary.
1682 +o Check the owner/group/mode of the directory
1683 + When the owner/group/mode of the adding directory differs from the
1684 + existing branch, aufs issues a warning because it may impose a
1686 + For example, when a upper writable branch has a world writable empty
1687 + top directory, a malicious user can create any files on the writable
1688 + branch directly, like copy-up and modify manually. If something like
1689 + /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1690 + writable branch, and the writable branch is world-writable, then a
1691 + malicious guy may create /etc/passwd on the writable branch directly
1692 + and the infected file will be valid in aufs.
1693 + I am afraid it can be a security issue, but aufs can do nothing except
1694 + producing a warning.
1698 +----------------------------------------------------------------------
1699 +o Confirm the deleting branch is not busy
1700 + To be general, there is one merit to adopt "remount" interface to
1701 + manipulate branches. It is to discard caches. At deleting a branch,
1702 + aufs checks the still cached (and connected) dentries and inodes. If
1703 + there are any, then they are all in-use. An inode without its
1704 + corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1706 + For the cached one, aufs checks whether the same named entry exists on
1708 + If the cached one is a directory, because aufs provides a merged view
1709 + to users, as long as one dir is left on any branch aufs can show the
1710 + dir to users. In this case, the branch can be removed from aufs.
1711 + Otherwise aufs rejects deleting the branch.
1713 + If any file on the deleting branch is opened by aufs, then aufs
1717 +Modify the Permission of a Branch
1718 +----------------------------------------------------------------------
1719 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1722 +o rw --> ro: Confirm the modifying branch is not busy
1723 + Aufs rejects the request if any of these conditions are true.
1724 + - a file on the branch is mmap-ed.
1725 + - a regular file on the branch is opened for write and there is no
1726 + same named entry on the upper branch.
1727 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1728 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
1729 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2022-11-05 23:02:18.959222617 +0100
1732 +# Copyright (C) 2005-2022 Junjiro R. Okajima
1734 +# This program is free software; you can redistribute it and/or modify
1735 +# it under the terms of the GNU General Public License as published by
1736 +# the Free Software Foundation; either version 2 of the License, or
1737 +# (at your option) any later version.
1739 +# This program is distributed in the hope that it will be useful,
1740 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1741 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1742 +# GNU General Public License for more details.
1744 +# You should have received a copy of the GNU General Public License
1745 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1747 +Policies to Select One among Multiple Writable Branches
1748 +----------------------------------------------------------------------
1749 +When the number of writable branch is more than one, aufs has to decide
1750 +the target branch for file creation or copy-up. By default, the highest
1751 +writable branch which has the parent (or ancestor) dir of the target
1752 +file is chosen (top-down-parent policy).
1753 +By user's request, aufs implements some other policies to select the
1754 +writable branch, for file creation several policies, round-robin,
1755 +most-free-space, and other policies. For copy-up, top-down-parent,
1756 +bottom-up-parent, bottom-up and others.
1758 +As expected, the round-robin policy selects the branch in circular. When
1759 +you have two writable branches and creates 10 new files, 5 files will be
1760 +created for each branch. mkdir(2) systemcall is an exception. When you
1761 +create 10 new directories, all will be created on the same branch.
1762 +And the most-free-space policy selects the one which has most free
1763 +space among the writable branches. The amount of free space will be
1764 +checked by aufs internally, and users can specify its time interval.
1766 +The policies for copy-up is more simple,
1767 +top-down-parent is equivalent to the same named on in create policy,
1768 +bottom-up-parent selects the writable branch where the parent dir
1769 +exists and the nearest upper one from the copyup-source,
1770 +bottom-up selects the nearest upper writable branch from the
1771 +copyup-source, regardless the existence of the parent dir.
1773 +There are some rules or exceptions to apply these policies.
1774 +- If there is a readonly branch above the policy-selected branch and
1775 + the parent dir is marked as opaque (a variation of whiteout), or the
1776 + target (creating) file is whiteout-ed on the upper readonly branch,
1777 + then the result of the policy is ignored and the target file will be
1778 + created on the nearest upper writable branch than the readonly branch.
1779 +- If there is a writable branch above the policy-selected branch and
1780 + the parent dir is marked as opaque or the target file is whiteouted
1781 + on the branch, then the result of the policy is ignored and the target
1782 + file will be created on the highest one among the upper writable
1783 + branches who has diropq or whiteout. In case of whiteout, aufs removes
1785 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1786 + They try selecting the branch where the source exists as possible
1787 + since copyup a large file will take long time. If it can't be,
1788 + ie. the branch where the source exists is readonly, then they will
1789 + follow the copyup policy.
1790 +- There is an exception for rename(2) when the target exists.
1791 + If the rename target exists, aufs compares the index of the branches
1792 + where the source and the target exists and selects the higher
1793 + one. If the selected branch is readonly, then aufs follows the
1795 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot linux/Documentation/filesystems/aufs/design/06dirren.dot
1796 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot 1970-01-01 01:00:00.000000000 +0100
1797 +++ linux/Documentation/filesystems/aufs/design/06dirren.dot 2022-11-05 23:02:18.959222617 +0100
1800 +// to view this graph, run dot(1) command in GRAPHVIZ.
1802 +// This program is free software; you can redistribute it and/or modify
1803 +// it under the terms of the GNU General Public License as published by
1804 +// the Free Software Foundation; either version 2 of the License, or
1805 +// (at your option) any later version.
1807 +// This program is distributed in the hope that it will be useful,
1808 +// but WITHOUT ANY WARRANTY; without even the implied warranty of
1809 +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1810 +// GNU General Public License for more details.
1812 +// You should have received a copy of the GNU General Public License
1813 +// along with this program. If not, see <http://www.gnu.org/licenses/>.
1817 +whinfo [label="detailed info file\n(lower_brid_root-hinum, h_inum, namelen, old name)"];
1821 +aufs_rename -> whinfo [label="store/remove"];
1824 +inode_list [label="h_inum list in branch\ncache"];
1827 +whinode [label="h_inum list file"];
1830 +brmgmt [label="br_add/del/mod/umount"];
1832 +brmgmt -> inode_list [label="create/remove"];
1833 +brmgmt -> whinode [label="load/store"];
1835 +inode_list -> whinode [style=dashed,dir=both];
1837 +aufs_rename -> inode_list [label="add/del"];
1839 +aufs_lookup -> inode_list [label="search"];
1841 +aufs_lookup -> whinfo [label="load/remove"];
1843 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt linux/Documentation/filesystems/aufs/design/06dirren.txt
1844 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt 1970-01-01 01:00:00.000000000 +0100
1845 +++ linux/Documentation/filesystems/aufs/design/06dirren.txt 2022-11-05 23:02:18.959222617 +0100
1848 +# Copyright (C) 2017-2022 Junjiro R. Okajima
1850 +# This program is free software; you can redistribute it and/or modify
1851 +# it under the terms of the GNU General Public License as published by
1852 +# the Free Software Foundation; either version 2 of the License, or
1853 +# (at your option) any later version.
1855 +# This program is distributed in the hope that it will be useful,
1856 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1857 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1858 +# GNU General Public License for more details.
1860 +# You should have received a copy of the GNU General Public License
1861 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1863 +Special handling for renaming a directory (DIRREN)
1864 +----------------------------------------------------------------------
1865 +First, let's assume we have a simple usecase.
1869 +- /ro/dirA and /ro/dirA/file exist too
1870 +- there is no dirB on both branches
1871 +- a user issues rename("dirA", "dirB")
1873 +Now, what should aufs behave against this rename(2)?
1874 +There are a few possible cases.
1877 + since dirA exists on a readonly branch which cannot be renamed.
1879 + it is possible to copy-up dirA (only the dir itself), but the child
1880 + entries ("file" in this case) should not be. it must be a bad
1881 + approach to copy-up recursively.
1882 +C. returns a success.
1883 + even the branch /ro is readonly, aufs tries renaming it. Obviously it
1884 + is a violation of aufs' policy.
1885 +D. construct an extra information which indicates that /ro/dirA should
1886 + be handled as the name of dirB.
1887 + overlayfs has a similar feature called REDIRECT.
1889 +Until now, aufs implements the case B only which returns EXDEV, and
1890 +expects the userspace application behaves like mv(1) which tries
1891 +issueing rename(2) recursively.
1893 +A new aufs feature called DIRREN is introduced which implements the case
1894 +D. There are several "extra information" added.
1896 +1. detailed info per renamed directory
1897 + path: /rw/dirB/$AUFS_WH_DR_INFO_PFX.<lower branch-id>
1898 +2. the inode-number list of directories on a branch
1899 + path: /rw/dirB/$AUFS_WH_DR_BRHINO
1901 +The filename of "detailed info per directory" represents the lower
1902 +branch, and its format is
1903 +- a type of the branch id
1905 + + uuid (not implemented yet)
1908 +- the inode-number of the branch root dir
1910 +And it contains these info in a single regular file.
1912 +- branch's inode-number of the logically renamed dir
1913 +- the name of the before-renamed dir
1915 +The "detailed info per directory" file is created in aufs rename(2), and
1916 +loaded in any lookup.
1917 +The info is considered in lookup for the matching case only. Here
1918 +"matching" means that the root of branch (in the info filename) is same
1919 +to the current looking-up branch. After looking-up the before-renamed
1920 +name, the inode-number is compared. And the matched dentry is used.
1922 +The "inode-number list of directories" is a regular file which contains
1923 +simply the inode-numbers on the branch. The file is created or updated
1924 +in removing the branch, and loaded in adding the branch. Its lifetime is
1925 +equal to the branch.
1926 +The list is referred in lookup, and when the current target inode is
1927 +found in the list, the aufs tries loading the "detailed info per
1928 +directory" and get the changed and valid name of the dir.
1930 +Theoretically these "extra informaiton" may be able to be put into XATTR
1931 +in the dir inode. But aufs doesn't choose this way because
1932 +1. XATTR may not be supported by the branch (or its configuration)
1933 +2. XATTR may have its size limit.
1934 +3. XATTR may be less easy to convert than a regular file, when the
1935 + format of the info is changed in the future.
1936 +At the same time, I agree that the regular file approach is much slower
1937 +than XATTR approach. So, in the future, aufs may take the XATTR or other
1940 +This DIRREN feature is enabled by aufs configuration, and is activated
1941 +by a new mount option.
1943 +For the more complicated case, there is a work with UDBA option, which
1944 +is to dected the direct access to the branches (by-passing aufs) and to
1945 +maintain the cashes in aufs. Since a single cached aufs dentry may
1946 +contains two names, before- and after-rename, the name comparision in
1947 +UDBA handler may not work correctly. In this case, the behaviour will be
1948 +equivalen to udba=reval case.
1949 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1950 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt 1970-01-01 01:00:00.000000000 +0100
1951 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt 2022-11-05 23:02:18.959222617 +0100
1954 +# Copyright (C) 2011-2022 Junjiro R. Okajima
1956 +# This program is free software; you can redistribute it and/or modify
1957 +# it under the terms of the GNU General Public License as published by
1958 +# the Free Software Foundation; either version 2 of the License, or
1959 +# (at your option) any later version.
1961 +# This program is distributed in the hope that it will be useful,
1962 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1963 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1964 +# GNU General Public License for more details.
1966 +# You should have received a copy of the GNU General Public License
1967 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1969 +File-based Hierarchical Storage Management (FHSM)
1970 +----------------------------------------------------------------------
1971 +Hierarchical Storage Management (or HSM) is a well-known feature in the
1972 +storage world. Aufs provides this feature as file-based with multiple
1973 +writable branches, based upon the principle of "Colder, the Lower".
1974 +Here the word "colder" means that the less used files, and "lower" means
1975 +that the position in the order of the stacked branches vertically.
1976 +These multiple writable branches are prioritized, ie. the topmost one
1977 +should be the fastest drive and be used heavily.
1979 +o Characters in aufs FHSM story
1980 +- aufs itself and a new branch attribute.
1981 +- a new ioctl interface to move-down and to establish a connection with
1982 + the daemon ("move-down" is a converse of "copy-up").
1983 +- userspace tool and daemon.
1985 +The userspace daemon establishes a connection with aufs and waits for
1986 +the notification. The notified information is very similar to struct
1987 +statfs containing the number of consumed blocks and inodes.
1988 +When the consumed blocks/inodes of a branch exceeds the user-specified
1989 +upper watermark, the daemon activates its move-down process until the
1990 +consumed blocks/inodes reaches the user-specified lower watermark.
1992 +The actual move-down is done by aufs based upon the request from
1993 +user-space since we need to maintain the inode number and the internal
1994 +pointer arrays in aufs.
1996 +Currently aufs FHSM handles the regular files only. Additionally they
1997 +must not be hard-linked nor pseudo-linked.
2000 +o Cowork of aufs and the user-space daemon
2001 + During the userspace daemon established the connection, aufs sends a
2002 + small notification to it whenever aufs writes something into the
2003 + writable branch. But it may cost high since aufs issues statfs(2)
2004 + internally. So user can specify a new option to cache the
2005 + info. Actually the notification is controlled by these factors.
2006 + + the specified cache time.
2007 + + classified as "force" by aufs internally.
2008 + Until the specified time expires, aufs doesn't send the info
2009 + except the forced cases. When aufs decide forcing, the info is always
2010 + notified to userspace.
2011 + For example, the number of free inodes is generally large enough and
2012 + the shortage of it happens rarely. So aufs doesn't force the
2013 + notification when creating a new file, directory and others. This is
2014 + the typical case which aufs doesn't force.
2015 + When aufs writes the actual filedata and the files consumes any of new
2016 + blocks, the aufs forces notifying.
2019 +o Interfaces in aufs
2020 +- New branch attribute.
2022 + Specifies that the branch is managed by FHSM feature. In other word,
2023 + participant in the FHSM.
2024 + When nofhsm is set to the branch, it will not be the source/target
2025 + branch of the move-down operation. This attribute is set
2026 + independently from coo and moo attributes, and if you want full
2027 + FHSM, you should specify them as well.
2028 +- New mount option.
2030 + Specifies a second to suppress many less important info to be
2033 + + AUFS_CTL_FHSM_FD
2034 + create a new file descriptor which userspace can read the notification
2035 + (a subset of struct statfs) from aufs.
2036 +- Module parameter 'brs'
2037 + It has to be set to 1. Otherwise the new mount option 'fhsm' will not
2039 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
2040 + When there are two or more branches with fhsm attributes,
2041 + /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
2042 + terminates it. As a result of remounting and branch-manipulation, the
2043 + number of branches with fhsm attribute can be one. In this case,
2044 + /sbin/mount.aufs will terminate the user-space daemon.
2047 +Finally the operation is done as these steps in kernel-space.
2049 + + no one else is using the file.
2050 + + the file is not hard-linked.
2051 + + the file is not pseudo-linked.
2052 + + the file is a regular file.
2053 + + the parent dir is not opaqued.
2054 +- find the target writable branch.
2055 +- make sure the file is not whiteout-ed by the upper (than the target)
2057 +- make the parent dir on the target branch.
2058 +- mutex lock the inode on the branch.
2059 +- unlink the whiteout on the target branch (if exists).
2060 +- lookup and create the whiteout-ed temporary name on the target branch.
2061 +- copy the file as the whiteout-ed temporary name on the target branch.
2062 +- rename the whiteout-ed temporary name to the original name.
2063 +- unlink the file on the source branch.
2064 +- maintain the internal pointer array and the external inode number
2066 +- maintain the timestamps and other attributes of the parent dir and the
2069 +And of course, in every step, an error may happen. So the operation
2070 +should restore the original file state after an error happens.
2071 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
2072 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
2073 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2022-11-05 23:02:18.959222617 +0100
2076 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2078 +# This program is free software; you can redistribute it and/or modify
2079 +# it under the terms of the GNU General Public License as published by
2080 +# the Free Software Foundation; either version 2 of the License, or
2081 +# (at your option) any later version.
2083 +# This program is distributed in the hope that it will be useful,
2084 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2085 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2086 +# GNU General Public License for more details.
2088 +# You should have received a copy of the GNU General Public License
2089 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2091 +mmap(2) -- File Memory Mapping
2092 +----------------------------------------------------------------------
2093 +In aufs, the file-mapped pages are handled by a branch fs directly, no
2094 +interaction with aufs. It means aufs_mmap() calls the branch fs's
2096 +This approach is simple and good, but there is one problem.
2097 +Under /proc, several entries show the mmapped files by its path (with
2098 +device and inode number), and the printed path will be the path on the
2099 +branch fs's instead of virtual aufs's.
2100 +This is not a problem in most cases, but some utilities lsof(1) (and its
2101 +user) may expect the path on aufs.
2103 +To address this issue, aufs adds a new member called vm_prfile in struct
2104 +vm_area_struct (and struct vm_region). The original vm_file points to
2105 +the file on the branch fs in order to handle everything correctly as
2106 +usual. The new vm_prfile points to a virtual file in aufs, and the
2107 +show-functions in procfs refers to vm_prfile if it is set.
2108 +Also we need to maintain several other places where touching vm_file
2110 +- fork()/clone() copies vma and the reference count of vm_file is
2112 +- merging vma maintains the ref count too.
2114 +This is not a good approach. It just fakes the printed path. But it
2115 +leaves all behaviour around f_mapping unchanged. This is surely an
2117 +Actually aufs had adopted another complicated approach which calls
2118 +generic_file_mmap() and handles struct vm_operations_struct. In this
2119 +approach, aufs met a hard problem and I could not solve it without
2120 +switching the approach.
2122 +There may be one more another approach which is
2123 +- bind-mount the branch-root onto the aufs-root internally
2124 +- grab the new vfsmount (ie. struct mount)
2125 +- lazy-umount the branch-root internally
2126 +- in open(2) the aufs-file, open the branch-file with the hidden
2127 + vfsmount (instead of the original branch's vfsmount)
2128 +- ideally this "bind-mount and lazy-umount" should be done atomically,
2129 + but it may be possible from userspace by the mount helper.
2131 +Adding the internal hidden vfsmount and using it in opening a file, the
2132 +file path under /proc will be printed correctly. This approach looks
2133 +smarter, but is not possible I am afraid.
2134 +- aufs-root may be bind-mount later. when it happens, another hidden
2135 + vfsmount will be required.
2136 +- it is hard to get the chance to bind-mount and lazy-umount
2137 + + in kernel-space, FS can have vfsmount in open(2) via
2138 + file->f_path, and aufs can know its vfsmount. But several locks are
2139 + already acquired, and if aufs tries to bind-mount and lazy-umount
2140 + here, then it may cause a deadlock.
2141 + + in user-space, bind-mount doesn't invoke the mount helper.
2142 +- since /proc shows dev and ino, aufs has to give vma these info. it
2143 + means a new member vm_prinode will be necessary. this is essentially
2144 + equivalent to vm_prfile described above.
2146 +I have to give up this "looks-smater" approach.
2147 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2148 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt 1970-01-01 01:00:00.000000000 +0100
2149 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt 2022-11-05 23:02:18.959222617 +0100
2152 +# Copyright (C) 2014-2022 Junjiro R. Okajima
2154 +# This program is free software; you can redistribute it and/or modify
2155 +# it under the terms of the GNU General Public License as published by
2156 +# the Free Software Foundation; either version 2 of the License, or
2157 +# (at your option) any later version.
2159 +# This program is distributed in the hope that it will be useful,
2160 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2161 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2162 +# GNU General Public License for more details.
2164 +# You should have received a copy of the GNU General Public License
2165 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2167 +Listing XATTR/EA and getting the value
2168 +----------------------------------------------------------------------
2169 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2170 +shows the values from the topmost existing file. This behaviour is good
2171 +for the non-dir entries since the bahaviour exactly matches the shown
2172 +information. But for the directories, aufs considers all the same named
2173 +entries on the lower branches. Which means, if one of the lower entry
2174 +rejects readdir call, then aufs returns an error even if the topmost
2175 +entry allows it. This behaviour is necessary to respect the branch fs's
2176 +security, but can make users confused since the user-visible standard
2177 +attributes don't match the behaviour.
2178 +To address this issue, aufs has a mount option called dirperm1 which
2179 +checks the permission for the topmost entry only, and ignores the lower
2180 +entry's permission.
2182 +A similar issue can happen around XATTR.
2183 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2184 +always set. Otherwise these very unpleasant situation would happen.
2185 +- listxattr(2) may return the duplicated entries.
2186 +- users may not be able to remove or reset the XATTR forever,
2189 +XATTR/EA support in the internal (copy,move)-(up,down)
2190 +----------------------------------------------------------------------
2191 +Generally the extended attributes of inode are categorized as these.
2192 +- "security" for LSM and capability.
2193 +- "system" for posix ACL, 'acl' mount option is required for the branch
2195 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2196 +- "user" for userspace, 'user_xattr' mount option is required for the
2197 + branch fs generally.
2199 +Moreover there are some other categories. Aufs handles these rather
2200 +unpopular categories as the ordinary ones, ie. there is no special
2201 +condition nor exception.
2203 +In copy-up, the support for XATTR on the dst branch may differ from the
2204 +src branch. In this case, the copy-up operation will get an error and
2205 +the original user operation which triggered the copy-up will fail. It
2206 +can happen that even all copy-up will fail.
2207 +When both of src and dst branches support XATTR and if an error occurs
2208 +during copying XATTR, then the copy-up should fail obviously. That is a
2209 +good reason and aufs should return an error to userspace. But when only
2210 +the src branch support that XATTR, aufs should not return an error.
2211 +For example, the src branch supports ACL but the dst branch doesn't
2212 +because the dst branch may natively un-support it or temporary
2213 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2214 +may NOT return an error even if the XATTR is not supported. It is
2215 +totally up to the branch fs.
2217 +Anyway when the aufs internal copy-up gets an error from the dst branch
2218 +fs, then aufs tries removing the just copied entry and returns the error
2219 +to the userspace. The worst case of this situation will be all copy-up
2222 +For the copy-up operation, there two basic approaches.
2223 +- copy the specified XATTR only (by category above), and return the
2224 + error unconditionally if it happens.
2225 +- copy all XATTR, and ignore the error on the specified category only.
2227 +In order to support XATTR and to implement the correct behaviour, aufs
2228 +chooses the latter approach and introduces some new branch attributes,
2229 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2230 +They correspond to the XATTR namespaces (see above). Additionally, to be
2231 +convenient, "icex" is also provided which means all "icex*" attributes
2232 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2234 +The meaning of these attributes is to ignore the error from setting
2235 +XATTR on that branch.
2236 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2237 +error from the dst branch according to the specified attributes.
2239 +Some XATTR may have its default value. The default value may come from
2240 +the parent dir or the environment. If the default value is set at the
2241 +file creating-time, it will be overwritten by copy-up.
2242 +Some contradiction may happen I am afraid.
2243 +Do we need another attribute to stop copying XATTR? I am unsure. For
2244 +now, aufs implements the branch attributes to ignore the error.
2245 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2246 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2247 +++ linux/Documentation/filesystems/aufs/design/07export.txt 2022-11-05 23:02:18.959222617 +0100
2250 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2252 +# This program is free software; you can redistribute it and/or modify
2253 +# it under the terms of the GNU General Public License as published by
2254 +# the Free Software Foundation; either version 2 of the License, or
2255 +# (at your option) any later version.
2257 +# This program is distributed in the hope that it will be useful,
2258 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2259 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2260 +# GNU General Public License for more details.
2262 +# You should have received a copy of the GNU General Public License
2263 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2265 +Export Aufs via NFS
2266 +----------------------------------------------------------------------
2267 +Here is an approach.
2268 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2270 +- iget_locked(): initialize aufs inode generation for a new inode, and
2271 + store it in xigen file.
2272 +- destroy_inode(): increment aufs inode generation and store it in xigen
2273 + file. it is necessary even if it is not unlinked, because any data of
2274 + inode may be changed by UDBA.
2275 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2276 + build file handle by
2277 + + branch id (4 bytes)
2278 + + superblock generation (4 bytes)
2279 + + inode number (4 or 8 bytes)
2280 + + parent dir inode number (4 or 8 bytes)
2281 + + inode generation (4 bytes))
2282 + + return value of exportfs_encode_fh() for the parent on a branch (4
2284 + + file handle for a branch (by exportfs_encode_fh())
2286 + + find the index of a branch from its id in handle, and check it is
2287 + still exist in aufs.
2288 + + 1st level: get the inode number from handle and search it in cache.
2289 + + 2nd level: if not found in cache, get the parent inode number from
2290 + the handle and search it in cache. and then open the found parent
2291 + dir, find the matching inode number by vfs_readdir() and get its
2292 + name, and call lookup_one_len() for the target dentry.
2293 + + 3rd level: if the parent dir is not cached, call
2294 + exportfs_decode_fh() for a branch and get the parent on a branch,
2295 + build a pathname of it, convert it a pathname in aufs, call
2296 + path_lookup(). now aufs gets a parent dir dentry, then handle it as
2298 + + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2299 + for every branch, but not itself. to get this, (currently) aufs
2300 + searches in current->nsproxy->mnt_ns list. it may not be a good
2301 + idea, but I didn't get other approach.
2302 + + test the generation of the gotten inode.
2303 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2304 + convert it into ESTALE for NFSD.
2305 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2306 + lookup_one_len(), vfs_getattr(), encode_fh() and others.
2307 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2308 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
2309 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2022-11-05 23:02:18.959222617 +0100
2312 +# Copyright (C) 2005-2022 Junjiro R. Okajima
2314 +# This program is free software; you can redistribute it and/or modify
2315 +# it under the terms of the GNU General Public License as published by
2316 +# the Free Software Foundation; either version 2 of the License, or
2317 +# (at your option) any later version.
2319 +# This program is distributed in the hope that it will be useful,
2320 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2321 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2322 +# GNU General Public License for more details.
2324 +# You should have received a copy of the GNU General Public License
2325 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2327 +Show Whiteout Mode (shwh)
2328 +----------------------------------------------------------------------
2329 +Generally aufs hides the name of whiteouts. But in some cases, to show
2330 +them is very useful for users. For instance, creating a new middle layer
2331 +(branch) by merging existing layers.
2333 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2334 +When you have three branches,
2335 +- Bottom: 'system', squashfs (underlying base system), read-only
2336 +- Middle: 'mods', squashfs, read-only
2337 +- Top: 'overlay', ram (tmpfs), read-write
2339 +The top layer is loaded at boot time and saved at shutdown, to preserve
2340 +the changes made to the system during the session.
2341 +When larger changes have been made, or smaller changes have accumulated,
2342 +the size of the saved top layer data grows. At this point, it would be
2343 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2344 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2345 +restoring save and load speed.
2347 +This merging is simplified by the use of another aufs mount, of just the
2348 +two overlay branches using the 'shwh' option.
2349 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2350 + aufs /livesys/merge_union
2352 +A merged view of these two branches is then available at
2353 +/livesys/merge_union, and the new feature is that the whiteouts are
2355 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2356 +writing to all branches. Also the default mode for all branches is 'ro'.
2357 +It is now possible to save the combined contents of the two overlay
2358 +branches to a new squashfs, e.g.:
2359 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2361 +This new squashfs archive can be stored on the boot device and the
2362 +initramfs will use it to replace the old one at the next boot.
2363 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2364 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
2365 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2022-11-05 23:02:18.959222617 +0100
2368 +# Copyright (C) 2010-2022 Junjiro R. Okajima
2370 +# This program is free software; you can redistribute it and/or modify
2371 +# it under the terms of the GNU General Public License as published by
2372 +# the Free Software Foundation; either version 2 of the License, or
2373 +# (at your option) any later version.
2375 +# This program is distributed in the hope that it will be useful,
2376 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2377 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2378 +# GNU General Public License for more details.
2380 +# You should have received a copy of the GNU General Public License
2381 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2383 +Dynamically customizable FS operations
2384 +----------------------------------------------------------------------
2385 +Generally FS operations (struct inode_operations, struct
2386 +address_space_operations, struct file_operations, etc.) are defined as
2387 +"static const", but it never means that FS have only one set of
2388 +operation. Some FS have multiple sets of them. For instance, ext2 has
2389 +three sets, one for XIP, for NOBH, and for normal.
2390 +Since aufs overrides and redirects these operations, sometimes aufs has
2391 +to change its behaviour according to the branch FS type. More importantly
2392 +VFS acts differently if a function (member in the struct) is set or
2393 +not. It means aufs should have several sets of operations and select one
2394 +among them according to the branch FS definition.
2396 +In order to solve this problem and not to affect the behaviour of VFS,
2397 +aufs defines these operations dynamically. For instance, aufs defines
2398 +dummy direct_IO function for struct address_space_operations, but it may
2399 +not be set to the address_space_operations actually. When the branch FS
2400 +doesn't have it, aufs doesn't set it to its address_space_operations
2401 +while the function definition itself is still alive. So the behaviour
2402 +itself will not change, and it will return an error when direct_IO is
2405 +The lifetime of these dynamically generated operation object is
2406 +maintained by aufs branch object. When the branch is removed from aufs,
2407 +the reference counter of the object is decremented. When it reaches
2408 +zero, the dynamically generated operation object will be freed.
2410 +This approach is designed to support AIO (io_submit), Direct I/O and
2412 +Currently this approach is applied to address_space_operations for
2413 +regular files only.
2414 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2415 --- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
2416 +++ linux/Documentation/filesystems/aufs/README 2023-07-04 23:43:55.846316200 +0200
2419 +Aufs6 -- advanced multi layered unification filesystem version 6.x
2425 +----------------------------------------
2426 +In the early days, aufs was entirely re-designed and re-implemented
2427 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2428 +improvements and implementations, it became totally different from
2429 +Unionfs while keeping the basic features.
2430 +Later, Unionfs Version 2.x series began taking some of the same
2431 +approaches to aufs1's.
2432 +Unionfs was being developed by Professor Erez Zadok at Stony Brook
2433 +University and his team.
2435 +Aufs6 supports linux-v6.0 and later, try aufs6.0 branch in
2436 +aufs-linux.git or aufs-standalone.git.
2437 +If you want older kernel version support,
2438 +- for linux-v5.x series, try aufs-linux.git or aufs-standalone.git
2439 +- for linux-v4.x series, try aufs4-linux.git or aufs4-standalone.git
2440 +- for linux-v3.x series, try aufs3-linux.git or aufs3-standalone.git
2441 +- for linux-v2.6.16 and later, try aufs2-2.6.git, aufs2-standalone.git
2442 + or aufs1 from CVS on SourceForge.
2444 +Note: the name of aufs5-linux.git and aufs5-standalone.git on github
2445 + were changed. Now they are aufs-linux.git and
2446 + aufs-standalone.git and they contain aufs5 and later branches.
2448 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2449 + According to Christoph Hellwig, linux rejects all union-type
2450 + filesystems but UnionMount.
2451 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2453 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2454 + UnionMount, and he pointed out an issue around a directory mutex
2455 + lock and aufs addressed it. But it is still unsure whether aufs will
2456 + be merged (or any other union solution).
2457 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2461 +----------------------------------------
2462 +- unite several directories into a single virtual filesystem. The member
2463 + directory is called as a branch.
2464 +- you can specify the permission flags to the branch, which are 'readonly',
2465 + 'readwrite' and 'whiteout-able.'
2466 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2467 + readonly branch are modifiable logically.
2468 +- dynamic branch manipulation, add, del.
2471 +Also there are many enhancements in aufs, such as:
2472 +- test only the highest one for the directory permission (dirperm1)
2473 +- copyup on open (coo=)
2474 +- 'move' policy for copy-up between two writable branches, after
2475 + checking free space.
2477 +- readdir(3) in userspace.
2478 +- keep inode number by external inode number table
2479 +- keep the timestamps of file/dir in internal copyup operation
2480 +- seekable directory, supporting NFS readdir.
2481 +- whiteout is hardlinked in order to reduce the consumption of inodes
2483 +- do not copyup, nor create a whiteout when it is unnecessary
2484 +- revert a single systemcall when an error occurs in aufs
2485 +- remount interface instead of ioctl
2486 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2487 +- loopback mounted filesystem as a branch
2488 +- kernel thread for removing the dir who has a plenty of whiteouts
2489 +- support copyup sparse file (a file which has a 'hole' in it)
2490 +- default permission flags for branches
2491 +- selectable permission flags for ro branch, whether whiteout can
2494 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2495 +- support multiple writable branches, some policies to select one
2496 + among multiple writable branches.
2497 +- a new semantics for link(2) and rename(2) to support multiple
2498 + writable branches.
2499 +- no glibc changes are required.
2500 +- pseudo hardlink (hardlink over branches)
2501 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2502 + including NFS or remote filesystem branch.
2503 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2506 +Currently these features are dropped temporary from aufs6.
2507 +See design/08plan.txt in detail.
2508 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2510 +- statistics of aufs thread (/sys/fs/aufs/stat)
2512 +Features or just an idea in the future (see also design/*.txt),
2513 +- reorder the branch index without del/re-add.
2514 +- permanent xino files for NFSD
2515 +- an option for refreshing the opened files after add/del branches
2516 +- light version, without branch manipulation. (unnecessary?)
2517 +- copyup in userspace
2518 +- inotify in userspace
2523 +----------------------------------------
2524 +There are three GIT trees for aufs6, aufs-linux.git,
2525 +aufs-standalone.git, and aufs-util.git.
2526 +While the aufs-util is always necessary, you need either of aufs-linux
2527 +or aufs-standalone.
2529 +The aufs-linux tree includes the whole linux mainline GIT tree,
2530 +git://git.kernel.org/.../torvalds/linux.git.
2531 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2532 +build aufs6 as an external kernel module.
2533 +Several extra patches are not included in this tree. Only
2534 +aufs-standalone tree contains them. They are described in the later
2535 +section "Configuration and Compilation."
2537 +On the other hand, the aufs-standalone tree has only aufs source files
2538 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2539 +But you need to apply all aufs patches manually.
2541 +You will find GIT branches whose name is in form of "aufs6.x" where "x"
2542 +represents the linux kernel version, "linux-6.x". For instance,
2543 +"aufs6.0" is for linux-6.0. For latest "linux-6.x-rcN", use
2544 +"aufs6.x-rcN" branch.
2547 +$ git clone --reference /your/linux/git/tree \
2548 + git://github.com/sfjro/aufs-linux.git aufs-linux.git
2549 +- if you don't have linux GIT tree, then remove "--reference ..."
2550 +$ cd aufs-linux.git
2551 +$ git checkout origin/aufs6.0
2553 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2554 +leave the patch-work to GIT.
2555 +$ cd /your/linux/git/tree
2556 +$ git remote add aufs git://github.com/sfjro/aufs-linux.git
2558 +$ git checkout -b my6.0 v6.0
2559 +$ (add your local change...)
2560 +$ git pull aufs aufs6.0
2561 +- now you have v6.0 + your_changes + aufs6.0 in you my6.0 branch.
2562 +- you may need to solve some conflicts between your_changes and
2563 + aufs6.0. in this case, git-rerere is recommended so that you can
2564 + solve the similar conflicts automatically when you upgrade to 6.1 or
2565 + later in the future.
2567 +o aufs-standalone tree
2568 +$ git clone git://github.com/sfjro/aufs-standalone.git aufs-standalone.git
2569 +$ cd aufs-standalone.git
2570 +$ git checkout origin/aufs6.0
2573 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2574 +- note that the public aufs-util.git is on SourceForge instead of
2577 +$ git checkout origin/aufs6.0
2579 +Note: The 6.x-rcN branch is to be used with `rc' kernel versions ONLY.
2580 +The minor version number, 'x' in '6.x', of aufs may not always
2581 +follow the minor version number of the kernel.
2582 +Because changes in the kernel that cause the use of a new
2583 +minor version number do not always require changes to aufs-util.
2585 +Since aufs-util has its own minor version number, you may not be
2586 +able to find a GIT branch in aufs-util for your kernel's
2587 +exact minor version number.
2588 +In this case, you should git-checkout the branch for the
2589 +nearest lower number.
2591 +For (an unreleased) example:
2592 +If you are using "linux-6.10" and the "aufs6.10" branch
2593 +does not exist in aufs-util repository, then "aufs6.9", "aufs6.8"
2594 +or something numerically smaller is the branch for your kernel.
2596 +Also you can view all branches by
2600 +3. Configuration and Compilation
2601 +----------------------------------------
2602 +Make sure you have git-checkout'ed the correct branch.
2604 +For aufs-linux tree,
2605 +- enable CONFIG_AUFS_FS.
2606 +- set other aufs configurations if necessary.
2607 +- for aufs5.13 and later
2608 + Because aufs is not only an ordinary filesystem (callee of VFS), but
2609 + also a caller of VFS functions for branch filesystems, subclassing of
2610 + the internal locks for LOCKDEP is necessary. LOCKDEP is a debugging
2611 + feature of linux kernel. If you enable CONFIG_LOCKDEP, then you will
2612 + need to customize some LOCKDEP numbers. Here are what I use on my
2614 + CONFIG_LOCKDEP_BITS=21
2615 + CONFIG_LOCKDEP_CHAINS_BITS=21
2616 + CONFIG_LOCKDEP_STACK_TRACE_BITS=24
2617 + Also you will need to expand some constant values in LOCKDEP. Refer
2618 + to lockdep-debug.patch in aufs-standalone.git.
2620 +For aufs-standalone tree,
2621 +There are several ways to build.
2624 +- apply ./aufs6-kbuild.patch to your kernel source files.
2625 +- apply ./aufs6-base.patch too.
2626 +- apply ./aufs6-mmap.patch too.
2627 +- apply ./aufs6-standalone.patch too, if you have a plan to set
2628 + CONFIG_AUFS_FS=m. otherwise you don't need ./aufs-standalone.patch.
2629 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2630 + kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2631 +- enable CONFIG_AUFS_FS, you can select either
2633 +- and build your kernel as usual.
2634 +- install the built kernel.
2635 +- install the header files too by "make headers_install" to the
2636 + directory where you specify. By default, it is $PWD/usr.
2637 + "make help" shows a brief note for headers_install.
2638 +- and reboot your system.
2641 +- module only (CONFIG_AUFS_FS=m).
2642 +- apply ./aufs6-base.patch to your kernel source files.
2643 +- apply ./aufs6-mmap.patch too.
2644 +- apply ./aufs6-standalone.patch too.
2645 +- build your kernel, don't forget "make headers_install", and reboot.
2646 +- edit ./config.mk and set other aufs configurations if necessary.
2647 + Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2648 + every aufs configurations.
2649 +- build the module by simple "make".
2650 +- you can specify ${KDIR} make variable which points to your kernel
2652 +- install the files
2653 + + run "make install" to install the aufs module, or copy the built
2654 + $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2655 + + run "make install_headers" (instead of headers_install) to install
2656 + the modified aufs header file (you can specify DESTDIR which is
2657 + available in aufs standalone version's Makefile only), or copy
2658 + $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2659 + you like manually. By default, the target directory is $PWD/usr.
2660 +- no need to apply aufs6-kbuild.patch, nor copying source files to your
2661 + kernel source tree.
2663 +Note: The header file aufs_type.h is necessary to build aufs-util
2664 + as well as "make headers_install" in the kernel source tree.
2665 + headers_install is subject to be forgotten, but it is essentially
2666 + necessary, not only for building aufs-util.
2667 + You may not meet problems without headers_install in some older
2671 +- read README in aufs-util, build and install it
2672 +- note that your distribution may contain an obsoleted version of
2673 + aufs_type.h in /usr/include/linux or something. When you build aufs
2674 + utilities, make sure that your compiler refers the correct aufs header
2675 + file which is built by "make headers_install."
2676 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2677 + then run "make install_ulib" too. And refer to the aufs manual in
2680 +There several other patches in aufs-standalone.git. They are all
2681 +optional. When you meet some problems, they will help you.
2682 +- aufs6-loopback.patch
2683 + Supports a nested loopback mount in a branch-fs. This patch is
2684 + unnecessary until aufs produces a message like "you may want to try
2685 + another patch for loopback file".
2687 + Modifies a system global kernel internal function get_next_ino() in
2688 + order to stop assigning 0 for an inode-number. Not directly related to
2689 + aufs, but recommended generally.
2691 + Keeps the tmpfs inode number as the lowest value. Effective to reduce
2692 + the size of aufs XINO files for tmpfs branch. Also it prevents the
2693 + duplication of inode number, which is important for backup tools and
2694 + other utilities. When you find aufs XINO files for tmpfs branch
2695 + growing too much, try this patch.
2696 +- lockdep-debug.patch
2697 + Similar to some kernel configurations for LOCKDEP (see the top of
2698 + this section), you will need expand some constants in LOCKDEP for
2699 + aufs if you enable CONFIG_LOCKDEP.
2703 +----------------------------------------
2704 +At first, make sure aufs-util are installed, and please read the aufs
2705 +manual, aufs.5 in aufs-util.git tree.
2709 +$ mkdir /tmp/rw /tmp/aufs
2710 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2712 +Here is another example. The result is equivalent.
2713 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2715 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2716 +# mount -o remount,append:${HOME} /tmp/aufs
2718 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2719 +you modify a file under /tmp/aufs, the one on your home directory is
2720 +not affected, instead the same named file will be newly created under
2721 +/tmp/rw. And all of your modification to a file will be applied to
2722 +the one under /tmp/rw. This is called the file based Copy on Write
2724 +Aufs mount options are described in aufs.5.
2725 +If you run chroot or something and make your aufs as a root directory,
2726 +then you need to customize the shutdown script. See the aufs manual in
2729 +Additionally, there are some sample usages of aufs which are a
2730 +diskless system with network booting, and LiveCD over NFS.
2731 +See sample dir in CVS tree on SourceForge.
2735 +----------------------------------------
2736 +When you have any problems or strange behaviour in aufs, please let me
2738 +- /proc/mounts (instead of the output of mount(8))
2739 +- /sys/module/aufs/*
2740 +- /sys/fs/aufs/* (if you have them)
2741 +- /debug/aufs/* (if you have them)
2742 +- linux kernel version
2743 + if your kernel is not plain, for example modified by distributor,
2744 + the url where i can download its source is necessary too.
2745 +- aufs version which was printed at loading the module or booting the
2746 + system, instead of the date you downloaded.
2747 +- configuration (define/undefine CONFIG_AUFS_xxx)
2748 +- kernel configuration or /proc/config.gz (if you have it)
2749 +- LSM (linux security module, if you are using)
2750 +- behaviour which you think to be incorrect
2751 +- actual operation, reproducible one is better
2752 +- mailto: aufs-users at lists.sourceforge.net
2754 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2755 +and Feature Requests) on SourceForge. Please join and write to
2759 +6. Acknowledgements
2760 +----------------------------------------
2761 +Thanks to everyone who have tried and are using aufs, whoever
2762 +have reported a bug or any feedback.
2764 +Especially donators:
2765 +Tomas Matejicek(slax.org) made a donation (much more than once).
2766 + Since Apr 2010, Tomas M (the author of Slax and Linux Live
2767 + scripts) is making "doubling" donations.
2768 + Unfortunately I cannot list all of the donators, but I really
2770 + It ends Aug 2010, but the ordinary donation URL is still available.
2771 + <http://sourceforge.net/donate/index.php?group_id=167503>
2772 +Dai Itasaka made a donation (2007/8).
2773 +Chuck Smith made a donation (2008/4, 10 and 12).
2774 +Henk Schoneveld made a donation (2008/9).
2775 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2776 +Francois Dupoux made a donation (2008/11).
2777 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2778 + aufs2 GIT tree (2009/2).
2779 +William Grant made a donation (2009/3).
2780 +Patrick Lane made a donation (2009/4).
2781 +The Mail Archive (mail-archive.com) made donations (2009/5).
2782 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2783 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2784 +Pavel Pronskiy made a donation (2011/2).
2785 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2786 + Networks (Ed Wildgoose) made a donation for hardware (2011/3).
2787 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2789 +Sam Liddicott made a donation (2011/9).
2790 +Era Scarecrow made a donation (2013/4).
2791 +Bor Ratajc made a donation (2013/4).
2792 +Alessandro Gorreta made a donation (2013/4).
2793 +POIRETTE Marc made a donation (2013/4).
2794 +Alessandro Gorreta made a donation (2013/4).
2795 +lauri kasvandik made a donation (2013/5).
2796 +"pemasu from Finland" made a donation (2013/7).
2797 +The Parted Magic Project made a donation (2013/9 and 11).
2798 +Pavel Barta made a donation (2013/10).
2799 +Nikolay Pertsev made a donation (2014/5).
2800 +James B made a donation (2014/7, 2015/7, and 2021/12).
2801 +Stefano Di Biase made a donation (2014/8).
2802 +Daniel Epellei made a donation (2015/1).
2803 +OmegaPhil made a donation (2016/1, 2018/4).
2804 +Tomasz Szewczyk made a donation (2016/4).
2805 +James Burry made a donation (2016/12).
2806 +Carsten Rose made a donation (2018/9).
2807 +Porteus Kiosk made a donation (2018/10).
2808 +huronOS team: Enya Quetzalli made donations (2022/5 and 2023/5).
2810 +Thank you very much.
2811 +Donations are always, including future donations, very important and
2812 +helpful for me to keep on developing aufs.
2816 +----------------------------------------
2817 +If you are an experienced user, no explanation is needed. Aufs is
2818 +just a linux filesystem.
2823 +# Local variables: ;
2826 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2827 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
2828 +++ linux/fs/aufs/aufs.h 2022-11-05 23:02:18.959222617 +0100
2830 +/* SPDX-License-Identifier: GPL-2.0 */
2832 + * Copyright (C) 2005-2022 Junjiro R. Okajima
2834 + * This program is free software; you can redistribute it and/or modify
2835 + * it under the terms of the GNU General Public License as published by
2836 + * the Free Software Foundation; either version 2 of the License, or
2837 + * (at your option) any later version.
2839 + * This program is distributed in the hope that it will be useful,
2840 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2841 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2842 + * GNU General Public License for more details.
2844 + * You should have received a copy of the GNU General Public License
2845 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2849 + * all header files
2857 +#define AuStub(type, name, body, ...) \
2858 + static inline type name(__VA_ARGS__) { body; }
2860 +#define AuStubVoid(name, ...) \
2861 + AuStub(void, name, , __VA_ARGS__)
2862 +#define AuStubInt0(name, ...) \
2863 + AuStub(int, name, return 0, __VA_ARGS__)
2867 +#include "branch.h"
2870 +#include "dbgaufs.h"
2871 +#include "dentry.h"
2873 +#include "dirren.h"
2876 +#include "fstype.h"
2881 +#include "module.h"
2885 +#include "sysaufs.h"
2890 +#endif /* __KERNEL__ */
2891 +#endif /* __AUFS_H__ */
2892 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2893 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
2894 +++ linux/fs/aufs/branch.c 2022-11-05 23:02:18.959222617 +0100
2896 +// SPDX-License-Identifier: GPL-2.0
2898 + * Copyright (C) 2005-2022 Junjiro R. Okajima
2900 + * This program is free software; you can redistribute it and/or modify
2901 + * it under the terms of the GNU General Public License as published by
2902 + * the Free Software Foundation; either version 2 of the License, or
2903 + * (at your option) any later version.
2905 + * This program is distributed in the hope that it will be useful,
2906 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2907 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2908 + * GNU General Public License for more details.
2910 + * You should have received a copy of the GNU General Public License
2911 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2915 + * branch management
2918 +#include <linux/compat.h>
2919 +#include <linux/statfs.h>
2923 + * free a single branch
2925 +static void au_br_do_free(struct au_branch *br)
2928 + struct au_wbr *wbr;
2929 + struct au_dykey **key;
2931 + au_hnotify_fin_br(br);
2932 + /* always, regardless the mount option */
2933 + au_dr_hino_free(&br->br_dirren);
2936 + AuLCntZero(au_lcnt_read(&br->br_nfiles, /*do_rev*/0));
2937 + au_lcnt_fin(&br->br_nfiles, /*do_sync*/0);
2938 + AuLCntZero(au_lcnt_read(&br->br_count, /*do_rev*/0));
2939 + au_lcnt_fin(&br->br_count, /*do_sync*/0);
2943 + for (i = 0; i < AuBrWh_Last; i++)
2944 + dput(wbr->wbr_wh[i]);
2945 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2946 + AuRwDestroy(&wbr->wbr_wh_rwsem);
2949 + if (br->br_fhsm) {
2950 + au_br_fhsm_fin(br->br_fhsm);
2951 + au_kfree_try_rcu(br->br_fhsm);
2954 + key = br->br_dykey;
2955 + for (i = 0; i < AuBrDynOp; i++, key++)
2961 + /* recursive lock, s_umount of branch's */
2962 + /* synchronize_rcu(); */ /* why? */
2964 + path_put(&br->br_path);
2966 + au_kfree_rcu(wbr);
2967 + au_lcnt_wait_for_fin(&br->br_nfiles);
2968 + au_lcnt_wait_for_fin(&br->br_count);
2969 + /* I don't know why, but percpu_refcount requires this */
2970 + /* synchronize_rcu(); */
2975 + * frees all branches
2977 +void au_br_free(struct au_sbinfo *sbinfo)
2979 + aufs_bindex_t bmax;
2980 + struct au_branch **br;
2982 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2984 + bmax = sbinfo->si_bbot + 1;
2985 + br = sbinfo->si_branch;
2987 + au_br_do_free(*br++);
2991 + * find the index of a branch which is specified by @br_id.
2993 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2995 + aufs_bindex_t bindex, bbot;
2997 + bbot = au_sbbot(sb);
2998 + for (bindex = 0; bindex <= bbot; bindex++)
2999 + if (au_sbr_id(sb, bindex) == br_id)
3004 +/* ---------------------------------------------------------------------- */
3010 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
3011 + struct dentry *h_root)
3013 + if (unlikely(h_adding == h_root
3014 + || au_test_loopback_overlap(sb, h_adding)))
3016 + if (h_adding->d_sb != h_root->d_sb)
3018 + return au_test_subdir(h_adding, h_root)
3019 + || au_test_subdir(h_root, h_adding);
3023 + * returns a newly allocated branch. @new_nbranch is a number of branches
3024 + * after adding a branch.
3026 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
3029 + struct au_branch *add_branch;
3030 + struct dentry *root;
3031 + struct inode *inode;
3035 + add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
3036 + if (unlikely(!add_branch))
3038 + add_branch->br_xino = au_xino_alloc(/*nfile*/1);
3039 + if (unlikely(!add_branch->br_xino))
3041 + err = au_hnotify_init_br(add_branch, perm);
3042 + if (unlikely(err))
3045 + if (au_br_writable(perm)) {
3046 + /* may be freed separately at changing the branch permission */
3047 + add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
3049 + if (unlikely(!add_branch->br_wbr))
3053 + if (au_br_fhsm(perm)) {
3054 + err = au_fhsm_br_alloc(add_branch);
3055 + if (unlikely(err))
3059 + root = sb->s_root;
3060 + err = au_sbr_realloc(au_sbi(sb), new_nbranch, /*may_shrink*/0);
3062 + err = au_di_realloc(au_di(root), new_nbranch, /*may_shrink*/0);
3064 + inode = d_inode(root);
3065 + err = au_hinode_realloc(au_ii(inode), new_nbranch,
3069 + return add_branch; /* success */
3072 + au_kfree_rcu(add_branch->br_wbr);
3074 + au_hnotify_fin_br(add_branch);
3076 + au_xino_put(add_branch);
3078 + au_kfree_rcu(add_branch);
3080 + return ERR_PTR(err);
3084 + * test if the branch permission is legal or not.
3086 +static int test_br(struct inode *inode, int brperm, char *path)
3090 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
3095 + pr_err("write permission for readonly mount or inode, %s\n", path);
3103 + * 0: success, the caller will add it
3104 + * plus: success, it is already unified, the caller should ignore it
3107 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
3110 + aufs_bindex_t bbot, bindex;
3111 + struct dentry *root, *h_dentry;
3112 + struct inode *inode, *h_inode;
3114 + root = sb->s_root;
3115 + bbot = au_sbbot(sb);
3116 + if (unlikely(bbot >= 0
3117 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
3121 + pr_err("%s duplicated\n", add->pathname);
3126 + err = -ENOSPC; /* -E2BIG; */
3127 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
3128 + || AUFS_BRANCH_MAX - 1 <= bbot)) {
3129 + pr_err("number of branches exceeded %s\n", add->pathname);
3134 + if (unlikely(add->bindex < 0 || bbot + 1 < add->bindex)) {
3135 + pr_err("bad index %d\n", add->bindex);
3139 + inode = d_inode(add->path.dentry);
3141 + if (unlikely(!inode->i_nlink)) {
3142 + pr_err("no existence %s\n", add->pathname);
3147 + if (unlikely(inode->i_sb == sb)) {
3148 + pr_err("%s must be outside\n", add->pathname);
3152 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
3153 + pr_err("unsupported filesystem, %s (%s)\n",
3154 + add->pathname, au_sbtype(inode->i_sb));
3158 + if (unlikely(inode->i_sb->s_stack_depth)) {
3159 + pr_err("already stacked, %s (%s)\n",
3160 + add->pathname, au_sbtype(inode->i_sb));
3164 + err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3165 + if (unlikely(err))
3169 + return 0; /* success */
3172 + for (bindex = 0; bindex <= bbot; bindex++)
3173 + if (unlikely(test_overlap(sb, add->path.dentry,
3174 + au_h_dptr(root, bindex)))) {
3175 + pr_err("%s is overlapped\n", add->pathname);
3180 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3181 + h_dentry = au_h_dptr(root, 0);
3182 + h_inode = d_inode(h_dentry);
3183 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3184 + || !uid_eq(h_inode->i_uid, inode->i_uid)
3185 + || !gid_eq(h_inode->i_gid, inode->i_gid))
3186 + pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3188 + i_uid_read(inode), i_gid_read(inode),
3189 + (inode->i_mode & S_IALLUGO),
3190 + i_uid_read(h_inode), i_gid_read(h_inode),
3191 + (h_inode->i_mode & S_IALLUGO));
3199 + * initialize or clean the whiteouts for an adding branch
3201 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3204 + int err, old_perm;
3205 + aufs_bindex_t bindex;
3206 + struct inode *h_inode;
3207 + struct au_wbr *wbr;
3208 + struct au_hinode *hdir;
3209 + struct dentry *h_dentry;
3211 + err = vfsub_mnt_want_write(au_br_mnt(br));
3212 + if (unlikely(err))
3216 + old_perm = br->br_perm;
3217 + br->br_perm = new_perm;
3220 + bindex = au_br_index(sb, br->br_id);
3221 + if (0 <= bindex) {
3222 + hdir = au_hi(d_inode(sb->s_root), bindex);
3223 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
3225 + h_dentry = au_br_dentry(br);
3226 + h_inode = d_inode(h_dentry);
3227 + inode_lock_nested(h_inode, AuLsc_I_PARENT);
3230 + err = au_wh_init(br, sb);
3232 + wbr_wh_write_lock(wbr);
3233 + err = au_wh_init(br, sb);
3234 + wbr_wh_write_unlock(wbr);
3237 + au_hn_inode_unlock(hdir);
3239 + inode_unlock(h_inode);
3240 + vfsub_mnt_drop_write(au_br_mnt(br));
3241 + br->br_perm = old_perm;
3243 + if (!err && wbr && !au_br_writable(new_perm)) {
3244 + au_kfree_rcu(wbr);
3245 + br->br_wbr = NULL;
3252 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3256 + struct kstatfs kst;
3257 + struct au_wbr *wbr;
3260 + au_rw_init(&wbr->wbr_wh_rwsem);
3261 + atomic_set(&wbr->wbr_wh_running, 0);
3264 + * a limit for rmdir/rename a dir
3265 + * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3267 + err = vfs_statfs(&br->br_path, &kst);
3268 + if (unlikely(err))
3271 + if (kst.f_namelen >= NAME_MAX)
3272 + err = au_br_init_wh(sb, br, perm);
3274 + pr_err("%pd(%s), unsupported namelen %ld\n",
3276 + au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3282 +/* initialize a new branch */
3283 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3284 + struct au_opt_add *add)
3287 + struct au_branch *brbase;
3289 + struct inode *h_inode;
3292 + br->br_perm = add->perm;
3293 + br->br_path = add->path; /* set first, path_get() later */
3294 + spin_lock_init(&br->br_dykey_lock);
3295 + au_lcnt_init(&br->br_nfiles, /*release*/NULL);
3296 + au_lcnt_init(&br->br_count, /*release*/NULL);
3297 + br->br_id = au_new_br_id(sb);
3298 + AuDebugOn(br->br_id < 0);
3300 + /* always, regardless the given option */
3301 + err = au_dr_br_init(sb, br, &add->path);
3302 + if (unlikely(err))
3305 + if (au_br_writable(add->perm)) {
3306 + err = au_wbr_init(br, sb, add->perm);
3307 + if (unlikely(err))
3311 + if (au_opt_test(au_mntflags(sb), XINO)) {
3312 + brbase = au_sbr(sb, 0);
3313 + xf = au_xino_file(brbase->br_xino, /*idx*/-1);
3315 + h_inode = d_inode(add->path.dentry);
3316 + err = au_xino_init_br(sb, br, h_inode->i_ino, &xf->f_path);
3317 + if (unlikely(err)) {
3318 + AuDebugOn(au_xino_file(br->br_xino, /*idx*/-1));
3323 + sysaufs_br_init(br);
3324 + path_get(&br->br_path);
3325 + goto out; /* success */
3328 + memset(&br->br_path, 0, sizeof(br->br_path));
3333 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3334 + struct au_branch *br, aufs_bindex_t bbot,
3335 + aufs_bindex_t amount)
3337 + struct au_branch **brp;
3339 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3341 + brp = sbinfo->si_branch + bindex;
3342 + memmove(brp + 1, brp, sizeof(*brp) * amount);
3344 + sbinfo->si_bbot++;
3345 + if (unlikely(bbot < 0))
3346 + sbinfo->si_bbot = 0;
3349 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3350 + aufs_bindex_t bbot, aufs_bindex_t amount)
3352 + struct au_hdentry *hdp;
3354 + AuRwMustWriteLock(&dinfo->di_rwsem);
3356 + hdp = au_hdentry(dinfo, bindex);
3357 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3358 + au_h_dentry_init(hdp);
3360 + if (unlikely(bbot < 0))
3361 + dinfo->di_btop = 0;
3364 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3365 + aufs_bindex_t bbot, aufs_bindex_t amount)
3367 + struct au_hinode *hip;
3369 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3371 + hip = au_hinode(iinfo, bindex);
3372 + memmove(hip + 1, hip, sizeof(*hip) * amount);
3373 + au_hinode_init(hip);
3375 + if (unlikely(bbot < 0))
3376 + iinfo->ii_btop = 0;
3379 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3380 + aufs_bindex_t bindex)
3382 + struct dentry *root, *h_dentry;
3383 + struct inode *root_inode, *h_inode;
3384 + aufs_bindex_t bbot, amount;
3386 + root = sb->s_root;
3387 + root_inode = d_inode(root);
3388 + bbot = au_sbbot(sb);
3389 + amount = bbot + 1 - bindex;
3390 + h_dentry = au_br_dentry(br);
3391 + au_sbilist_lock();
3392 + au_br_do_add_brp(au_sbi(sb), bindex, br, bbot, amount);
3393 + au_br_do_add_hdp(au_di(root), bindex, bbot, amount);
3394 + au_br_do_add_hip(au_ii(root_inode), bindex, bbot, amount);
3395 + au_set_h_dptr(root, bindex, dget(h_dentry));
3396 + h_inode = d_inode(h_dentry);
3397 + au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3398 + au_sbilist_unlock();
3401 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3404 + aufs_bindex_t bbot, add_bindex;
3405 + struct dentry *root, *h_dentry;
3406 + struct inode *root_inode;
3407 + struct au_branch *add_branch;
3409 + root = sb->s_root;
3410 + root_inode = d_inode(root);
3411 + IMustLock(root_inode);
3412 + IiMustWriteLock(root_inode);
3413 + err = test_add(sb, add, remount);
3414 + if (unlikely(err < 0))
3418 + goto out; /* success */
3421 + bbot = au_sbbot(sb);
3422 + add_branch = au_br_alloc(sb, bbot + 2, add->perm);
3423 + err = PTR_ERR(add_branch);
3424 + if (IS_ERR(add_branch))
3427 + err = au_br_init(add_branch, sb, add);
3428 + if (unlikely(err)) {
3429 + au_br_do_free(add_branch);
3433 + add_bindex = add->bindex;
3434 + sysaufs_brs_del(sb, add_bindex); /* remove successors */
3435 + au_br_do_add(sb, add_branch, add_bindex);
3436 + sysaufs_brs_add(sb, add_bindex); /* append successors */
3437 + dbgaufs_brs_add(sb, add_bindex, /*topdown*/0); /* rename successors */
3439 + h_dentry = add->path.dentry;
3440 + if (!add_bindex) {
3441 + au_cpup_attr_all(root_inode, /*force*/1);
3442 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3444 + au_add_nlink(root_inode, d_inode(h_dentry));
3450 +/* ---------------------------------------------------------------------- */
3452 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3453 + unsigned long long max __maybe_unused,
3456 + unsigned long long n;
3457 + struct file **p, *f;
3458 + struct hlist_bl_head *files;
3459 + struct hlist_bl_node *pos;
3460 + struct au_finfo *finfo;
3464 + files = &au_sbi(sb)->si_files;
3465 + hlist_bl_lock(files);
3466 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
3467 + f = finfo->fi_file;
3469 + && !special_file(file_inode(f)->i_mode)) {
3473 + AuDebugOn(n > max);
3476 + hlist_bl_unlock(files);
3481 +static struct file **au_farray_alloc(struct super_block *sb,
3482 + unsigned long long *max)
3484 + struct au_sbinfo *sbi;
3487 + *max = au_lcnt_read(&sbi->si_nfiles, /*do_rev*/1);
3488 + return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3491 +static void au_farray_free(struct file **a, unsigned long long max)
3493 + unsigned long long ull;
3495 + for (ull = 0; ull < max; ull++)
3501 +/* ---------------------------------------------------------------------- */
3507 +/* to show the line number, do not make it inlined function */
3508 +#define AuVerbose(do_info, fmt, ...) do { \
3510 + pr_info(fmt, ##__VA_ARGS__); \
3513 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t btop,
3514 + aufs_bindex_t bbot)
3516 + return (inode && !S_ISDIR(inode->i_mode)) || btop == bbot;
3519 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t btop,
3520 + aufs_bindex_t bbot)
3522 + return au_test_ibusy(d_inode(dentry), btop, bbot);
3526 + * test if the branch is deletable or not.
3528 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3529 + unsigned int sigen, const unsigned int verbose)
3531 + int err, i, j, ndentry;
3532 + aufs_bindex_t btop, bbot;
3533 + struct au_dcsub_pages dpages;
3534 + struct au_dpage *dpage;
3537 + err = au_dpages_init(&dpages, GFP_NOFS);
3538 + if (unlikely(err))
3540 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
3541 + if (unlikely(err))
3544 + for (i = 0; !err && i < dpages.ndpage; i++) {
3545 + dpage = dpages.dpages + i;
3546 + ndentry = dpage->ndentry;
3547 + for (j = 0; !err && j < ndentry; j++) {
3548 + d = dpage->dentries[j];
3549 + AuDebugOn(au_dcount(d) <= 0);
3550 + if (!au_digen_test(d, sigen)) {
3551 + di_read_lock_child(d, AuLock_IR);
3552 + if (unlikely(au_dbrange_test(d))) {
3553 + di_read_unlock(d, AuLock_IR);
3557 + di_write_lock_child(d);
3558 + if (unlikely(au_dbrange_test(d))) {
3559 + di_write_unlock(d);
3562 + err = au_reval_dpath(d, sigen);
3564 + di_downgrade_lock(d, AuLock_IR);
3566 + di_write_unlock(d);
3571 + /* AuDbgDentry(d); */
3572 + btop = au_dbtop(d);
3573 + bbot = au_dbbot(d);
3574 + if (btop <= bindex
3576 + && au_h_dptr(d, bindex)
3577 + && au_test_dbusy(d, btop, bbot)) {
3579 + AuVerbose(verbose, "busy %pd\n", d);
3582 + di_read_unlock(d, AuLock_IR);
3587 + au_dpages_free(&dpages);
3592 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3593 + unsigned int sigen, const unsigned int verbose)
3596 + unsigned long long max, ull;
3597 + struct inode *i, **array;
3598 + aufs_bindex_t btop, bbot;
3600 + array = au_iarray_alloc(sb, &max);
3601 + err = PTR_ERR(array);
3602 + if (IS_ERR(array))
3606 + AuDbg("b%d\n", bindex);
3607 + for (ull = 0; !err && ull < max; ull++) {
3611 + if (i->i_ino == AUFS_ROOT_INO)
3614 + /* AuDbgInode(i); */
3615 + if (au_iigen(i, NULL) == sigen)
3616 + ii_read_lock_child(i);
3618 + ii_write_lock_child(i);
3619 + err = au_refresh_hinode_self(i);
3622 + ii_downgrade_lock(i);
3624 + ii_write_unlock(i);
3629 + btop = au_ibtop(i);
3630 + bbot = au_ibbot(i);
3631 + if (btop <= bindex
3633 + && au_h_iptr(i, bindex)
3634 + && au_test_ibusy(i, btop, bbot)) {
3636 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3639 + ii_read_unlock(i);
3641 + au_iarray_free(array, max);
3647 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3648 + const unsigned int verbose)
3651 + unsigned int sigen;
3653 + sigen = au_sigen(root->d_sb);
3654 + DiMustNoWaiters(root);
3655 + IiMustNoWaiters(d_inode(root));
3656 + di_write_unlock(root);
3657 + err = test_dentry_busy(root, bindex, sigen, verbose);
3659 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3660 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3665 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3666 + struct file **to_free, int *idx)
3669 + unsigned char matched, root;
3670 + aufs_bindex_t bindex, bbot;
3671 + struct au_fidir *fidir;
3672 + struct au_hfile *hfile;
3675 + root = IS_ROOT(file->f_path.dentry);
3678 + to_free[*idx] = file;
3684 + fidir = au_fi(file)->fi_hdir;
3685 + AuDebugOn(!fidir);
3686 + bbot = au_fbbot_dir(file);
3687 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++) {
3688 + hfile = fidir->fd_hfile + bindex;
3689 + if (!hfile->hf_file)
3692 + if (hfile->hf_br->br_id == br_id) {
3704 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3705 + struct file **to_free, int opened)
3708 + unsigned long long ull, max;
3709 + aufs_bindex_t btop;
3710 + struct file *file, **array;
3711 + struct dentry *root;
3712 + struct au_hfile *hfile;
3714 + array = au_farray_alloc(sb, &max);
3715 + err = PTR_ERR(array);
3716 + if (IS_ERR(array))
3721 + root = sb->s_root;
3722 + di_write_unlock(root);
3723 + for (ull = 0; ull < max; ull++) {
3724 + file = array[ull];
3725 + if (unlikely(!file))
3728 + /* AuDbg("%pD\n", file); */
3729 + fi_read_lock(file);
3730 + btop = au_fbtop(file);
3731 + if (!d_is_dir(file->f_path.dentry)) {
3732 + hfile = &au_fi(file)->fi_htop;
3733 + if (hfile->hf_br->br_id == br_id)
3736 + err = test_dir_busy(file, br_id, to_free, &idx);
3737 + fi_read_unlock(file);
3738 + if (unlikely(err))
3741 + di_write_lock_child(root);
3742 + au_farray_free(array, max);
3743 + AuDebugOn(idx > opened);
3749 +static void br_del_file(struct file **to_free, unsigned long long opened,
3750 + aufs_bindex_t br_id)
3752 + unsigned long long ull;
3753 + aufs_bindex_t bindex, btop, bbot, bfound;
3754 + struct file *file;
3755 + struct au_fidir *fidir;
3756 + struct au_hfile *hfile;
3758 + for (ull = 0; ull < opened; ull++) {
3759 + file = to_free[ull];
3760 + if (unlikely(!file))
3763 + /* AuDbg("%pD\n", file); */
3764 + AuDebugOn(!d_is_dir(file->f_path.dentry));
3766 + fidir = au_fi(file)->fi_hdir;
3767 + AuDebugOn(!fidir);
3768 + fi_write_lock(file);
3769 + btop = au_fbtop(file);
3770 + bbot = au_fbbot_dir(file);
3771 + for (bindex = btop; bindex <= bbot; bindex++) {
3772 + hfile = fidir->fd_hfile + bindex;
3773 + if (!hfile->hf_file)
3776 + if (hfile->hf_br->br_id == br_id) {
3781 + AuDebugOn(bfound < 0);
3782 + au_set_h_fptr(file, bfound, NULL);
3783 + if (bfound == btop) {
3784 + for (btop++; btop <= bbot; btop++)
3785 + if (au_hf_dir(file, btop)) {
3786 + au_set_fbtop(file, btop);
3790 + fi_write_unlock(file);
3794 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3795 + const aufs_bindex_t bindex,
3796 + const aufs_bindex_t bbot)
3798 + struct au_branch **brp, **p;
3800 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3802 + brp = sbinfo->si_branch + bindex;
3803 + if (bindex < bbot)
3804 + memmove(brp, brp + 1, sizeof(*brp) * (bbot - bindex));
3805 + sbinfo->si_branch[0 + bbot] = NULL;
3806 + sbinfo->si_bbot--;
3808 + p = au_krealloc(sbinfo->si_branch, sizeof(*p) * bbot, AuGFP_SBILIST,
3811 + sbinfo->si_branch = p;
3812 + /* harmless error */
3815 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3816 + const aufs_bindex_t bbot)
3818 + struct au_hdentry *hdp, *p;
3820 + AuRwMustWriteLock(&dinfo->di_rwsem);
3822 + hdp = au_hdentry(dinfo, bindex);
3823 + if (bindex < bbot)
3824 + memmove(hdp, hdp + 1, sizeof(*hdp) * (bbot - bindex));
3825 + /* au_h_dentry_init(au_hdentry(dinfo, bbot); */
3828 + p = au_krealloc(dinfo->di_hdentry, sizeof(*p) * bbot, AuGFP_SBILIST,
3831 + dinfo->di_hdentry = p;
3832 + /* harmless error */
3835 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3836 + const aufs_bindex_t bbot)
3838 + struct au_hinode *hip, *p;
3840 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3842 + hip = au_hinode(iinfo, bindex);
3843 + if (bindex < bbot)
3844 + memmove(hip, hip + 1, sizeof(*hip) * (bbot - bindex));
3845 + /* au_hinode_init(au_hinode(iinfo, bbot)); */
3848 + p = au_krealloc(iinfo->ii_hinode, sizeof(*p) * bbot, AuGFP_SBILIST,
3851 + iinfo->ii_hinode = p;
3852 + /* harmless error */
3855 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3856 + struct au_branch *br)
3858 + aufs_bindex_t bbot;
3859 + struct au_sbinfo *sbinfo;
3860 + struct dentry *root, *h_root;
3861 + struct inode *inode, *h_inode;
3862 + struct au_hinode *hinode;
3864 + SiMustWriteLock(sb);
3866 + root = sb->s_root;
3867 + inode = d_inode(root);
3868 + sbinfo = au_sbi(sb);
3869 + bbot = sbinfo->si_bbot;
3871 + h_root = au_h_dptr(root, bindex);
3872 + hinode = au_hi(inode, bindex);
3873 + h_inode = au_igrab(hinode->hi_inode);
3876 + au_sbilist_lock();
3877 + au_br_do_del_brp(sbinfo, bindex, bbot);
3878 + au_br_do_del_hdp(au_di(root), bindex, bbot);
3879 + au_br_do_del_hip(au_ii(inode), bindex, bbot);
3880 + au_sbilist_unlock();
3882 + /* ignore an error */
3883 + au_dr_br_fin(sb, br); /* always, regardless the mount option */
3887 + au_br_do_free(br);
3890 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3891 + unsigned long long max, void *arg)
3896 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3899 + unsigned long long opened;
3900 + unsigned int mnt_flags;
3901 + aufs_bindex_t bindex, bbot, br_id;
3902 + unsigned char do_wh, verbose;
3903 + struct au_branch *br;
3904 + struct au_wbr *wbr;
3905 + struct dentry *root;
3906 + struct file **to_free;
3911 + root = sb->s_root;
3912 + bindex = au_find_dbindex(root, del->h_path.dentry);
3915 + goto out; /* success */
3917 + pr_err("%s no such branch\n", del->pathname);
3920 + AuDbg("bindex b%d\n", bindex);
3923 + mnt_flags = au_mntflags(sb);
3924 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3925 + bbot = au_sbbot(sb);
3926 + if (unlikely(!bbot)) {
3927 + AuVerbose(verbose, "no more branches left\n");
3931 + br = au_sbr(sb, bindex);
3932 + AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3933 + if (unlikely(au_lcnt_read(&br->br_count, /*do_rev*/1))) {
3934 + AuVerbose(verbose, "br %pd2 is busy now\n", del->h_path.dentry);
3938 + br_id = br->br_id;
3939 + opened = au_lcnt_read(&br->br_nfiles, /*do_rev*/1);
3940 + if (unlikely(opened)) {
3941 + to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3942 + err = PTR_ERR(to_free);
3943 + if (IS_ERR(to_free))
3946 + err = test_file_busy(sb, br_id, to_free, opened);
3947 + if (unlikely(err)) {
3948 + AuVerbose(verbose, "%llu file(s) opened\n", opened);
3954 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3956 + /* instead of WbrWhMustWriteLock(wbr) */
3957 + SiMustWriteLock(sb);
3958 + for (i = 0; i < AuBrWh_Last; i++) {
3959 + dput(wbr->wbr_wh[i]);
3960 + wbr->wbr_wh[i] = NULL;
3964 + err = test_children_busy(root, bindex, verbose);
3965 + if (unlikely(err)) {
3974 + * now we confirmed the branch is deletable.
3975 + * let's free the remaining opened dirs on the branch.
3977 + di_write_unlock(root);
3978 + br_del_file(to_free, opened, br_id);
3979 + di_write_lock_child(root);
3982 + sysaufs_brs_del(sb, bindex); /* remove successors */
3983 + dbgaufs_xino_del(br); /* remove one */
3984 + au_br_do_del(sb, bindex, br);
3985 + sysaufs_brs_add(sb, bindex); /* append successors */
3986 + dbgaufs_brs_add(sb, bindex, /*topdown*/1); /* rename successors */
3989 + au_cpup_attr_all(d_inode(root), /*force*/1);
3990 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3992 + au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
3993 + if (au_opt_test(mnt_flags, PLINK))
3994 + au_plink_half_refresh(sb, br_id);
3996 + goto out; /* success */
4000 + rerr = au_br_init_wh(sb, br, br->br_perm);
4002 + pr_warn("failed re-creating base whiteout, %s. (%d)\n",
4003 + del->pathname, rerr);
4006 + au_farray_free(to_free, opened);
4010 +/* ---------------------------------------------------------------------- */
4012 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
4015 + aufs_bindex_t btop, bbot;
4016 + struct aufs_ibusy ibusy;
4017 + struct inode *inode, *h_inode;
4020 + if (unlikely(!capable(CAP_SYS_ADMIN)))
4023 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
4025 + /* VERIFY_WRITE */
4026 + err = !access_ok(&arg->h_ino, sizeof(arg->h_ino));
4027 + if (unlikely(err)) {
4034 + si_read_lock(sb, AuLock_FLUSH);
4035 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbbot(sb)))
4039 + ibusy.h_ino = 0; /* invalid */
4040 + inode = ilookup(sb, ibusy.ino);
4042 + || inode->i_ino == AUFS_ROOT_INO
4043 + || au_is_bad_inode(inode))
4046 + ii_read_lock_child(inode);
4047 + btop = au_ibtop(inode);
4048 + bbot = au_ibbot(inode);
4049 + if (btop <= ibusy.bindex && ibusy.bindex <= bbot) {
4050 + h_inode = au_h_iptr(inode, ibusy.bindex);
4051 + if (h_inode && au_test_ibusy(inode, btop, bbot))
4052 + ibusy.h_ino = h_inode->i_ino;
4054 + ii_read_unlock(inode);
4058 + si_read_unlock(sb);
4060 + err = __put_user(ibusy.h_ino, &arg->h_ino);
4061 + if (unlikely(err)) {
4070 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
4072 + return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
4075 +#ifdef CONFIG_COMPAT
4076 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
4078 + return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
4082 +/* ---------------------------------------------------------------------- */
4085 + * change a branch permission
4088 +static void au_warn_ima(void)
4091 + /* since it doesn't support mark_files_ro() */
4092 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
4096 +static int do_need_sigen_inc(int a, int b)
4098 + return au_br_whable(a) && !au_br_whable(b);
4101 +static int need_sigen_inc(int old, int new)
4103 + return do_need_sigen_inc(old, new)
4104 + || do_need_sigen_inc(new, old);
4107 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
4110 + unsigned int mnt_flags;
4111 + unsigned long long ull, max;
4112 + aufs_bindex_t br_id;
4113 + unsigned char verbose, writer;
4114 + struct file *file, *hf, **array;
4115 + struct au_hfile *hfile;
4116 + struct inode *h_inode;
4118 + mnt_flags = au_mntflags(sb);
4119 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4121 + array = au_farray_alloc(sb, &max);
4122 + err = PTR_ERR(array);
4123 + if (IS_ERR(array))
4127 + br_id = au_sbr_id(sb, bindex);
4128 + for (ull = 0; ull < max; ull++) {
4129 + file = array[ull];
4130 + if (unlikely(!file))
4133 + /* AuDbg("%pD\n", file); */
4134 + fi_read_lock(file);
4135 + if (unlikely(au_test_mmapped(file))) {
4137 + AuVerbose(verbose, "mmapped %pD\n", file);
4139 + FiMustNoWaiters(file);
4140 + fi_read_unlock(file);
4144 + hfile = &au_fi(file)->fi_htop;
4145 + hf = hfile->hf_file;
4146 + if (!d_is_reg(file->f_path.dentry)
4147 + || !(file->f_mode & FMODE_WRITE)
4148 + || hfile->hf_br->br_id != br_id
4149 + || !(hf->f_mode & FMODE_WRITE))
4150 + array[ull] = NULL;
4156 + FiMustNoWaiters(file);
4157 + fi_read_unlock(file);
4165 + for (ull = 0; ull < max; ull++) {
4166 + file = array[ull];
4170 + /* todo: already flushed? */
4172 + * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4173 + * approach which resets f_mode and calls mnt_drop_write() and
4174 + * file_release_write() for each file, because the branch
4175 + * attribute in aufs world is totally different from the native
4178 + /* fi_read_lock(file); */
4179 + hfile = &au_fi(file)->fi_htop;
4180 + hf = hfile->hf_file;
4181 + /* fi_read_unlock(file); */
4182 + spin_lock(&hf->f_lock);
4183 + writer = !!(hf->f_mode & FMODE_WRITER);
4184 + hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4185 + spin_unlock(&hf->f_lock);
4187 + h_inode = file_inode(hf);
4188 + if (hf->f_mode & FMODE_READ)
4189 + i_readcount_inc(h_inode);
4190 + put_write_access(h_inode);
4191 + __mnt_drop_write(hf->f_path.mnt);
4196 + au_farray_free(array, max);
4202 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4206 + aufs_bindex_t bindex;
4207 + struct dentry *root;
4208 + struct au_branch *br;
4209 + struct au_br_fhsm *bf;
4211 + root = sb->s_root;
4212 + bindex = au_find_dbindex(root, mod->h_root);
4215 + return 0; /* success */
4217 + pr_err("%s no such branch\n", mod->path);
4220 + AuDbg("bindex b%d\n", bindex);
4222 + err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4223 + if (unlikely(err))
4226 + br = au_sbr(sb, bindex);
4227 + AuDebugOn(mod->h_root != au_br_dentry(br));
4228 + if (br->br_perm == mod->perm)
4229 + return 0; /* success */
4231 + /* pre-allocate for non-fhsm --> fhsm */
4233 + if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4234 + err = au_fhsm_br_alloc(br);
4235 + if (unlikely(err))
4238 + br->br_fhsm = NULL;
4241 + if (au_br_writable(br->br_perm)) {
4242 + /* remove whiteout base */
4243 + err = au_br_init_wh(sb, br, mod->perm);
4244 + if (unlikely(err))
4247 + if (!au_br_writable(mod->perm)) {
4248 + /* rw --> ro, file might be mmapped */
4249 + DiMustNoWaiters(root);
4250 + IiMustNoWaiters(d_inode(root));
4251 + di_write_unlock(root);
4252 + err = au_br_mod_files_ro(sb, bindex);
4253 + /* aufs_write_lock() calls ..._child() */
4254 + di_write_lock_child(root);
4256 + if (unlikely(err)) {
4258 + br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4261 + rerr = au_wbr_init(br, sb, br->br_perm);
4262 + if (unlikely(rerr)) {
4263 + AuIOErr("nested error %d (%d)\n",
4265 + br->br_perm = mod->perm;
4269 + } else if (au_br_writable(mod->perm)) {
4272 + br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4274 + err = au_wbr_init(br, sb, mod->perm);
4275 + if (unlikely(err)) {
4276 + au_kfree_rcu(br->br_wbr);
4277 + br->br_wbr = NULL;
4281 + if (unlikely(err))
4284 + if (au_br_fhsm(br->br_perm)) {
4285 + if (!au_br_fhsm(mod->perm)) {
4286 + /* fhsm --> non-fhsm */
4287 + au_br_fhsm_fin(br->br_fhsm);
4288 + au_kfree_rcu(br->br_fhsm);
4289 + br->br_fhsm = NULL;
4291 + } else if (au_br_fhsm(mod->perm))
4292 + /* non-fhsm --> fhsm */
4295 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4296 + br->br_perm = mod->perm;
4297 + goto out; /* success */
4300 + au_kfree_try_rcu(bf);
4306 +/* ---------------------------------------------------------------------- */
4308 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4311 + struct kstatfs kstfs;
4313 + err = vfs_statfs(&br->br_path, &kstfs);
4315 + stfs->f_blocks = kstfs.f_blocks;
4316 + stfs->f_bavail = kstfs.f_bavail;
4317 + stfs->f_files = kstfs.f_files;
4318 + stfs->f_ffree = kstfs.f_ffree;
4323 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4324 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
4325 +++ linux/fs/aufs/branch.h 2023-04-24 08:13:40.969576115 +0200
4327 +/* SPDX-License-Identifier: GPL-2.0 */
4329 + * Copyright (C) 2005-2022 Junjiro R. Okajima
4331 + * This program is free software; you can redistribute it and/or modify
4332 + * it under the terms of the GNU General Public License as published by
4333 + * the Free Software Foundation; either version 2 of the License, or
4334 + * (at your option) any later version.
4336 + * This program is distributed in the hope that it will be useful,
4337 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4338 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4339 + * GNU General Public License for more details.
4341 + * You should have received a copy of the GNU General Public License
4342 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4346 + * branch filesystems and xino for them
4349 +#ifndef __AUFS_BRANCH_H__
4350 +#define __AUFS_BRANCH_H__
4354 +#include <linux/mount.h>
4355 +#include "dirren.h"
4361 +/* ---------------------------------------------------------------------- */
4365 + struct file **xi_file;
4366 + unsigned int xi_nfile;
4372 + /* reserved for future use */
4373 + /* unsigned long *bitmap; */
4374 + wait_queue_head_t wqh;
4377 + struct mutex xi_mtx; /* protects xi_file array */
4378 + struct hlist_bl_head xi_writing;
4380 + atomic_t xi_truncating;
4382 + struct kref xi_kref;
4385 +/* File-based Hierarchical Storage Management */
4386 +struct au_br_fhsm {
4387 +#ifdef CONFIG_AUFS_FHSM
4388 + struct mutex bf_lock;
4389 + unsigned long bf_jiffy;
4390 + struct aufs_stfs bf_stfs;
4395 +/* members for writable branch only */
4396 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4398 + struct au_rwsem wbr_wh_rwsem;
4399 + struct dentry *wbr_wh[AuBrWh_Last];
4400 + atomic_t wbr_wh_running;
4401 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4402 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4403 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4406 + unsigned long long wbr_bytes;
4409 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4410 +#define AuBrDynOp (AuDyLast * 4)
4412 +#ifdef CONFIG_AUFS_HFSNOTIFY
4413 +/* support for asynchronous destruction */
4414 +struct au_br_hfsnotify {
4415 + struct fsnotify_group *hfsn_group;
4419 +/* sysfs entries */
4420 +struct au_brsysfs {
4422 + struct attribute attr;
4431 +/* protected by superblock rwsem */
4433 + struct au_xino *br_xino;
4435 + aufs_bindex_t br_id;
4438 + struct path br_path;
4439 + spinlock_t br_dykey_lock;
4440 + struct au_dykey *br_dykey[AuBrDynOp];
4441 + au_lcnt_t br_nfiles; /* opened files */
4442 + au_lcnt_t br_count; /* in-use for other */
4444 + struct au_wbr *br_wbr;
4445 + struct au_br_fhsm *br_fhsm;
4447 +#ifdef CONFIG_AUFS_HFSNOTIFY
4448 + struct au_br_hfsnotify *br_hfsn;
4451 +#ifdef CONFIG_SYSFS
4452 + /* entries under sysfs per mount-point */
4453 + struct au_brsysfs br_sysfs[AuBrSysfs_Last];
4456 +#ifdef CONFIG_DEBUG_FS
4457 + struct dentry *br_dbgaufs; /* xino */
4460 + struct au_dr_br br_dirren;
4463 +/* ---------------------------------------------------------------------- */
4465 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4467 + return br->br_path.mnt;
4470 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4472 + return br->br_path.dentry;
4475 +static inline struct mnt_idmap *au_br_idmap(struct au_branch *br)
4477 + return mnt_idmap(br->br_path.mnt);
4480 +static inline struct super_block *au_br_sb(struct au_branch *br)
4482 + return au_br_mnt(br)->mnt_sb;
4485 +static inline int au_br_rdonly(struct au_branch *br)
4487 + return (sb_rdonly(au_br_sb(br))
4488 + || !au_br_writable(br->br_perm))
4492 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4494 +#ifdef CONFIG_AUFS_HNOTIFY
4495 + return !(brperm & AuBrPerm_RR);
4501 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4503 + int err, exec_flag;
4506 + exec_flag = oflag & __FMODE_EXEC;
4507 + if (unlikely(exec_flag && path_noexec(&br->br_path)))
4513 +static inline void au_xino_get(struct au_branch *br)
4515 + struct au_xino *xi;
4519 + kref_get(&xi->xi_kref);
4522 +static inline int au_xino_count(struct au_branch *br)
4525 + struct au_xino *xi;
4530 + v = kref_read(&xi->xi_kref);
4535 +/* ---------------------------------------------------------------------- */
4539 +void au_br_free(struct au_sbinfo *sinfo);
4540 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4542 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4544 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4545 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4546 +#ifdef CONFIG_COMPAT
4547 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4550 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4553 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4556 +static const loff_t au_loff_max = LLONG_MAX;
4558 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry);
4559 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
4561 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
4562 + struct file *copy_src);
4564 + struct au_xino *xi; /* switch between xino and xigen */
4566 + struct path *base;
4567 + struct file *copy_src;
4569 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew);
4571 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4573 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4575 +ssize_t xino_fread(struct file *file, void *buf, size_t size, loff_t *pos);
4576 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos);
4578 +int au_xib_trunc(struct super_block *sb);
4579 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin);
4581 +struct au_xino *au_xino_alloc(unsigned int nfile);
4582 +int au_xino_put(struct au_branch *br);
4583 +struct file *au_xino_file1(struct au_xino *xi);
4585 +struct au_opt_xino;
4586 +void au_xino_clr(struct super_block *sb);
4587 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount);
4588 +struct file *au_xino_def(struct super_block *sb);
4589 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4590 + struct path *base);
4592 +ino_t au_xino_new_ino(struct super_block *sb);
4593 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4595 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
4596 + ino_t h_ino, int idx);
4597 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4600 +int au_xino_path(struct seq_file *seq, struct file *file);
4602 +/* ---------------------------------------------------------------------- */
4604 +/* @idx is signed to accept -1 meaning the first file */
4605 +static inline struct file *au_xino_file(struct au_xino *xi, int idx)
4607 + struct file *file;
4614 + if (idx < xi->xi_nfile)
4615 + file = xi->xi_file[idx];
4617 + file = au_xino_file1(xi);
4623 +/* ---------------------------------------------------------------------- */
4625 +/* Superblock to branch */
4627 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4629 + return au_sbr(sb, bindex)->br_id;
4633 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4635 + return au_br_mnt(au_sbr(sb, bindex));
4639 +struct mnt_idmap *au_sbr_idmap(struct super_block *sb, aufs_bindex_t bindex)
4641 + return au_br_idmap(au_sbr(sb, bindex));
4645 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4647 + return au_br_sb(au_sbr(sb, bindex));
4650 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4652 + return au_sbr(sb, bindex)->br_perm;
4655 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4657 + return au_br_whable(au_sbr_perm(sb, bindex));
4660 +/* ---------------------------------------------------------------------- */
4662 +#define wbr_wh_read_lock(wbr) au_rw_read_lock(&(wbr)->wbr_wh_rwsem)
4663 +#define wbr_wh_write_lock(wbr) au_rw_write_lock(&(wbr)->wbr_wh_rwsem)
4664 +#define wbr_wh_read_trylock(wbr) au_rw_read_trylock(&(wbr)->wbr_wh_rwsem)
4665 +#define wbr_wh_write_trylock(wbr) au_rw_write_trylock(&(wbr)->wbr_wh_rwsem)
4667 +#define wbr_wh_read_trylock_nested(wbr) \
4668 + au_rw_read_trylock_nested(&(wbr)->wbr_wh_rwsem)
4669 +#define wbr_wh_write_trylock_nested(wbr) \
4670 + au_rw_write_trylock_nested(&(wbr)->wbr_wh_rwsem)
4673 +#define wbr_wh_read_unlock(wbr) au_rw_read_unlock(&(wbr)->wbr_wh_rwsem)
4674 +#define wbr_wh_write_unlock(wbr) au_rw_write_unlock(&(wbr)->wbr_wh_rwsem)
4675 +#define wbr_wh_downgrade_lock(wbr) au_rw_dgrade_lock(&(wbr)->wbr_wh_rwsem)
4677 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&(wbr)->wbr_wh_rwsem)
4678 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&(wbr)->wbr_wh_rwsem)
4679 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&(wbr)->wbr_wh_rwsem)
4681 +/* ---------------------------------------------------------------------- */
4683 +#ifdef CONFIG_AUFS_FHSM
4684 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4686 + mutex_init(&brfhsm->bf_lock);
4687 + brfhsm->bf_jiffy = 0;
4688 + brfhsm->bf_readable = 0;
4691 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4693 + mutex_destroy(&brfhsm->bf_lock);
4696 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4697 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4700 +#endif /* __KERNEL__ */
4701 +#endif /* __AUFS_BRANCH_H__ */
4702 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4703 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
4704 +++ linux/fs/aufs/conf.mk 2022-11-05 23:02:18.959222617 +0100
4706 +# SPDX-License-Identifier: GPL-2.0
4708 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4712 +AuConfStr += ${1}=${${1}}
4716 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4718 + HNOTIFY HFSNOTIFY \
4730 +$(foreach i, ${AuConfAll}, \
4731 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
4733 +AuConfName = ${obj}/conf.str
4734 +${AuConfName}.tmp: FORCE
4735 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4736 +${AuConfName}: ${AuConfName}.tmp
4737 + @diff -q $< $@ > /dev/null 2>&1 || { \
4738 + echo ' GEN ' $@; \
4742 +clean-files += ${AuConfName} ${AuConfName}.tmp
4743 +${obj}/sysfs.o: ${AuConfName}
4745 +-include ${srctree}/${src}/conf_priv.mk
4746 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4747 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
4748 +++ linux/fs/aufs/cpup.c 2023-04-24 08:13:40.969576115 +0200
4750 +// SPDX-License-Identifier: GPL-2.0
4752 + * Copyright (C) 2005-2022 Junjiro R. Okajima
4754 + * This program is free software; you can redistribute it and/or modify
4755 + * it under the terms of the GNU General Public License as published by
4756 + * the Free Software Foundation; either version 2 of the License, or
4757 + * (at your option) any later version.
4759 + * This program is distributed in the hope that it will be useful,
4760 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4761 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4762 + * GNU General Public License for more details.
4764 + * You should have received a copy of the GNU General Public License
4765 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4769 + * copy-up functions, see wbr_policy.c for copy-down
4772 +#include <linux/fs_stack.h>
4773 +#include <linux/mm.h>
4774 +#include <linux/task_work.h>
4777 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4779 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4780 + | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4782 + BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4784 + dst->i_flags |= iflags & ~mask;
4785 + if (au_test_fs_notime(dst->i_sb))
4786 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
4789 +void au_cpup_attr_timesizes(struct inode *inode)
4791 + struct inode *h_inode;
4793 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4794 + fsstack_copy_attr_times(inode, h_inode);
4795 + fsstack_copy_inode_size(inode, h_inode);
4798 +void au_cpup_attr_nlink(struct inode *inode, int force)
4800 + struct inode *h_inode;
4801 + struct super_block *sb;
4802 + aufs_bindex_t bindex, bbot;
4805 + bindex = au_ibtop(inode);
4806 + h_inode = au_h_iptr(inode, bindex);
4808 + && !S_ISDIR(h_inode->i_mode)
4809 + && au_opt_test(au_mntflags(sb), PLINK)
4810 + && au_plink_test(inode))
4814 + * 0 can happen in revalidating.
4815 + * h_inode->i_mutex may not be held here, but it is harmless since once
4816 + * i_nlink reaches 0, it will never become positive except O_TMPFILE
4818 + * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4819 + * the incorrect link count.
4821 + set_nlink(inode, h_inode->i_nlink);
4824 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4825 + * it may includes whplink directory.
4827 + if (S_ISDIR(h_inode->i_mode)) {
4828 + bbot = au_ibbot(inode);
4829 + for (bindex++; bindex <= bbot; bindex++) {
4830 + h_inode = au_h_iptr(inode, bindex);
4832 + au_add_nlink(inode, h_inode);
4837 +void au_cpup_attr_changeable(struct inode *inode)
4839 + struct inode *h_inode;
4841 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4842 + inode->i_mode = h_inode->i_mode;
4843 + inode->i_uid = h_inode->i_uid;
4844 + inode->i_gid = h_inode->i_gid;
4845 + au_cpup_attr_timesizes(inode);
4846 + au_cpup_attr_flags(inode, h_inode->i_flags);
4849 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4851 + struct au_iinfo *iinfo = au_ii(inode);
4853 + IiMustWriteLock(inode);
4855 + iinfo->ii_higen = h_inode->i_generation;
4856 + iinfo->ii_hsb1 = h_inode->i_sb;
4859 +void au_cpup_attr_all(struct inode *inode, int force)
4861 + struct inode *h_inode;
4863 + h_inode = au_h_iptr(inode, au_ibtop(inode));
4864 + au_cpup_attr_changeable(inode);
4865 + if (inode->i_nlink > 0)
4866 + au_cpup_attr_nlink(inode, force);
4867 + inode->i_rdev = h_inode->i_rdev;
4868 + inode->i_blkbits = h_inode->i_blkbits;
4869 + au_cpup_igen(inode, h_inode);
4872 +/* ---------------------------------------------------------------------- */
4874 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4876 +/* keep the timestamps of the parent dir when cpup */
4877 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4878 + struct path *h_path)
4880 + struct inode *h_inode;
4882 + dt->dt_dentry = dentry;
4883 + dt->dt_h_path = *h_path;
4884 + h_inode = d_inode(h_path->dentry);
4885 + dt->dt_atime = h_inode->i_atime;
4886 + dt->dt_mtime = h_inode->i_mtime;
4890 +void au_dtime_revert(struct au_dtime *dt)
4892 + struct iattr attr;
4895 + attr.ia_atime = dt->dt_atime;
4896 + attr.ia_mtime = dt->dt_mtime;
4897 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4898 + | ATTR_ATIME | ATTR_ATIME_SET;
4900 + /* no delegation since this is a directory */
4901 + err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4902 + if (unlikely(err))
4903 + pr_warn("restoring timestamps failed(%d). ignored\n", err);
4906 +/* ---------------------------------------------------------------------- */
4908 +/* internal use only */
4909 +struct au_cpup_reg_attr {
4912 + unsigned int iflags; /* inode->i_flags */
4915 +static noinline_for_stack
4916 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct path *h_src,
4917 + struct au_cpup_reg_attr *h_src_attr)
4919 + int err, sbits, icex;
4920 + unsigned int mnt_flags;
4921 + unsigned char verbose;
4923 + struct path h_path;
4924 + struct inode *h_isrc, *h_idst;
4925 + struct kstat *h_st;
4926 + struct au_branch *br;
4928 + br = au_sbr(dst->d_sb, bindex);
4929 + h_path.mnt = au_br_mnt(br);
4930 + h_path.dentry = au_h_dptr(dst, bindex);
4931 + h_idst = d_inode(h_path.dentry);
4932 + h_isrc = d_inode(h_src->dentry);
4933 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4934 + | ATTR_ATIME | ATTR_MTIME
4935 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
4936 + if (h_src_attr && h_src_attr->valid) {
4937 + h_st = &h_src_attr->st;
4938 + ia.ia_uid = h_st->uid;
4939 + ia.ia_gid = h_st->gid;
4940 + ia.ia_atime = h_st->atime;
4941 + ia.ia_mtime = h_st->mtime;
4942 + if (h_idst->i_mode != h_st->mode
4943 + && !S_ISLNK(h_idst->i_mode)) {
4944 + ia.ia_valid |= ATTR_MODE;
4945 + ia.ia_mode = h_st->mode;
4947 + sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4948 + au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4950 + ia.ia_uid = h_isrc->i_uid;
4951 + ia.ia_gid = h_isrc->i_gid;
4952 + ia.ia_atime = h_isrc->i_atime;
4953 + ia.ia_mtime = h_isrc->i_mtime;
4954 + if (h_idst->i_mode != h_isrc->i_mode
4955 + && !S_ISLNK(h_idst->i_mode)) {
4956 + ia.ia_valid |= ATTR_MODE;
4957 + ia.ia_mode = h_isrc->i_mode;
4959 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4960 + au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4962 + /* no delegation since it is just created */
4963 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4965 + /* is this nfs only? */
4966 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4967 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4968 + ia.ia_mode = h_isrc->i_mode;
4969 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4972 + icex = br->br_perm & AuBrAttr_ICEX;
4974 + mnt_flags = au_mntflags(dst->d_sb);
4975 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4976 + err = au_cpup_xattr(&h_path, h_src, icex, verbose);
4982 +/* ---------------------------------------------------------------------- */
4984 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4985 + char *buf, unsigned long blksize)
4988 + size_t sz, rbytes, wbytes;
4989 + unsigned char all_zero;
4991 + struct inode *h_inode;
4992 + /* reduce stack usage */
4995 + zp = page_address(ZERO_PAGE(0));
4996 + if (unlikely(!zp))
4997 + return -ENOMEM; /* possible? */
5002 + AuDbg("len %lld\n", len);
5004 + if (len < blksize)
5008 + /* todo: signal_pending? */
5009 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
5010 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
5013 + if (unlikely(err < 0))
5017 + if (len >= rbytes && rbytes == blksize)
5018 + all_zero = !memcmp(buf, zp, rbytes);
5025 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
5027 + /* todo: signal_pending? */
5028 + if (unlikely(err == -EAGAIN || err == -EINTR))
5030 + if (unlikely(err < 0))
5035 + if (unlikely(err < 0))
5041 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
5043 + if (unlikely(res < 0))
5050 + /* the last block may be a hole */
5051 + if (!err && all_zero) {
5052 + AuLabel(last hole);
5055 + if (au_test_nfs(dst->f_path.dentry->d_sb)) {
5056 + /* nfs requires this step to make last hole */
5057 + /* is this only nfs? */
5059 + /* todo: signal_pending? */
5060 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
5061 + } while (err == -EAGAIN || err == -EINTR);
5068 + ia->ia_size = dst->f_pos;
5069 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
5070 + ia->ia_file = dst;
5071 + h_inode = file_inode(dst);
5072 + inode_lock_nested(h_inode, AuLsc_I_CHILD2);
5073 + /* no delegation since it is just created */
5074 + err = vfsub_notify_change(&dst->f_path, ia,
5075 + /*delegated*/NULL);
5076 + inode_unlock(h_inode);
5083 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
5086 + unsigned long blksize;
5087 + unsigned char do_kfree;
5089 + struct super_block *h_sb;
5092 + h_sb = file_inode(dst)->i_sb;
5093 + blksize = h_sb->s_blocksize;
5094 + if (!blksize || PAGE_SIZE < blksize)
5095 + blksize = PAGE_SIZE;
5096 + AuDbg("blksize %lu\n", blksize);
5097 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
5099 + buf = kmalloc(blksize, GFP_NOFS);
5101 + buf = (void *)__get_free_page(GFP_NOFS);
5102 + if (unlikely(!buf))
5105 + if (len > (1 << 22))
5106 + AuDbg("copying a large file %lld\n", (long long)len);
5110 + err = au_do_copy_file(dst, src, len, buf, blksize);
5112 + AuDebugOn(!au_kfree_do_sz_test(blksize));
5113 + au_kfree_do_rcu(buf);
5115 + free_page((unsigned long)buf);
5121 +static int au_do_copy(struct file *dst, struct file *src, loff_t len)
5124 + struct super_block *h_src_sb;
5125 + struct inode *h_src_inode;
5127 + h_src_inode = file_inode(src);
5128 + h_src_sb = h_src_inode->i_sb;
5130 + /* XFS acquires inode_lock */
5131 + if (!au_test_xfs(h_src_sb))
5132 + err = au_copy_file(dst, src, len);
5134 + inode_unlock_shared(h_src_inode);
5135 + err = au_copy_file(dst, src, len);
5136 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5142 +static int au_clone_or_copy(struct file *dst, struct file *src, loff_t len)
5146 + struct super_block *h_src_sb;
5147 + struct inode *h_src_inode;
5149 + h_src_inode = file_inode(src);
5150 + h_src_sb = h_src_inode->i_sb;
5151 + if (h_src_sb != file_inode(dst)->i_sb
5152 + || !dst->f_op->remap_file_range) {
5153 + err = au_do_copy(dst, src, len);
5157 + if (!au_test_nfs(h_src_sb)) {
5158 + inode_unlock_shared(h_src_inode);
5159 + lo = vfsub_clone_file_range(src, dst, len);
5160 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5162 + lo = vfsub_clone_file_range(src, dst, len);
5165 + goto out; /* success */
5166 + } else if (lo >= 0)
5167 + /* todo: possible? */
5168 + /* paritially succeeded */
5169 + AuDbg("lo %lld, len %lld. Retrying.\n", lo, len);
5170 + else if (lo != -EOPNOTSUPP) {
5171 + /* older XFS has a condition in cloning */
5176 + /* the backend fs on NFS may not support cloning */
5177 + err = au_do_copy(dst, src, len);
5185 + * to support a sparse file which is opened with O_APPEND,
5186 + * we need to close the file.
5188 +static int au_cp_regular(struct au_cp_generic *cpg)
5191 + enum { SRC, DST };
5193 + aufs_bindex_t bindex;
5194 + unsigned int flags;
5195 + struct dentry *dentry;
5197 + struct file *file;
5200 + .bindex = cpg->bsrc,
5201 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
5204 + .bindex = cpg->bdst,
5205 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
5206 + .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
5209 + struct au_branch *br;
5210 + struct super_block *sb, *h_src_sb;
5211 + struct inode *h_src_inode;
5212 + struct task_struct *tsk = current;
5214 + /* bsrc branch can be ro/rw. */
5215 + sb = cpg->dentry->d_sb;
5217 + for (i = 0; i < 2; i++, f++) {
5218 + f->dentry = au_h_dptr(cpg->dentry, f->bindex);
5219 + f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
5220 + /*file*/NULL, f->force_wr);
5221 + if (IS_ERR(f->file)) {
5222 + err = PTR_ERR(f->file);
5230 + /* try stopping to update while we copyup */
5231 + h_src_inode = d_inode(file[SRC].dentry);
5232 + h_src_sb = h_src_inode->i_sb;
5233 + if (!au_test_nfs(h_src_sb))
5234 + IMustLock(h_src_inode);
5235 + err = au_clone_or_copy(file[DST].file, file[SRC].file, cpg->len);
5237 + /* i wonder if we had O_NO_DELAY_FPUT flag */
5238 + if (tsk->flags & PF_KTHREAD)
5239 + __fput_sync(file[DST].file);
5241 + /* it happened actually */
5242 + fput(file[DST].file);
5245 + * we have to call both since we don't know which place the file
5249 + flush_delayed_fput();
5251 + br = au_sbr(sb, file[DST].bindex);
5252 + au_lcnt_dec(&br->br_nfiles);
5255 + fput(file[SRC].file);
5256 + br = au_sbr(sb, file[SRC].bindex);
5257 + au_lcnt_dec(&br->br_nfiles);
5262 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
5263 + struct au_cpup_reg_attr *h_src_attr)
5267 + struct path h_path;
5268 + struct inode *h_src_inode, *h_dst_inode;
5271 + h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
5272 + l = i_size_read(h_src_inode);
5273 + if (cpg->len == -1 || l < cpg->len)
5276 + /* try stopping to update while we are referencing */
5277 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5278 + au_pin_hdir_unlock(cpg->pin);
5280 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5281 + h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
5282 + h_src_attr->iflags = h_src_inode->i_flags;
5283 + if (!au_test_nfs(h_src_inode->i_sb))
5284 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5286 + inode_unlock_shared(h_src_inode);
5287 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5288 + inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5290 + if (unlikely(err)) {
5291 + inode_unlock_shared(h_src_inode);
5294 + h_src_attr->valid = 1;
5295 + if (!au_test_nfs(h_src_inode->i_sb)) {
5296 + err = au_cp_regular(cpg);
5297 + inode_unlock_shared(h_src_inode);
5299 + inode_unlock_shared(h_src_inode);
5300 + err = au_cp_regular(cpg);
5302 + rerr = au_pin_hdir_relock(cpg->pin);
5306 + if (!err && (h_src_inode->i_state & I_LINKABLE)) {
5307 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5308 + h_dst_inode = d_inode(h_path.dentry);
5309 + spin_lock(&h_dst_inode->i_lock);
5310 + h_dst_inode->i_state |= I_LINKABLE;
5311 + spin_unlock(&h_dst_inode->i_lock);
5318 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
5319 + struct inode *h_dir)
5322 + DEFINE_DELAYED_CALL(done);
5325 + sym = vfs_get_link(h_src, &done);
5326 + err = PTR_ERR(sym);
5330 + err = vfsub_symlink(h_dir, h_path, sym);
5333 + do_delayed_call(&done);
5338 + * regardless 'acl' option, reset all ACL.
5339 + * All ACL will be copied up later from the original entry on the lower branch.
5341 +static int au_reset_acl(struct path *h_path, umode_t mode)
5344 + struct dentry *h_dentry;
5345 + /* struct inode *h_inode; */
5346 + struct mnt_idmap *h_idmap;
5348 + h_idmap = mnt_idmap(h_path->mnt);
5349 + h_dentry = h_path->dentry;
5350 + /* h_inode = d_inode(h_dentry); */
5351 + /* forget_all_cached_acls(h_inode)); */
5352 + err = vfsub_remove_acl(h_idmap, h_dentry, XATTR_NAME_POSIX_ACL_ACCESS);
5354 + if (err == -EOPNOTSUPP)
5357 + err = vfsub_acl_chmod(h_idmap, h_dentry, mode);
5363 +static int au_do_cpup_dir(struct au_cp_generic *cpg, struct dentry *dst_parent,
5364 + struct inode *h_dir, struct path *h_path)
5367 + struct inode *dir, *inode;
5368 + struct mnt_idmap *h_idmap;
5370 + h_idmap = mnt_idmap(h_path->mnt);
5371 + err = vfsub_remove_acl(h_idmap, h_path->dentry,
5372 + XATTR_NAME_POSIX_ACL_DEFAULT);
5374 + if (err == -EOPNOTSUPP)
5376 + if (unlikely(err))
5380 + * strange behaviour from the users view,
5381 + * particularly setattr case
5383 + dir = d_inode(dst_parent);
5384 + if (au_ibtop(dir) == cpg->bdst)
5385 + au_cpup_attr_nlink(dir, /*force*/1);
5386 + inode = d_inode(cpg->dentry);
5387 + au_cpup_attr_nlink(inode, /*force*/1);
5393 +static noinline_for_stack
5394 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5395 + struct au_cpup_reg_attr *h_src_attr)
5399 + unsigned int mnt_flags;
5400 + unsigned char isdir, isreg, force;
5401 + const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5402 + struct au_dtime dt;
5403 + struct path h_path;
5404 + struct dentry *h_src, *h_dst, *h_parent;
5405 + struct inode *h_inode, *h_dir;
5406 + struct super_block *sb;
5408 + /* bsrc branch can be ro/rw. */
5409 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5410 + h_inode = d_inode(h_src);
5411 + AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5413 + /* try stopping to be referenced while we are creating */
5414 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5415 + if (au_ftest_cpup(cpg->flags, RENAME))
5416 + AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5417 + AUFS_WH_PFX_LEN));
5418 + h_parent = h_dst->d_parent; /* dir inode is locked */
5419 + h_dir = d_inode(h_parent);
5421 + AuDebugOn(h_parent != h_dst->d_parent);
5423 + sb = cpg->dentry->d_sb;
5424 + h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5426 + h_path.dentry = h_parent;
5427 + au_dtime_store(&dt, dst_parent, &h_path);
5429 + h_path.dentry = h_dst;
5433 + mode = h_inode->i_mode;
5434 + switch (mode & S_IFMT) {
5437 + err = vfsub_create(h_dir, &h_path, 0600, /*want_excl*/true);
5439 + err = au_do_cpup_regular(cpg, h_src_attr);
5443 + err = vfsub_mkdir(h_dir, &h_path, mode);
5445 + err = au_do_cpup_dir(cpg, dst_parent, h_dir, &h_path);
5448 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5452 + AuDebugOn(!capable(CAP_MKNOD));
5456 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5459 + AuIOErr("Unknown inode type 0%o\n", mode);
5463 + err = au_reset_acl(&h_path, mode);
5465 + mnt_flags = au_mntflags(sb);
5466 + if (!au_opt_test(mnt_flags, UDBA_NONE)
5468 + && au_opt_test(mnt_flags, XINO)
5469 + && (h_inode->i_nlink == 1
5470 + || (h_inode->i_state & I_LINKABLE))
5471 + /* todo: unnecessary? */
5472 + /* && d_inode(cpg->dentry)->i_nlink == 1 */
5473 + && cpg->bdst < cpg->bsrc
5474 + && !au_ftest_cpup(cpg->flags, KEEPLINO))
5475 + au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5476 + /* ignore this error */
5481 + force = !!cpg->len;
5482 + if (cpg->len == -1)
5483 + force = !!i_size_read(h_inode);
5485 + au_fhsm_wrote(sb, cpg->bdst, force);
5489 + au_dtime_revert(&dt);
5493 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5496 + struct dentry *dentry, *h_dentry, *h_parent, *parent;
5497 + struct path h_ppath;
5498 + struct inode *h_dir;
5499 + aufs_bindex_t bdst;
5501 + dentry = cpg->dentry;
5503 + h_ppath.mnt = au_sbr_mnt(dentry->d_sb, bdst);
5504 + h_dentry = au_h_dptr(dentry, bdst);
5505 + if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5507 + au_set_h_dptr(dentry, bdst, NULL);
5508 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
5510 + h_path->dentry = dget(au_h_dptr(dentry, bdst));
5511 + au_set_h_dptr(dentry, bdst, h_dentry);
5514 + parent = dget_parent(dentry);
5515 + h_ppath.dentry = au_h_dptr(parent, bdst);
5517 + h_path->dentry = vfsub_lkup_one(&dentry->d_name, &h_ppath);
5518 + if (IS_ERR(h_path->dentry))
5519 + err = PTR_ERR(h_path->dentry);
5521 + if (unlikely(err))
5524 + h_parent = h_dentry->d_parent; /* dir inode is locked */
5525 + h_dir = d_inode(h_parent);
5527 + AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5528 + /* no delegation since it is just created */
5529 + err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL,
5531 + dput(h_path->dentry);
5538 + * copyup the @dentry from @bsrc to @bdst.
5539 + * the caller must set the both of lower dentries.
5540 + * @len is for truncating when it is -1 copyup the entire file.
5541 + * in link/rename cases, @dst_parent may be different from the real one.
5542 + * basic->bsrc can be larger than basic->bdst.
5543 + * aufs doesn't touch the credential so
5544 + * security_inode_copy_up{,_xattr}() are unnecessary.
5546 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5549 + aufs_bindex_t old_ibtop;
5550 + unsigned char isdir, plink;
5551 + struct dentry *h_src, *h_dst, *h_parent;
5552 + struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5553 + struct super_block *sb;
5554 + struct au_branch *br;
5555 + struct path h_src_path;
5556 + /* to reduce stack size */
5558 + struct au_dtime dt;
5559 + struct path h_path;
5560 + struct au_cpup_reg_attr h_src_attr;
5564 + a = kmalloc(sizeof(*a), GFP_NOFS);
5567 + a->h_src_attr.valid = 0;
5569 + sb = cpg->dentry->d_sb;
5570 + br = au_sbr(sb, cpg->bdst);
5571 + a->h_path.mnt = au_br_mnt(br);
5572 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5573 + h_parent = h_dst->d_parent; /* dir inode is locked */
5574 + h_dir = d_inode(h_parent);
5577 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5578 + inode = d_inode(cpg->dentry);
5581 + dst_parent = dget_parent(cpg->dentry);
5585 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
5586 + dst_inode = au_h_iptr(inode, cpg->bdst);
5588 + if (unlikely(!plink)) {
5590 + AuIOErr("hi%lu(i%lu) exists on b%d "
5591 + "but plink is disabled\n",
5592 + dst_inode->i_ino, inode->i_ino, cpg->bdst);
5596 + if (dst_inode->i_nlink) {
5597 + const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5599 + h_src = au_plink_lkup(inode, cpg->bdst);
5600 + err = PTR_ERR(h_src);
5601 + if (IS_ERR(h_src))
5603 + if (unlikely(d_is_negative(h_src))) {
5605 + AuIOErr("i%lu exists on b%d "
5606 + "but not pseudo-linked\n",
5607 + inode->i_ino, cpg->bdst);
5613 + a->h_path.dentry = h_parent;
5614 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5617 + a->h_path.dentry = h_dst;
5619 + err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5620 + if (!err && au_ftest_cpup(cpg->flags, RENAME))
5621 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5623 + au_dtime_revert(&a->dt);
5624 + if (unlikely(err == -EWOULDBLOCK)) {
5625 + pr_warn("cannot retry for NFSv4 delegation"
5626 + " for an internal link\n");
5632 + /* todo: cpup_wh_file? */
5634 + au_update_ibrange(inode, /*do_put_zero*/1);
5637 + isdir = S_ISDIR(inode->i_mode);
5638 + old_ibtop = au_ibtop(inode);
5639 + err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5640 + if (unlikely(err))
5642 + dst_inode = d_inode(h_dst);
5643 + inode_lock_nested(dst_inode, AuLsc_I_CHILD2);
5644 + /* todo: necessary? */
5645 + /* au_pin_hdir_unlock(cpg->pin); */
5647 + h_src_path.dentry = h_src;
5648 + h_src_path.mnt = au_sbr_mnt(sb, cpg->bsrc);
5649 + err = cpup_iattr(cpg->dentry, cpg->bdst, &h_src_path, &a->h_src_attr);
5650 + if (unlikely(err)) {
5651 + /* todo: necessary? */
5652 + /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5653 + inode_unlock(dst_inode);
5657 + if (cpg->bdst < old_ibtop) {
5658 + if (S_ISREG(inode->i_mode)) {
5659 + err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5660 + if (unlikely(err)) {
5661 + /* ignore an error */
5662 + /* au_pin_hdir_relock(cpg->pin); */
5663 + inode_unlock(dst_inode);
5667 + au_set_ibtop(inode, cpg->bdst);
5669 + au_set_ibbot(inode, cpg->bdst);
5670 + au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5671 + au_hi_flags(inode, isdir));
5673 + /* todo: necessary? */
5674 + /* err = au_pin_hdir_relock(cpg->pin); */
5675 + inode_unlock(dst_inode);
5676 + if (unlikely(err))
5679 + src_inode = d_inode(h_src);
5681 + && (src_inode->i_nlink > 1
5682 + || src_inode->i_state & I_LINKABLE)
5684 + au_plink_append(inode, cpg->bdst, h_dst);
5686 + if (au_ftest_cpup(cpg->flags, RENAME)) {
5687 + a->h_path.dentry = h_dst;
5688 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5691 + goto out_parent; /* success */
5695 + a->h_path.dentry = h_parent;
5696 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5697 + a->h_path.dentry = h_dst;
5699 + if (d_is_positive(h_dst)) {
5701 + /* no delegation since it is just created */
5702 + rerr = vfsub_unlink(h_dir, &a->h_path,
5703 + /*delegated*/NULL, /*force*/0);
5705 + rerr = vfsub_rmdir(h_dir, &a->h_path);
5707 + au_dtime_revert(&a->dt);
5709 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5719 +#if 0 /* reserved */
5720 +struct au_cpup_single_args {
5722 + struct au_cp_generic *cpg;
5723 + struct dentry *dst_parent;
5726 +static void au_call_cpup_single(void *args)
5728 + struct au_cpup_single_args *a = args;
5730 + au_pin_hdir_acquire_nest(a->cpg->pin);
5731 + *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5732 + au_pin_hdir_release(a->cpg->pin);
5737 + * prevent SIGXFSZ in copy-up.
5738 + * testing CAP_MKNOD is for generic fs,
5739 + * but CAP_FSETID is for xfs only, currently.
5741 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5744 + struct super_block *sb;
5745 + struct inode *h_dir;
5748 + sb = au_pinned_parent(pin)->d_sb;
5749 + if (!au_wkq_test()
5750 + && (!au_sbi(sb)->si_plink_maint_pid
5751 + || au_plink_maint(sb, AuLock_NOPLM))) {
5752 + switch (mode & S_IFMT) {
5754 + /* no condition about RLIMIT_FSIZE and the file size */
5759 + do_sio = !capable(CAP_MKNOD);
5763 + do_sio = ((mode & (S_ISUID | S_ISGID))
5764 + && !capable(CAP_FSETID));
5765 + /* this workaround may be removed in the future */
5767 + h_dir = au_pinned_h_dir(pin);
5768 + do_sio = h_dir->i_mode & S_ISVTX;
5775 +#if 0 /* reserved */
5776 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5779 + struct dentry *h_dentry;
5781 + h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5782 + if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5783 + err = au_cpup_single(cpg, dst_parent);
5785 + struct au_cpup_single_args args = {
5788 + .dst_parent = dst_parent
5790 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5791 + if (unlikely(wkq_err))
5800 + * copyup the @dentry from the first active lower branch to @bdst,
5801 + * using au_cpup_single().
5803 +static int au_cpup_simple(struct au_cp_generic *cpg)
5806 + unsigned int flags_orig;
5807 + struct dentry *dentry;
5809 + AuDebugOn(cpg->bsrc < 0);
5811 + dentry = cpg->dentry;
5812 + DiMustWriteLock(dentry);
5814 + err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5816 + flags_orig = cpg->flags;
5817 + au_fset_cpup(cpg->flags, RENAME);
5818 + err = au_cpup_single(cpg, NULL);
5819 + cpg->flags = flags_orig;
5821 + return 0; /* success */
5824 + au_set_h_dptr(dentry, cpg->bdst, NULL);
5825 + au_set_dbtop(dentry, cpg->bsrc);
5831 +struct au_cpup_simple_args {
5833 + struct au_cp_generic *cpg;
5836 +static void au_call_cpup_simple(void *args)
5838 + struct au_cpup_simple_args *a = args;
5840 + au_pin_hdir_acquire_nest(a->cpg->pin);
5841 + *a->errp = au_cpup_simple(a->cpg);
5842 + au_pin_hdir_release(a->cpg->pin);
5845 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5848 + struct dentry *dentry, *parent;
5849 + struct file *h_file;
5850 + struct inode *h_dir;
5851 + struct mnt_idmap *h_idmap;
5853 + dentry = cpg->dentry;
5855 + if (au_ftest_cpup(cpg->flags, HOPEN)) {
5856 + AuDebugOn(cpg->bsrc < 0);
5857 + h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5858 + err = PTR_ERR(h_file);
5859 + if (IS_ERR(h_file))
5863 + parent = dget_parent(dentry);
5864 + h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5865 + h_idmap = au_sbr_idmap(dentry->d_sb, cpg->bdst);
5866 + if (!au_test_h_perm_sio(h_idmap, h_dir, MAY_EXEC | MAY_WRITE)
5867 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5868 + err = au_cpup_simple(cpg);
5870 + struct au_cpup_simple_args args = {
5874 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5875 + if (unlikely(wkq_err))
5881 + au_h_open_post(dentry, cpg->bsrc, h_file);
5887 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5889 + aufs_bindex_t bsrc, bbot;
5890 + struct dentry *dentry, *h_dentry;
5892 + if (cpg->bsrc < 0) {
5893 + dentry = cpg->dentry;
5894 + bbot = au_dbbot(dentry);
5895 + for (bsrc = cpg->bdst + 1; bsrc <= bbot; bsrc++) {
5896 + h_dentry = au_h_dptr(dentry, bsrc);
5898 + AuDebugOn(d_is_negative(h_dentry));
5902 + AuDebugOn(bsrc > bbot);
5905 + AuDebugOn(cpg->bsrc <= cpg->bdst);
5906 + return au_do_sio_cpup_simple(cpg);
5909 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5911 + AuDebugOn(cpg->bdst <= cpg->bsrc);
5912 + return au_do_sio_cpup_simple(cpg);
5915 +/* ---------------------------------------------------------------------- */
5918 + * copyup the deleted file for writing.
5920 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5921 + struct file *file)
5924 + unsigned int flags_orig;
5925 + aufs_bindex_t bsrc_orig;
5926 + struct au_dinfo *dinfo;
5928 + struct au_hdentry *hd;
5929 + struct dentry *h_dentry;
5932 + dinfo = au_di(cpg->dentry);
5933 + AuRwMustWriteLock(&dinfo->di_rwsem);
5935 + bsrc_orig = cpg->bsrc;
5936 + cpg->bsrc = dinfo->di_btop;
5937 + hdst.hd = au_hdentry(dinfo, cpg->bdst);
5938 + hdst.h_dentry = hdst.hd->hd_dentry;
5939 + hdst.hd->hd_dentry = wh_dentry;
5940 + dinfo->di_btop = cpg->bdst;
5942 + hsrc.h_dentry = NULL;
5944 + hsrc.hd = au_hdentry(dinfo, cpg->bsrc);
5945 + hsrc.h_dentry = hsrc.hd->hd_dentry;
5946 + hsrc.hd->hd_dentry = au_hf_top(file)->f_path.dentry;
5948 + flags_orig = cpg->flags;
5949 + cpg->flags = !AuCpup_DTIME;
5950 + err = au_cpup_single(cpg, /*h_parent*/NULL);
5951 + cpg->flags = flags_orig;
5954 + err = au_reopen_nondir(file);
5955 + hsrc.hd->hd_dentry = hsrc.h_dentry;
5957 + hdst.hd->hd_dentry = hdst.h_dentry;
5958 + dinfo->di_btop = cpg->bsrc;
5959 + cpg->bsrc = bsrc_orig;
5964 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5967 + aufs_bindex_t bdst;
5968 + struct au_dtime dt;
5969 + struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5970 + struct au_branch *br;
5971 + struct path h_path;
5973 + dentry = cpg->dentry;
5975 + br = au_sbr(dentry->d_sb, bdst);
5976 + parent = dget_parent(dentry);
5977 + h_parent = au_h_dptr(parent, bdst);
5978 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5979 + err = PTR_ERR(wh_dentry);
5980 + if (IS_ERR(wh_dentry))
5983 + h_path.dentry = h_parent;
5984 + h_path.mnt = au_br_mnt(br);
5985 + au_dtime_store(&dt, parent, &h_path);
5986 + err = au_do_cpup_wh(cpg, wh_dentry, file);
5987 + if (unlikely(err))
5991 + h_path.dentry = wh_dentry;
5992 + if (!d_is_dir(wh_dentry)) {
5993 + /* no delegation since it is just created */
5994 + err = vfsub_unlink(d_inode(h_parent), &h_path,
5995 + /*delegated*/NULL, /*force*/0);
5997 + err = vfsub_rmdir(d_inode(h_parent), &h_path);
5998 + if (unlikely(err)) {
5999 + AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
6003 + au_dtime_revert(&dt);
6004 + au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
6013 +struct au_cpup_wh_args {
6015 + struct au_cp_generic *cpg;
6016 + struct file *file;
6019 +static void au_call_cpup_wh(void *args)
6021 + struct au_cpup_wh_args *a = args;
6023 + au_pin_hdir_acquire_nest(a->cpg->pin);
6024 + *a->errp = au_cpup_wh(a->cpg, a->file);
6025 + au_pin_hdir_release(a->cpg->pin);
6028 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6031 + aufs_bindex_t bdst;
6032 + struct dentry *dentry, *parent, *h_orph, *h_parent;
6033 + struct inode *dir, *h_dir, *h_tmpdir;
6034 + struct au_wbr *wbr;
6035 + struct au_pin wh_pin, *pin_orig;
6036 + struct mnt_idmap *h_idmap;
6038 + dentry = cpg->dentry;
6040 + parent = dget_parent(dentry);
6041 + dir = d_inode(parent);
6044 + h_dir = au_igrab(au_h_iptr(dir, bdst));
6047 + if (!h_dir->i_nlink) {
6048 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
6049 + h_orph = wbr->wbr_orph;
6051 + h_parent = dget(au_h_dptr(parent, bdst));
6052 + au_set_h_dptr(parent, bdst, dget(h_orph));
6053 + h_tmpdir = d_inode(h_orph);
6054 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
6056 + inode_lock_nested(h_tmpdir, AuLsc_I_PARENT3);
6057 + /* todo: au_h_open_pre()? */
6059 + pin_orig = cpg->pin;
6060 + au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
6061 + AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
6062 + cpg->pin = &wh_pin;
6065 + h_idmap = au_sbr_idmap(dentry->d_sb, bdst);
6066 + if (!au_test_h_perm_sio(h_idmap, h_tmpdir, MAY_EXEC | MAY_WRITE)
6067 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6068 + err = au_cpup_wh(cpg, file);
6070 + struct au_cpup_wh_args args = {
6075 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
6076 + if (unlikely(wkq_err))
6081 + inode_unlock(h_tmpdir);
6082 + /* todo: au_h_open_post()? */
6083 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
6084 + au_set_h_dptr(parent, bdst, h_parent);
6085 + AuDebugOn(!pin_orig);
6086 + cpg->pin = pin_orig;
6094 +/* ---------------------------------------------------------------------- */
6097 + * generic routine for both of copy-up and copy-down.
6099 +/* cf. revalidate function in file.c */
6100 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6101 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6102 + struct au_pin *pin,
6103 + struct dentry *h_parent, void *arg),
6107 + struct au_pin pin;
6108 + struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
6111 + parent = dget_parent(dentry);
6112 + if (IS_ROOT(parent))
6115 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
6116 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
6118 + /* do not use au_dpage */
6119 + real_parent = parent;
6122 + parent = dget_parent(dentry);
6123 + h_parent = au_h_dptr(parent, bdst);
6125 + goto out; /* success */
6127 + /* find top dir which is necessary to cpup */
6131 + parent = dget_parent(d);
6132 + di_read_lock_parent3(parent, !AuLock_IR);
6133 + h_parent = au_h_dptr(parent, bdst);
6134 + di_read_unlock(parent, !AuLock_IR);
6135 + } while (!h_parent);
6137 + if (d != real_parent)
6138 + di_write_lock_child3(d);
6140 + /* somebody else might create while we were sleeping */
6141 + h_dentry = au_h_dptr(d, bdst);
6142 + if (!h_dentry || d_is_negative(h_dentry)) {
6144 + au_update_dbtop(d);
6146 + au_pin_set_dentry(&pin, d);
6147 + err = au_do_pin(&pin);
6149 + err = cp(d, bdst, &pin, h_parent, arg);
6154 + if (d != real_parent)
6155 + di_write_unlock(d);
6156 + if (unlikely(err))
6165 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
6166 + struct au_pin *pin,
6167 + struct dentry *h_parent __maybe_unused,
6168 + void *arg __maybe_unused)
6170 + struct au_cp_generic cpg = {
6176 + .flags = AuCpup_DTIME
6178 + return au_sio_cpup_simple(&cpg);
6181 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6183 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
6186 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6189 + struct dentry *parent;
6190 + struct inode *dir;
6192 + parent = dget_parent(dentry);
6193 + dir = d_inode(parent);
6195 + if (au_h_iptr(dir, bdst))
6198 + di_read_unlock(parent, AuLock_IR);
6199 + di_write_lock_parent(parent);
6200 + /* someone else might change our inode while we were sleeping */
6201 + if (!au_h_iptr(dir, bdst))
6202 + err = au_cpup_dirs(dentry, bdst);
6203 + di_downgrade_lock(parent, AuLock_IR);
6209 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
6210 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
6211 +++ linux/fs/aufs/cpup.h 2022-11-05 23:02:18.962555950 +0100
6213 +/* SPDX-License-Identifier: GPL-2.0 */
6215 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6217 + * This program is free software; you can redistribute it and/or modify
6218 + * it under the terms of the GNU General Public License as published by
6219 + * the Free Software Foundation; either version 2 of the License, or
6220 + * (at your option) any later version.
6222 + * This program is distributed in the hope that it will be useful,
6223 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6224 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6225 + * GNU General Public License for more details.
6227 + * You should have received a copy of the GNU General Public License
6228 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6232 + * copy-up/down functions
6235 +#ifndef __AUFS_CPUP_H__
6236 +#define __AUFS_CPUP_H__
6240 +#include <linux/path.h>
6246 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
6247 +void au_cpup_attr_timesizes(struct inode *inode);
6248 +void au_cpup_attr_nlink(struct inode *inode, int force);
6249 +void au_cpup_attr_changeable(struct inode *inode);
6250 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
6251 +void au_cpup_attr_all(struct inode *inode, int force);
6253 +/* ---------------------------------------------------------------------- */
6255 +struct au_cp_generic {
6256 + struct dentry *dentry;
6257 + aufs_bindex_t bdst, bsrc;
6259 + struct au_pin *pin;
6260 + unsigned int flags;
6264 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
6265 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
6267 +#define AuCpup_RENAME (1 << 2) /* rename after cpup */
6268 +#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
6270 +#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
6272 +#define AuCpup_RWDST (1 << 5) /* force write target even if
6273 + the branch is marked as RO */
6275 +#ifndef CONFIG_AUFS_BR_HFSPLUS
6276 +#undef AuCpup_HOPEN
6277 +#define AuCpup_HOPEN 0
6280 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
6281 +#define au_fset_cpup(flags, name) \
6282 + do { (flags) |= AuCpup_##name; } while (0)
6283 +#define au_fclr_cpup(flags, name) \
6284 + do { (flags) &= ~AuCpup_##name; } while (0)
6286 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
6287 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
6288 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
6289 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
6291 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6292 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6293 + struct au_pin *pin,
6294 + struct dentry *h_parent, void *arg),
6296 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6297 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6299 +/* ---------------------------------------------------------------------- */
6301 +/* keep timestamps when copyup */
6303 + struct dentry *dt_dentry;
6304 + struct path dt_h_path;
6305 + struct timespec64 dt_atime, dt_mtime;
6307 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
6308 + struct path *h_path);
6309 +void au_dtime_revert(struct au_dtime *dt);
6311 +#endif /* __KERNEL__ */
6312 +#endif /* __AUFS_CPUP_H__ */
6313 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
6314 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
6315 +++ linux/fs/aufs/dbgaufs.c 2022-11-05 23:02:18.962555950 +0100
6317 +// SPDX-License-Identifier: GPL-2.0
6319 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6321 + * This program is free software; you can redistribute it and/or modify
6322 + * it under the terms of the GNU General Public License as published by
6323 + * the Free Software Foundation; either version 2 of the License, or
6324 + * (at your option) any later version.
6326 + * This program is distributed in the hope that it will be useful,
6327 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6328 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6329 + * GNU General Public License for more details.
6331 + * You should have received a copy of the GNU General Public License
6332 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6336 + * debugfs interface
6339 +#include <linux/debugfs.h>
6342 +#ifndef CONFIG_SYSFS
6343 +#error DEBUG_FS depends upon SYSFS
6346 +static struct dentry *dbgaufs;
6347 +static const mode_t dbgaufs_mode = 0444;
6349 +/* 20 is max digits length of ulong 64 */
6350 +struct dbgaufs_arg {
6356 + * common function for all XINO files
6358 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
6359 + struct file *file)
6363 + p = file->private_data;
6365 + /* this is struct dbgaufs_arg */
6366 + AuDebugOn(!au_kfree_sz_test(p));
6367 + au_kfree_do_rcu(p);
6372 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt,
6377 + struct dbgaufs_arg *p;
6380 + p = kmalloc(sizeof(*p), GFP_NOFS);
6386 + file->private_data = p;
6390 + err = vfsub_getattr(&xf->f_path, &st);
6394 + (p->a, sizeof(p->a), "%d, %llux%u %lld\n",
6395 + cnt, st.blocks, st.blksize,
6396 + (long long)st.size);
6398 + p->n = snprintf(p->a, sizeof(p->a), "%llux%u %lld\n",
6399 + st.blocks, st.blksize,
6400 + (long long)st.size);
6401 + AuDebugOn(p->n >= sizeof(p->a));
6403 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
6411 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
6412 + size_t count, loff_t *ppos)
6414 + struct dbgaufs_arg *p;
6416 + p = file->private_data;
6417 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6420 +/* ---------------------------------------------------------------------- */
6422 +struct dbgaufs_plink_arg {
6427 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6428 + struct file *file)
6430 + free_page((unsigned long)file->private_data);
6434 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6436 + int err, i, limit;
6437 + unsigned long n, sum;
6438 + struct dbgaufs_plink_arg *p;
6439 + struct au_sbinfo *sbinfo;
6440 + struct super_block *sb;
6441 + struct hlist_bl_head *hbl;
6444 + p = (void *)get_zeroed_page(GFP_NOFS);
6449 + sbinfo = inode->i_private;
6450 + sb = sbinfo->si_sb;
6451 + si_noflush_read_lock(sb);
6452 + if (au_opt_test(au_mntflags(sb), PLINK)) {
6453 + limit = PAGE_SIZE - sizeof(p->n);
6455 + /* the number of buckets */
6456 + n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6461 + for (i = 0, hbl = sbinfo->si_plink; i < AuPlink_NHASH;
6463 + n = au_hbl_count(hbl);
6466 + n = snprintf(p->a + p->n, limit, "%lu ", n);
6469 + if (unlikely(limit <= 0))
6472 + p->a[p->n - 1] = '\n';
6474 + /* the sum of plinks */
6475 + n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6478 + if (unlikely(limit <= 0))
6481 +#define str "1\n0\n0\n"
6482 + p->n = sizeof(str) - 1;
6483 + strcpy(p->a, str);
6486 + si_read_unlock(sb);
6489 + file->private_data = p;
6490 + goto out; /* success */
6493 + free_page((unsigned long)p);
6498 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6499 + size_t count, loff_t *ppos)
6501 + struct dbgaufs_plink_arg *p;
6503 + p = file->private_data;
6504 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6507 +static const struct file_operations dbgaufs_plink_fop = {
6508 + .owner = THIS_MODULE,
6509 + .open = dbgaufs_plink_open,
6510 + .release = dbgaufs_plink_release,
6511 + .read = dbgaufs_plink_read
6514 +/* ---------------------------------------------------------------------- */
6516 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6519 + struct au_sbinfo *sbinfo;
6520 + struct super_block *sb;
6522 + sbinfo = inode->i_private;
6523 + sb = sbinfo->si_sb;
6524 + si_noflush_read_lock(sb);
6525 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0, /*cnt*/0);
6526 + si_read_unlock(sb);
6530 +static const struct file_operations dbgaufs_xib_fop = {
6531 + .owner = THIS_MODULE,
6532 + .open = dbgaufs_xib_open,
6533 + .release = dbgaufs_xi_release,
6534 + .read = dbgaufs_xi_read
6537 +/* ---------------------------------------------------------------------- */
6539 +#define DbgaufsXi_PREFIX "xi"
6541 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6545 + aufs_bindex_t bindex;
6546 + char *p, a[sizeof(DbgaufsXi_PREFIX) + 8];
6547 + struct au_sbinfo *sbinfo;
6548 + struct super_block *sb;
6549 + struct au_xino *xi;
6551 + struct qstr *name;
6552 + struct au_branch *br;
6555 + name = &file->f_path.dentry->d_name;
6556 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6557 + || memcmp(name->name, DbgaufsXi_PREFIX,
6558 + sizeof(DbgaufsXi_PREFIX) - 1)))
6561 + AuDebugOn(name->len >= sizeof(a));
6562 + memcpy(a, name->name, name->len);
6563 + a[name->len] = '\0';
6564 + p = strchr(a, '-');
6567 + err = kstrtol(a + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6568 + if (unlikely(err))
6573 + err = kstrtol(p + 1, 10, &l);
6574 + if (unlikely(err))
6580 + sbinfo = inode->i_private;
6581 + sb = sbinfo->si_sb;
6582 + si_noflush_read_lock(sb);
6583 + if (unlikely(bindex < 0 || bindex > au_sbbot(sb)))
6585 + br = au_sbr(sb, bindex);
6587 + if (unlikely(idx >= xi->xi_nfile))
6589 + xf = au_xino_file(xi, idx);
6591 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1,
6592 + au_xino_count(br));
6595 + si_read_unlock(sb);
6601 +static const struct file_operations dbgaufs_xino_fop = {
6602 + .owner = THIS_MODULE,
6603 + .open = dbgaufs_xino_open,
6604 + .release = dbgaufs_xi_release,
6605 + .read = dbgaufs_xi_read
6608 +void dbgaufs_xino_del(struct au_branch *br)
6610 + struct dentry *dbgaufs;
6612 + dbgaufs = br->br_dbgaufs;
6616 + br->br_dbgaufs = NULL;
6617 + /* debugfs acquires the parent i_mutex */
6619 + debugfs_remove(dbgaufs);
6623 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6625 + aufs_bindex_t bbot;
6626 + struct au_branch *br;
6628 + if (!au_sbi(sb)->si_dbgaufs)
6631 + bbot = au_sbbot(sb);
6632 + for (; bindex <= bbot; bindex++) {
6633 + br = au_sbr(sb, bindex);
6634 + dbgaufs_xino_del(br);
6638 +static void dbgaufs_br_do_add(struct super_block *sb, aufs_bindex_t bindex,
6639 + unsigned int idx, struct dentry *parent,
6640 + struct au_sbinfo *sbinfo)
6642 + struct au_branch *br;
6644 + /* "xi" bindex(5) "-" idx(2) NULL */
6645 + char name[sizeof(DbgaufsXi_PREFIX) + 8];
6648 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6650 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d-%u",
6652 + br = au_sbr(sb, bindex);
6653 + if (br->br_dbgaufs) {
6654 + struct qstr qstr = QSTR_INIT(name, strlen(name));
6656 + if (!au_qstreq(&br->br_dbgaufs->d_name, &qstr)) {
6657 + /* debugfs acquires the parent i_mutex */
6659 + d = debugfs_rename(parent, br->br_dbgaufs, parent,
6663 + pr_warn("failed renaming %pd/%s, ignored.\n",
6668 + br->br_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6669 + sbinfo, &dbgaufs_xino_fop);
6671 + if (unlikely(!br->br_dbgaufs))
6672 + pr_warn("failed creating %pd/%s, ignored.\n",
6677 +static void dbgaufs_br_add(struct super_block *sb, aufs_bindex_t bindex,
6678 + struct dentry *parent, struct au_sbinfo *sbinfo)
6680 + struct au_branch *br;
6681 + struct au_xino *xi;
6684 + br = au_sbr(sb, bindex);
6686 + for (u = 0; u < xi->xi_nfile; u++)
6687 + dbgaufs_br_do_add(sb, bindex, u, parent, sbinfo);
6690 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown)
6692 + struct au_sbinfo *sbinfo;
6693 + struct dentry *parent;
6694 + aufs_bindex_t bbot;
6696 + if (!au_opt_test(au_mntflags(sb), XINO))
6699 + sbinfo = au_sbi(sb);
6700 + parent = sbinfo->si_dbgaufs;
6704 + bbot = au_sbbot(sb);
6706 + for (; bindex <= bbot; bindex++)
6707 + dbgaufs_br_add(sb, bindex, parent, sbinfo);
6709 + for (; bbot >= bindex; bbot--)
6710 + dbgaufs_br_add(sb, bbot, parent, sbinfo);
6713 +/* ---------------------------------------------------------------------- */
6715 +#ifdef CONFIG_AUFS_EXPORT
6716 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6719 + struct au_sbinfo *sbinfo;
6720 + struct super_block *sb;
6722 + sbinfo = inode->i_private;
6723 + sb = sbinfo->si_sb;
6724 + si_noflush_read_lock(sb);
6725 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0, /*cnt*/0);
6726 + si_read_unlock(sb);
6730 +static const struct file_operations dbgaufs_xigen_fop = {
6731 + .owner = THIS_MODULE,
6732 + .open = dbgaufs_xigen_open,
6733 + .release = dbgaufs_xi_release,
6734 + .read = dbgaufs_xi_read
6737 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6742 + * This function is a dynamic '__init' function actually,
6743 + * so the tiny check for si_rwsem is unnecessary.
6745 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6748 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
6749 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6750 + &dbgaufs_xigen_fop);
6751 + if (sbinfo->si_dbgaufs_xigen)
6757 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6761 +#endif /* CONFIG_AUFS_EXPORT */
6763 +/* ---------------------------------------------------------------------- */
6765 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6768 + * This function is a dynamic '__fin' function actually,
6769 + * so the tiny check for si_rwsem is unnecessary.
6771 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6773 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
6774 + sbinfo->si_dbgaufs = NULL;
6777 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6780 + char name[SysaufsSiNameLen];
6783 + * This function is a dynamic '__init' function actually,
6784 + * so the tiny check for si_rwsem is unnecessary.
6786 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6790 + AuErr1("/debug/aufs is uninitialized\n");
6795 + sysaufs_name(sbinfo, name);
6796 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6797 + if (unlikely(!sbinfo->si_dbgaufs))
6800 + /* regardless plink/noplink option */
6801 + sbinfo->si_dbgaufs_plink = debugfs_create_file
6802 + ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6803 + &dbgaufs_plink_fop);
6804 + if (unlikely(!sbinfo->si_dbgaufs_plink))
6807 + /* regardless xino/noxino option */
6808 + sbinfo->si_dbgaufs_xib = debugfs_create_file
6809 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6810 + &dbgaufs_xib_fop);
6811 + if (unlikely(!sbinfo->si_dbgaufs_xib))
6814 + err = dbgaufs_xigen_init(sbinfo);
6816 + goto out; /* success */
6819 + dbgaufs_si_fin(sbinfo);
6821 + if (unlikely(err))
6822 + pr_err("debugfs/aufs failed\n");
6826 +/* ---------------------------------------------------------------------- */
6828 +void dbgaufs_fin(void)
6830 + debugfs_remove(dbgaufs);
6833 +int __init dbgaufs_init(void)
6838 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6843 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6844 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
6845 +++ linux/fs/aufs/dbgaufs.h 2022-11-05 23:02:18.962555950 +0100
6847 +/* SPDX-License-Identifier: GPL-2.0 */
6849 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6851 + * This program is free software; you can redistribute it and/or modify
6852 + * it under the terms of the GNU General Public License as published by
6853 + * the Free Software Foundation; either version 2 of the License, or
6854 + * (at your option) any later version.
6856 + * This program is distributed in the hope that it will be useful,
6857 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6858 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6859 + * GNU General Public License for more details.
6861 + * You should have received a copy of the GNU General Public License
6862 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6866 + * debugfs interface
6869 +#ifndef __DBGAUFS_H__
6870 +#define __DBGAUFS_H__
6874 +struct super_block;
6878 +#ifdef CONFIG_DEBUG_FS
6880 +void dbgaufs_xino_del(struct au_branch *br);
6881 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6882 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex, int topdown);
6883 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6884 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6885 +void dbgaufs_fin(void);
6886 +int __init dbgaufs_init(void);
6888 +AuStubVoid(dbgaufs_xino_del, struct au_branch *br)
6889 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6890 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex,
6892 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6893 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6894 +AuStubVoid(dbgaufs_fin, void)
6895 +AuStubInt0(__init dbgaufs_init, void)
6896 +#endif /* CONFIG_DEBUG_FS */
6898 +#endif /* __KERNEL__ */
6899 +#endif /* __DBGAUFS_H__ */
6900 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6901 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
6902 +++ linux/fs/aufs/dcsub.c 2022-11-05 23:02:18.962555950 +0100
6904 +// SPDX-License-Identifier: GPL-2.0
6906 + * Copyright (C) 2005-2022 Junjiro R. Okajima
6908 + * This program is free software; you can redistribute it and/or modify
6909 + * it under the terms of the GNU General Public License as published by
6910 + * the Free Software Foundation; either version 2 of the License, or
6911 + * (at your option) any later version.
6913 + * This program is distributed in the hope that it will be useful,
6914 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6915 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6916 + * GNU General Public License for more details.
6918 + * You should have received a copy of the GNU General Public License
6919 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6923 + * sub-routines for dentry cache
6928 +static void au_dpage_free(struct au_dpage *dpage)
6931 + struct dentry **p;
6933 + p = dpage->dentries;
6934 + for (i = 0; i < dpage->ndentry; i++)
6936 + free_page((unsigned long)dpage->dentries);
6939 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6945 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6946 + if (unlikely(!dpages->dpages))
6949 + p = (void *)__get_free_page(gfp);
6953 + dpages->dpages[0].ndentry = 0;
6954 + dpages->dpages[0].dentries = p;
6955 + dpages->ndpage = 1;
6956 + return 0; /* success */
6959 + au_kfree_try_rcu(dpages->dpages);
6964 +void au_dpages_free(struct au_dcsub_pages *dpages)
6967 + struct au_dpage *p;
6969 + p = dpages->dpages;
6970 + for (i = 0; i < dpages->ndpage; i++)
6971 + au_dpage_free(p++);
6972 + au_kfree_try_rcu(dpages->dpages);
6975 +static int au_dpages_append(struct au_dcsub_pages *dpages,
6976 + struct dentry *dentry, gfp_t gfp)
6979 + struct au_dpage *dpage;
6982 + dpage = dpages->dpages + dpages->ndpage - 1;
6983 + sz = PAGE_SIZE / sizeof(dentry);
6984 + if (unlikely(dpage->ndentry >= sz)) {
6985 + AuLabel(new dpage);
6987 + sz = dpages->ndpage * sizeof(*dpages->dpages);
6988 + p = au_kzrealloc(dpages->dpages, sz,
6989 + sz + sizeof(*dpages->dpages), gfp,
6994 + dpages->dpages = p;
6995 + dpage = dpages->dpages + dpages->ndpage;
6996 + p = (void *)__get_free_page(gfp);
7000 + dpage->ndentry = 0;
7001 + dpage->dentries = p;
7005 + AuDebugOn(au_dcount(dentry) <= 0);
7006 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
7007 + return 0; /* success */
7013 +/* todo: BAD approach */
7014 +/* copied from linux/fs/dcache.c */
7022 +extern void d_walk(struct dentry *parent, void *data,
7023 + enum d_walk_ret (*enter)(void *, struct dentry *));
7025 +struct ac_dpages_arg {
7027 + struct au_dcsub_pages *dpages;
7028 + struct super_block *sb;
7029 + au_dpages_test test;
7033 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
7035 + enum d_walk_ret ret;
7036 + struct ac_dpages_arg *arg = _arg;
7038 + ret = D_WALK_CONTINUE;
7039 + if (dentry->d_sb == arg->sb
7040 + && !IS_ROOT(dentry)
7041 + && au_dcount(dentry) > 0
7043 + && (!arg->test || arg->test(dentry, arg->arg))) {
7044 + arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
7045 + if (unlikely(arg->err))
7046 + ret = D_WALK_QUIT;
7052 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7053 + au_dpages_test test, void *arg)
7055 + struct ac_dpages_arg args = {
7063 + d_walk(root, &args, au_call_dpages_append);
7068 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7069 + int do_include, au_dpages_test test, void *arg)
7074 + write_seqlock(&rename_lock);
7075 + spin_lock(&dentry->d_lock);
7077 + && au_dcount(dentry) > 0
7078 + && (!test || test(dentry, arg)))
7079 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7080 + spin_unlock(&dentry->d_lock);
7081 + if (unlikely(err))
7085 + * RCU for vfsmount is unnecessary since this is a traverse in a single
7088 + while (!IS_ROOT(dentry)) {
7089 + dentry = dentry->d_parent; /* rename_lock is locked */
7090 + spin_lock(&dentry->d_lock);
7091 + if (au_dcount(dentry) > 0
7092 + && (!test || test(dentry, arg)))
7093 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7094 + spin_unlock(&dentry->d_lock);
7095 + if (unlikely(err))
7100 + write_sequnlock(&rename_lock);
7104 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
7106 + return au_di(dentry) && dentry->d_sb == arg;
7109 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7110 + struct dentry *dentry, int do_include)
7112 + return au_dcsub_pages_rev(dpages, dentry, do_include,
7113 + au_dcsub_dpages_aufs, dentry->d_sb);
7116 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
7118 + struct path path[2] = {
7127 + return path_is_under(path + 0, path + 1);
7129 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
7130 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
7131 +++ linux/fs/aufs/dcsub.h 2022-11-05 23:02:18.962555950 +0100
7133 +/* SPDX-License-Identifier: GPL-2.0 */
7135 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7137 + * This program is free software; you can redistribute it and/or modify
7138 + * it under the terms of the GNU General Public License as published by
7139 + * the Free Software Foundation; either version 2 of the License, or
7140 + * (at your option) any later version.
7142 + * This program is distributed in the hope that it will be useful,
7143 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7144 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7145 + * GNU General Public License for more details.
7147 + * You should have received a copy of the GNU General Public License
7148 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7152 + * sub-routines for dentry cache
7155 +#ifndef __AUFS_DCSUB_H__
7156 +#define __AUFS_DCSUB_H__
7160 +#include <linux/dcache.h>
7161 +#include <linux/fs.h>
7165 + struct dentry **dentries;
7168 +struct au_dcsub_pages {
7170 + struct au_dpage *dpages;
7173 +/* ---------------------------------------------------------------------- */
7176 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
7177 +void au_dpages_free(struct au_dcsub_pages *dpages);
7178 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
7179 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7180 + au_dpages_test test, void *arg);
7181 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7182 + int do_include, au_dpages_test test, void *arg);
7183 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7184 + struct dentry *dentry, int do_include);
7185 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
7187 +/* ---------------------------------------------------------------------- */
7190 + * todo: in linux-3.13, several similar (but faster) helpers are added to
7191 + * include/linux/dcache.h. Try them (in the future).
7194 +static inline int au_d_hashed_positive(struct dentry *d)
7197 + struct inode *inode = d_inode(d);
7200 + if (unlikely(d_unhashed(d)
7201 + || d_is_negative(d)
7202 + || !inode->i_nlink))
7207 +static inline int au_d_linkable(struct dentry *d)
7210 + struct inode *inode = d_inode(d);
7212 + err = au_d_hashed_positive(d);
7214 + && d_is_positive(d)
7215 + && (inode->i_state & I_LINKABLE))
7220 +static inline int au_d_alive(struct dentry *d)
7223 + struct inode *inode;
7227 + err = au_d_hashed_positive(d);
7229 + inode = d_inode(d);
7230 + if (unlikely(d_unlinked(d)
7231 + || d_is_negative(d)
7232 + || !inode->i_nlink))
7238 +static inline int au_alive_dir(struct dentry *d)
7242 + err = au_d_alive(d);
7243 + if (unlikely(err || IS_DEADDIR(d_inode(d))))
7248 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
7250 + return a->len == b->len
7251 + && !memcmp(a->name, b->name, a->len);
7256 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
7258 + * the type of d_lockref.count became int, but the inlined function d_count()
7259 + * still returns unsigned int.
7260 + * I don't know why. Maybe it is for every d_count() users?
7261 + * Anyway au_dcount() lives on.
7263 +static inline int au_dcount(struct dentry *d)
7265 + return (int)d_count(d);
7268 +#endif /* __KERNEL__ */
7269 +#endif /* __AUFS_DCSUB_H__ */
7270 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
7271 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
7272 +++ linux/fs/aufs/debug.c 2023-02-20 21:05:51.959693785 +0100
7274 +// SPDX-License-Identifier: GPL-2.0
7276 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7278 + * This program is free software; you can redistribute it and/or modify
7279 + * it under the terms of the GNU General Public License as published by
7280 + * the Free Software Foundation; either version 2 of the License, or
7281 + * (at your option) any later version.
7283 + * This program is distributed in the hope that it will be useful,
7284 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7285 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7286 + * GNU General Public License for more details.
7288 + * You should have received a copy of the GNU General Public License
7289 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7293 + * debug print functions
7296 +#include <linux/iversion.h>
7299 +/* Returns 0, or -errno. arg is in kp->arg. */
7300 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
7304 + err = kstrtoint(val, 0, &n);
7314 +/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
7315 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
7320 + return sprintf(buffer, "%d", atomic_read(a));
7323 +static const struct kernel_param_ops param_ops_atomic_t = {
7324 + .set = param_atomic_t_set,
7325 + .get = param_atomic_t_get
7326 + /* void (*free)(void *arg) */
7329 +atomic_t aufs_debug = ATOMIC_INIT(0);
7330 +MODULE_PARM_DESC(debug, "debug print");
7331 +module_param_named(debug, aufs_debug, atomic_t, 0664);
7333 +DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
7334 +char *au_plevel = KERN_DEBUG;
7335 +#define dpri(fmt, ...) do { \
7337 + && strcmp(au_plevel, KERN_DEBUG)) \
7338 + || au_debug_test()) \
7339 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
7342 +/* ---------------------------------------------------------------------- */
7344 +void au_dpri_whlist(struct au_nhash *whlist)
7346 + unsigned long ul, n;
7347 + struct hlist_head *head;
7348 + struct au_vdir_wh *pos;
7350 + n = whlist->nh_num;
7351 + head = whlist->nh_head;
7352 + for (ul = 0; ul < n; ul++) {
7353 + hlist_for_each_entry(pos, head, wh_hash)
7354 + dpri("b%d, %.*s, %d\n",
7356 + pos->wh_str.len, pos->wh_str.name,
7362 +void au_dpri_vdir(struct au_vdir *vdir)
7365 + union au_vdir_deblk_p p;
7368 + if (!vdir || IS_ERR(vdir)) {
7369 + dpri("err %ld\n", PTR_ERR(vdir));
7373 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %llu\n",
7374 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
7375 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
7376 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
7377 + p.deblk = vdir->vd_deblk[ul];
7379 + dpri("[%lu]: %p\n", ul, o);
7383 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
7384 + struct dentry *wh)
7389 + if (!inode || IS_ERR(inode)) {
7390 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
7394 + /* the type of i_blocks depends upon CONFIG_LBDAF */
7395 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
7396 + && sizeof(inode->i_blocks) != sizeof(u64));
7398 + n = (void *)wh->d_name.name;
7399 + l = wh->d_name.len;
7402 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
7403 + " acl %p, def_acl %p,"
7404 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
7406 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
7407 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
7408 + i_size_read(inode), (unsigned long long)inode->i_blocks,
7409 + inode->i_acl, inode->i_default_acl,
7410 + hn, (long long)timespec64_to_ns(&inode->i_ctime) & 0x0ffff,
7411 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
7412 + inode->i_state, inode->i_flags, inode_peek_iversion(inode),
7413 + inode->i_generation,
7414 + l ? ", wh " : "", l, n);
7418 +void au_dpri_inode(struct inode *inode)
7420 + struct au_iinfo *iinfo;
7421 + struct au_hinode *hi;
7422 + aufs_bindex_t bindex;
7425 + err = do_pri_inode(-1, inode, -1, NULL);
7426 + if (err || !au_test_aufs(inode->i_sb) || au_is_bad_inode(inode))
7429 + iinfo = au_ii(inode);
7430 + dpri("i-1: btop %d, bbot %d, gen %d\n",
7431 + iinfo->ii_btop, iinfo->ii_bbot, au_iigen(inode, NULL));
7432 + if (iinfo->ii_btop < 0)
7435 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot; bindex++) {
7436 + hi = au_hinode(iinfo, bindex);
7438 + do_pri_inode(bindex, hi->hi_inode, hn, hi->hi_whdentry);
7442 +void au_dpri_dalias(struct inode *inode)
7446 + spin_lock(&inode->i_lock);
7447 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
7448 + au_dpri_dentry(d);
7449 + spin_unlock(&inode->i_lock);
7452 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
7454 + struct dentry *wh = NULL;
7456 + struct inode *inode;
7457 + struct au_iinfo *iinfo;
7458 + struct au_hinode *hi;
7460 + if (!dentry || IS_ERR(dentry)) {
7461 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
7464 + /* do not call dget_parent() here */
7465 + /* note: access d_xxx without d_lock */
7466 + dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
7467 + bindex, dentry, dentry,
7468 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
7469 + au_dcount(dentry), dentry->d_flags,
7470 + d_unhashed(dentry) ? "un" : "");
7473 + if (d_is_positive(dentry))
7474 + inode = d_inode(dentry);
7476 + && au_test_aufs(dentry->d_sb)
7478 + && !au_is_bad_inode(inode)) {
7479 + iinfo = au_ii(inode);
7480 + hi = au_hinode(iinfo, bindex);
7482 + wh = hi->hi_whdentry;
7484 + do_pri_inode(bindex, inode, hn, wh);
7488 +void au_dpri_dentry(struct dentry *dentry)
7490 + struct au_dinfo *dinfo;
7491 + aufs_bindex_t bindex;
7494 + err = do_pri_dentry(-1, dentry);
7495 + if (err || !au_test_aufs(dentry->d_sb))
7498 + dinfo = au_di(dentry);
7501 + dpri("d-1: btop %d, bbot %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
7502 + dinfo->di_btop, dinfo->di_bbot,
7503 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
7504 + dinfo->di_tmpfile);
7505 + if (dinfo->di_btop < 0)
7507 + for (bindex = dinfo->di_btop; bindex <= dinfo->di_bbot; bindex++)
7508 + do_pri_dentry(bindex, au_hdentry(dinfo, bindex)->hd_dentry);
7511 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7515 + if (!file || IS_ERR(file)) {
7516 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7521 + && !IS_ERR_OR_NULL(file->f_path.dentry)
7522 + && au_test_aufs(file->f_path.dentry->d_sb)
7524 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7525 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7526 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7527 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
7528 + file->f_version, file->f_pos, a);
7529 + if (!IS_ERR_OR_NULL(file->f_path.dentry))
7530 + do_pri_dentry(bindex, file->f_path.dentry);
7534 +void au_dpri_file(struct file *file)
7536 + struct au_finfo *finfo;
7537 + struct au_fidir *fidir;
7538 + struct au_hfile *hfile;
7539 + aufs_bindex_t bindex;
7542 + err = do_pri_file(-1, file);
7544 + || IS_ERR_OR_NULL(file->f_path.dentry)
7545 + || !au_test_aufs(file->f_path.dentry->d_sb))
7548 + finfo = au_fi(file);
7551 + if (finfo->fi_btop < 0)
7553 + fidir = finfo->fi_hdir;
7555 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7557 + for (bindex = finfo->fi_btop;
7558 + bindex >= 0 && bindex <= fidir->fd_bbot;
7560 + hfile = fidir->fd_hfile + bindex;
7561 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7565 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7567 + struct vfsmount *mnt;
7568 + struct super_block *sb;
7570 + if (!br || IS_ERR(br))
7572 + mnt = au_br_mnt(br);
7573 + if (!mnt || IS_ERR(mnt))
7576 + if (!sb || IS_ERR(sb))
7579 + dpri("s%d: {perm 0x%x, id %d, wbr %p}, "
7580 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7582 + bindex, br->br_perm, br->br_id, br->br_wbr,
7583 + au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7584 + sb->s_flags, sb->s_count,
7585 + atomic_read(&sb->s_active),
7586 + !!au_xino_file(br->br_xino, /*idx*/-1));
7590 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7594 +void au_dpri_sb(struct super_block *sb)
7596 + struct au_sbinfo *sbinfo;
7597 + aufs_bindex_t bindex;
7599 + /* to reduce stack size */
7601 + struct vfsmount mnt;
7602 + struct au_branch fake;
7605 + /* this function can be called from magic sysrq */
7606 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
7607 + if (unlikely(!a)) {
7608 + dpri("no memory\n");
7612 + a->mnt.mnt_sb = sb;
7613 + a->fake.br_path.mnt = &a->mnt;
7614 + err = do_pri_br(-1, &a->fake);
7616 + dpri("dev 0x%x\n", sb->s_dev);
7617 + if (err || !au_test_aufs(sb))
7620 + sbinfo = au_sbi(sb);
7623 + dpri("nw %d, gen %u, kobj %d\n",
7624 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7625 + kref_read(&sbinfo->si_kobj.kref));
7626 + for (bindex = 0; bindex <= sbinfo->si_bbot; bindex++)
7627 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7630 +/* ---------------------------------------------------------------------- */
7632 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7634 + struct inode *h_inode, *inode = d_inode(dentry);
7635 + struct dentry *h_dentry;
7636 + aufs_bindex_t bindex, bbot, bi;
7638 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7641 + bbot = au_dbbot(dentry);
7642 + bi = au_ibbot(inode);
7645 + bindex = au_dbtop(dentry);
7646 + bi = au_ibtop(inode);
7650 + for (; bindex <= bbot; bindex++) {
7651 + h_dentry = au_h_dptr(dentry, bindex);
7654 + h_inode = au_h_iptr(inode, bindex);
7655 + if (unlikely(h_inode != d_inode(h_dentry))) {
7657 + AuDbg("b%d, %s:%d\n", bindex, func, line);
7658 + AuDbgDentry(dentry);
7659 + AuDbgInode(inode);
7661 + if (au_test_fuse(h_inode->i_sb))
7669 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7672 + struct au_dcsub_pages dpages;
7673 + struct au_dpage *dpage;
7674 + struct dentry **dentries;
7676 + err = au_dpages_init(&dpages, GFP_NOFS);
7678 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7680 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7681 + dpage = dpages.dpages + i;
7682 + dentries = dpage->dentries;
7683 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7684 + AuDebugOn(au_digen_test(dentries[j], sigen));
7686 + au_dpages_free(&dpages);
7689 +void au_dbg_verify_kthread(void)
7691 + if (au_wkq_test()) {
7694 + * It may be recursive, but udba=notify between two aufs mounts,
7695 + * where a single ro branch is shared, is not a problem.
7701 +/* ---------------------------------------------------------------------- */
7703 +int __init au_debug_init(void)
7705 + aufs_bindex_t bindex;
7706 + struct au_vdir_destr destr;
7709 + AuDebugOn(bindex >= 0);
7712 + AuDebugOn(destr.len < NAME_MAX);
7714 +#ifdef CONFIG_4KSTACKS
7715 + pr_warn("CONFIG_4KSTACKS is defined.\n");
7720 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7721 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
7722 +++ linux/fs/aufs/debug.h 2022-11-05 23:02:18.962555950 +0100
7724 +/* SPDX-License-Identifier: GPL-2.0 */
7726 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7728 + * This program is free software; you can redistribute it and/or modify
7729 + * it under the terms of the GNU General Public License as published by
7730 + * the Free Software Foundation; either version 2 of the License, or
7731 + * (at your option) any later version.
7733 + * This program is distributed in the hope that it will be useful,
7734 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7735 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7736 + * GNU General Public License for more details.
7738 + * You should have received a copy of the GNU General Public License
7739 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7743 + * debug print functions
7746 +#ifndef __AUFS_DEBUG_H__
7747 +#define __AUFS_DEBUG_H__
7751 +#include <linux/atomic.h>
7752 +#include <linux/module.h>
7753 +#include <linux/kallsyms.h>
7754 +#include <linux/sysrq.h>
7756 +#ifdef CONFIG_AUFS_DEBUG
7757 +#define AuDebugOn(a) BUG_ON(a)
7759 +/* module parameter */
7760 +extern atomic_t aufs_debug;
7761 +static inline void au_debug_on(void)
7763 + atomic_inc(&aufs_debug);
7765 +static inline void au_debug_off(void)
7767 + atomic_dec_if_positive(&aufs_debug);
7770 +static inline int au_debug_test(void)
7772 + return atomic_read(&aufs_debug) > 0;
7775 +#define AuDebugOn(a) do {} while (0)
7776 +AuStubVoid(au_debug_on, void)
7777 +AuStubVoid(au_debug_off, void)
7778 +AuStubInt0(au_debug_test, void)
7779 +#endif /* CONFIG_AUFS_DEBUG */
7781 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7783 +/* ---------------------------------------------------------------------- */
7787 +#define AuDbg(fmt, ...) do { \
7788 + if (au_debug_test()) \
7789 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7791 +#define AuLabel(l) AuDbg(#l "\n")
7792 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7793 +#define AuWarn1(fmt, ...) do { \
7794 + static unsigned char _c; \
7796 + pr_warn(fmt, ##__VA_ARGS__); \
7799 +#define AuErr1(fmt, ...) do { \
7800 + static unsigned char _c; \
7802 + pr_err(fmt, ##__VA_ARGS__); \
7805 +#define AuIOErr1(fmt, ...) do { \
7806 + static unsigned char _c; \
7808 + AuIOErr(fmt, ##__VA_ARGS__); \
7811 +#define AuUnsupportMsg "This operation is not supported." \
7812 + " Please report this application to aufs-users ML."
7813 +#define AuUnsupport(fmt, ...) do { \
7814 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7818 +#define AuTraceErr(e) do { \
7819 + if (unlikely((e) < 0)) \
7820 + AuDbg("err %d\n", (int)(e)); \
7823 +#define AuTraceErrPtr(p) do { \
7825 + AuDbg("err %ld\n", PTR_ERR(p)); \
7828 +/* dirty macros for debug print, use with "%.*s" and caution */
7829 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
7831 +/* ---------------------------------------------------------------------- */
7834 +#ifdef CONFIG_AUFS_DEBUG
7835 +extern struct mutex au_dbg_mtx;
7836 +extern char *au_plevel;
7838 +void au_dpri_whlist(struct au_nhash *whlist);
7840 +void au_dpri_vdir(struct au_vdir *vdir);
7842 +void au_dpri_inode(struct inode *inode);
7843 +void au_dpri_dalias(struct inode *inode);
7844 +void au_dpri_dentry(struct dentry *dentry);
7846 +void au_dpri_file(struct file *filp);
7847 +struct super_block;
7848 +void au_dpri_sb(struct super_block *sb);
7850 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7851 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7852 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7853 +void au_dbg_verify_kthread(void);
7855 +int __init au_debug_init(void);
7857 +#define AuDbgWhlist(w) do { \
7858 + mutex_lock(&au_dbg_mtx); \
7860 + au_dpri_whlist(w); \
7861 + mutex_unlock(&au_dbg_mtx); \
7864 +#define AuDbgVdir(v) do { \
7865 + mutex_lock(&au_dbg_mtx); \
7867 + au_dpri_vdir(v); \
7868 + mutex_unlock(&au_dbg_mtx); \
7871 +#define AuDbgInode(i) do { \
7872 + mutex_lock(&au_dbg_mtx); \
7874 + au_dpri_inode(i); \
7875 + mutex_unlock(&au_dbg_mtx); \
7878 +#define AuDbgDAlias(i) do { \
7879 + mutex_lock(&au_dbg_mtx); \
7881 + au_dpri_dalias(i); \
7882 + mutex_unlock(&au_dbg_mtx); \
7885 +#define AuDbgDentry(d) do { \
7886 + mutex_lock(&au_dbg_mtx); \
7888 + au_dpri_dentry(d); \
7889 + mutex_unlock(&au_dbg_mtx); \
7892 +#define AuDbgFile(f) do { \
7893 + mutex_lock(&au_dbg_mtx); \
7895 + au_dpri_file(f); \
7896 + mutex_unlock(&au_dbg_mtx); \
7899 +#define AuDbgSb(sb) do { \
7900 + mutex_lock(&au_dbg_mtx); \
7901 + AuDbg(#sb "\n"); \
7903 + mutex_unlock(&au_dbg_mtx); \
7906 +#define AuDbgSym(addr) do { \
7907 + char sym[KSYM_SYMBOL_LEN]; \
7908 + sprint_symbol(sym, (unsigned long)addr); \
7909 + AuDbg("%s\n", sym); \
7912 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7913 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7914 +AuStubVoid(au_dbg_verify_kthread, void)
7915 +AuStubInt0(__init au_debug_init, void)
7917 +#define AuDbgWhlist(w) do {} while (0)
7918 +#define AuDbgVdir(v) do {} while (0)
7919 +#define AuDbgInode(i) do {} while (0)
7920 +#define AuDbgDAlias(i) do {} while (0)
7921 +#define AuDbgDentry(d) do {} while (0)
7922 +#define AuDbgFile(f) do {} while (0)
7923 +#define AuDbgSb(sb) do {} while (0)
7924 +#define AuDbgSym(addr) do {} while (0)
7925 +#endif /* CONFIG_AUFS_DEBUG */
7927 +/* ---------------------------------------------------------------------- */
7929 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7930 +int __init au_sysrq_init(void);
7931 +void au_sysrq_fin(void);
7933 +#ifdef CONFIG_HW_CONSOLE
7934 +#define au_dbg_blocked() do { \
7936 + handle_sysrq('w'); \
7939 +AuStubVoid(au_dbg_blocked, void)
7943 +AuStubInt0(__init au_sysrq_init, void)
7944 +AuStubVoid(au_sysrq_fin, void)
7945 +AuStubVoid(au_dbg_blocked, void)
7946 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7948 +#endif /* __KERNEL__ */
7949 +#endif /* __AUFS_DEBUG_H__ */
7950 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7951 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
7952 +++ linux/fs/aufs/dentry.c 2023-04-24 08:13:40.969576115 +0200
7954 +// SPDX-License-Identifier: GPL-2.0
7956 + * Copyright (C) 2005-2022 Junjiro R. Okajima
7958 + * This program is free software; you can redistribute it and/or modify
7959 + * it under the terms of the GNU General Public License as published by
7960 + * the Free Software Foundation; either version 2 of the License, or
7961 + * (at your option) any later version.
7963 + * This program is distributed in the hope that it will be useful,
7964 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7965 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7966 + * GNU General Public License for more details.
7968 + * You should have received a copy of the GNU General Public License
7969 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7973 + * lookup and dentry operations
7976 +#include <linux/iversion.h>
7980 + * returns positive/negative dentry, NULL or an error.
7981 + * NULL means whiteout-ed or not-found.
7983 +static struct dentry*
7984 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7985 + aufs_bindex_t bindex, struct au_do_lookup_args *args)
7987 + struct dentry *h_dentry;
7988 + struct inode *h_inode;
7989 + struct au_branch *br;
7990 + struct mnt_idmap *h_idmap;
7991 + struct path h_path;
7992 + int wh_found, opq;
7993 + unsigned char wh_able;
7994 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
7995 + const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
7999 + br = au_sbr(dentry->d_sb, bindex);
8000 + h_path.dentry = h_parent;
8001 + h_path.mnt = au_br_mnt(br);
8002 + h_idmap = au_br_idmap(br);
8003 + wh_able = !!au_br_whable(br->br_perm);
8005 + wh_found = au_wh_test(h_idmap, &h_path, &args->whname,
8007 + h_dentry = ERR_PTR(wh_found);
8010 + if (unlikely(wh_found < 0))
8013 + /* We found a whiteout */
8014 + /* au_set_dbbot(dentry, bindex); */
8015 + au_set_dbwh(dentry, bindex);
8017 + return NULL; /* success */
8021 + h_dentry = vfsub_lkup_one(args->name, &h_path);
8023 + h_dentry = au_sio_lkup_one(h_idmap, args->name, &h_path);
8024 + if (IS_ERR(h_dentry)) {
8025 + if (PTR_ERR(h_dentry) == -ENAMETOOLONG
8031 + h_inode = d_inode(h_dentry);
8032 + if (d_is_negative(h_dentry)) {
8035 + } else if (wh_found
8036 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
8038 + else if (au_ftest_lkup(args->flags, DIRREN)
8040 + && !au_dr_lkup_h_ino(args, bindex, h_inode->i_ino)) {
8041 + AuDbg("b%d %pd ignored hi%llu\n", bindex, h_dentry,
8042 + (unsigned long long)h_inode->i_ino);
8046 + if (au_dbbot(dentry) <= bindex)
8047 + au_set_dbbot(dentry, bindex);
8048 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
8049 + au_set_dbtop(dentry, bindex);
8050 + au_set_h_dptr(dentry, bindex, h_dentry);
8052 + if (!d_is_dir(h_dentry)
8054 + || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
8055 + goto out; /* success */
8057 + h_path.dentry = h_dentry;
8058 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
8059 + opq = au_diropq_test(h_idmap, &h_path);
8060 + inode_unlock_shared(h_inode);
8062 + au_set_dbdiropq(dentry, bindex);
8063 + else if (unlikely(opq < 0)) {
8064 + au_set_h_dptr(dentry, bindex, NULL);
8065 + h_dentry = ERR_PTR(opq);
8076 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
8078 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
8079 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
8085 + * returns the number of lower positive dentries,
8086 + * otherwise an error.
8087 + * can be called at unlinking with @type is zero.
8089 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
8090 + unsigned int flags)
8092 + int npositive, err;
8093 + aufs_bindex_t bindex, btail, bdiropq;
8094 + unsigned char isdir, dirperm1, dirren;
8095 + struct au_do_lookup_args args = {
8097 + .name = &dentry->d_name
8099 + struct dentry *parent;
8100 + struct super_block *sb;
8102 + sb = dentry->d_sb;
8103 + err = au_test_shwh(sb, args.name);
8104 + if (unlikely(err))
8107 + err = au_wh_name_alloc(&args.whname, args.name);
8108 + if (unlikely(err))
8111 + isdir = !!d_is_dir(dentry);
8112 + dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
8113 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
8115 + au_fset_lkup(args.flags, DIRREN);
8118 + parent = dget_parent(dentry);
8119 + btail = au_dbtaildir(parent);
8120 + for (bindex = btop; bindex <= btail; bindex++) {
8121 + struct dentry *h_parent, *h_dentry;
8122 + struct inode *h_inode, *h_dir;
8123 + struct au_branch *br;
8125 + h_dentry = au_h_dptr(dentry, bindex);
8127 + if (d_is_positive(h_dentry))
8131 + h_parent = au_h_dptr(parent, bindex);
8132 + if (!h_parent || !d_is_dir(h_parent))
8136 + /* if the inum matches, then use the prepared name */
8137 + err = au_dr_lkup_name(&args, bindex);
8138 + if (unlikely(err))
8142 + h_dir = d_inode(h_parent);
8143 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
8144 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &args);
8145 + inode_unlock_shared(h_dir);
8146 + err = PTR_ERR(h_dentry);
8147 + if (IS_ERR(h_dentry))
8150 + au_fclr_lkup(args.flags, ALLOW_NEG);
8152 + au_fset_lkup(args.flags, IGNORE_PERM);
8154 + if (au_dbwh(dentry) == bindex)
8158 + if (d_is_negative(h_dentry))
8160 + h_inode = d_inode(h_dentry);
8163 + args.type = h_inode->i_mode & S_IFMT;
8164 + if (args.type != S_IFDIR)
8167 + /* the type of lower may be different */
8168 + bdiropq = au_dbdiropq(dentry);
8169 + if (bdiropq >= 0 && bdiropq <= bindex)
8172 + br = au_sbr(sb, bindex);
8174 + && au_dr_hino_test_add(&br->br_dirren, h_inode->i_ino,
8175 + /*add_ent*/NULL)) {
8176 + /* prepare next name to lookup */
8177 + err = au_dr_lkup(&args, dentry, bindex);
8178 + if (unlikely(err))
8184 + AuLabel(positive);
8185 + au_update_dbtop(dentry);
8188 + if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
8189 + && au_dbtop(dentry) < 0)) {
8191 + AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
8197 + au_kfree_try_rcu(args.whname.name);
8199 + au_dr_lkup_fin(&args);
8204 +struct dentry *au_sio_lkup_one(struct mnt_idmap *idmap, struct qstr *name,
8205 + struct path *ppath)
8207 + struct dentry *dentry;
8210 + if (!au_test_h_perm_sio(idmap, d_inode(ppath->dentry), MAY_EXEC))
8211 + dentry = vfsub_lkup_one(name, ppath);
8213 + struct vfsub_lkup_one_args args = {
8219 + wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
8220 + if (unlikely(wkq_err))
8221 + dentry = ERR_PTR(wkq_err);
8228 + * lookup @dentry on @bindex which should be negative.
8230 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
8233 + struct dentry *parent, *h_dentry;
8234 + struct au_branch *br;
8235 + struct mnt_idmap *h_idmap;
8236 + struct path h_ppath;
8238 + parent = dget_parent(dentry);
8239 + br = au_sbr(dentry->d_sb, bindex);
8240 + h_ppath.dentry = au_h_dptr(parent, bindex);
8241 + h_ppath.mnt = au_br_mnt(br);
8242 + h_idmap = au_br_idmap(br);
8244 + h_dentry = au_whtmp_lkup(h_ppath.dentry, br, &dentry->d_name);
8246 + h_dentry = au_sio_lkup_one(h_idmap, &dentry->d_name, &h_ppath);
8247 + err = PTR_ERR(h_dentry);
8248 + if (IS_ERR(h_dentry))
8250 + if (unlikely(d_is_positive(h_dentry))) {
8252 + AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
8258 + if (bindex < au_dbtop(dentry))
8259 + au_set_dbtop(dentry, bindex);
8260 + if (au_dbbot(dentry) < bindex)
8261 + au_set_dbbot(dentry, bindex);
8262 + au_set_h_dptr(dentry, bindex, h_dentry);
8269 +/* ---------------------------------------------------------------------- */
8271 +/* subset of struct inode */
8273 + unsigned long i_ino;
8274 + /* unsigned int i_nlink; */
8280 + blkcnt_t i_blocks;
8285 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
8287 + ia->i_ino = h_inode->i_ino;
8288 + /* ia->i_nlink = h_inode->i_nlink; */
8289 + ia->i_uid = h_inode->i_uid;
8290 + ia->i_gid = h_inode->i_gid;
8291 + ia->i_version = inode_query_iversion(h_inode);
8293 + ia->i_size = h_inode->i_size;
8294 + ia->i_blocks = h_inode->i_blocks;
8296 + ia->i_mode = (h_inode->i_mode & S_IFMT);
8299 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
8301 + return ia->i_ino != h_inode->i_ino
8302 + /* || ia->i_nlink != h_inode->i_nlink */
8303 + || !uid_eq(ia->i_uid, h_inode->i_uid)
8304 + || !gid_eq(ia->i_gid, h_inode->i_gid)
8305 + || !inode_eq_iversion(h_inode, ia->i_version)
8307 + || ia->i_size != h_inode->i_size
8308 + || ia->i_blocks != h_inode->i_blocks
8310 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
8313 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
8314 + struct au_branch *br)
8317 + struct au_iattr ia;
8318 + struct inode *h_inode;
8319 + struct dentry *h_d;
8320 + struct super_block *h_sb;
8321 + struct path h_ppath;
8324 + memset(&ia, -1, sizeof(ia));
8325 + h_sb = h_dentry->d_sb;
8327 + if (d_is_positive(h_dentry)) {
8328 + h_inode = d_inode(h_dentry);
8329 + au_iattr_save(&ia, h_inode);
8330 + } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
8331 + /* nfs d_revalidate may return 0 for negative dentry */
8332 + /* fuse d_revalidate always return 0 for negative dentry */
8335 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
8336 + h_ppath.dentry = h_parent;
8337 + h_ppath.mnt = au_br_mnt(br);
8338 + h_d = vfsub_lkup_one(&h_dentry->d_name, &h_ppath);
8339 + err = PTR_ERR(h_d);
8344 + if (unlikely(h_d != h_dentry
8345 + || d_inode(h_d) != h_inode
8346 + || (h_inode && au_iattr_test(&ia, h_inode))))
8347 + err = au_busy_or_stale();
8355 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8356 + struct dentry *h_parent, struct au_branch *br)
8361 + if (udba == AuOpt_UDBA_REVAL
8362 + && !au_test_fs_remote(h_dentry->d_sb)) {
8364 + err = (d_inode(h_dentry->d_parent) != h_dir);
8365 + } else if (udba != AuOpt_UDBA_NONE)
8366 + err = au_h_verify_dentry(h_dentry, h_parent, br);
8371 +/* ---------------------------------------------------------------------- */
8373 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
8376 + aufs_bindex_t new_bindex, bindex, bbot, bwh, bdiropq;
8377 + struct au_hdentry tmp, *p, *q;
8378 + struct au_dinfo *dinfo;
8379 + struct super_block *sb;
8381 + DiMustWriteLock(dentry);
8383 + sb = dentry->d_sb;
8384 + dinfo = au_di(dentry);
8385 + bbot = dinfo->di_bbot;
8386 + bwh = dinfo->di_bwh;
8387 + bdiropq = dinfo->di_bdiropq;
8388 + bindex = dinfo->di_btop;
8389 + p = au_hdentry(dinfo, bindex);
8390 + for (; bindex <= bbot; bindex++, p++) {
8391 + if (!p->hd_dentry)
8394 + new_bindex = au_br_index(sb, p->hd_id);
8395 + if (new_bindex == bindex)
8398 + if (dinfo->di_bwh == bindex)
8400 + if (dinfo->di_bdiropq == bindex)
8401 + bdiropq = new_bindex;
8402 + if (new_bindex < 0) {
8404 + p->hd_dentry = NULL;
8408 + /* swap two lower dentries, and loop again */
8409 + q = au_hdentry(dinfo, new_bindex);
8413 + if (tmp.hd_dentry) {
8419 + dinfo->di_bwh = -1;
8420 + if (bwh >= 0 && bwh <= au_sbbot(sb) && au_sbr_whable(sb, bwh))
8421 + dinfo->di_bwh = bwh;
8423 + dinfo->di_bdiropq = -1;
8425 + && bdiropq <= au_sbbot(sb)
8426 + && au_sbr_whable(sb, bdiropq))
8427 + dinfo->di_bdiropq = bdiropq;
8430 + dinfo->di_btop = -1;
8431 + dinfo->di_bbot = -1;
8432 + bbot = au_dbbot(parent);
8434 + p = au_hdentry(dinfo, bindex);
8435 + for (; bindex <= bbot; bindex++, p++)
8436 + if (p->hd_dentry) {
8437 + dinfo->di_btop = bindex;
8441 + if (dinfo->di_btop >= 0) {
8443 + p = au_hdentry(dinfo, bindex);
8444 + for (; bindex >= 0; bindex--, p--)
8445 + if (p->hd_dentry) {
8446 + dinfo->di_bbot = bindex;
8455 +static void au_do_hide(struct dentry *dentry)
8457 + struct inode *inode;
8459 + if (d_really_is_positive(dentry)) {
8460 + inode = d_inode(dentry);
8461 + if (!d_is_dir(dentry)) {
8462 + if (inode->i_nlink && !d_unhashed(dentry))
8463 + drop_nlink(inode);
8465 + clear_nlink(inode);
8466 + /* stop next lookup */
8467 + inode->i_flags |= S_DEAD;
8469 + smp_mb(); /* necessary? */
8474 +static int au_hide_children(struct dentry *parent)
8476 + int err, i, j, ndentry;
8477 + struct au_dcsub_pages dpages;
8478 + struct au_dpage *dpage;
8479 + struct dentry *dentry;
8481 + err = au_dpages_init(&dpages, GFP_NOFS);
8482 + if (unlikely(err))
8484 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
8485 + if (unlikely(err))
8488 + /* in reverse order */
8489 + for (i = dpages.ndpage - 1; i >= 0; i--) {
8490 + dpage = dpages.dpages + i;
8491 + ndentry = dpage->ndentry;
8492 + for (j = ndentry - 1; j >= 0; j--) {
8493 + dentry = dpage->dentries[j];
8494 + if (dentry != parent)
8495 + au_do_hide(dentry);
8500 + au_dpages_free(&dpages);
8505 +static void au_hide(struct dentry *dentry)
8509 + AuDbgDentry(dentry);
8510 + if (d_is_dir(dentry)) {
8511 + /* shrink_dcache_parent(dentry); */
8512 + err = au_hide_children(dentry);
8513 + if (unlikely(err))
8514 + AuIOErr("%pd, failed hiding children, ignored %d\n",
8517 + au_do_hide(dentry);
8521 + * By adding a dirty branch, a cached dentry may be affected in various ways.
8523 + * a dirty branch is added
8524 + * - on the top of layers
8525 + * - in the middle of layers
8526 + * - to the bottom of layers
8528 + * on the added branch there exists
8531 + * - a same named entry
8533 + * * negative --> positive
8534 + * * positive --> positive
8535 + * - type is unchanged
8536 + * - type is changed
8538 + * * negative --> negative
8539 + * * positive --> negative (rejected by au_br_del() for non-dir case)
8542 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8543 + struct au_dinfo *tmp)
8546 + aufs_bindex_t bindex, bbot;
8548 + struct dentry *dentry;
8549 + struct inode *inode;
8551 + } orig_h, tmp_h = {
8554 + struct au_hdentry *hd;
8555 + struct inode *inode, *h_inode;
8556 + struct dentry *h_dentry;
8559 + AuDebugOn(dinfo->di_btop < 0);
8561 + orig_h.dentry = au_hdentry(dinfo, dinfo->di_btop)->hd_dentry;
8562 + orig_h.inode = NULL;
8563 + if (d_is_positive(orig_h.dentry)) {
8564 + orig_h.inode = d_inode(orig_h.dentry);
8565 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8567 + if (tmp->di_btop >= 0) {
8568 + tmp_h.dentry = au_hdentry(tmp, tmp->di_btop)->hd_dentry;
8569 + if (d_is_positive(tmp_h.dentry)) {
8570 + tmp_h.inode = d_inode(tmp_h.dentry);
8571 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8576 + if (d_really_is_positive(dentry))
8577 + inode = d_inode(dentry);
8578 + if (!orig_h.inode) {
8579 + AuDbg("negative originally\n");
8585 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8586 + AuDebugOn(dinfo->di_bdiropq != -1);
8588 + if (!tmp_h.inode) {
8589 + AuDbg("negative --> negative\n");
8590 + /* should have only one negative lower */
8591 + if (tmp->di_btop >= 0
8592 + && tmp->di_btop < dinfo->di_btop) {
8593 + AuDebugOn(tmp->di_btop != tmp->di_bbot);
8594 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8595 + au_set_h_dptr(dentry, dinfo->di_btop, NULL);
8596 + au_di_cp(dinfo, tmp);
8597 + hd = au_hdentry(tmp, tmp->di_btop);
8598 + au_set_h_dptr(dentry, tmp->di_btop,
8599 + dget(hd->hd_dentry));
8601 + au_dbg_verify_dinode(dentry);
8603 + AuDbg("negative --> positive\n");
8605 + * similar to the behaviour of creating with bypassing
8607 + * unhash it in order to force an error in the
8608 + * succeeding create operation.
8609 + * we should not set S_DEAD here.
8612 + /* au_di_swap(tmp, dinfo); */
8613 + au_dbg_verify_dinode(dentry);
8616 + AuDbg("positive originally\n");
8617 + /* inode may be NULL */
8618 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8619 + if (!tmp_h.inode) {
8620 + AuDbg("positive --> negative\n");
8621 + /* or bypassing aufs */
8623 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_btop)
8624 + dinfo->di_bwh = tmp->di_bwh;
8626 + err = au_refresh_hinode_self(inode);
8627 + au_dbg_verify_dinode(dentry);
8628 + } else if (orig_h.mode == tmp_h.mode) {
8629 + AuDbg("positive --> positive, same type\n");
8630 + if (!S_ISDIR(orig_h.mode)
8631 + && dinfo->di_btop > tmp->di_btop) {
8633 + * similar to the behaviour of removing and
8638 + err = au_refresh_hinode_self(inode);
8639 + au_dbg_verify_dinode(dentry);
8641 + /* fill empty slots */
8642 + if (dinfo->di_btop > tmp->di_btop)
8643 + dinfo->di_btop = tmp->di_btop;
8644 + if (dinfo->di_bbot < tmp->di_bbot)
8645 + dinfo->di_bbot = tmp->di_bbot;
8646 + dinfo->di_bwh = tmp->di_bwh;
8647 + dinfo->di_bdiropq = tmp->di_bdiropq;
8648 + bbot = dinfo->di_bbot;
8649 + bindex = tmp->di_btop;
8650 + hd = au_hdentry(tmp, bindex);
8651 + for (; bindex <= bbot; bindex++, hd++) {
8652 + if (au_h_dptr(dentry, bindex))
8654 + h_dentry = hd->hd_dentry;
8657 + AuDebugOn(d_is_negative(h_dentry));
8658 + h_inode = d_inode(h_dentry);
8659 + AuDebugOn(orig_h.mode
8660 + != (h_inode->i_mode
8662 + au_set_h_dptr(dentry, bindex,
8666 + err = au_refresh_hinode(inode, dentry);
8667 + au_dbg_verify_dinode(dentry);
8670 + AuDbg("positive --> positive, different type\n");
8671 + /* similar to the behaviour of removing and creating */
8674 + err = au_refresh_hinode_self(inode);
8675 + au_dbg_verify_dinode(dentry);
8683 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8685 + const struct dentry_operations *dop
8686 + = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8687 + static const unsigned int mask
8688 + = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8690 + BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8692 + if (dentry->d_op == dop)
8695 + AuDbg("%pd\n", dentry);
8696 + spin_lock(&dentry->d_lock);
8697 + if (dop == &aufs_dop)
8698 + dentry->d_flags |= mask;
8700 + dentry->d_flags &= ~mask;
8701 + dentry->d_op = dop;
8702 + spin_unlock(&dentry->d_lock);
8705 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8707 + int err, ebrange, nbr;
8708 + unsigned int sigen;
8709 + struct au_dinfo *dinfo, *tmp;
8710 + struct super_block *sb;
8711 + struct inode *inode;
8713 + DiMustWriteLock(dentry);
8714 + AuDebugOn(IS_ROOT(dentry));
8715 + AuDebugOn(d_really_is_negative(parent));
8717 + sb = dentry->d_sb;
8718 + sigen = au_sigen(sb);
8719 + err = au_digen_test(parent, sigen);
8720 + if (unlikely(err))
8723 + nbr = au_sbbot(sb) + 1;
8724 + dinfo = au_di(dentry);
8725 + err = au_di_realloc(dinfo, nbr, /*may_shrink*/0);
8726 + if (unlikely(err))
8728 + ebrange = au_dbrange_test(dentry);
8730 + ebrange = au_do_refresh_hdentry(dentry, parent);
8732 + if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8733 + AuDebugOn(au_dbtop(dentry) < 0 && au_dbbot(dentry) >= 0);
8734 + if (d_really_is_positive(dentry)) {
8735 + inode = d_inode(dentry);
8736 + err = au_refresh_hinode_self(inode);
8738 + au_dbg_verify_dinode(dentry);
8740 + goto out_dgen; /* success */
8744 + /* temporary dinfo */
8745 + AuDbgDentry(dentry);
8747 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8748 + if (unlikely(!tmp))
8750 + au_di_swap(tmp, dinfo);
8751 + /* returns the number of positive dentries */
8753 + * if current working dir is removed, it returns an error.
8754 + * but the dentry is legal.
8756 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
8757 + AuDbgDentry(dentry);
8758 + au_di_swap(tmp, dinfo);
8759 + if (err == -ENOENT)
8762 + /* compare/refresh by dinfo */
8763 + AuDbgDentry(dentry);
8764 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8765 + au_dbg_verify_dinode(dentry);
8768 + au_di_realloc(dinfo, nbr, /*may_shrink*/1); /* harmless if err */
8769 + au_rw_write_unlock(&tmp->di_rwsem);
8771 + if (unlikely(err))
8775 + au_update_digen(dentry);
8777 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8778 + AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8779 + AuDbgDentry(dentry);
8785 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8786 + struct dentry *dentry, aufs_bindex_t bindex)
8791 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8794 + AuDbg("b%d\n", bindex);
8796 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8797 + * due to whiteout and branch permission.
8799 + flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8800 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
8801 + /* it may return tri-state */
8802 + valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8804 + if (unlikely(valid < 0))
8814 +/* todo: remove this */
8815 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8816 + unsigned int flags, int do_udba, int dirren)
8819 + umode_t mode, h_mode;
8820 + aufs_bindex_t bindex, btail, btop, ibs, ibe;
8821 + unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8822 + struct inode *h_inode, *h_cached_inode;
8823 + struct dentry *h_dentry;
8824 + struct qstr *name, *h_name;
8831 + unhashed = !!d_unhashed(dentry);
8832 + is_root = !!IS_ROOT(dentry);
8833 + name = &dentry->d_name;
8834 + tmpfile = au_di(dentry)->di_tmpfile;
8837 + * Theoretically, REVAL test should be unnecessary in case of
8839 + * But {fs,i}notify doesn't fire some necessary events,
8840 + * IN_ATTRIB for atime/nlink/pageio
8841 + * Let's do REVAL test too.
8843 + if (do_udba && inode) {
8844 + mode = (inode->i_mode & S_IFMT);
8845 + plus = (inode->i_nlink > 0);
8846 + ibs = au_ibtop(inode);
8847 + ibe = au_ibbot(inode);
8850 + btop = au_dbtop(dentry);
8852 + if (inode && S_ISDIR(inode->i_mode))
8853 + btail = au_dbtaildir(dentry);
8854 + for (bindex = btop; bindex <= btail; bindex++) {
8855 + h_dentry = au_h_dptr(dentry, bindex);
8859 + AuDbg("b%d, %pd\n", bindex, h_dentry);
8860 + h_nfs = !!au_test_nfs(h_dentry->d_sb);
8861 + spin_lock(&h_dentry->d_lock);
8862 + h_name = &h_dentry->d_name;
8863 + if (unlikely(do_udba
8866 + && (unhashed != !!d_unhashed(h_dentry)
8867 + || (!tmpfile && !dirren
8868 + && !au_qstreq(name, h_name))
8871 + && !(flags & LOOKUP_OPEN)
8872 + && (h_dentry->d_flags
8873 + & DCACHE_NFSFS_RENAMED)))
8877 + h_unhashed = d_unhashed(h_dentry);
8878 + spin_unlock(&h_dentry->d_lock);
8879 + AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8880 + unhashed, h_unhashed, dentry, h_dentry);
8883 + spin_unlock(&h_dentry->d_lock);
8885 + err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8886 + if (unlikely(err))
8887 + /* do not goto err, to keep the errno */
8890 + /* todo: plink too? */
8895 + if (unlikely(!!inode != d_is_positive(h_dentry)))
8899 + if (d_is_positive(h_dentry))
8900 + h_inode = d_inode(h_dentry);
8903 + h_cached_inode = h_inode;
8905 + h_mode = (h_inode->i_mode & S_IFMT);
8906 + h_plus = (h_inode->i_nlink > 0);
8908 + if (inode && ibs <= bindex && bindex <= ibe)
8909 + h_cached_inode = au_h_iptr(inode, bindex);
8912 + if (unlikely(plus != h_plus && !tmpfile))
8915 + if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8917 + && !IS_ROOT(h_dentry)
8918 + && unhashed != d_unhashed(h_dentry)))
8921 + if (unlikely(mode != h_mode
8922 + || h_cached_inode != h_inode))
8935 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8936 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8939 + struct dentry *parent;
8941 + if (!au_digen_test(dentry, sigen))
8944 + parent = dget_parent(dentry);
8945 + di_read_lock_parent(parent, AuLock_IR);
8946 + AuDebugOn(au_digen_test(parent, sigen));
8947 + au_dbg_verify_gen(parent, sigen);
8948 + err = au_refresh_dentry(dentry, parent);
8949 + di_read_unlock(parent, AuLock_IR);
8955 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8958 + struct dentry *d, *parent;
8960 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8961 + return simple_reval_dpath(dentry, sigen);
8963 + /* slow loop, keep it simple and stupid */
8964 + /* cf: au_cpup_dirs() */
8967 + while (au_digen_test(dentry, sigen)) {
8971 + parent = dget_parent(d);
8972 + if (!au_digen_test(parent, sigen))
8978 + di_write_lock_child2(d);
8980 + /* someone might update our dentry while we were sleeping */
8981 + if (au_digen_test(d, sigen)) {
8983 + * todo: consolidate with simple_reval_dpath(),
8984 + * do_refresh() and au_reval_for_attr().
8986 + di_read_lock_parent(parent, AuLock_IR);
8987 + err = au_refresh_dentry(d, parent);
8988 + di_read_unlock(parent, AuLock_IR);
8992 + di_write_unlock(d);
8994 + if (unlikely(err))
9002 + * if valid returns 1, otherwise 0.
9004 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
9007 + unsigned int sigen;
9008 + unsigned char do_udba, dirren;
9009 + struct super_block *sb;
9010 + struct inode *inode;
9012 + /* todo: support rcu-walk? */
9013 + if (flags & LOOKUP_RCU)
9017 + if (unlikely(!au_di(dentry)))
9021 + sb = dentry->d_sb;
9024 + * i_mutex of parent dir may be held,
9025 + * but we should not return 'invalid' due to busy.
9027 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
9028 + if (unlikely(err)) {
9034 + if (d_really_is_positive(dentry))
9035 + inode = d_inode(dentry);
9036 + if (unlikely(inode && au_is_bad_inode(inode))) {
9041 + if (unlikely(au_dbrange_test(dentry))) {
9047 + sigen = au_sigen(sb);
9048 + if (au_digen_test(dentry, sigen)) {
9049 + AuDebugOn(IS_ROOT(dentry));
9050 + err = au_reval_dpath(dentry, sigen);
9051 + if (unlikely(err)) {
9056 + di_downgrade_lock(dentry, AuLock_IR);
9059 + if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
9061 + && !(inode->i_state && I_LINKABLE)
9062 + && (IS_DEADDIR(inode) || !inode->i_nlink)) {
9067 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
9068 + if (do_udba && inode) {
9069 + aufs_bindex_t btop = au_ibtop(inode);
9070 + struct inode *h_inode;
9073 + h_inode = au_h_iptr(inode, btop);
9074 + if (h_inode && au_test_higen(inode, h_inode)) {
9081 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
9082 + err = h_d_revalidate(dentry, inode, flags, do_udba, dirren);
9083 + if (unlikely(!err && do_udba && au_dbtop(dentry) < 0)) {
9085 + AuDbg("both of real entry and whiteout found, %p, err %d\n",
9091 + di_downgrade_lock(dentry, AuLock_IR);
9093 + aufs_read_unlock(dentry, AuLock_IR);
9098 + AuDbg("%pd invalid, %d\n", dentry, valid);
9104 +static void aufs_d_release(struct dentry *dentry)
9106 + if (au_di(dentry)) {
9107 + au_di_fin(dentry);
9108 + au_hn_di_reinit(dentry);
9112 +const struct dentry_operations aufs_dop = {
9113 + .d_revalidate = aufs_d_revalidate,
9114 + .d_weak_revalidate = aufs_d_revalidate,
9115 + .d_release = aufs_d_release
9118 +/* aufs_dop without d_revalidate */
9119 +const struct dentry_operations aufs_dop_noreval = {
9120 + .d_release = aufs_d_release
9122 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
9123 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
9124 +++ linux/fs/aufs/dentry.h 2023-04-24 08:13:40.969576115 +0200
9126 +/* SPDX-License-Identifier: GPL-2.0 */
9128 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9130 + * This program is free software; you can redistribute it and/or modify
9131 + * it under the terms of the GNU General Public License as published by
9132 + * the Free Software Foundation; either version 2 of the License, or
9133 + * (at your option) any later version.
9135 + * This program is distributed in the hope that it will be useful,
9136 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9137 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9138 + * GNU General Public License for more details.
9140 + * You should have received a copy of the GNU General Public License
9141 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9145 + * lookup and dentry operations
9148 +#ifndef __AUFS_DENTRY_H__
9149 +#define __AUFS_DENTRY_H__
9153 +#include <linux/dcache.h>
9154 +#include "dirren.h"
9157 +struct au_hdentry {
9158 + struct dentry *hd_dentry;
9159 + aufs_bindex_t hd_id;
9163 + atomic_t di_generation;
9165 + struct au_rwsem di_rwsem;
9166 + aufs_bindex_t di_btop, di_bbot, di_bwh, di_bdiropq;
9167 + unsigned char di_tmpfile; /* to allow the different name */
9168 + struct au_hdentry *di_hdentry;
9169 + struct file *di_htmpfile;
9170 + struct rcu_head rcu;
9171 +} ____cacheline_aligned_in_smp;
9173 +/* ---------------------------------------------------------------------- */
9175 +/* flags for au_lkup_dentry() */
9176 +#define AuLkup_ALLOW_NEG 1
9177 +#define AuLkup_IGNORE_PERM (1 << 1)
9178 +#define AuLkup_DIRREN (1 << 2)
9179 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
9180 +#define au_fset_lkup(flags, name) \
9181 + do { (flags) |= AuLkup_##name; } while (0)
9182 +#define au_fclr_lkup(flags, name) \
9183 + do { (flags) &= ~AuLkup_##name; } while (0)
9185 +#ifndef CONFIG_AUFS_DIRREN
9186 +#undef AuLkup_DIRREN
9187 +#define AuLkup_DIRREN 0
9190 +struct au_do_lookup_args {
9191 + unsigned int flags;
9193 + struct qstr whname, *name;
9194 + struct au_dr_lookup dirren;
9197 +/* ---------------------------------------------------------------------- */
9200 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
9202 +struct dentry *au_sio_lkup_one(struct mnt_idmap *idmap, struct qstr *name,
9203 + struct path *ppath);
9204 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
9205 + struct dentry *h_parent, struct au_branch *br);
9207 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
9208 + unsigned int flags);
9209 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
9210 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
9211 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
9212 +void au_refresh_dop(struct dentry *dentry, int force_reval);
9215 +void au_di_init_once(void *_di);
9216 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
9217 +void au_di_free(struct au_dinfo *dinfo);
9218 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
9219 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
9220 +int au_di_init(struct dentry *dentry);
9221 +void au_di_fin(struct dentry *dentry);
9222 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink);
9224 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
9225 +void di_read_unlock(struct dentry *d, int flags);
9226 +void di_downgrade_lock(struct dentry *d, int flags);
9227 +void di_write_lock(struct dentry *d, unsigned int lsc);
9228 +void di_write_unlock(struct dentry *d);
9229 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
9230 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
9231 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
9233 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
9234 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
9235 +aufs_bindex_t au_dbtail(struct dentry *dentry);
9236 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
9238 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9239 + struct dentry *h_dentry);
9240 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
9241 +int au_dbrange_test(struct dentry *dentry);
9242 +void au_update_digen(struct dentry *dentry);
9243 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
9244 +void au_update_dbtop(struct dentry *dentry);
9245 +void au_update_dbbot(struct dentry *dentry);
9246 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
9248 +/* ---------------------------------------------------------------------- */
9250 +static inline struct au_dinfo *au_di(struct dentry *dentry)
9252 + return dentry->d_fsdata;
9255 +/* ---------------------------------------------------------------------- */
9257 +/* lock subclass for dinfo */
9259 + AuLsc_DI_CHILD, /* child first */
9260 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
9261 + AuLsc_DI_CHILD3, /* copyup dirs */
9265 + AuLsc_DI_TMP /* temp for replacing dinfo */
9269 + * di_read_lock_child, di_write_lock_child,
9270 + * di_read_lock_child2, di_write_lock_child2,
9271 + * di_read_lock_child3, di_write_lock_child3,
9272 + * di_read_lock_parent, di_write_lock_parent,
9273 + * di_read_lock_parent2, di_write_lock_parent2,
9274 + * di_read_lock_parent3, di_write_lock_parent3,
9276 +#define AuReadLockFunc(name, lsc) \
9277 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
9278 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
9280 +#define AuWriteLockFunc(name, lsc) \
9281 +static inline void di_write_lock_##name(struct dentry *d) \
9282 +{ di_write_lock(d, AuLsc_DI_##lsc); }
9284 +#define AuRWLockFuncs(name, lsc) \
9285 + AuReadLockFunc(name, lsc) \
9286 + AuWriteLockFunc(name, lsc)
9288 +AuRWLockFuncs(child, CHILD);
9289 +AuRWLockFuncs(child2, CHILD2);
9290 +AuRWLockFuncs(child3, CHILD3);
9291 +AuRWLockFuncs(parent, PARENT);
9292 +AuRWLockFuncs(parent2, PARENT2);
9293 +AuRWLockFuncs(parent3, PARENT3);
9295 +#undef AuReadLockFunc
9296 +#undef AuWriteLockFunc
9297 +#undef AuRWLockFuncs
9299 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
9300 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
9301 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
9303 +/* ---------------------------------------------------------------------- */
9305 +/* todo: memory barrier? */
9306 +static inline unsigned int au_digen(struct dentry *d)
9308 + return atomic_read(&au_di(d)->di_generation);
9311 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
9313 + hdentry->hd_dentry = NULL;
9316 +static inline struct au_hdentry *au_hdentry(struct au_dinfo *di,
9317 + aufs_bindex_t bindex)
9319 + return di->di_hdentry + bindex;
9322 +static inline void au_hdput(struct au_hdentry *hd)
9325 + dput(hd->hd_dentry);
9328 +static inline aufs_bindex_t au_dbtop(struct dentry *dentry)
9330 + DiMustAnyLock(dentry);
9331 + return au_di(dentry)->di_btop;
9334 +static inline aufs_bindex_t au_dbbot(struct dentry *dentry)
9336 + DiMustAnyLock(dentry);
9337 + return au_di(dentry)->di_bbot;
9340 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
9342 + DiMustAnyLock(dentry);
9343 + return au_di(dentry)->di_bwh;
9346 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
9348 + DiMustAnyLock(dentry);
9349 + return au_di(dentry)->di_bdiropq;
9352 +/* todo: hard/soft set? */
9353 +static inline void au_set_dbtop(struct dentry *dentry, aufs_bindex_t bindex)
9355 + DiMustWriteLock(dentry);
9356 + au_di(dentry)->di_btop = bindex;
9359 +static inline void au_set_dbbot(struct dentry *dentry, aufs_bindex_t bindex)
9361 + DiMustWriteLock(dentry);
9362 + au_di(dentry)->di_bbot = bindex;
9365 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
9367 + DiMustWriteLock(dentry);
9368 + /* dbwh can be outside of btop - bbot range */
9369 + au_di(dentry)->di_bwh = bindex;
9372 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
9374 + DiMustWriteLock(dentry);
9375 + au_di(dentry)->di_bdiropq = bindex;
9378 +/* ---------------------------------------------------------------------- */
9380 +#ifdef CONFIG_AUFS_HNOTIFY
9381 +static inline void au_digen_dec(struct dentry *d)
9383 + atomic_dec(&au_di(d)->di_generation);
9386 +static inline void au_hn_di_reinit(struct dentry *dentry)
9388 + dentry->d_fsdata = NULL;
9391 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
9392 +#endif /* CONFIG_AUFS_HNOTIFY */
9394 +#endif /* __KERNEL__ */
9395 +#endif /* __AUFS_DENTRY_H__ */
9396 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
9397 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
9398 +++ linux/fs/aufs/dinfo.c 2022-12-17 09:21:34.796521861 +0100
9400 +// SPDX-License-Identifier: GPL-2.0
9402 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9404 + * This program is free software; you can redistribute it and/or modify
9405 + * it under the terms of the GNU General Public License as published by
9406 + * the Free Software Foundation; either version 2 of the License, or
9407 + * (at your option) any later version.
9409 + * This program is distributed in the hope that it will be useful,
9410 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9411 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9412 + * GNU General Public License for more details.
9414 + * You should have received a copy of the GNU General Public License
9415 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9419 + * dentry private data
9424 +void au_di_init_once(void *_dinfo)
9426 + struct au_dinfo *dinfo = _dinfo;
9428 + au_rw_init(&dinfo->di_rwsem);
9431 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
9433 + struct au_dinfo *dinfo;
9436 + dinfo = au_cache_alloc_dinfo();
9437 + if (unlikely(!dinfo))
9440 + nbr = au_sbbot(sb) + 1;
9443 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
9444 + if (dinfo->di_hdentry) {
9445 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
9446 + dinfo->di_btop = -1;
9447 + dinfo->di_bbot = -1;
9448 + dinfo->di_bwh = -1;
9449 + dinfo->di_bdiropq = -1;
9450 + dinfo->di_tmpfile = 0;
9451 + for (i = 0; i < nbr; i++)
9452 + dinfo->di_hdentry[i].hd_id = -1;
9453 + dinfo->di_htmpfile = NULL;
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 2023-04-24 08:13:40.969576115 +0200
9959 +// SPDX-License-Identifier: GPL-2.0
9961 + * Copyright (C) 2005-2022 Junjiro R. Okajima
9963 + * This program is free software; you can redistribute it and/or modify
9964 + * it under the terms of the GNU General Public License as published by
9965 + * the Free Software Foundation; either version 2 of the License, or
9966 + * (at your option) any later version.
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 bool 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);
10543 + return !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 mnt_idmap *h_idmap;
10600 + h_idmap = au_sbr_idmap(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_idmap, 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 2022-11-05 23:02:18.962555950 +0100
10728 +/* SPDX-License-Identifier: GPL-2.0 */
10730 + * Copyright (C) 2005-2022 Junjiro R. Okajima
10732 + * This program is free software; you can redistribute it and/or modify
10733 + * it under the terms of the GNU General Public License as published by
10734 + * the Free Software Foundation; either version 2 of the License, or
10735 + * (at your option) any later version.
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 2022-11-05 23:02:18.962555950 +0100
10866 +// SPDX-License-Identifier: GPL-2.0
10868 + * Copyright (C) 2017-2022 Junjiro R. Okajima
10870 + * This program is free software; you can redistribute it and/or modify
10871 + * it under the terms of the GNU General Public License as published by
10872 + * the Free Software Foundation; either version 2 of the License, or
10873 + * (at your option) any later version.
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 2022-11-05 23:02:18.962555950 +0100
12185 +/* SPDX-License-Identifier: GPL-2.0 */
12187 + * Copyright (C) 2017-2022 Junjiro R. Okajima
12189 + * This program is free software; you can redistribute it and/or modify
12190 + * it under the terms of the GNU General Public License as published by
12191 + * the Free Software Foundation; either version 2 of the License, or
12192 + * (at your option) any later version.
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 2022-11-05 23:02:18.962555950 +0100
12329 +// SPDX-License-Identifier: GPL-2.0
12331 + * Copyright (C) 2010-2022 Junjiro R. Okajima
12333 + * This program is free software; you can redistribute it and/or modify
12334 + * it under the terms of the GNU General Public License as published by
12335 + * the Free Software Foundation; either version 2 of the License, or
12336 + * (at your option) any later version.
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(read_folio); /* force */
12508 + DySetAop(writepages);
12509 + DySetAop(dirty_folio);
12510 + DySetAop(invalidate_folio);
12511 + DySetAop(readahead);
12512 + DySetAop(write_begin);
12513 + DySetAop(write_end);
12515 + DySetAop(release_folio);
12516 + DySetAop(free_folio);
12517 + /* this one will be changed according to an aufs mount option */
12518 + DySetAop(direct_IO);
12519 + DySetAop(migrate_folio);
12520 + DySetAop(launder_folio);
12521 + DySetAop(is_partially_uptodate);
12522 + DySetAop(is_dirty_writeback);
12523 + DySetAop(error_remove_page);
12524 + DySetAop(swap_activate);
12525 + DySetAop(swap_deactivate);
12526 + DySetAop(swap_rw);
12528 + DyDbgSize(cnt, *h_aop);
12531 +/* ---------------------------------------------------------------------- */
12533 +static void dy_bug(struct kref *kref)
12538 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
12540 + struct au_dykey *key, *old;
12541 + struct hlist_bl_head *hbl;
12544 + void (*set)(struct au_dykey *key, const void *h_op,
12545 + struct super_block *h_sb __maybe_unused);
12547 + static const struct op a[] = {
12549 + .sz = sizeof(struct au_dyaop),
12553 + const struct op *p;
12555 + hbl = dynop + op->dy_type;
12556 + key = dy_gfind_get(hbl, op->dy_hop);
12558 + goto out_add; /* success */
12560 + p = a + op->dy_type;
12561 + key = kzalloc(p->sz, GFP_NOFS);
12562 + if (unlikely(!key)) {
12563 + key = ERR_PTR(-ENOMEM);
12567 + key->dk_op.dy_hop = op->dy_hop;
12568 + kref_init(&key->dk_kref);
12569 + p->set(key, op->dy_hop, au_br_sb(br));
12570 + old = dy_gadd(hbl, key);
12572 + au_kfree_rcu(key);
12577 + old = dy_bradd(br, key);
12579 + /* its ref-count should never be zero here */
12580 + kref_put(&key->dk_kref, dy_bug);
12585 +/* ---------------------------------------------------------------------- */
12587 + * Aufs prohibits O_DIRECT by default even if the branch supports it.
12588 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
12589 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
12590 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
12591 + * See the aufs manual in detail.
12593 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
12596 + dyaop->da_op.direct_IO = NULL;
12598 + dyaop->da_op.direct_IO = aufs_aop.direct_IO;
12601 +static struct au_dyaop *dy_aget(struct au_branch *br,
12602 + const struct address_space_operations *h_aop,
12605 + struct au_dyaop *dyaop;
12606 + struct au_dynop op;
12608 + op.dy_type = AuDy_AOP;
12609 + op.dy_haop = h_aop;
12610 + dyaop = (void *)dy_get(&op, br);
12611 + if (IS_ERR(dyaop))
12613 + dy_adx(dyaop, do_dx);
12619 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12620 + struct inode *h_inode)
12623 + struct super_block *sb;
12624 + struct au_branch *br;
12625 + struct au_dyaop *dyaop;
12627 + AuDebugOn(!S_ISREG(h_inode->i_mode));
12628 + IiMustWriteLock(inode);
12630 + sb = inode->i_sb;
12631 + br = au_sbr(sb, bindex);
12632 + do_dx = !!au_opt_test(au_mntflags(sb), DIO);
12633 + dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
12634 + err = PTR_ERR(dyaop);
12635 + if (IS_ERR(dyaop))
12636 + /* unnecessary to call dy_fput() */
12640 + inode->i_mapping->a_ops = &dyaop->da_op;
12647 + * Is it safe to replace a_ops during the inode/file is in operation?
12648 + * Yes, I hope so.
12650 +int au_dy_irefresh(struct inode *inode)
12653 + aufs_bindex_t btop;
12654 + struct inode *h_inode;
12657 + if (S_ISREG(inode->i_mode)) {
12658 + btop = au_ibtop(inode);
12659 + h_inode = au_h_iptr(inode, btop);
12660 + err = au_dy_iaop(inode, btop, h_inode);
12665 +void au_dy_arefresh(int do_dx)
12667 + struct hlist_bl_head *hbl;
12668 + struct hlist_bl_node *pos;
12669 + struct au_dykey *key;
12671 + hbl = dynop + AuDy_AOP;
12672 + hlist_bl_lock(hbl);
12673 + hlist_bl_for_each_entry(key, pos, hbl, dk_hnode)
12674 + dy_adx((void *)key, do_dx);
12675 + hlist_bl_unlock(hbl);
12678 +/* ---------------------------------------------------------------------- */
12680 +void __init au_dy_init(void)
12684 + for (i = 0; i < AuDyLast; i++)
12685 + INIT_HLIST_BL_HEAD(dynop + i);
12688 +void au_dy_fin(void)
12692 + for (i = 0; i < AuDyLast; i++)
12693 + WARN_ON(!hlist_bl_empty(dynop + i));
12695 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
12696 --- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
12697 +++ linux/fs/aufs/dynop.h 2022-11-05 23:02:18.962555950 +0100
12699 +/* SPDX-License-Identifier: GPL-2.0 */
12701 + * Copyright (C) 2010-2022 Junjiro R. Okajima
12703 + * This program is free software; you can redistribute it and/or modify
12704 + * it under the terms of the GNU General Public License as published by
12705 + * the Free Software Foundation; either version 2 of the License, or
12706 + * (at your option) any later version.
12708 + * This program is distributed in the hope that it will be useful,
12709 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12710 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12711 + * GNU General Public License for more details.
12713 + * You should have received a copy of the GNU General Public License
12714 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12718 + * dynamically customizable operations (for regular files only)
12721 +#ifndef __AUFS_DYNOP_H__
12722 +#define __AUFS_DYNOP_H__
12726 +#include <linux/fs.h>
12727 +#include <linux/kref.h>
12729 +enum {AuDy_AOP, AuDyLast};
12734 + const void *dy_hop;
12735 + const struct address_space_operations *dy_haop;
12741 + struct hlist_bl_node dk_hnode;
12742 + struct rcu_head dk_rcu;
12744 + struct au_dynop dk_op;
12747 + * during I am in the branch local array, kref is gotten. when the
12748 + * branch is removed, kref is put.
12750 + struct kref dk_kref;
12753 +/* stop unioning since their sizes are very different from each other */
12755 + struct au_dykey da_key;
12756 + struct address_space_operations da_op; /* not const */
12758 +/* make sure that 'struct au_dykey *' can be any type */
12759 +static_assert(!offsetof(struct au_dyaop, da_key));
12761 +/* ---------------------------------------------------------------------- */
12765 +void au_dy_put(struct au_dykey *key);
12766 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12767 + struct inode *h_inode);
12768 +int au_dy_irefresh(struct inode *inode);
12769 +void au_dy_arefresh(int do_dio);
12771 +void __init au_dy_init(void);
12772 +void au_dy_fin(void);
12774 +#endif /* __KERNEL__ */
12775 +#endif /* __AUFS_DYNOP_H__ */
12776 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
12777 --- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
12778 +++ linux/fs/aufs/export.c 2022-12-17 09:21:34.796521861 +0100
12780 +// SPDX-License-Identifier: GPL-2.0
12782 + * Copyright (C) 2005-2022 Junjiro R. Okajima
12784 + * This program is free software; you can redistribute it and/or modify
12785 + * it under the terms of the GNU General Public License as published by
12786 + * the Free Software Foundation; either version 2 of the License, or
12787 + * (at your option) any later version.
12789 + * This program is distributed in the hope that it will be useful,
12790 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12791 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12792 + * GNU General Public License for more details.
12794 + * You should have received a copy of the GNU General Public License
12795 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12802 +#include <linux/exportfs.h>
12803 +#include <linux/fs_struct.h>
12804 +#include <linux/nsproxy.h>
12805 +#include <linux/random.h>
12806 +#include <linux/writeback.h>
12810 +#ifdef CONFIG_AUFS_INO_T_64
12818 +static ino_t decode_ino(__u32 *a)
12822 + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
12824 +#ifdef CONFIG_AUFS_INO_T_64
12830 +static void encode_ino(__u32 *a, ino_t ino)
12836 +#ifdef CONFIG_AUFS_INO_T_64
12841 +/* NFS file handle */
12845 +#ifdef CONFIG_AUFS_INO_T_64
12846 + /* support 64bit inode number */
12859 + Fh_ino = Fh_ino1,
12860 + Fh_dir_ino = Fh_dir_ino1
12863 +static int au_test_anon(struct dentry *dentry)
12865 + /* note: read d_flags without d_lock */
12866 + return !!(dentry->d_flags & DCACHE_DISCONNECTED);
12869 +int au_test_nfsd(void)
12872 + struct task_struct *tsk = current;
12873 + char comm[sizeof(tsk->comm)];
12876 + if (tsk->flags & PF_KTHREAD) {
12877 + get_task_comm(comm, tsk);
12878 + ret = !strcmp(comm, "nfsd");
12884 +/* ---------------------------------------------------------------------- */
12885 +/* inode generation external table */
12887 +void au_xigen_inc(struct inode *inode)
12892 + struct super_block *sb;
12893 + struct au_sbinfo *sbinfo;
12895 + sb = inode->i_sb;
12896 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
12898 + sbinfo = au_sbi(sb);
12899 + pos = inode->i_ino;
12900 + pos *= sizeof(igen);
12901 + igen = inode->i_generation + 1;
12902 + sz = xino_fwrite(sbinfo->si_xigen, &igen, sizeof(igen), &pos);
12903 + if (sz == sizeof(igen))
12904 + return; /* success */
12906 + if (unlikely(sz >= 0))
12907 + AuIOErr("xigen error (%zd)\n", sz);
12910 +int au_xigen_new(struct inode *inode)
12915 + struct super_block *sb;
12916 + struct au_sbinfo *sbinfo;
12917 + struct file *file;
12920 + /* todo: dirty, at mount time */
12921 + if (inode->i_ino == AUFS_ROOT_INO)
12923 + sb = inode->i_sb;
12924 + SiMustAnyLock(sb);
12925 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
12929 + pos = inode->i_ino;
12930 + if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
12931 + AuIOErr1("too large i%lld\n", pos);
12934 + pos *= sizeof(inode->i_generation);
12937 + sbinfo = au_sbi(sb);
12938 + file = sbinfo->si_xigen;
12941 + if (vfsub_f_size_read(file)
12942 + < pos + sizeof(inode->i_generation)) {
12943 + inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
12944 + sz = xino_fwrite(file, &inode->i_generation,
12945 + sizeof(inode->i_generation), &pos);
12947 + sz = xino_fread(file, &inode->i_generation,
12948 + sizeof(inode->i_generation), &pos);
12949 + if (sz == sizeof(inode->i_generation))
12950 + goto out; /* success */
12953 + if (unlikely(sz >= 0)) {
12955 + AuIOErr("xigen error (%zd)\n", sz);
12962 +int au_xigen_set(struct super_block *sb, struct path *path)
12965 + struct au_sbinfo *sbinfo;
12966 + struct file *file;
12968 + SiMustWriteLock(sb);
12970 + sbinfo = au_sbi(sb);
12971 + file = au_xino_create2(sb, path, sbinfo->si_xigen);
12972 + err = PTR_ERR(file);
12973 + if (IS_ERR(file))
12976 + if (sbinfo->si_xigen)
12977 + fput(sbinfo->si_xigen);
12978 + sbinfo->si_xigen = file;
12985 +void au_xigen_clr(struct super_block *sb)
12987 + struct au_sbinfo *sbinfo;
12989 + SiMustWriteLock(sb);
12991 + sbinfo = au_sbi(sb);
12992 + if (sbinfo->si_xigen) {
12993 + fput(sbinfo->si_xigen);
12994 + sbinfo->si_xigen = NULL;
12998 +/* ---------------------------------------------------------------------- */
13000 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
13003 + struct dentry *dentry, *d;
13004 + struct inode *inode;
13005 + unsigned int sigen;
13008 + inode = ilookup(sb, ino);
13012 + dentry = ERR_PTR(-ESTALE);
13013 + sigen = au_sigen(sb);
13014 + if (unlikely(au_is_bad_inode(inode)
13015 + || IS_DEADDIR(inode)
13016 + || sigen != au_iigen(inode, NULL)))
13020 + if (!dir_ino || S_ISDIR(inode->i_mode))
13021 + dentry = d_find_alias(inode);
13023 + spin_lock(&inode->i_lock);
13024 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
13025 + spin_lock(&d->d_lock);
13026 + if (!au_test_anon(d)
13027 + && d_inode(d->d_parent)->i_ino == dir_ino) {
13028 + dentry = dget_dlock(d);
13029 + spin_unlock(&d->d_lock);
13032 + spin_unlock(&d->d_lock);
13034 + spin_unlock(&inode->i_lock);
13036 + if (unlikely(dentry && au_digen_test(dentry, sigen))) {
13037 + /* need to refresh */
13045 + AuTraceErrPtr(dentry);
13049 +/* ---------------------------------------------------------------------- */
13051 +/* todo: dirty? */
13052 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
13054 +struct au_compare_mnt_args {
13056 + struct super_block *sb;
13059 + struct vfsmount *mnt;
13062 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
13064 + struct au_compare_mnt_args *a = arg;
13066 + if (mnt->mnt_sb != a->sb)
13068 + a->mnt = mntget(mnt);
13072 +static struct vfsmount *au_mnt_get(struct super_block *sb)
13075 + struct path root;
13076 + struct au_compare_mnt_args args = {
13080 + get_fs_root(current->fs, &root);
13082 + err = iterate_mounts(au_compare_mnt, &args, root.mnt);
13083 + rcu_read_unlock();
13086 + AuDebugOn(!args.mnt);
13090 +struct au_nfsd_si_lock {
13091 + unsigned int sigen;
13092 + aufs_bindex_t bindex, br_id;
13093 + unsigned char force_lock;
13096 +static int si_nfsd_read_lock(struct super_block *sb,
13097 + struct au_nfsd_si_lock *nsi_lock)
13100 + aufs_bindex_t bindex;
13102 + si_read_lock(sb, AuLock_FLUSH);
13104 + /* branch id may be wrapped around */
13106 + bindex = au_br_index(sb, nsi_lock->br_id);
13107 + if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
13108 + goto out; /* success */
13112 + if (!nsi_lock->force_lock)
13113 + si_read_unlock(sb);
13116 + nsi_lock->bindex = bindex;
13120 +struct find_name_by_ino {
13121 + struct dir_context ctx;
13122 + int called, found;
13129 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
13130 + loff_t offset, u64 ino, unsigned int d_type)
13132 + struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
13136 + if (a->ino != ino)
13139 + memcpy(a->name, name, namelen);
13140 + a->namelen = namelen;
13145 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
13146 + struct au_nfsd_si_lock *nsi_lock)
13148 + struct dentry *dentry, *parent;
13149 + struct file *file;
13150 + struct inode *dir;
13151 + struct find_name_by_ino arg = {
13153 + .actor = find_name_by_ino
13158 + parent = path->dentry;
13160 + si_read_unlock(parent->d_sb);
13161 + file = vfsub_dentry_open(path, au_dir_roflags);
13162 + dentry = (void *)file;
13163 + if (IS_ERR(file))
13166 + dentry = ERR_PTR(-ENOMEM);
13167 + arg.name = (void *)__get_free_page(GFP_NOFS);
13168 + if (unlikely(!arg.name))
13175 + err = vfsub_iterate_dir(file, &arg.ctx);
13176 + } while (!err && !arg.found && arg.called);
13177 + dentry = ERR_PTR(err);
13178 + if (unlikely(err))
13180 + /* instead of ENOENT */
13181 + dentry = ERR_PTR(-ESTALE);
13185 + /* do not call vfsub_lkup_one() */
13186 + dir = d_inode(parent);
13187 + dentry = vfsub_lookup_one_len_unlocked(arg.name, path, arg.namelen);
13188 + AuTraceErrPtr(dentry);
13189 + if (IS_ERR(dentry))
13191 + AuDebugOn(au_test_anon(dentry));
13192 + if (unlikely(d_really_is_negative(dentry))) {
13194 + dentry = ERR_PTR(-ENOENT);
13198 + free_page((unsigned long)arg.name);
13202 + if (unlikely(nsi_lock
13203 + && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
13204 + if (!IS_ERR(dentry)) {
13206 + dentry = ERR_PTR(-ESTALE);
13208 + AuTraceErrPtr(dentry);
13212 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
13214 + struct au_nfsd_si_lock *nsi_lock)
13216 + struct dentry *dentry;
13217 + struct path path;
13219 + if (dir_ino != AUFS_ROOT_INO) {
13220 + path.dentry = decode_by_ino(sb, dir_ino, 0);
13221 + dentry = path.dentry;
13222 + if (!path.dentry || IS_ERR(path.dentry))
13224 + AuDebugOn(au_test_anon(path.dentry));
13226 + path.dentry = dget(sb->s_root);
13228 + path.mnt = au_mnt_get(sb);
13229 + dentry = au_lkup_by_ino(&path, ino, nsi_lock);
13233 + AuTraceErrPtr(dentry);
13237 +/* ---------------------------------------------------------------------- */
13239 +static int h_acceptable(void *expv, struct dentry *dentry)
13244 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
13245 + char *buf, int len, struct super_block *sb)
13249 + struct path path;
13251 + p = d_path(h_rootpath, buf, len);
13256 + path.mnt = h_rootpath->mnt;
13257 + path.dentry = h_parent;
13258 + p = d_path(&path, buf, len);
13264 + path.mnt = au_mnt_get(sb);
13265 + path.dentry = sb->s_root;
13266 + p = d_path(&path, buf, len - strlen(p));
13267 + mntput(path.mnt);
13271 + p[strlen(p)] = '/';
13274 + AuTraceErrPtr(p);
13279 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
13280 + int fh_len, struct au_nfsd_si_lock *nsi_lock)
13282 + struct dentry *dentry, *h_parent, *root;
13283 + struct super_block *h_sb;
13284 + char *pathname, *p;
13285 + struct vfsmount *h_mnt;
13286 + struct au_branch *br;
13288 + struct path path;
13290 + br = au_sbr(sb, nsi_lock->bindex);
13291 + h_mnt = au_br_mnt(br);
13292 + h_sb = h_mnt->mnt_sb;
13293 + /* todo: call lower fh_to_dentry()? fh_to_parent()? */
13295 + h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
13296 + fh_len - Fh_tail, fh[Fh_h_type],
13297 + h_acceptable, /*context*/NULL);
13299 + dentry = h_parent;
13300 + if (unlikely(!h_parent || IS_ERR(h_parent))) {
13301 + AuWarn1("%s decode_fh failed, %ld\n",
13302 + au_sbtype(h_sb), PTR_ERR(h_parent));
13306 + if (unlikely(au_test_anon(h_parent))) {
13307 + AuWarn1("%s decode_fh returned a disconnected dentry\n",
13308 + au_sbtype(h_sb));
13309 + goto out_h_parent;
13312 + dentry = ERR_PTR(-ENOMEM);
13313 + pathname = (void *)__get_free_page(GFP_NOFS);
13314 + if (unlikely(!pathname))
13315 + goto out_h_parent;
13317 + root = sb->s_root;
13318 + path.mnt = h_mnt;
13319 + di_read_lock_parent(root, !AuLock_IR);
13320 + path.dentry = au_h_dptr(root, nsi_lock->bindex);
13321 + di_read_unlock(root, !AuLock_IR);
13322 + p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
13323 + dentry = (void *)p;
13325 + goto out_pathname;
13327 + si_read_unlock(sb);
13328 + err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
13329 + dentry = ERR_PTR(err);
13330 + if (unlikely(err))
13333 + dentry = ERR_PTR(-ENOENT);
13334 + AuDebugOn(au_test_anon(path.dentry));
13335 + if (unlikely(d_really_is_negative(path.dentry)))
13338 + if (ino != d_inode(path.dentry)->i_ino)
13339 + dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
13341 + dentry = dget(path.dentry);
13346 + if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
13347 + if (!IS_ERR(dentry)) {
13349 + dentry = ERR_PTR(-ESTALE);
13352 + free_page((unsigned long)pathname);
13356 + AuTraceErrPtr(dentry);
13360 +/* ---------------------------------------------------------------------- */
13362 +static struct dentry *
13363 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
13366 + struct dentry *dentry;
13367 + __u32 *fh = fid->raw;
13368 + struct au_branch *br;
13369 + ino_t ino, dir_ino;
13370 + struct au_nfsd_si_lock nsi_lock = {
13374 + dentry = ERR_PTR(-ESTALE);
13375 + /* it should never happen, but the file handle is unreliable */
13376 + if (unlikely(fh_len < Fh_tail))
13378 + nsi_lock.sigen = fh[Fh_sigen];
13379 + nsi_lock.br_id = fh[Fh_br_id];
13381 + /* branch id may be wrapped around */
13383 + if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
13385 + nsi_lock.force_lock = 1;
13387 + /* is this inode still cached? */
13388 + ino = decode_ino(fh + Fh_ino);
13389 + /* it should never happen */
13390 + if (unlikely(ino == AUFS_ROOT_INO))
13393 + dir_ino = decode_ino(fh + Fh_dir_ino);
13394 + dentry = decode_by_ino(sb, ino, dir_ino);
13395 + if (IS_ERR(dentry))
13400 + /* is the parent dir cached? */
13401 + br = au_sbr(sb, nsi_lock.bindex);
13402 + au_lcnt_inc(&br->br_nfiles);
13403 + dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
13404 + if (IS_ERR(dentry))
13409 + /* lookup path */
13410 + dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
13411 + if (IS_ERR(dentry))
13413 + if (unlikely(!dentry))
13414 + /* todo?: make it ESTALE */
13418 + if (!au_digen_test(dentry, au_sigen(sb))
13419 + && d_inode(dentry)->i_generation == fh[Fh_igen])
13420 + goto out_unlock; /* success */
13423 + dentry = ERR_PTR(-ESTALE);
13426 + au_lcnt_dec(&br->br_nfiles);
13427 + si_read_unlock(sb);
13429 + AuTraceErrPtr(dentry);
13433 +#if 0 /* reserved for future use */
13434 +/* support subtreecheck option */
13435 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
13436 + int fh_len, int fh_type)
13438 + struct dentry *parent;
13439 + __u32 *fh = fid->raw;
13442 + dir_ino = decode_ino(fh + Fh_dir_ino);
13443 + parent = decode_by_ino(sb, dir_ino, 0);
13444 + if (IS_ERR(parent))
13447 + parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
13448 + dir_ino, fh, fh_len);
13451 + AuTraceErrPtr(parent);
13456 +/* ---------------------------------------------------------------------- */
13458 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
13459 + struct inode *dir)
13462 + aufs_bindex_t bindex;
13463 + struct super_block *sb, *h_sb;
13464 + struct dentry *dentry, *parent, *h_parent;
13465 + struct inode *h_dir;
13466 + struct au_branch *br;
13469 + if (unlikely(*max_len <= Fh_tail)) {
13470 + AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
13474 + err = FILEID_ROOT;
13475 + if (inode->i_ino == AUFS_ROOT_INO) {
13476 + AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
13481 + sb = inode->i_sb;
13482 + err = si_read_lock(sb, AuLock_FLUSH);
13483 + if (unlikely(err))
13486 +#ifdef CONFIG_AUFS_DEBUG
13487 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
13488 + AuWarn1("NFS-exporting requires xino\n");
13492 + ii_read_lock_child(inode);
13493 + bindex = au_ibtop(inode);
13495 + dentry = d_find_any_alias(inode);
13496 + if (unlikely(!dentry))
13498 + AuDebugOn(au_test_anon(dentry));
13499 + parent = dget_parent(dentry);
13501 + if (unlikely(!parent))
13503 + if (d_really_is_positive(parent))
13504 + dir = d_inode(parent);
13507 + ii_read_lock_parent(dir);
13508 + h_dir = au_h_iptr(dir, bindex);
13509 + ii_read_unlock(dir);
13510 + if (unlikely(!h_dir))
13512 + h_parent = d_find_any_alias(h_dir);
13513 + if (unlikely(!h_parent))
13514 + goto out_hparent;
13517 + br = au_sbr(sb, bindex);
13518 + h_sb = au_br_sb(br);
13519 + if (unlikely(!h_sb->s_export_op)) {
13520 + AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
13521 + goto out_hparent;
13524 + fh[Fh_br_id] = br->br_id;
13525 + fh[Fh_sigen] = au_sigen(sb);
13526 + encode_ino(fh + Fh_ino, inode->i_ino);
13527 + encode_ino(fh + Fh_dir_ino, dir->i_ino);
13528 + fh[Fh_igen] = inode->i_generation;
13530 + *max_len -= Fh_tail;
13531 + fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
13533 + /*connectable or subtreecheck*/0);
13534 + err = fh[Fh_h_type];
13535 + *max_len += Fh_tail;
13536 + /* todo: macros? */
13537 + if (err != FILEID_INVALID)
13540 + AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
13547 + ii_read_unlock(inode);
13548 + si_read_unlock(sb);
13550 + if (unlikely(err < 0))
13551 + err = FILEID_INVALID;
13555 +/* ---------------------------------------------------------------------- */
13557 +static int aufs_commit_metadata(struct inode *inode)
13560 + aufs_bindex_t bindex;
13561 + struct super_block *sb;
13562 + struct inode *h_inode;
13563 + int (*f)(struct inode *inode);
13565 + sb = inode->i_sb;
13566 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13567 + ii_write_lock_child(inode);
13568 + bindex = au_ibtop(inode);
13569 + AuDebugOn(bindex < 0);
13570 + h_inode = au_h_iptr(inode, bindex);
13572 + f = h_inode->i_sb->s_export_op->commit_metadata;
13574 + err = f(h_inode);
13576 + err = sync_inode_metadata(h_inode, /*wait*/1);
13578 + au_cpup_attr_timesizes(inode);
13579 + ii_write_unlock(inode);
13580 + si_read_unlock(sb);
13584 +/* ---------------------------------------------------------------------- */
13586 +static struct export_operations aufs_export_op = {
13587 + .fh_to_dentry = aufs_fh_to_dentry,
13588 + /* .fh_to_parent = aufs_fh_to_parent, */
13589 + .encode_fh = aufs_encode_fh,
13590 + .commit_metadata = aufs_commit_metadata
13593 +void au_export_init(struct super_block *sb)
13595 + struct au_sbinfo *sbinfo;
13598 + BUILD_BUG_ON_MSG(IS_BUILTIN(CONFIG_AUFS_FS)
13599 + && IS_MODULE(CONFIG_EXPORTFS),
13600 + AUFS_NAME ": unsupported configuration "
13601 + "CONFIG_EXPORTFS=m and CONFIG_AUFS_FS=y");
13603 + sb->s_export_op = &aufs_export_op;
13604 + sbinfo = au_sbi(sb);
13605 + sbinfo->si_xigen = NULL;
13606 + get_random_bytes(&u, sizeof(u));
13607 + BUILD_BUG_ON(sizeof(u) != sizeof(int));
13608 + atomic_set(&sbinfo->si_xigen_next, u);
13610 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
13611 --- /usr/share/empty/fs/aufs/fhsm.c 1970-01-01 01:00:00.000000000 +0100
13612 +++ linux/fs/aufs/fhsm.c 2022-11-05 23:02:18.962555950 +0100
13614 +// SPDX-License-Identifier: GPL-2.0
13616 + * Copyright (C) 2011-2022 Junjiro R. Okajima
13618 + * This program is free software; you can redistribute it and/or modify
13619 + * it under the terms of the GNU General Public License as published by
13620 + * the Free Software Foundation; either version 2 of the License, or
13621 + * (at your option) any later version.
13623 + * This program is distributed in the hope that it will be useful,
13624 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13625 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13626 + * GNU General Public License for more details.
13628 + * You should have received a copy of the GNU General Public License
13629 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
13633 + * File-based Hierarchy Storage Management
13636 +#include <linux/anon_inodes.h>
13637 +#include <linux/poll.h>
13638 +#include <linux/seq_file.h>
13639 +#include <linux/statfs.h>
13642 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
13644 + struct au_sbinfo *sbinfo;
13645 + struct au_fhsm *fhsm;
13647 + SiMustAnyLock(sb);
13649 + sbinfo = au_sbi(sb);
13650 + fhsm = &sbinfo->si_fhsm;
13651 + AuDebugOn(!fhsm);
13652 + return fhsm->fhsm_bottom;
13655 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
13657 + struct au_sbinfo *sbinfo;
13658 + struct au_fhsm *fhsm;
13660 + SiMustWriteLock(sb);
13662 + sbinfo = au_sbi(sb);
13663 + fhsm = &sbinfo->si_fhsm;
13664 + AuDebugOn(!fhsm);
13665 + fhsm->fhsm_bottom = bindex;
13668 +/* ---------------------------------------------------------------------- */
13670 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
13672 + struct au_br_fhsm *bf;
13674 + bf = br->br_fhsm;
13675 + MtxMustLock(&bf->bf_lock);
13677 + return !bf->bf_readable
13678 + || time_after(jiffies,
13679 + bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
13682 +/* ---------------------------------------------------------------------- */
13684 +static void au_fhsm_notify(struct super_block *sb, int val)
13686 + struct au_sbinfo *sbinfo;
13687 + struct au_fhsm *fhsm;
13689 + SiMustAnyLock(sb);
13691 + sbinfo = au_sbi(sb);
13692 + fhsm = &sbinfo->si_fhsm;
13693 + if (au_fhsm_pid(fhsm)
13694 + && atomic_read(&fhsm->fhsm_readable) != -1) {
13695 + atomic_set(&fhsm->fhsm_readable, val);
13697 + wake_up(&fhsm->fhsm_wqh);
13701 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
13702 + struct aufs_stfs *rstfs, int do_lock, int do_notify)
13705 + struct au_branch *br;
13706 + struct au_br_fhsm *bf;
13708 + br = au_sbr(sb, bindex);
13709 + AuDebugOn(au_br_rdonly(br));
13710 + bf = br->br_fhsm;
13714 + mutex_lock(&bf->bf_lock);
13716 + MtxMustLock(&bf->bf_lock);
13718 + /* sb->s_root for NFS is unreliable */
13719 + err = au_br_stfs(br, &bf->bf_stfs);
13720 + if (unlikely(err)) {
13721 + AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
13725 + bf->bf_jiffy = jiffies;
13726 + bf->bf_readable = 1;
13728 + au_fhsm_notify(sb, /*val*/1);
13730 + *rstfs = bf->bf_stfs;
13734 + mutex_unlock(&bf->bf_lock);
13735 + au_fhsm_notify(sb, /*val*/1);
13740 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
13743 + struct au_sbinfo *sbinfo;
13744 + struct au_fhsm *fhsm;
13745 + struct au_branch *br;
13746 + struct au_br_fhsm *bf;
13748 + AuDbg("b%d, force %d\n", bindex, force);
13749 + SiMustAnyLock(sb);
13751 + sbinfo = au_sbi(sb);
13752 + fhsm = &sbinfo->si_fhsm;
13753 + if (!au_ftest_si(sbinfo, FHSM)
13754 + || fhsm->fhsm_bottom == bindex)
13757 + br = au_sbr(sb, bindex);
13758 + bf = br->br_fhsm;
13760 + mutex_lock(&bf->bf_lock);
13762 + || au_fhsm_pid(fhsm)
13763 + || au_fhsm_test_jiffy(sbinfo, br))
13764 + err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
13766 + mutex_unlock(&bf->bf_lock);
13769 +void au_fhsm_wrote_all(struct super_block *sb, int force)
13771 + aufs_bindex_t bindex, bbot;
13772 + struct au_branch *br;
13774 + /* exclude the bottom */
13775 + bbot = au_fhsm_bottom(sb);
13776 + for (bindex = 0; bindex < bbot; bindex++) {
13777 + br = au_sbr(sb, bindex);
13778 + if (au_br_fhsm(br->br_perm))
13779 + au_fhsm_wrote(sb, bindex, force);
13783 +/* ---------------------------------------------------------------------- */
13785 +static __poll_t au_fhsm_poll(struct file *file, struct poll_table_struct *wait)
13788 + struct au_sbinfo *sbinfo;
13789 + struct au_fhsm *fhsm;
13792 + sbinfo = file->private_data;
13793 + fhsm = &sbinfo->si_fhsm;
13794 + poll_wait(file, &fhsm->fhsm_wqh, wait);
13795 + if (atomic_read(&fhsm->fhsm_readable))
13796 + mask = EPOLLIN /* | EPOLLRDNORM */;
13799 + AuDbg("mask 0x%x\n", mask);
13803 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
13804 + struct aufs_stfs *stfs, __s16 brid)
13808 + err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
13810 + err = __put_user(brid, &stbr->brid);
13811 + if (unlikely(err))
13817 +static ssize_t au_fhsm_do_read(struct super_block *sb,
13818 + struct aufs_stbr __user *stbr, size_t count)
13822 + aufs_bindex_t bindex, bbot;
13823 + struct au_branch *br;
13824 + struct au_br_fhsm *bf;
13826 + /* except the bottom branch */
13829 + bbot = au_fhsm_bottom(sb);
13830 + for (bindex = 0; !err && bindex < bbot; bindex++) {
13831 + br = au_sbr(sb, bindex);
13832 + if (!au_br_fhsm(br->br_perm))
13835 + bf = br->br_fhsm;
13836 + mutex_lock(&bf->bf_lock);
13837 + if (bf->bf_readable) {
13839 + if (count >= sizeof(*stbr))
13840 + err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
13843 + bf->bf_readable = 0;
13844 + count -= sizeof(*stbr);
13848 + mutex_unlock(&bf->bf_lock);
13851 + err = sizeof(*stbr) * nstbr;
13856 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
13861 + aufs_bindex_t nfhsm, bindex, bbot;
13862 + struct au_sbinfo *sbinfo;
13863 + struct au_fhsm *fhsm;
13864 + struct au_branch *br;
13865 + struct super_block *sb;
13868 + sbinfo = file->private_data;
13869 + fhsm = &sbinfo->si_fhsm;
13871 + spin_lock_irq(&fhsm->fhsm_wqh.lock);
13872 + if (!atomic_read(&fhsm->fhsm_readable)) {
13873 + if (vfsub_file_flags(file) & O_NONBLOCK)
13876 + err = wait_event_interruptible_locked_irq
13878 + atomic_read(&fhsm->fhsm_readable));
13880 + spin_unlock_irq(&fhsm->fhsm_wqh.lock);
13881 + if (unlikely(err))
13884 + /* sb may already be dead */
13885 + au_rw_read_lock(&sbinfo->si_rwsem);
13886 + readable = atomic_read(&fhsm->fhsm_readable);
13887 + if (readable > 0) {
13888 + sb = sbinfo->si_sb;
13890 + /* exclude the bottom branch */
13892 + bbot = au_fhsm_bottom(sb);
13893 + for (bindex = 0; bindex < bbot; bindex++) {
13894 + br = au_sbr(sb, bindex);
13895 + if (au_br_fhsm(br->br_perm))
13899 + if (nfhsm * sizeof(struct aufs_stbr) <= count) {
13900 + atomic_set(&fhsm->fhsm_readable, 0);
13901 + err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
13905 + au_rw_read_unlock(&sbinfo->si_rwsem);
13913 +static int au_fhsm_release(struct inode *inode, struct file *file)
13915 + struct au_sbinfo *sbinfo;
13916 + struct au_fhsm *fhsm;
13918 + /* sb may already be dead */
13919 + sbinfo = file->private_data;
13920 + fhsm = &sbinfo->si_fhsm;
13921 + spin_lock(&fhsm->fhsm_spin);
13922 + fhsm->fhsm_pid = 0;
13923 + spin_unlock(&fhsm->fhsm_spin);
13924 + kobject_put(&sbinfo->si_kobj);
13929 +static const struct file_operations au_fhsm_fops = {
13930 + .owner = THIS_MODULE,
13931 + .llseek = noop_llseek,
13932 + .read = au_fhsm_read,
13933 + .poll = au_fhsm_poll,
13934 + .release = au_fhsm_release
13937 +int au_fhsm_fd(struct super_block *sb, int oflags)
13940 + struct au_sbinfo *sbinfo;
13941 + struct au_fhsm *fhsm;
13944 + if (unlikely(!capable(CAP_SYS_ADMIN)))
13948 + if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
13952 + sbinfo = au_sbi(sb);
13953 + fhsm = &sbinfo->si_fhsm;
13954 + spin_lock(&fhsm->fhsm_spin);
13955 + if (!fhsm->fhsm_pid)
13956 + fhsm->fhsm_pid = current->pid;
13959 + spin_unlock(&fhsm->fhsm_spin);
13960 + if (unlikely(err))
13963 + oflags |= O_RDONLY;
13964 + /* oflags |= FMODE_NONOTIFY; */
13965 + fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
13967 + if (unlikely(fd < 0))
13970 + /* succeed regardless 'fhsm' status */
13971 + kobject_get(&sbinfo->si_kobj);
13972 + si_noflush_read_lock(sb);
13973 + if (au_ftest_si(sbinfo, FHSM))
13974 + au_fhsm_wrote_all(sb, /*force*/0);
13975 + si_read_unlock(sb);
13976 + goto out; /* success */
13979 + spin_lock(&fhsm->fhsm_spin);
13980 + fhsm->fhsm_pid = 0;
13981 + spin_unlock(&fhsm->fhsm_spin);
13987 +/* ---------------------------------------------------------------------- */
13989 +int au_fhsm_br_alloc(struct au_branch *br)
13994 + br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
13996 + au_br_fhsm_init(br->br_fhsm);
14003 +/* ---------------------------------------------------------------------- */
14005 +void au_fhsm_fin(struct super_block *sb)
14007 + au_fhsm_notify(sb, /*val*/-1);
14010 +void au_fhsm_init(struct au_sbinfo *sbinfo)
14012 + struct au_fhsm *fhsm;
14014 + fhsm = &sbinfo->si_fhsm;
14015 + spin_lock_init(&fhsm->fhsm_spin);
14016 + init_waitqueue_head(&fhsm->fhsm_wqh);
14017 + atomic_set(&fhsm->fhsm_readable, 0);
14018 + fhsm->fhsm_expire
14019 + = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
14020 + fhsm->fhsm_bottom = -1;
14023 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
14025 + sbinfo->si_fhsm.fhsm_expire
14026 + = msecs_to_jiffies(sec * MSEC_PER_SEC);
14029 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
14033 + if (!au_ftest_si(sbinfo, FHSM))
14036 + u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
14037 + if (u != AUFS_FHSM_CACHE_DEF_SEC)
14038 + seq_printf(seq, ",fhsm_sec=%u", u);
14040 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
14041 --- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
14042 +++ linux/fs/aufs/file.c 2022-12-17 09:21:34.799855195 +0100
14044 +// SPDX-License-Identifier: GPL-2.0
14046 + * Copyright (C) 2005-2022 Junjiro R. Okajima
14048 + * This program is free software; you can redistribute it and/or modify
14049 + * it under the terms of the GNU General Public License as published by
14050 + * the Free Software Foundation; either version 2 of the License, or
14051 + * (at your option) any later version.
14053 + * This program is distributed in the hope that it will be useful,
14054 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14055 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14056 + * GNU General Public License for more details.
14058 + * You should have received a copy of the GNU General Public License
14059 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14063 + * handling file/dir, and address_space operation
14066 +#ifdef CONFIG_AUFS_DEBUG
14067 +#include <linux/migrate.h>
14069 +#include <linux/pagemap.h>
14072 +/* drop flags for writing */
14073 +unsigned int au_file_roflags(unsigned int flags)
14075 + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
14076 + flags |= O_RDONLY | O_NOATIME;
14080 +/* common functions to regular file and dir */
14081 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14082 + struct file *file, int force_wr)
14084 + struct file *h_file;
14085 + struct dentry *h_dentry;
14086 + struct inode *h_inode;
14087 + struct super_block *sb;
14088 + struct au_branch *br;
14089 + struct path h_path;
14092 + /* a race condition can happen between open and unlink/rmdir */
14093 + h_file = ERR_PTR(-ENOENT);
14094 + h_dentry = au_h_dptr(dentry, bindex);
14095 + if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
14097 + h_inode = d_inode(h_dentry);
14098 + spin_lock(&h_dentry->d_lock);
14099 + err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
14100 + /* || !d_inode(dentry)->i_nlink */
14102 + spin_unlock(&h_dentry->d_lock);
14103 + if (unlikely(err))
14106 + sb = dentry->d_sb;
14107 + br = au_sbr(sb, bindex);
14108 + err = au_br_test_oflag(flags, br);
14109 + h_file = ERR_PTR(err);
14110 + if (unlikely(err))
14113 + /* drop flags for writing */
14114 + if (au_test_ro(sb, bindex, d_inode(dentry))) {
14115 + if (force_wr && !(flags & O_WRONLY))
14117 + flags = au_file_roflags(flags);
14119 + h_file = ERR_PTR(-EROFS);
14120 + flags = au_file_roflags(flags);
14121 + if (unlikely(vfsub_native_ro(h_inode)
14122 + || IS_APPEND(h_inode)))
14124 + flags &= ~O_ACCMODE;
14125 + flags |= O_WRONLY;
14128 + flags &= ~O_CREAT;
14129 + au_lcnt_inc(&br->br_nfiles);
14130 + h_path.dentry = h_dentry;
14131 + h_path.mnt = au_br_mnt(br);
14132 + h_file = vfsub_dentry_open(&h_path, flags);
14133 + if (IS_ERR(h_file))
14136 + if (flags & __FMODE_EXEC) {
14137 + err = deny_write_access(h_file);
14138 + if (unlikely(err)) {
14140 + h_file = ERR_PTR(err);
14144 + fsnotify_open(h_file);
14145 + goto out; /* success */
14148 + au_lcnt_dec(&br->br_nfiles);
14153 +static int au_cmoo(struct dentry *dentry)
14155 + int err, cmoo, matched;
14156 + unsigned int udba;
14157 + struct path h_path;
14158 + struct au_pin pin;
14159 + struct au_cp_generic cpg = {
14160 + .dentry = dentry,
14165 + .flags = AuCpup_DTIME | AuCpup_HOPEN
14167 + struct inode *delegated;
14168 + struct super_block *sb;
14169 + struct au_sbinfo *sbinfo;
14170 + struct au_fhsm *fhsm;
14172 + struct au_branch *br;
14173 + struct dentry *parent;
14174 + struct au_hinode *hdir;
14176 + DiMustWriteLock(dentry);
14177 + IiMustWriteLock(d_inode(dentry));
14180 + if (IS_ROOT(dentry))
14182 + cpg.bsrc = au_dbtop(dentry);
14186 + sb = dentry->d_sb;
14187 + sbinfo = au_sbi(sb);
14188 + fhsm = &sbinfo->si_fhsm;
14189 + pid = au_fhsm_pid(fhsm);
14192 + && (current->pid == pid
14193 + || rcu_dereference(current->real_parent)->pid == pid));
14194 + rcu_read_unlock();
14198 + br = au_sbr(sb, cpg.bsrc);
14199 + cmoo = au_br_cmoo(br->br_perm);
14202 + if (!d_is_reg(dentry))
14203 + cmoo &= AuBrAttr_COO_ALL;
14207 + parent = dget_parent(dentry);
14208 + di_write_lock_parent(parent);
14209 + err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
14211 + if (unlikely(err < 0)) {
14212 + err = 0; /* there is no upper writable branch */
14215 + AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
14217 + /* do not respect the coo attrib for the target branch */
14218 + err = au_cpup_dirs(dentry, cpg.bdst);
14219 + if (unlikely(err))
14222 + di_downgrade_lock(parent, AuLock_IR);
14223 + udba = au_opt_udba(sb);
14224 + err = au_pin(&pin, dentry, cpg.bdst, udba,
14225 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14226 + if (unlikely(err))
14229 + err = au_sio_cpup_simple(&cpg);
14231 + if (unlikely(err))
14233 + if (!(cmoo & AuBrWAttr_MOO))
14234 + goto out_parent; /* success */
14236 + err = au_pin(&pin, dentry, cpg.bsrc, udba,
14237 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14238 + if (unlikely(err))
14241 + h_path.mnt = au_br_mnt(br);
14242 + h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
14243 + hdir = au_hi(d_inode(parent), cpg.bsrc);
14244 + delegated = NULL;
14245 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
14247 + /* todo: keep h_dentry or not? */
14248 + if (unlikely(err == -EWOULDBLOCK)) {
14249 + pr_warn("cannot retry for NFSv4 delegation"
14250 + " for an internal unlink\n");
14253 + if (unlikely(err)) {
14254 + pr_err("unlink %pd after coo failed (%d), ignored\n",
14258 + goto out_parent; /* success */
14261 + di_downgrade_lock(parent, AuLock_IR);
14263 + di_read_unlock(parent, AuLock_IR);
14270 +int au_do_open(struct file *file, struct au_do_open_args *args)
14272 + int err, aopen = args->aopen;
14273 + struct dentry *dentry;
14274 + struct au_finfo *finfo;
14277 + err = au_finfo_init(file, args->fidir);
14280 + err = au_finfo_init(file, args->fidir);
14283 + if (unlikely(err))
14286 + dentry = file->f_path.dentry;
14287 + AuDebugOn(IS_ERR_OR_NULL(dentry));
14288 + di_write_lock_child(dentry);
14289 + err = au_cmoo(dentry);
14292 + err = args->open(file, vfsub_file_flags(file),
14293 + au_di(dentry)->di_htmpfile);
14294 + di_write_unlock(dentry);
14296 + di_downgrade_lock(dentry, AuLock_IR);
14298 + err = args->open(file, vfsub_file_flags(file),
14301 + di_read_unlock(dentry, AuLock_IR);
14305 + finfo = au_fi(file);
14307 + finfo->fi_file = file;
14308 + au_hbl_add(&finfo->fi_hlist,
14309 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
14312 + fi_write_unlock(file);
14315 + fi_write_unlock(file);
14318 + if (unlikely(err)) {
14319 + finfo->fi_hdir = NULL;
14320 + au_finfo_fin(file);
14328 +int au_reopen_nondir(struct file *file)
14331 + aufs_bindex_t btop;
14332 + struct dentry *dentry;
14333 + struct au_branch *br;
14334 + struct file *h_file, *h_file_tmp;
14336 + dentry = file->f_path.dentry;
14337 + btop = au_dbtop(dentry);
14338 + br = au_sbr(dentry->d_sb, btop);
14339 + h_file_tmp = NULL;
14340 + if (au_fbtop(file) == btop) {
14341 + h_file = au_hf_top(file);
14342 + if (file->f_mode == h_file->f_mode)
14343 + return 0; /* success */
14344 + h_file_tmp = h_file;
14345 + get_file(h_file_tmp);
14346 + au_lcnt_inc(&br->br_nfiles);
14347 + au_set_h_fptr(file, btop, NULL);
14349 + AuDebugOn(au_fi(file)->fi_hdir);
14352 + * file exists on both of rw and ro
14353 + * open --> dbtop and fbtop are both 0
14354 + * prepend a branch as rw, "rw" become ro
14356 + * delete the top branch, "rw" becomes rw again
14357 + * --> dbtop is 1, fbtop is still 0
14358 + * write --> fbtop is 0 but dbtop is 1
14360 + /* AuDebugOn(au_fbtop(file) < btop); */
14362 + h_file = au_h_open(dentry, btop, vfsub_file_flags(file) & ~O_TRUNC,
14363 + file, /*force_wr*/0);
14364 + err = PTR_ERR(h_file);
14365 + if (IS_ERR(h_file)) {
14366 + if (h_file_tmp) {
14368 + au_set_h_fptr(file, btop, h_file_tmp);
14369 + h_file_tmp = NULL;
14371 + goto out; /* todo: close all? */
14375 + au_set_fbtop(file, btop);
14376 + au_set_h_fptr(file, btop, h_file);
14377 + au_update_figen(file);
14378 + /* todo: necessary? */
14379 + /* file->f_ra = h_file->f_ra; */
14382 + if (h_file_tmp) {
14383 + fput(h_file_tmp);
14384 + au_lcnt_dec(&br->br_nfiles);
14389 +/* ---------------------------------------------------------------------- */
14391 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
14392 + struct dentry *hi_wh)
14395 + aufs_bindex_t btop;
14396 + struct au_dinfo *dinfo;
14397 + struct dentry *h_dentry;
14398 + struct au_hdentry *hdp;
14400 + dinfo = au_di(file->f_path.dentry);
14401 + AuRwMustWriteLock(&dinfo->di_rwsem);
14403 + btop = dinfo->di_btop;
14404 + dinfo->di_btop = btgt;
14405 + hdp = au_hdentry(dinfo, btgt);
14406 + h_dentry = hdp->hd_dentry;
14407 + hdp->hd_dentry = hi_wh;
14408 + err = au_reopen_nondir(file);
14409 + hdp->hd_dentry = h_dentry;
14410 + dinfo->di_btop = btop;
14415 +static int au_ready_to_write_wh(struct file *file, loff_t len,
14416 + aufs_bindex_t bcpup, struct au_pin *pin)
14419 + struct inode *inode, *h_inode;
14420 + struct dentry *h_dentry, *hi_wh;
14421 + struct au_cp_generic cpg = {
14422 + .dentry = file->f_path.dentry,
14429 + au_update_dbtop(cpg.dentry);
14430 + inode = d_inode(cpg.dentry);
14432 + if (au_dbtop(cpg.dentry) <= bcpup
14433 + && au_dbbot(cpg.dentry) >= bcpup) {
14434 + h_dentry = au_h_dptr(cpg.dentry, bcpup);
14435 + if (h_dentry && d_is_positive(h_dentry))
14436 + h_inode = d_inode(h_dentry);
14438 + hi_wh = au_hi_wh(inode, bcpup);
14439 + if (!hi_wh && !h_inode)
14440 + err = au_sio_cpup_wh(&cpg, file);
14442 + /* already copied-up after unlink */
14443 + err = au_reopen_wh(file, bcpup, hi_wh);
14446 + && (inode->i_nlink > 1
14447 + || (inode->i_state & I_LINKABLE))
14448 + && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
14449 + au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
14455 + * prepare the @file for writing.
14457 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
14460 + aufs_bindex_t dbtop;
14461 + struct dentry *parent;
14462 + struct inode *inode;
14463 + struct super_block *sb;
14464 + struct file *h_file;
14465 + struct au_cp_generic cpg = {
14466 + .dentry = file->f_path.dentry,
14471 + .flags = AuCpup_DTIME
14474 + sb = cpg.dentry->d_sb;
14475 + inode = d_inode(cpg.dentry);
14476 + cpg.bsrc = au_fbtop(file);
14477 + err = au_test_ro(sb, cpg.bsrc, inode);
14478 + if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
14479 + err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
14484 + /* need to cpup or reopen */
14485 + parent = dget_parent(cpg.dentry);
14486 + di_write_lock_parent(parent);
14487 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14489 + if (unlikely(err < 0))
14493 + if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
14494 + err = au_cpup_dirs(cpg.dentry, cpg.bdst);
14495 + if (unlikely(err))
14499 + err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14500 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14501 + if (unlikely(err))
14504 + dbtop = au_dbtop(cpg.dentry);
14505 + if (dbtop <= cpg.bdst)
14506 + cpg.bsrc = cpg.bdst;
14508 + if (dbtop <= cpg.bdst /* just reopen */
14509 + || !d_unhashed(cpg.dentry) /* copyup and reopen */
14511 + h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
14512 + if (IS_ERR(h_file))
14513 + err = PTR_ERR(h_file);
14515 + di_downgrade_lock(parent, AuLock_IR);
14516 + if (dbtop > cpg.bdst)
14517 + err = au_sio_cpup_simple(&cpg);
14519 + err = au_reopen_nondir(file);
14520 + au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
14522 + } else { /* copyup as wh and reopen */
14524 + * since writable hfsplus branch is not supported,
14525 + * h_open_pre/post() are unnecessary.
14527 + err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
14528 + di_downgrade_lock(parent, AuLock_IR);
14532 + au_pin_set_parent_lflag(pin, /*lflag*/0);
14533 + goto out_dput; /* success */
14539 + di_downgrade_lock(parent, AuLock_IR);
14541 + di_read_unlock(parent, AuLock_IR);
14548 +/* ---------------------------------------------------------------------- */
14550 +int au_do_flush(struct file *file, fl_owner_t id,
14551 + int (*flush)(struct file *file, fl_owner_t id))
14554 + struct super_block *sb;
14555 + struct inode *inode;
14557 + inode = file_inode(file);
14558 + sb = inode->i_sb;
14559 + si_noflush_read_lock(sb);
14560 + fi_read_lock(file);
14561 + ii_read_lock_child(inode);
14563 + err = flush(file, id);
14564 + au_cpup_attr_timesizes(inode);
14566 + ii_read_unlock(inode);
14567 + fi_read_unlock(file);
14568 + si_read_unlock(sb);
14572 +/* ---------------------------------------------------------------------- */
14574 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
14577 + struct au_pin pin;
14578 + struct au_finfo *finfo;
14579 + struct dentry *parent, *hi_wh;
14580 + struct inode *inode;
14581 + struct super_block *sb;
14582 + struct au_cp_generic cpg = {
14583 + .dentry = file->f_path.dentry,
14588 + .flags = AuCpup_DTIME
14591 + FiMustWriteLock(file);
14594 + finfo = au_fi(file);
14595 + sb = cpg.dentry->d_sb;
14596 + inode = d_inode(cpg.dentry);
14597 + cpg.bdst = au_ibtop(inode);
14598 + if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
14601 + parent = dget_parent(cpg.dentry);
14602 + if (au_test_ro(sb, cpg.bdst, inode)) {
14603 + di_read_lock_parent(parent, !AuLock_IR);
14604 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14606 + di_read_unlock(parent, !AuLock_IR);
14607 + if (unlikely(err < 0))
14612 + di_read_lock_parent(parent, AuLock_IR);
14613 + hi_wh = au_hi_wh(inode, cpg.bdst);
14614 + if (!S_ISDIR(inode->i_mode)
14615 + && au_opt_test(au_mntflags(sb), PLINK)
14616 + && au_plink_test(inode)
14617 + && !d_unhashed(cpg.dentry)
14618 + && cpg.bdst < au_dbtop(cpg.dentry)) {
14619 + err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
14620 + if (unlikely(err))
14623 + /* always superio. */
14624 + err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14625 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14627 + err = au_sio_cpup_simple(&cpg);
14630 + } else if (hi_wh) {
14631 + /* already copied-up after unlink */
14632 + err = au_reopen_wh(file, cpg.bdst, hi_wh);
14633 + *need_reopen = 0;
14637 + di_read_unlock(parent, AuLock_IR);
14644 +static void au_do_refresh_dir(struct file *file)
14646 + aufs_bindex_t bindex, bbot, new_bindex, brid;
14647 + struct au_hfile *p, tmp, *q;
14648 + struct au_finfo *finfo;
14649 + struct super_block *sb;
14650 + struct au_fidir *fidir;
14652 + FiMustWriteLock(file);
14654 + sb = file->f_path.dentry->d_sb;
14655 + finfo = au_fi(file);
14656 + fidir = finfo->fi_hdir;
14657 + AuDebugOn(!fidir);
14658 + p = fidir->fd_hfile + finfo->fi_btop;
14659 + brid = p->hf_br->br_id;
14660 + bbot = fidir->fd_bbot;
14661 + for (bindex = finfo->fi_btop; bindex <= bbot; bindex++, p++) {
14665 + new_bindex = au_br_index(sb, p->hf_br->br_id);
14666 + if (new_bindex == bindex)
14668 + if (new_bindex < 0) {
14669 + au_set_h_fptr(file, bindex, NULL);
14673 + /* swap two lower inode, and loop again */
14674 + q = fidir->fd_hfile + new_bindex;
14678 + if (tmp.hf_file) {
14684 + p = fidir->fd_hfile;
14685 + if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
14686 + bbot = au_sbbot(sb);
14687 + for (finfo->fi_btop = 0; finfo->fi_btop <= bbot;
14688 + finfo->fi_btop++, p++)
14689 + if (p->hf_file) {
14690 + if (file_inode(p->hf_file))
14692 + au_hfput(p, /*execed*/0);
14695 + bbot = au_br_index(sb, brid);
14696 + for (finfo->fi_btop = 0; finfo->fi_btop < bbot;
14697 + finfo->fi_btop++, p++)
14699 + au_hfput(p, /*execed*/0);
14700 + bbot = au_sbbot(sb);
14703 + p = fidir->fd_hfile + bbot;
14704 + for (fidir->fd_bbot = bbot; fidir->fd_bbot >= finfo->fi_btop;
14705 + fidir->fd_bbot--, p--)
14706 + if (p->hf_file) {
14707 + if (file_inode(p->hf_file))
14709 + au_hfput(p, /*execed*/0);
14711 + AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
14715 + * after branch manipulating, refresh the file.
14717 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
14719 + int err, need_reopen, nbr;
14720 + aufs_bindex_t bbot, bindex;
14721 + struct dentry *dentry;
14722 + struct super_block *sb;
14723 + struct au_finfo *finfo;
14724 + struct au_hfile *hfile;
14726 + dentry = file->f_path.dentry;
14727 + sb = dentry->d_sb;
14728 + nbr = au_sbbot(sb) + 1;
14729 + finfo = au_fi(file);
14730 + if (!finfo->fi_hdir) {
14731 + hfile = &finfo->fi_htop;
14732 + AuDebugOn(!hfile->hf_file);
14733 + bindex = au_br_index(sb, hfile->hf_br->br_id);
14734 + AuDebugOn(bindex < 0);
14735 + if (bindex != finfo->fi_btop)
14736 + au_set_fbtop(file, bindex);
14738 + err = au_fidir_realloc(finfo, nbr, /*may_shrink*/0);
14739 + if (unlikely(err))
14741 + au_do_refresh_dir(file);
14746 + if (!au_test_mmapped(file))
14747 + err = au_file_refresh_by_inode(file, &need_reopen);
14748 + if (finfo->fi_hdir)
14749 + /* harmless if err */
14750 + au_fidir_realloc(finfo, nbr, /*may_shrink*/1);
14751 + if (!err && need_reopen && !d_unlinked(dentry))
14752 + err = reopen(file);
14754 + au_update_figen(file);
14755 + goto out; /* success */
14758 + /* error, close all lower files */
14759 + if (finfo->fi_hdir) {
14760 + bbot = au_fbbot_dir(file);
14761 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++)
14762 + au_set_h_fptr(file, bindex, NULL);
14769 +/* common function to regular file and dir */
14770 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14771 + int wlock, unsigned int fi_lsc)
14774 + unsigned int sigen, figen;
14775 + aufs_bindex_t btop;
14776 + unsigned char pseudo_link;
14777 + struct dentry *dentry;
14778 + struct inode *inode;
14781 + dentry = file->f_path.dentry;
14782 + inode = d_inode(dentry);
14783 + sigen = au_sigen(dentry->d_sb);
14784 + fi_write_lock_nested(file, fi_lsc);
14785 + figen = au_figen(file);
14787 + di_write_lock_child(dentry);
14789 + di_write_lock_child2(dentry);
14790 + btop = au_dbtop(dentry);
14791 + pseudo_link = (btop != au_ibtop(inode));
14792 + if (sigen == figen && !pseudo_link && au_fbtop(file) == btop) {
14794 + di_downgrade_lock(dentry, AuLock_IR);
14795 + fi_downgrade_lock(file);
14797 + goto out; /* success */
14800 + AuDbg("sigen %d, figen %d\n", sigen, figen);
14801 + if (au_digen_test(dentry, sigen)) {
14802 + err = au_reval_dpath(dentry, sigen);
14803 + AuDebugOn(!err && au_digen_test(dentry, sigen));
14807 + err = refresh_file(file, reopen);
14810 + di_downgrade_lock(dentry, AuLock_IR);
14811 + fi_downgrade_lock(file);
14814 + di_write_unlock(dentry);
14815 + fi_write_unlock(file);
14822 +/* ---------------------------------------------------------------------- */
14824 +/* cf. aufs_nopage() */
14825 +/* for madvise(2) */
14826 +static int aufs_read_folio(struct file *file __maybe_unused, struct folio *folio)
14828 + folio_unlock(folio);
14832 +/* it will never be called, but necessary to support O_DIRECT */
14833 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
14834 +{ BUG(); return 0; }
14836 +/* they will never be called. */
14837 +#ifdef CONFIG_AUFS_DEBUG
14838 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
14839 + loff_t pos, unsigned len,
14840 + struct page **pagep, void **fsdata)
14841 +{ AuUnsupport(); return 0; }
14842 +static int aufs_write_end(struct file *file, struct address_space *mapping,
14843 + loff_t pos, unsigned len, unsigned copied,
14844 + struct page *page, void *fsdata)
14845 +{ AuUnsupport(); return 0; }
14846 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
14847 +{ AuUnsupport(); return 0; }
14849 +static bool aufs_dirty_folio(struct address_space *mapping, struct folio *folio)
14850 +{ AuUnsupport(); return true; }
14851 +static void aufs_invalidate_folio(struct folio *folio, size_t offset, size_t len)
14852 +{ AuUnsupport(); }
14853 +static bool aufs_release_folio(struct folio *folio, gfp_t gfp)
14854 +{ AuUnsupport(); return true; }
14855 +#if 0 /* called by memory compaction regardless file */
14856 +static int aufs_migrate_folio(struct address_space *mapping, struct folio *dst,
14857 + struct folio *src, enum migrate_mode mode)
14858 +{ AuUnsupport(); return 0; }
14860 +static int aufs_launder_folio(struct folio *folio)
14861 +{ AuUnsupport(); return 0; }
14862 +static bool aufs_is_partially_uptodate(struct folio *folio, size_t from,
14864 +{ AuUnsupport(); return true; }
14865 +static void aufs_is_dirty_writeback(struct folio *folio, bool *dirty,
14867 +{ AuUnsupport(); }
14868 +static int aufs_error_remove_page(struct address_space *mapping,
14869 + struct page *page)
14870 +{ AuUnsupport(); return 0; }
14871 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
14873 +{ AuUnsupport(); return 0; }
14874 +static void aufs_swap_deactivate(struct file *file)
14875 +{ AuUnsupport(); }
14876 +static int aufs_swap_rw(struct kiocb *iocb, struct iov_iter *iter)
14877 +{ AuUnsupport(); return 0; }
14878 +#endif /* CONFIG_AUFS_DEBUG */
14880 +const struct address_space_operations aufs_aop = {
14881 + .read_folio = aufs_read_folio,
14882 + .direct_IO = aufs_direct_IO,
14883 +#ifdef CONFIG_AUFS_DEBUG
14884 + .writepage = aufs_writepage,
14885 + /* no writepages, because of writepage */
14886 + .dirty_folio = aufs_dirty_folio,
14887 + /* no readpages, because of readpage */
14888 + .write_begin = aufs_write_begin,
14889 + .write_end = aufs_write_end,
14890 + /* no bmap, no block device */
14891 + .invalidate_folio = aufs_invalidate_folio,
14892 + .release_folio = aufs_release_folio,
14893 + /* is fallback_migrate_page ok? */
14894 + /* .migrate_folio = aufs_migrate_folio, */
14895 + .launder_folio = aufs_launder_folio,
14896 + .is_partially_uptodate = aufs_is_partially_uptodate,
14897 + .is_dirty_writeback = aufs_is_dirty_writeback,
14898 + .error_remove_page = aufs_error_remove_page,
14899 + .swap_activate = aufs_swap_activate,
14900 + .swap_deactivate = aufs_swap_deactivate,
14901 + .swap_rw = aufs_swap_rw
14902 +#endif /* CONFIG_AUFS_DEBUG */
14904 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
14905 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
14906 +++ linux/fs/aufs/file.h 2022-11-05 23:02:18.965889284 +0100
14908 +/* SPDX-License-Identifier: GPL-2.0 */
14910 + * Copyright (C) 2005-2022 Junjiro R. Okajima
14912 + * This program is free software; you can redistribute it and/or modify
14913 + * it under the terms of the GNU General Public License as published by
14914 + * the Free Software Foundation; either version 2 of the License, or
14915 + * (at your option) any later version.
14917 + * This program is distributed in the hope that it will be useful,
14918 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14919 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14920 + * GNU General Public License for more details.
14922 + * You should have received a copy of the GNU General Public License
14923 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14927 + * file operations
14930 +#ifndef __AUFS_FILE_H__
14931 +#define __AUFS_FILE_H__
14935 +#include <linux/file.h>
14936 +#include <linux/fs.h>
14937 +#include <linux/mm_types.h>
14938 +#include <linux/poll.h>
14939 +#include "rwsem.h"
14943 + struct file *hf_file;
14944 + struct au_branch *hf_br;
14949 + aufs_bindex_t fd_bbot;
14950 + aufs_bindex_t fd_nent;
14951 + struct au_vdir *fd_vdir_cache;
14952 + struct au_hfile fd_hfile[];
14955 +static inline int au_fidir_sz(int nent)
14957 + AuDebugOn(nent < 0);
14958 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
14962 + atomic_t fi_generation;
14964 + struct au_rwsem fi_rwsem;
14965 + aufs_bindex_t fi_btop;
14967 + /* do not union them */
14968 + struct { /* for non-dir */
14969 + struct au_hfile fi_htop;
14970 + atomic_t fi_mmapped;
14972 + struct au_fidir *fi_hdir; /* for dir only */
14974 + struct hlist_bl_node fi_hlist;
14975 + struct file *fi_file; /* very ugly */
14976 + struct rcu_head rcu;
14977 +} ____cacheline_aligned_in_smp;
14979 +/* ---------------------------------------------------------------------- */
14982 +extern const struct address_space_operations aufs_aop;
14983 +unsigned int au_file_roflags(unsigned int flags);
14984 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14985 + struct file *file, int force_wr);
14986 +struct au_do_open_args {
14988 + int (*open)(struct file *file, int flags,
14989 + struct file *h_file);
14990 + struct au_fidir *fidir;
14991 + struct file *h_file;
14993 +int au_do_open(struct file *file, struct au_do_open_args *args);
14994 +int au_reopen_nondir(struct file *file);
14996 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
14997 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14998 + int wlock, unsigned int fi_lsc);
14999 +int au_do_flush(struct file *file, fl_owner_t id,
15000 + int (*flush)(struct file *file, fl_owner_t id));
15003 +#ifdef CONFIG_AUFS_POLL
15004 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt);
15007 +#ifdef CONFIG_AUFS_BR_HFSPLUS
15009 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
15011 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
15012 + struct file *h_file);
15014 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
15015 + aufs_bindex_t bindex, int force_wr)
15016 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
15017 + struct file *h_file);
15021 +extern const struct file_operations aufs_file_fop;
15022 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
15023 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
15024 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc);
15027 +void au_hfput(struct au_hfile *hf, int execed);
15028 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
15029 + struct file *h_file);
15031 +void au_update_figen(struct file *file);
15032 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
15033 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink);
15035 +void au_fi_init_once(void *_fi);
15036 +void au_finfo_fin(struct file *file);
15037 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
15040 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
15041 +#ifdef CONFIG_COMPAT
15042 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15043 + unsigned long arg);
15044 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15045 + unsigned long arg);
15048 +/* ---------------------------------------------------------------------- */
15050 +static inline struct au_finfo *au_fi(struct file *file)
15052 + return file->private_data;
15055 +/* ---------------------------------------------------------------------- */
15057 +#define fi_read_lock(f) au_rw_read_lock(&au_fi(f)->fi_rwsem)
15058 +#define fi_write_lock(f) au_rw_write_lock(&au_fi(f)->fi_rwsem)
15059 +#define fi_read_trylock(f) au_rw_read_trylock(&au_fi(f)->fi_rwsem)
15060 +#define fi_write_trylock(f) au_rw_write_trylock(&au_fi(f)->fi_rwsem)
15062 +#define fi_read_trylock_nested(f) \
15063 + au_rw_read_trylock_nested(&au_fi(f)->fi_rwsem)
15064 +#define fi_write_trylock_nested(f) \
15065 + au_rw_write_trylock_nested(&au_fi(f)->fi_rwsem)
15068 +#define fi_read_unlock(f) au_rw_read_unlock(&au_fi(f)->fi_rwsem)
15069 +#define fi_write_unlock(f) au_rw_write_unlock(&au_fi(f)->fi_rwsem)
15070 +#define fi_downgrade_lock(f) au_rw_dgrade_lock(&au_fi(f)->fi_rwsem)
15072 +/* lock subclass for finfo */
15078 +static inline void fi_read_lock_nested(struct file *f, unsigned int lsc)
15080 + au_rw_read_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15083 +static inline void fi_write_lock_nested(struct file *f, unsigned int lsc)
15085 + au_rw_write_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15089 + * fi_read_lock_1, fi_write_lock_1,
15090 + * fi_read_lock_2, fi_write_lock_2
15092 +#define AuReadLockFunc(name) \
15093 +static inline void fi_read_lock_##name(struct file *f) \
15094 +{ fi_read_lock_nested(f, AuLsc_FI_##name); }
15096 +#define AuWriteLockFunc(name) \
15097 +static inline void fi_write_lock_##name(struct file *f) \
15098 +{ fi_write_lock_nested(f, AuLsc_FI_##name); }
15100 +#define AuRWLockFuncs(name) \
15101 + AuReadLockFunc(name) \
15102 + AuWriteLockFunc(name)
15107 +#undef AuReadLockFunc
15108 +#undef AuWriteLockFunc
15109 +#undef AuRWLockFuncs
15111 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
15112 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
15113 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
15115 +/* ---------------------------------------------------------------------- */
15117 +/* todo: hard/soft set? */
15118 +static inline aufs_bindex_t au_fbtop(struct file *file)
15120 + FiMustAnyLock(file);
15121 + return au_fi(file)->fi_btop;
15124 +static inline aufs_bindex_t au_fbbot_dir(struct file *file)
15126 + FiMustAnyLock(file);
15127 + AuDebugOn(!au_fi(file)->fi_hdir);
15128 + return au_fi(file)->fi_hdir->fd_bbot;
15131 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
15133 + FiMustAnyLock(file);
15134 + AuDebugOn(!au_fi(file)->fi_hdir);
15135 + return au_fi(file)->fi_hdir->fd_vdir_cache;
15138 +static inline void au_set_fbtop(struct file *file, aufs_bindex_t bindex)
15140 + FiMustWriteLock(file);
15141 + au_fi(file)->fi_btop = bindex;
15144 +static inline void au_set_fbbot_dir(struct file *file, aufs_bindex_t bindex)
15146 + FiMustWriteLock(file);
15147 + AuDebugOn(!au_fi(file)->fi_hdir);
15148 + au_fi(file)->fi_hdir->fd_bbot = bindex;
15151 +static inline void au_set_fvdir_cache(struct file *file,
15152 + struct au_vdir *vdir_cache)
15154 + FiMustWriteLock(file);
15155 + AuDebugOn(!au_fi(file)->fi_hdir);
15156 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
15159 +static inline struct file *au_hf_top(struct file *file)
15161 + FiMustAnyLock(file);
15162 + AuDebugOn(au_fi(file)->fi_hdir);
15163 + return au_fi(file)->fi_htop.hf_file;
15166 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
15168 + FiMustAnyLock(file);
15169 + AuDebugOn(!au_fi(file)->fi_hdir);
15170 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
15173 +/* todo: memory barrier? */
15174 +static inline unsigned int au_figen(struct file *f)
15176 + return atomic_read(&au_fi(f)->fi_generation);
15179 +static inline void au_set_mmapped(struct file *f)
15181 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
15183 + pr_warn("fi_mmapped wrapped around\n");
15184 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
15188 +static inline void au_unset_mmapped(struct file *f)
15190 + atomic_dec(&au_fi(f)->fi_mmapped);
15193 +static inline int au_test_mmapped(struct file *f)
15195 + return atomic_read(&au_fi(f)->fi_mmapped);
15198 +/* customize vma->vm_file */
15200 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
15201 + struct file *file)
15205 + f = vma->vm_file;
15207 + vma->vm_file = file;
15212 +#define AuDbgVmRegion(file, vma) do {} while (0)
15214 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15215 + struct file *file)
15217 + au_do_vm_file_reset(vma, file);
15220 +#define AuDbgVmRegion(file, vma) \
15221 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
15223 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15224 + struct file *file)
15228 + au_do_vm_file_reset(vma, file);
15229 + f = vma->vm_region->vm_file;
15231 + vma->vm_region->vm_file = file;
15234 +#endif /* CONFIG_MMU */
15236 +/* handle vma->vm_prfile */
15237 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
15238 + struct file *file)
15241 + vma->vm_prfile = file;
15242 +#ifndef CONFIG_MMU
15244 + vma->vm_region->vm_prfile = file;
15248 +#endif /* __KERNEL__ */
15249 +#endif /* __AUFS_FILE_H__ */
15250 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
15251 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
15252 +++ linux/fs/aufs/finfo.c 2022-11-05 23:02:18.965889284 +0100
15254 +// SPDX-License-Identifier: GPL-2.0
15256 + * Copyright (C) 2005-2022 Junjiro R. Okajima
15258 + * This program is free software; you can redistribute it and/or modify
15259 + * it under the terms of the GNU General Public License as published by
15260 + * the Free Software Foundation; either version 2 of the License, or
15261 + * (at your option) any later version.
15263 + * This program is distributed in the hope that it will be useful,
15264 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15265 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15266 + * GNU General Public License for more details.
15268 + * You should have received a copy of the GNU General Public License
15269 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15273 + * file private data
15278 +void au_hfput(struct au_hfile *hf, int execed)
15281 + allow_write_access(hf->hf_file);
15282 + fput(hf->hf_file);
15283 + hf->hf_file = NULL;
15284 + au_lcnt_dec(&hf->hf_br->br_nfiles);
15285 + hf->hf_br = NULL;
15288 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
15290 + struct au_finfo *finfo = au_fi(file);
15291 + struct au_hfile *hf;
15292 + struct au_fidir *fidir;
15294 + fidir = finfo->fi_hdir;
15296 + AuDebugOn(finfo->fi_btop != bindex);
15297 + hf = &finfo->fi_htop;
15299 + hf = fidir->fd_hfile + bindex;
15301 + if (hf && hf->hf_file)
15302 + au_hfput(hf, vfsub_file_execed(file));
15304 + FiMustWriteLock(file);
15305 + AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
15306 + hf->hf_file = val;
15307 + hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
15311 +void au_update_figen(struct file *file)
15313 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
15314 + /* smp_mb(); */ /* atomic_set */
15317 +/* ---------------------------------------------------------------------- */
15319 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
15321 + struct au_fidir *fidir;
15324 + nbr = au_sbbot(sb) + 1;
15326 + nbr = 2; /* initial allocate for 2 branches */
15327 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
15329 + fidir->fd_bbot = -1;
15330 + fidir->fd_nent = nbr;
15336 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink)
15339 + struct au_fidir *fidir, *p;
15341 + AuRwMustWriteLock(&finfo->fi_rwsem);
15342 + fidir = finfo->fi_hdir;
15343 + AuDebugOn(!fidir);
15346 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
15347 + GFP_NOFS, may_shrink);
15349 + p->fd_nent = nbr;
15350 + finfo->fi_hdir = p;
15357 +/* ---------------------------------------------------------------------- */
15359 +void au_finfo_fin(struct file *file)
15361 + struct au_finfo *finfo;
15363 + au_lcnt_dec(&au_sbi(file->f_path.dentry->d_sb)->si_nfiles);
15365 + finfo = au_fi(file);
15366 + AuDebugOn(finfo->fi_hdir);
15367 + AuRwDestroy(&finfo->fi_rwsem);
15368 + au_cache_free_finfo(finfo);
15371 +void au_fi_init_once(void *_finfo)
15373 + struct au_finfo *finfo = _finfo;
15375 + au_rw_init(&finfo->fi_rwsem);
15378 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
15381 + struct au_finfo *finfo;
15382 + struct dentry *dentry;
15385 + dentry = file->f_path.dentry;
15386 + finfo = au_cache_alloc_finfo();
15387 + if (unlikely(!finfo))
15391 + au_lcnt_inc(&au_sbi(dentry->d_sb)->si_nfiles);
15392 + au_rw_write_lock(&finfo->fi_rwsem);
15393 + finfo->fi_btop = -1;
15394 + finfo->fi_hdir = fidir;
15395 + atomic_set(&finfo->fi_generation, au_digen(dentry));
15396 + /* smp_mb(); */ /* atomic_set */
15398 + file->private_data = finfo;
15403 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
15404 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
15405 +++ linux/fs/aufs/f_op.c 2023-02-20 21:05:51.959693785 +0100
15407 +// SPDX-License-Identifier: GPL-2.0
15409 + * Copyright (C) 2005-2022 Junjiro R. Okajima
15411 + * This program is free software; you can redistribute it and/or modify
15412 + * it under the terms of the GNU General Public License as published by
15413 + * the Free Software Foundation; either version 2 of the License, or
15414 + * (at your option) any later version.
15416 + * This program is distributed in the hope that it will be useful,
15417 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15418 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15419 + * GNU General Public License for more details.
15421 + * You should have received a copy of the GNU General Public License
15422 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15426 + * file and vm operations
15429 +#include <linux/aio.h>
15430 +#include <linux/fs_stack.h>
15431 +#include <linux/mman.h>
15432 +#include <linux/security.h>
15435 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
15438 + aufs_bindex_t bindex;
15439 + struct dentry *dentry, *h_dentry;
15440 + struct au_finfo *finfo;
15441 + struct inode *h_inode;
15443 + FiMustWriteLock(file);
15446 + dentry = file->f_path.dentry;
15447 + AuDebugOn(IS_ERR_OR_NULL(dentry));
15448 + finfo = au_fi(file);
15449 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
15450 + atomic_set(&finfo->fi_mmapped, 0);
15451 + bindex = au_dbtop(dentry);
15453 + h_dentry = au_h_dptr(dentry, bindex);
15454 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15455 + if (unlikely(err))
15457 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
15458 + if (IS_ERR(h_file)) {
15459 + err = PTR_ERR(h_file);
15463 + h_dentry = h_file->f_path.dentry;
15464 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15465 + if (unlikely(err))
15467 + /* br ref is already inc-ed */
15470 + if (flags & __O_TMPFILE) {
15471 + AuDebugOn(!h_file);
15472 + AuDebugOn(h_file != au_di(dentry)->di_htmpfile);
15473 + au_di(dentry)->di_htmpfile = NULL;
15475 + if (!(flags & O_EXCL)) {
15476 + h_inode = file_inode(h_file);
15477 + spin_lock(&h_inode->i_lock);
15478 + h_inode->i_state |= I_LINKABLE;
15479 + spin_unlock(&h_inode->i_lock);
15482 + au_set_fbtop(file, bindex);
15483 + au_set_h_fptr(file, bindex, h_file);
15484 + au_update_figen(file);
15485 + /* todo: necessary? */
15486 + /* file->f_ra = h_file->f_ra; */
15492 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
15493 + struct file *file)
15496 + struct super_block *sb;
15497 + struct au_do_open_args args = {
15498 + .open = au_do_open_nondir
15501 + AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
15502 + file, vfsub_file_flags(file), file->f_mode);
15504 + sb = file->f_path.dentry->d_sb;
15505 + si_read_lock(sb, AuLock_FLUSH);
15506 + err = au_do_open(file, &args);
15507 + si_read_unlock(sb);
15511 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
15513 + struct au_finfo *finfo;
15514 + aufs_bindex_t bindex;
15516 + finfo = au_fi(file);
15517 + au_hbl_del(&finfo->fi_hlist,
15518 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
15519 + bindex = finfo->fi_btop;
15521 + au_set_h_fptr(file, bindex, NULL);
15523 + au_finfo_fin(file);
15527 +/* ---------------------------------------------------------------------- */
15529 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
15532 + struct file *h_file;
15535 + h_file = au_hf_top(file);
15537 + err = vfsub_flush(h_file, id);
15541 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
15543 + return au_do_flush(file, id, au_do_flush_nondir);
15546 +/* ---------------------------------------------------------------------- */
15548 + * read and write functions acquire [fdi]_rwsem once, but release before
15549 + * mmap_sem. This is because to stop a race condition between mmap(2).
15550 + * Releasing these aufs-rwsem should be safe, no branch-management (by keeping
15551 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
15552 + * read functions after [fdi]_rwsem are released, but it should be harmless.
15555 +/* Callers should call au_read_post() or fput() in the end */
15556 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc)
15558 + struct file *h_file;
15561 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0, lsc);
15563 + di_read_unlock(file->f_path.dentry, AuLock_IR);
15564 + h_file = au_hf_top(file);
15565 + get_file(h_file);
15567 + fi_read_unlock(file);
15569 + h_file = ERR_PTR(err);
15574 +static void au_read_post(struct inode *inode, struct file *h_file)
15576 + /* update without lock, I don't think it a problem */
15577 + fsstack_copy_attr_atime(inode, file_inode(h_file));
15581 +struct au_write_pre {
15583 + unsigned int lsc;
15587 + aufs_bindex_t btop;
15591 + * return with iinfo is write-locked
15592 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
15595 +static struct file *au_write_pre(struct file *file, int do_ready,
15596 + struct au_write_pre *wpre)
15598 + struct file *h_file;
15599 + struct dentry *dentry;
15601 + unsigned int lsc;
15602 + struct au_pin pin;
15607 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1, lsc);
15608 + h_file = ERR_PTR(err);
15609 + if (unlikely(err))
15612 + dentry = file->f_path.dentry;
15614 + err = au_ready_to_write(file, -1, &pin);
15615 + if (unlikely(err)) {
15616 + h_file = ERR_PTR(err);
15617 + di_write_unlock(dentry);
15622 + di_downgrade_lock(dentry, /*flags*/0);
15624 + wpre->btop = au_fbtop(file);
15625 + h_file = au_hf_top(file);
15626 + get_file(h_file);
15628 + wpre->blks = file_inode(h_file)->i_blocks;
15631 + di_read_unlock(dentry, /*flags*/0);
15634 + fi_write_unlock(file);
15639 +static void au_write_post(struct inode *inode, struct file *h_file,
15640 + struct au_write_pre *wpre, ssize_t written)
15642 + struct inode *h_inode;
15644 + au_cpup_attr_timesizes(inode);
15645 + AuDebugOn(au_ibtop(inode) != wpre->btop);
15646 + h_inode = file_inode(h_file);
15647 + inode->i_mode = h_inode->i_mode;
15648 + ii_write_unlock(inode);
15649 + /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
15651 + au_fhsm_wrote(inode->i_sb, wpre->btop,
15652 + /*force*/h_inode->i_blocks > wpre->blks);
15657 + * todo: very ugly
15658 + * it locks both of i_mutex and si_rwsem for read in safe.
15659 + * if the plink maintenance mode continues forever (that is the problem),
15660 + * may loop forever.
15662 +static void au_mtx_and_read_lock(struct inode *inode)
15665 + struct super_block *sb = inode->i_sb;
15668 + inode_lock(inode);
15669 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
15672 + inode_unlock(inode);
15673 + si_read_lock(sb, AuLock_NOPLMW);
15674 + si_read_unlock(sb);
15678 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
15679 + struct iov_iter *iov_iter)
15682 + struct file *file;
15683 + ssize_t (*iter)(struct kiocb *, struct iov_iter *);
15685 + err = security_file_permission(h_file, rw);
15686 + if (unlikely(err))
15689 + err = -ENOSYS; /* the branch doesn't have its ->(read|write)_iter() */
15691 + if (rw == MAY_READ)
15692 + iter = h_file->f_op->read_iter;
15693 + else if (rw == MAY_WRITE)
15694 + iter = h_file->f_op->write_iter;
15696 + file = kio->ki_filp;
15697 + kio->ki_filp = h_file;
15700 + err = iter(kio, iov_iter);
15703 + /* currently there is no such fs */
15705 + kio->ki_filp = file;
15711 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15714 + struct file *file, *h_file;
15715 + struct inode *inode;
15716 + struct super_block *sb;
15718 + file = kio->ki_filp;
15719 + inode = file_inode(file);
15720 + sb = inode->i_sb;
15721 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15723 + h_file = au_read_pre(file, /*keep_fi*/1, /*lsc*/0);
15724 + err = PTR_ERR(h_file);
15725 + if (IS_ERR(h_file))
15728 + if (au_test_loopback_kthread()) {
15729 + au_warn_loopback(h_file->f_path.dentry->d_sb);
15730 + if (file->f_mapping != h_file->f_mapping) {
15731 + file->f_mapping = h_file->f_mapping;
15732 + smp_mb(); /* unnecessary? */
15735 + fi_read_unlock(file);
15737 + err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
15738 + /* todo: necessary? */
15739 + /* file->f_ra = h_file->f_ra; */
15740 + au_read_post(inode, h_file);
15743 + si_read_unlock(sb);
15747 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15750 + struct au_write_pre wpre;
15751 + struct inode *inode;
15752 + struct file *file, *h_file;
15754 + file = kio->ki_filp;
15755 + inode = file_inode(file);
15756 + au_mtx_and_read_lock(inode);
15759 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15760 + err = PTR_ERR(h_file);
15761 + if (IS_ERR(h_file))
15764 + err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
15765 + au_write_post(inode, h_file, &wpre, err);
15768 + si_read_unlock(inode->i_sb);
15769 + inode_unlock(inode);
15774 + * We may be able to remove aufs_splice_{read,write}() since almost all FSes
15775 + * don't have their own .splice_{read,write} implimentations, and they use
15776 + * generic_file_splice_read() and iter_file_splice_write() who can act like the
15777 + * simple converters to f_op->iter_read() and ->iter_write().
15778 + * But we keep our own implementations because some non-mainlined FSes may have
15779 + * their own .splice_{read,write} implimentations and aufs doesn't want to take
15780 + * away an opportunity to co-work with aufs from them.
15782 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
15783 + struct pipe_inode_info *pipe, size_t len,
15784 + unsigned int flags)
15787 + struct file *h_file;
15788 + struct inode *inode;
15789 + struct super_block *sb;
15791 + inode = file_inode(file);
15792 + sb = inode->i_sb;
15793 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15795 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15796 + err = PTR_ERR(h_file);
15797 + if (IS_ERR(h_file))
15800 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
15801 + /* todo: necessary? */
15802 + /* file->f_ra = h_file->f_ra; */
15803 + au_read_post(inode, h_file);
15806 + si_read_unlock(sb);
15811 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
15812 + size_t len, unsigned int flags)
15815 + struct au_write_pre wpre;
15816 + struct inode *inode;
15817 + struct file *h_file;
15819 + inode = file_inode(file);
15820 + au_mtx_and_read_lock(inode);
15823 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15824 + err = PTR_ERR(h_file);
15825 + if (IS_ERR(h_file))
15828 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
15829 + au_write_post(inode, h_file, &wpre, err);
15832 + si_read_unlock(inode->i_sb);
15833 + inode_unlock(inode);
15837 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
15841 + struct au_write_pre wpre;
15842 + struct inode *inode;
15843 + struct file *h_file;
15845 + inode = file_inode(file);
15846 + au_mtx_and_read_lock(inode);
15849 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15850 + err = PTR_ERR(h_file);
15851 + if (IS_ERR(h_file))
15855 + err = vfs_fallocate(h_file, mode, offset, len);
15858 + * we don't need to call file_modifed() here since au_write_post()
15859 + * is equivalent and copies-up all timestamps and permission bits.
15861 + au_write_post(inode, h_file, &wpre, /*written*/1);
15864 + si_read_unlock(inode->i_sb);
15865 + inode_unlock(inode);
15869 +static ssize_t aufs_copy_file_range(struct file *src, loff_t src_pos,
15870 + struct file *dst, loff_t dst_pos,
15871 + size_t len, unsigned int flags)
15874 + struct au_write_pre wpre;
15875 + enum { SRC, DST };
15877 + struct inode *inode;
15878 + struct file *h_file;
15879 + struct super_block *h_sb;
15881 +#define a_src a[SRC]
15882 +#define a_dst a[DST]
15885 + a_src.inode = file_inode(src);
15886 + if (unlikely(!S_ISREG(a_src.inode->i_mode)))
15888 + a_dst.inode = file_inode(dst);
15889 + if (unlikely(!S_ISREG(a_dst.inode->i_mode)))
15892 + au_mtx_and_read_lock(a_dst.inode);
15894 + * in order to match the order in di_write_lock2_{child,parent}(),
15895 + * use f_path.dentry for this comparison.
15897 + if (src->f_path.dentry < dst->f_path.dentry) {
15898 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_1);
15899 + err = PTR_ERR(a_src.h_file);
15900 + if (IS_ERR(a_src.h_file))
15903 + wpre.lsc = AuLsc_FI_2;
15904 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15905 + err = PTR_ERR(a_dst.h_file);
15906 + if (IS_ERR(a_dst.h_file)) {
15907 + au_read_post(a_src.inode, a_src.h_file);
15911 + wpre.lsc = AuLsc_FI_1;
15912 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15913 + err = PTR_ERR(a_dst.h_file);
15914 + if (IS_ERR(a_dst.h_file))
15917 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_2);
15918 + err = PTR_ERR(a_src.h_file);
15919 + if (IS_ERR(a_src.h_file)) {
15920 + au_write_post(a_dst.inode, a_dst.h_file, &wpre,
15927 + a_src.h_sb = file_inode(a_src.h_file)->i_sb;
15928 + a_dst.h_sb = file_inode(a_dst.h_file)->i_sb;
15929 + if (unlikely(a_src.h_sb != a_dst.h_sb)) {
15935 + err = vfsub_copy_file_range(a_src.h_file, src_pos, a_dst.h_file,
15936 + dst_pos, len, flags);
15939 + au_write_post(a_dst.inode, a_dst.h_file, &wpre, err);
15940 + fi_read_unlock(src);
15941 + au_read_post(a_src.inode, a_src.h_file);
15943 + si_read_unlock(a_dst.inode->i_sb);
15944 + inode_unlock(a_dst.inode);
15951 +/* ---------------------------------------------------------------------- */
15954 + * The locking order around current->mmap_sem.
15955 + * - in most and regular cases
15956 + * file I/O syscall -- aufs_read() or something
15957 + * -- si_rwsem for read -- mmap_sem
15958 + * (Note that [fdi]i_rwsem are released before mmap_sem).
15960 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
15961 + * This AB-BA order is definitely bad, but is not a problem since "si_rwsem for
15962 + * read" allows multiple processes to acquire it and [fdi]i_rwsem are not held
15963 + * in file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
15964 + * It means that when aufs acquires si_rwsem for write, the process should never
15965 + * acquire mmap_sem.
15967 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
15968 + * problem either since any directory is not able to be mmap-ed.
15969 + * The similar scenario is applied to aufs_readlink() too.
15972 +#if 0 /* stop calling security_file_mmap() */
15973 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
15974 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
15976 +static unsigned long au_arch_prot_conv(unsigned long flags)
15978 + /* currently ppc64 only */
15979 +#ifdef CONFIG_PPC64
15980 + /* cf. linux/arch/powerpc/include/asm/mman.h */
15981 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
15982 + return AuConv_VM_PROT(flags, SAO);
15984 + AuDebugOn(arch_calc_vm_prot_bits(-1));
15989 +static unsigned long au_prot_conv(unsigned long flags)
15991 + return AuConv_VM_PROT(flags, READ)
15992 + | AuConv_VM_PROT(flags, WRITE)
15993 + | AuConv_VM_PROT(flags, EXEC)
15994 + | au_arch_prot_conv(flags);
15997 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
15998 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
16000 +static unsigned long au_flag_conv(unsigned long flags)
16002 + return AuConv_VM_MAP(flags, GROWSDOWN)
16003 + | AuConv_VM_MAP(flags, DENYWRITE)
16004 + | AuConv_VM_MAP(flags, LOCKED);
16008 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
16011 + const unsigned char wlock
16012 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
16013 + struct super_block *sb;
16014 + struct file *h_file;
16015 + struct inode *inode;
16017 + AuDbgVmRegion(file, vma);
16019 + inode = file_inode(file);
16020 + sb = inode->i_sb;
16022 + si_read_lock(sb, AuLock_NOPLMW);
16024 + h_file = au_write_pre(file, wlock, /*wpre*/NULL);
16026 + err = PTR_ERR(h_file);
16027 + if (IS_ERR(h_file))
16031 + au_set_mmapped(file);
16032 + au_vm_file_reset(vma, h_file);
16034 + * we cannot call security_mmap_file() here since it may acquire
16035 + * mmap_sem or i_mutex.
16037 + * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
16038 + * au_flag_conv(vma->vm_flags));
16041 + err = call_mmap(h_file, vma);
16043 + au_vm_prfile_set(vma, file);
16044 + fsstack_copy_attr_atime(inode, file_inode(h_file));
16045 + goto out_fput; /* success */
16047 + au_unset_mmapped(file);
16048 + au_vm_file_reset(vma, file);
16052 + ii_write_unlock(inode);
16057 + si_read_unlock(sb);
16063 +/* ---------------------------------------------------------------------- */
16065 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
16069 + struct au_write_pre wpre;
16070 + struct inode *inode;
16071 + struct file *h_file;
16073 + err = 0; /* -EBADF; */ /* posix? */
16074 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
16077 + inode = file_inode(file);
16078 + au_mtx_and_read_lock(inode);
16081 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
16082 + err = PTR_ERR(h_file);
16083 + if (IS_ERR(h_file))
16086 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
16087 + au_write_post(inode, h_file, &wpre, /*written*/0);
16090 + si_read_unlock(inode->i_sb);
16091 + inode_unlock(inode);
16096 +static int aufs_fasync(int fd, struct file *file, int flag)
16099 + struct file *h_file;
16100 + struct super_block *sb;
16102 + sb = file->f_path.dentry->d_sb;
16103 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16105 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16106 + err = PTR_ERR(h_file);
16107 + if (IS_ERR(h_file))
16110 + if (h_file->f_op->fasync)
16111 + err = h_file->f_op->fasync(fd, h_file, flag);
16112 + fput(h_file); /* instead of au_read_post() */
16115 + si_read_unlock(sb);
16119 +static int aufs_setfl(struct file *file, unsigned long arg)
16122 + struct file *h_file;
16123 + struct super_block *sb;
16125 + sb = file->f_path.dentry->d_sb;
16126 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16128 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16129 + err = PTR_ERR(h_file);
16130 + if (IS_ERR(h_file))
16133 + /* stop calling h_file->fasync */
16134 + arg |= vfsub_file_flags(file) & FASYNC;
16135 + err = setfl(/*unused fd*/-1, h_file, arg);
16136 + fput(h_file); /* instead of au_read_post() */
16139 + si_read_unlock(sb);
16143 +/* ---------------------------------------------------------------------- */
16145 +/* no one supports this operation, currently */
16146 +#if 0 /* reserved for future use */
16147 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
16148 + size_t len, loff_t *pos, int more)
16153 +/* ---------------------------------------------------------------------- */
16155 +const struct file_operations aufs_file_fop = {
16156 + .owner = THIS_MODULE,
16158 + .llseek = default_llseek,
16160 + .read_iter = aufs_read_iter,
16161 + .write_iter = aufs_write_iter,
16163 +#ifdef CONFIG_AUFS_POLL
16164 + .poll = aufs_poll,
16166 + .unlocked_ioctl = aufs_ioctl_nondir,
16167 +#ifdef CONFIG_COMPAT
16168 + .compat_ioctl = aufs_compat_ioctl_nondir,
16170 + .mmap = aufs_mmap,
16171 + .open = aufs_open_nondir,
16172 + .flush = aufs_flush_nondir,
16173 + .release = aufs_release_nondir,
16174 + .fsync = aufs_fsync_nondir,
16175 + .fasync = aufs_fasync,
16176 + /* .sendpage = aufs_sendpage, */
16177 + .setfl = aufs_setfl,
16178 + .splice_write = aufs_splice_write,
16179 + .splice_read = aufs_splice_read,
16180 +#if 0 /* reserved for future use */
16181 + .aio_splice_write = aufs_aio_splice_write,
16182 + .aio_splice_read = aufs_aio_splice_read,
16184 + .fallocate = aufs_fallocate,
16185 + .copy_file_range = aufs_copy_file_range
16187 diff -urN /usr/share/empty/fs/aufs/fsctx.c linux/fs/aufs/fsctx.c
16188 --- /usr/share/empty/fs/aufs/fsctx.c 1970-01-01 01:00:00.000000000 +0100
16189 +++ linux/fs/aufs/fsctx.c 2022-11-05 23:02:18.965889284 +0100
16191 +// SPDX-License-Identifier: GPL-2.0
16193 + * Copyright (C) 2022 Junjiro R. Okajima
16195 + * This program is free software; you can redistribute it and/or modify
16196 + * it under the terms of the GNU General Public License as published by
16197 + * the Free Software Foundation; either version 2 of the License, or
16198 + * (at your option) any later version.
16200 + * This program is distributed in the hope that it will be useful,
16201 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16202 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16203 + * GNU General Public License for more details.
16205 + * You should have received a copy of the GNU General Public License
16206 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16210 + * fs context, aka new mount api
16213 +#include <linux/fs_context.h>
16216 +struct au_fsctx_opts {
16217 + aufs_bindex_t bindex;
16218 + unsigned char skipped;
16219 + struct au_opt *opt, *opt_tail;
16220 + struct super_block *sb;
16221 + struct au_sbinfo *sbinfo;
16222 + struct au_opts opts;
16225 +/* stop extra interpretation of errno in mount(8), and strange error messages */
16226 +static int cvt_err(int err)
16240 +static int au_fsctx_reconfigure(struct fs_context *fc)
16243 + unsigned int mntflags;
16244 + struct dentry *root;
16245 + struct super_block *sb;
16246 + struct inode *inode;
16247 + struct au_fsctx_opts *a = fc->fs_private;
16249 + AuDbg("fc %p\n", fc);
16253 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16255 + di_write_lock_child(root);
16256 + err = au_opts_verify(sb, fc->sb_flags, /*pending*/0);
16257 + aufs_write_unlock(root);
16260 + inode = d_inode(root);
16261 + inode_lock(inode);
16262 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16263 + if (unlikely(err))
16265 + di_write_lock_child(root);
16267 + /* au_opts_remount() may return an error */
16268 + err = au_opts_remount(sb, &a->opts);
16270 + if (au_ftest_opts(a->opts.flags, REFRESH))
16271 + au_remount_refresh(sb, au_ftest_opts(a->opts.flags,
16274 + if (au_ftest_opts(a->opts.flags, REFRESH_DYAOP)) {
16275 + mntflags = au_mntflags(sb);
16276 + do_dx = !!au_opt_test(mntflags, DIO);
16277 + au_dy_arefresh(do_dx);
16280 + au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
16281 + aufs_write_unlock(root);
16284 + inode_unlock(inode);
16285 + err = cvt_err(err);
16291 +/* ---------------------------------------------------------------------- */
16293 +static int au_fsctx_fill_super(struct super_block *sb, struct fs_context *fc)
16296 + struct au_fsctx_opts *a = fc->fs_private;
16297 + struct au_sbinfo *sbinfo = a->sbinfo;
16298 + struct dentry *root;
16299 + struct inode *inode;
16301 + sbinfo->si_sb = sb;
16302 + sb->s_fs_info = sbinfo;
16303 + kobject_get(&sbinfo->si_kobj);
16305 + __si_write_lock(sb);
16307 + au_sbilist_add(sb);
16309 + /* all timestamps always follow the ones on the branch */
16310 + sb->s_flags |= SB_NOATIME | SB_NODIRATIME;
16311 + sb->s_flags |= SB_I_VERSION; /* do we really need this? */
16312 + sb->s_op = &aufs_sop;
16313 + sb->s_d_op = &aufs_dop;
16314 + sb->s_magic = AUFS_SUPER_MAGIC;
16315 + sb->s_maxbytes = 0;
16316 + sb->s_stack_depth = 1;
16317 + au_export_init(sb);
16318 + au_xattr_init(sb);
16320 + err = au_alloc_root(sb);
16321 + if (unlikely(err)) {
16322 + si_write_unlock(sb);
16325 + root = sb->s_root;
16326 + inode = d_inode(root);
16327 + ii_write_lock_parent(inode);
16328 + aufs_write_unlock(root);
16330 + /* lock vfs_inode first, then aufs. */
16331 + inode_lock(inode);
16332 + aufs_write_lock(root);
16333 + err = au_opts_mount(sb, &a->opts);
16335 + if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
16336 + sb->s_d_op = &aufs_dop_noreval;
16337 + /* infofc(fc, "%ps", sb->s_d_op); */
16338 + pr_info("%ps\n", sb->s_d_op);
16339 + au_refresh_dop(root, /*force_reval*/0);
16340 + sbinfo->si_iop_array = aufs_iop_nogetattr;
16341 + au_refresh_iop(inode, /*force_getattr*/0);
16343 + aufs_write_unlock(root);
16344 + inode_unlock(inode);
16346 + goto out; /* success */
16349 + sb->s_root = NULL;
16352 + if (unlikely(err))
16353 + kobject_put(&sbinfo->si_kobj);
16355 + err = cvt_err(err);
16360 +static int au_fsctx_get_tree(struct fs_context *fc)
16364 + AuDbg("fc %p\n", fc);
16365 + err = get_tree_nodev(fc, au_fsctx_fill_super);
16371 +/* ---------------------------------------------------------------------- */
16373 +static void au_fsctx_dump(struct au_opts *opts)
16375 +#ifdef CONFIG_AUFS_DEBUG
16376 + /* reduce stack space */
16378 + struct au_opt_add *add;
16379 + struct au_opt_del *del;
16380 + struct au_opt_mod *mod;
16381 + struct au_opt_xino *xino;
16382 + struct au_opt_xino_itrunc *xino_itrunc;
16383 + struct au_opt_wbr_create *create;
16385 + struct au_opt *opt;
16388 + while (opt->type != Opt_tail) {
16389 + switch (opt->type) {
16391 + u.add = &opt->add;
16392 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
16393 + u.add->bindex, u.add->pathname, u.add->perm,
16394 + u.add->path.dentry);
16399 + u.del = &opt->del;
16400 + AuDbg("del {%s, %p}\n",
16401 + u.del->pathname, u.del->h_path.dentry);
16406 + u.mod = &opt->mod;
16407 + AuDbg("mod {%s, 0x%x, %p}\n",
16408 + u.mod->path, u.mod->perm, u.mod->h_root);
16411 + u.add = &opt->add;
16412 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
16413 + u.add->bindex, u.add->pathname, u.add->perm,
16414 + u.add->path.dentry);
16416 + case Opt_prepend:
16417 + u.add = &opt->add;
16418 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
16419 + u.add->bindex, u.add->pathname, u.add->perm,
16420 + u.add->path.dentry);
16424 + AuDbg("dirwh %d\n", opt->dirwh);
16426 + case Opt_rdcache:
16427 + AuDbg("rdcache %d\n", opt->rdcache);
16430 + AuDbg("rdblk %d\n", opt->rdblk);
16433 + AuDbg("rdhash %u\n", opt->rdhash);
16437 + u.xino = &opt->xino;
16438 + AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
16441 +#define au_fsctx_TF(name) \
16442 + case Opt_##name: \
16446 + AuLabel(no##name); \
16449 + /* simple true/false flag */
16450 + au_fsctx_TF(trunc_xino);
16451 + au_fsctx_TF(trunc_xib);
16452 + au_fsctx_TF(dirperm1);
16453 + au_fsctx_TF(plink);
16454 + au_fsctx_TF(shwh);
16455 + au_fsctx_TF(dio);
16456 + au_fsctx_TF(warn_perm);
16457 + au_fsctx_TF(verbose);
16458 + au_fsctx_TF(sum);
16459 + au_fsctx_TF(dirren);
16460 + au_fsctx_TF(acl);
16461 +#undef au_fsctx_TF
16463 + case Opt_trunc_xino_path:
16465 + case Opt_itrunc_xino:
16466 + u.xino_itrunc = &opt->xino_itrunc;
16467 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
16473 + case Opt_list_plink:
16474 + AuLabel(list_plink);
16477 + AuDbg("udba %d, %s\n",
16478 + opt->udba, au_optstr_udba(opt->udba));
16480 + case Opt_diropq_a:
16481 + AuLabel(diropq_a);
16483 + case Opt_diropq_w:
16484 + AuLabel(diropq_w);
16489 + case Opt_wbr_create:
16490 + u.create = &opt->wbr_create;
16491 + AuDbg("create %d, %s\n", u.create->wbr_create,
16492 + au_optstr_wbr_create(u.create->wbr_create));
16493 + switch (u.create->wbr_create) {
16494 + case AuWbrCreate_MFSV:
16496 + case AuWbrCreate_PMFSV:
16497 + AuDbg("%d sec\n", u.create->mfs_second);
16499 + case AuWbrCreate_MFSRR:
16501 + case AuWbrCreate_TDMFS:
16502 + AuDbg("%llu watermark\n",
16503 + u.create->mfsrr_watermark);
16505 + case AuWbrCreate_MFSRRV:
16507 + case AuWbrCreate_TDMFSV:
16509 + case AuWbrCreate_PMFSRRV:
16510 + AuDbg("%llu watermark, %d sec\n",
16511 + u.create->mfsrr_watermark,
16512 + u.create->mfs_second);
16516 + case Opt_wbr_copyup:
16517 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
16518 + au_optstr_wbr_copyup(opt->wbr_copyup));
16520 + case Opt_fhsm_sec:
16521 + AuDbg("fhsm_sec %u\n", opt->fhsm_second);
16525 + AuDbg("type %d\n", opt->type);
16533 +/* ---------------------------------------------------------------------- */
16536 + * For conditionally compiled mount options.
16537 + * Instead of fsparam_flag_no(), use this macro to distinguish ignore_silent.
16539 +#define au_ignore_flag(name, action) \
16540 + fsparam_flag(name, action), \
16541 + fsparam_flag("no" name, Opt_ignore_silent)
16543 +const struct fs_parameter_spec aufs_fsctx_paramspec[] = {
16544 + fsparam_string("br", Opt_br),
16546 + /* "add=%d:%s" or "ins=%d:%s" */
16547 + fsparam_string("add", Opt_add),
16548 + fsparam_string("ins", Opt_add),
16549 + fsparam_path("append", Opt_append),
16550 + fsparam_path("prepend", Opt_prepend),
16552 + fsparam_path("del", Opt_del),
16553 + /* fsparam_s32("idel", Opt_idel), */
16554 + fsparam_path("mod", Opt_mod),
16555 + /* fsparam_string("imod", Opt_imod), */
16557 + fsparam_s32("dirwh", Opt_dirwh),
16559 + fsparam_path("xino", Opt_xino),
16560 + fsparam_flag("noxino", Opt_noxino),
16561 + fsparam_flag_no("trunc_xino", Opt_trunc_xino),
16562 + /* "trunc_xino_v=%d:%d" */
16563 + /* fsparam_string("trunc_xino_v", Opt_trunc_xino_v), */
16564 + fsparam_path("trunc_xino", Opt_trunc_xino_path),
16565 + fsparam_s32("itrunc_xino", Opt_itrunc_xino),
16566 + /* fsparam_path("zxino", Opt_zxino), */
16567 + fsparam_flag_no("trunc_xib", Opt_trunc_xib),
16569 +#ifdef CONFIG_PROC_FS
16570 + fsparam_flag_no("plink", Opt_plink),
16572 + au_ignore_flag("plink", Opt_ignore),
16575 +#ifdef CONFIG_AUFS_DEBUG
16576 + fsparam_flag("list_plink", Opt_list_plink),
16579 + fsparam_string("udba", Opt_udba),
16581 + fsparam_flag_no("dio", Opt_dio),
16583 +#ifdef CONFIG_AUFS_DIRREN
16584 + fsparam_flag_no("dirren", Opt_dirren),
16586 + au_ignore_flag("dirren", Opt_ignore),
16589 +#ifdef CONFIG_AUFS_FHSM
16590 + fsparam_s32("fhsm_sec", Opt_fhsm_sec),
16592 + fsparam_s32("fhsm_sec", Opt_ignore),
16595 + /* always | a | whiteouted | w */
16596 + fsparam_string("diropq", Opt_diropq),
16598 + fsparam_flag_no("warn_perm", Opt_warn_perm),
16600 +#ifdef CONFIG_AUFS_SHWH
16601 + fsparam_flag_no("shwh", Opt_shwh),
16603 + au_ignore_flag("shwh", Opt_err),
16606 + fsparam_flag_no("dirperm1", Opt_dirperm1),
16608 + fsparam_flag_no("verbose", Opt_verbose),
16609 + fsparam_flag("v", Opt_verbose),
16610 + fsparam_flag("quiet", Opt_noverbose),
16611 + fsparam_flag("q", Opt_noverbose),
16612 + /* user-space may handle this */
16613 + fsparam_flag("silent", Opt_noverbose),
16615 + fsparam_flag_no("sum", Opt_sum),
16616 + fsparam_flag("wsum", Opt_wsum),
16618 + fsparam_s32("rdcache", Opt_rdcache),
16619 + /* "def" or s32 */
16620 + fsparam_string("rdblk", Opt_rdblk),
16621 + /* "def" or s32 */
16622 + fsparam_string("rdhash", Opt_rdhash),
16624 + fsparam_string("create", Opt_wbr_create),
16625 + fsparam_string("create_policy", Opt_wbr_create),
16626 + fsparam_string("cpup", Opt_wbr_copyup),
16627 + fsparam_string("copyup", Opt_wbr_copyup),
16628 + fsparam_string("copyup_policy", Opt_wbr_copyup),
16630 + /* generic VFS flag */
16631 +#ifdef CONFIG_FS_POSIX_ACL
16632 + fsparam_flag_no("acl", Opt_acl),
16634 + au_ignore_flag("acl", Opt_ignore),
16637 + /* internal use for the scripts */
16638 + fsparam_string("si", Opt_ignore_silent),
16640 + /* obsoleted, keep them temporary */
16641 + fsparam_flag("nodlgt", Opt_ignore_silent),
16642 + fsparam_flag("clean_plink", Opt_ignore),
16643 + fsparam_string("dirs", Opt_br),
16644 + fsparam_u32("debug", Opt_ignore),
16645 + /* "whiteout" or "all" */
16646 + fsparam_string("delete", Opt_ignore),
16647 + fsparam_string("imap", Opt_ignore),
16649 + /* temporary workaround, due to old mount(8)? */
16650 + fsparam_flag("relatime", Opt_ignore_silent),
16655 +static int au_fsctx_parse_do_add(struct fs_context *fc, struct au_opt *opt,
16656 + char *brspec, size_t speclen,
16657 + aufs_bindex_t bindex)
16662 + AuDbg("brspec %s\n", brspec);
16666 + speclen = strlen(brspec);
16667 + /* will be freed by au_fsctx_free() */
16668 + p = kmemdup_nul(brspec, speclen, GFP_NOFS);
16669 + if (unlikely(!p)) {
16670 + errorfc(fc, "failed in %s", brspec);
16673 + err = au_opt_add(opt, p, fc->sb_flags, bindex);
16680 +static int au_fsctx_parse_br(struct fs_context *fc, char *brspec)
16684 + struct au_fsctx_opts *a = fc->fs_private;
16685 + struct au_opt *opt = a->opt;
16686 + aufs_bindex_t bindex = a->bindex;
16688 + AuDbg("brspec %s\n", brspec);
16691 + while ((p = strsep(&brspec, ":")) && *p) {
16692 + err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, bindex);
16694 + if (unlikely(err))
16698 + if (unlikely(opt > a->opt_tail)) {
16704 + opt->type = Opt_tail;
16707 + a->bindex = bindex;
16714 +static int au_fsctx_parse_add(struct fs_context *fc, char *addspec)
16718 + struct au_fsctx_opts *a = fc->fs_private;
16719 + struct au_opt *opt = a->opt;
16722 + p = strchr(addspec, ':');
16723 + if (unlikely(!p)) {
16724 + errorfc(fc, "bad arg in %s", addspec);
16728 + err = kstrtoint(addspec, 0, &n);
16729 + if (unlikely(err)) {
16730 + errorfc(fc, "bad integer in %s", addspec);
16733 + AuDbg("n %d\n", n);
16734 + err = au_fsctx_parse_do_add(fc, opt, p, /*len*/0, n);
16741 +static int au_fsctx_parse_del(struct fs_context *fc, struct au_opt_del *del,
16742 + struct fs_parameter *param)
16747 + /* will be freed by au_fsctx_free() */
16748 + del->pathname = kmemdup_nul(param->string, param->size, GFP_NOFS);
16749 + if (unlikely(!del->pathname))
16751 + AuDbg("del %s\n", del->pathname);
16752 + err = vfsub_kern_path(del->pathname, AuOpt_LkupDirFlags, &del->h_path);
16753 + if (unlikely(err))
16754 + errorfc(fc, "lookup failed %s (%d)", del->pathname, err);
16761 +#if 0 /* reserved for future use */
16762 +static int au_fsctx_parse_idel(struct fs_context *fc, struct au_opt_del *del,
16763 + aufs_bindex_t bindex)
16766 + struct super_block *sb;
16767 + struct dentry *root;
16768 + struct au_fsctx_opts *a = fc->fs_private;
16774 + root = sb->s_root;
16775 + aufs_read_lock(root, AuLock_FLUSH);
16776 + if (bindex < 0 || au_sbbot(sb) < bindex) {
16777 + errorfc(fc, "out of bounds, %d", bindex);
16782 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
16783 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
16786 + aufs_read_unlock(root, !AuLock_IR);
16792 +static int au_fsctx_parse_mod(struct fs_context *fc, struct au_opt_mod *mod,
16793 + struct fs_parameter *param)
16796 + struct path path;
16800 + /* will be freed by au_fsctx_free() */
16801 + mod->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16802 + if (unlikely(!mod->path))
16806 + p = strchr(mod->path, '=');
16807 + if (unlikely(!p)) {
16808 + errorfc(fc, "no permission %s", mod->path);
16813 + err = vfsub_kern_path(mod->path, AuOpt_LkupDirFlags, &path);
16814 + if (unlikely(err)) {
16815 + errorfc(fc, "lookup failed %s (%d)", mod->path, err);
16819 + mod->perm = au_br_perm_val(p);
16820 + AuDbg("mod path %s, perm 0x%x, %s", mod->path, mod->perm, p);
16821 + mod->h_root = dget(path.dentry);
16829 +#if 0 /* reserved for future use */
16830 +static int au_fsctx_parse_imod(struct fs_context *fc, struct au_opt_mod *mod,
16835 + struct super_block *sb;
16836 + struct dentry *root;
16837 + struct au_fsctx_opts *a = fc->fs_private;
16843 + p = strchr(ibrspec, ':');
16844 + if (unlikely(!p)) {
16845 + errorfc(fc, "no index, %s", ibrspec);
16849 + err = kstrtoint(ibrspec, 0, &n);
16850 + if (unlikely(err)) {
16851 + errorfc(fc, "bad integer in %s", ibrspec);
16854 + AuDbg("n %d\n", n);
16856 + root = sb->s_root;
16857 + aufs_read_lock(root, AuLock_FLUSH);
16858 + if (n < 0 || au_sbbot(sb) < n) {
16859 + errorfc(fc, "out of bounds, %d", bindex);
16864 + mod->perm = au_br_perm_val(p);
16865 + AuDbg("mod path %s, perm 0x%x, %s\n",
16866 + mod->path, mod->perm, p);
16867 + mod->h_root = dget(au_h_dptr(root, bindex));
16870 + aufs_read_unlock(root, !AuLock_IR);
16877 +static int au_fsctx_parse_xino(struct fs_context *fc,
16878 + struct au_opt_xino *xino,
16879 + struct fs_parameter *param)
16882 + struct au_fsctx_opts *a = fc->fs_private;
16885 + /* will be freed by au_opts_free() */
16886 + xino->path = kmemdup_nul(param->string, param->size, GFP_NOFS);
16887 + if (unlikely(!xino->path))
16889 + AuDbg("path %s\n", xino->path);
16891 + xino->file = au_xino_create(a->sb, xino->path, /*silent*/0,
16893 + err = PTR_ERR(xino->file);
16894 + if (IS_ERR(xino->file)) {
16895 + xino->file = NULL;
16900 + if (unlikely(a->sb && xino->file->f_path.dentry->d_sb == a->sb)) {
16902 + errorfc(fc, "%s must be outside", xino->path);
16911 +int au_fsctx_parse_xino_itrunc_path(struct fs_context *fc,
16912 + struct au_opt_xino_itrunc *xino_itrunc,
16916 + aufs_bindex_t bbot, bindex;
16917 + struct path path;
16918 + struct dentry *root;
16919 + struct au_fsctx_opts *a = fc->fs_private;
16921 + AuDebugOn(!a->sb);
16923 + err = vfsub_kern_path(pathname, AuOpt_LkupDirFlags, &path);
16924 + if (unlikely(err)) {
16925 + errorfc(fc, "lookup failed %s (%d)", pathname, err);
16929 + xino_itrunc->bindex = -1;
16930 + root = a->sb->s_root;
16931 + aufs_read_lock(root, AuLock_FLUSH);
16932 + bbot = au_sbbot(a->sb);
16933 + for (bindex = 0; bindex <= bbot; bindex++) {
16934 + if (au_h_dptr(root, bindex) == path.dentry) {
16935 + xino_itrunc->bindex = bindex;
16939 + aufs_read_unlock(root, !AuLock_IR);
16942 + if (unlikely(xino_itrunc->bindex < 0)) {
16944 + errorfc(fc, "no such branch %s", pathname);
16952 +static int au_fsctx_parse_xino_itrunc(struct fs_context *fc,
16953 + struct au_opt_xino_itrunc *xino_itrunc,
16954 + unsigned int bindex)
16957 + aufs_bindex_t bbot;
16958 + struct super_block *sb;
16959 + struct au_fsctx_opts *a = fc->fs_private;
16965 + si_noflush_read_lock(sb);
16966 + bbot = au_sbbot(sb);
16967 + si_read_unlock(sb);
16968 + if (bindex <= bbot)
16969 + xino_itrunc->bindex = bindex;
16972 + errorfc(fc, "out of bounds, %u", bindex);
16979 +static int au_fsctx_parse_param(struct fs_context *fc, struct fs_parameter *param)
16982 + struct fs_parse_result result;
16983 + struct au_fsctx_opts *a = fc->fs_private;
16984 + struct au_opt *opt = a->opt;
16986 + AuDbg("fc %p, param {key %s, string %s}\n",
16987 + fc, param->key, param->string);
16988 + err = fs_parse(fc, aufs_fsctx_paramspec, param, &result);
16989 + if (unlikely(err < 0))
16992 + AuDbg("token %d, res{negated %d, uint64 %llu}\n",
16993 + token, result.negated, result.uint_64);
16999 + err = au_fsctx_parse_br(fc, param->string);
17002 + err = au_fsctx_parse_add(fc, param->string);
17005 + err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17006 + /*dummy bindex*/1);
17008 + case Opt_prepend:
17009 + err = au_fsctx_parse_do_add(fc, opt, param->string, param->size,
17014 + err = au_fsctx_parse_del(fc, &opt->del, param);
17016 +#if 0 /* reserved for future use */
17023 + del->pathname = "(indexed)";
17024 + err = au_opts_parse_idel(fc, &opt->del, result.uint_32);
17029 + err = au_fsctx_parse_mod(fc, &opt->mod, param);
17031 +#ifdef IMOD /* reserved for future use */
17038 + u.mod->path = "(indexed)";
17039 + err = au_opts_parse_imod(fc, &opt->mod, param->string);
17044 + err = au_fsctx_parse_xino(fc, &opt->xino, param);
17046 + case Opt_trunc_xino_path:
17048 + errorfc(fc, "no such branch %s", param->string);
17051 + err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17055 + case Opt_trunc_xino_v:
17061 + err = au_fsctx_parse_xino_itrunc_path(fc, &opt->xino_itrunc,
17065 + case Opt_itrunc_xino:
17067 + errorfc(fc, "out of bounds %s", param->string);
17070 + err = au_fsctx_parse_xino_itrunc(fc, &opt->xino_itrunc,
17076 + opt->dirwh = result.int_32;
17079 + case Opt_rdcache:
17080 + if (unlikely(result.int_32 > AUFS_RDCACHE_MAX)) {
17081 + errorfc(fc, "rdcache must be smaller than %d",
17082 + AUFS_RDCACHE_MAX);
17086 + opt->rdcache = result.int_32;
17091 + opt->rdblk = AUFS_RDBLK_DEF;
17092 + if (!strcmp(param->string, "def"))
17095 + err = kstrtoint(param->string, 0, &result.int_32);
17096 + if (unlikely(err)) {
17097 + errorfc(fc, "bad value in %s", param->key);
17101 + if (unlikely(result.int_32 < 0
17102 + || result.int_32 > KMALLOC_MAX_SIZE)) {
17103 + errorfc(fc, "bad value in %s", param->key);
17106 + if (unlikely(result.int_32 && result.int_32 < NAME_MAX)) {
17107 + errorfc(fc, "rdblk must be larger than %d", NAME_MAX);
17111 + opt->rdblk = result.int_32;
17116 + opt->rdhash = AUFS_RDHASH_DEF;
17117 + if (!strcmp(param->string, "def"))
17120 + err = kstrtoint(param->string, 0, &result.int_32);
17121 + if (unlikely(err)) {
17122 + errorfc(fc, "bad value in %s", param->key);
17125 + /* how about zero? */
17126 + if (result.int_32 < 0
17127 + || result.int_32 * sizeof(struct hlist_head)
17128 + > KMALLOC_MAX_SIZE) {
17130 + errorfc(fc, "bad integer in %s", param->key);
17133 + opt->rdhash = result.int_32;
17138 + * As other options, fs/aufs/opts.c can handle these strings by
17139 + * match_token(). But "diropq=" is deprecated now and will
17140 + * never have other value. So simple strcmp() is enough here.
17142 + if (!strcmp(param->string, "a") ||
17143 + !strcmp(param->string, "always")) {
17145 + opt->type = Opt_diropq_a;
17146 + } else if (!strcmp(param->string, "w") ||
17147 + !strcmp(param->string, "whiteouted")) {
17149 + opt->type = Opt_diropq_w;
17151 + errorfc(fc, "unknown value %s", param->string);
17155 + opt->udba = au_udba_val(param->string);
17156 + if (opt->udba >= 0)
17159 + errorf(fc, "wrong value, %s", param->string);
17162 + case Opt_wbr_create:
17163 + opt->wbr_create.wbr_create
17164 + = au_wbr_create_val(param->string, &opt->wbr_create);
17165 + if (opt->wbr_create.wbr_create >= 0)
17168 + errorf(fc, "wrong value, %s", param->key);
17171 + case Opt_wbr_copyup:
17172 + opt->wbr_copyup = au_wbr_copyup_val(param->string);
17173 + if (opt->wbr_copyup >= 0)
17176 + errorfc(fc, "wrong value, %s", param->key);
17179 + case Opt_fhsm_sec:
17180 + if (unlikely(result.int_32 < 0)) {
17181 + errorfc(fc, "bad integer in %s\n", param->key);
17186 + opt->fhsm_second = result.int_32;
17188 + warnfc(fc, "ignored %s %s", param->key, param->string);
17191 + /* simple true/false flag */
17192 +#define au_fsctx_TF(name) \
17193 + case Opt_##name: \
17195 + opt->tf = !result.negated; \
17197 + au_fsctx_TF(trunc_xino);
17198 + au_fsctx_TF(trunc_xib);
17199 + au_fsctx_TF(dirperm1);
17200 + au_fsctx_TF(plink);
17201 + au_fsctx_TF(shwh);
17202 + au_fsctx_TF(dio);
17203 + au_fsctx_TF(warn_perm);
17204 + au_fsctx_TF(verbose);
17205 + au_fsctx_TF(sum);
17206 + au_fsctx_TF(dirren);
17207 + au_fsctx_TF(acl);
17208 +#undef au_fsctx_TF
17210 + case Opt_noverbose:
17212 + opt->type = Opt_verbose;
17218 + case Opt_list_plink:
17225 + warnfc(fc, "ignored %s", param->key);
17227 + case Opt_ignore_silent:
17236 + if (unlikely(err))
17244 + case Opt_noverbose:
17249 + opt->type = token;
17253 + if (unlikely(opt > a->opt_tail)) {
17257 + opt->type = Opt_tail;
17265 + * these options accept both 'name=val' and 'name:val' form.
17266 + * some accept optional '=' in its value.
17267 + * eg. br:/br1=rw:/br2=ro and br=/br1=rw:/br2=ro
17269 +static inline unsigned int is_colonopt(char *str)
17271 +#define do_test(name) \
17272 + if (!strncmp(str, name ":", sizeof(name))) \
17273 + return sizeof(name) - 1
17277 + do_test("append");
17278 + do_test("prepend");
17280 + /* do_test("idel"); */
17282 + /* do_test("imod"); */
17288 +static int au_fsctx_parse_monolithic(struct fs_context *fc, void *data)
17293 + struct au_fsctx_opts *a = fc->fs_private;
17296 + AuDbg("str %s\n", str);
17298 + u = is_colonopt(str);
17301 + str = strchr(str, ',');
17307 + AuDbg("str %s\n", str);
17309 + err = generic_parse_monolithic(fc, str);
17311 + au_fsctx_dump(&a->opts);
17316 +/* ---------------------------------------------------------------------- */
17318 +static void au_fsctx_opts_free(struct au_opts *opts)
17320 + struct au_opt *opt;
17323 + while (opt->type != Opt_tail) {
17324 + switch (opt->type) {
17329 + case Opt_prepend:
17330 + kfree(opt->add.pathname);
17331 + path_put(&opt->add.path);
17334 + kfree(opt->del.pathname);
17337 + path_put(&opt->del.h_path);
17340 + kfree(opt->mod.path);
17343 + dput(opt->mod.h_root);
17346 + kfree(opt->xino.path);
17347 + fput(opt->xino.file);
17354 +static void au_fsctx_free(struct fs_context *fc)
17356 + struct au_fsctx_opts *a = fc->fs_private;
17358 + /* fs_type=%p, root=%pD */
17359 + AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17360 + fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17362 + kobject_put(&a->sbinfo->si_kobj);
17363 + au_fsctx_opts_free(&a->opts);
17364 + free_page((unsigned long)a->opts.opt);
17368 +static const struct fs_context_operations au_fsctx_ops = {
17369 + .free = au_fsctx_free,
17370 + .parse_param = au_fsctx_parse_param,
17371 + .parse_monolithic = au_fsctx_parse_monolithic,
17372 + .get_tree = au_fsctx_get_tree,
17373 + .reconfigure = au_fsctx_reconfigure
17375 + * nfs4 requires ->dup()? No.
17376 + * I don't know what is this ->dup() for.
17380 +int aufs_fsctx_init(struct fs_context *fc)
17383 + struct au_fsctx_opts *a;
17385 + /* fs_type=%p, root=%pD */
17386 + AuDbg("fc %p{sb_flags 0x%x, sb_flags_mask 0x%x, purpose %u\n",
17387 + fc, fc->sb_flags, fc->sb_flags_mask, fc->purpose);
17389 + /* they will be freed by au_fsctx_free() */
17391 + a = kzalloc(sizeof(*a), GFP_NOFS);
17392 + if (unlikely(!a))
17395 + a->opts.opt = (void *)__get_free_page(GFP_NOFS);
17396 + if (unlikely(!a->opts.opt))
17398 + a->opt = a->opts.opt;
17399 + a->opt->type = Opt_tail;
17400 + a->opts.max_opt = PAGE_SIZE / sizeof(*a->opts.opt);
17401 + a->opt_tail = a->opt + a->opts.max_opt - 1;
17402 + a->opts.sb_flags = fc->sb_flags;
17406 + AuDebugOn(fc->purpose != FS_CONTEXT_FOR_RECONFIGURE);
17407 + a->opts.flags = AuOpts_REMOUNT;
17408 + a->sb = fc->root->d_sb;
17409 + a->sbinfo = au_sbi(a->sb);
17410 + kobject_get(&a->sbinfo->si_kobj);
17412 + a->sbinfo = au_si_alloc(a->sb);
17413 + AuDebugOn(!a->sbinfo);
17414 + err = PTR_ERR(a->sbinfo);
17415 + if (IS_ERR(a->sbinfo))
17417 + au_rw_write_unlock(&a->sbinfo->si_rwsem);
17421 + fc->fs_private = a;
17422 + fc->ops = &au_fsctx_ops;
17423 + goto out; /* success */
17426 + free_page((unsigned long)a->opts.opt);
17433 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
17434 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
17435 +++ linux/fs/aufs/fstype.h 2022-11-05 23:02:18.965889284 +0100
17437 +/* SPDX-License-Identifier: GPL-2.0 */
17439 + * Copyright (C) 2005-2022 Junjiro R. Okajima
17441 + * This program is free software; you can redistribute it and/or modify
17442 + * it under the terms of the GNU General Public License as published by
17443 + * the Free Software Foundation; either version 2 of the License, or
17444 + * (at your option) any later version.
17446 + * This program is distributed in the hope that it will be useful,
17447 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17448 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17449 + * GNU General Public License for more details.
17451 + * You should have received a copy of the GNU General Public License
17452 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17456 + * judging filesystem type
17459 +#ifndef __AUFS_FSTYPE_H__
17460 +#define __AUFS_FSTYPE_H__
17464 +#include <linux/fs.h>
17465 +#include <linux/magic.h>
17466 +#include <linux/nfs_fs.h>
17467 +#include <linux/romfs_fs.h>
17469 +static inline int au_test_aufs(struct super_block *sb)
17471 + return sb->s_magic == AUFS_SUPER_MAGIC;
17474 +static inline const char *au_sbtype(struct super_block *sb)
17476 + return sb->s_type->name;
17479 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
17481 +#if IS_ENABLED(CONFIG_ISO9660_FS)
17482 + return sb->s_magic == ISOFS_SUPER_MAGIC;
17488 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
17490 +#if IS_ENABLED(CONFIG_ROMFS_FS)
17491 + return sb->s_magic == ROMFS_MAGIC;
17497 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
17499 +#if IS_ENABLED(CONFIG_CRAMFS)
17500 + return sb->s_magic == CRAMFS_MAGIC;
17505 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
17507 +#if IS_ENABLED(CONFIG_NFS_FS)
17508 + return sb->s_magic == NFS_SUPER_MAGIC;
17514 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
17516 +#if IS_ENABLED(CONFIG_FUSE_FS)
17517 + return sb->s_magic == FUSE_SUPER_MAGIC;
17523 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
17525 +#if IS_ENABLED(CONFIG_XFS_FS)
17526 + return sb->s_magic == XFS_SB_MAGIC;
17532 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
17534 +#ifdef CONFIG_TMPFS
17535 + return sb->s_magic == TMPFS_MAGIC;
17541 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
17543 +#if IS_ENABLED(CONFIG_ECRYPT_FS)
17544 + return !strcmp(au_sbtype(sb), "ecryptfs");
17550 +static inline int au_test_ramfs(struct super_block *sb)
17552 + return sb->s_magic == RAMFS_MAGIC;
17555 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
17557 +#if IS_ENABLED(CONFIG_UBIFS_FS)
17558 + return sb->s_magic == UBIFS_SUPER_MAGIC;
17564 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
17566 +#ifdef CONFIG_PROC_FS
17567 + return sb->s_magic == PROC_SUPER_MAGIC;
17573 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
17575 +#ifdef CONFIG_SYSFS
17576 + return sb->s_magic == SYSFS_MAGIC;
17582 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
17584 +#if IS_ENABLED(CONFIG_CONFIGFS_FS)
17585 + return sb->s_magic == CONFIGFS_MAGIC;
17591 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
17593 +#if IS_ENABLED(CONFIG_MINIX_FS)
17594 + return sb->s_magic == MINIX3_SUPER_MAGIC
17595 + || sb->s_magic == MINIX2_SUPER_MAGIC
17596 + || sb->s_magic == MINIX2_SUPER_MAGIC2
17597 + || sb->s_magic == MINIX_SUPER_MAGIC
17598 + || sb->s_magic == MINIX_SUPER_MAGIC2;
17604 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
17606 +#if IS_ENABLED(CONFIG_FAT_FS)
17607 + return sb->s_magic == MSDOS_SUPER_MAGIC;
17613 +static inline int au_test_msdos(struct super_block *sb)
17615 + return au_test_fat(sb);
17618 +static inline int au_test_vfat(struct super_block *sb)
17620 + return au_test_fat(sb);
17623 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
17625 +#ifdef CONFIG_SECURITYFS
17626 + return sb->s_magic == SECURITYFS_MAGIC;
17632 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
17634 +#if IS_ENABLED(CONFIG_SQUASHFS)
17635 + return sb->s_magic == SQUASHFS_MAGIC;
17641 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
17643 +#if IS_ENABLED(CONFIG_BTRFS_FS)
17644 + return sb->s_magic == BTRFS_SUPER_MAGIC;
17650 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
17652 +#if IS_ENABLED(CONFIG_XENFS)
17653 + return sb->s_magic == XENFS_SUPER_MAGIC;
17659 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
17661 +#ifdef CONFIG_DEBUG_FS
17662 + return sb->s_magic == DEBUGFS_MAGIC;
17668 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
17670 +#if IS_ENABLED(CONFIG_NILFS)
17671 + return sb->s_magic == NILFS_SUPER_MAGIC;
17677 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
17679 +#if IS_ENABLED(CONFIG_HFSPLUS_FS)
17680 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
17686 +/* ---------------------------------------------------------------------- */
17688 + * they can't be an aufs branch.
17690 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
17693 +#ifndef CONFIG_AUFS_BR_RAMFS
17694 + au_test_ramfs(sb) ||
17696 + au_test_procfs(sb)
17697 + || au_test_sysfs(sb)
17698 + || au_test_configfs(sb)
17699 + || au_test_debugfs(sb)
17700 + || au_test_securityfs(sb)
17701 + || au_test_xenfs(sb)
17702 + || au_test_ecryptfs(sb)
17703 + /* || !strcmp(au_sbtype(sb), "unionfs") */
17704 + || au_test_aufs(sb); /* will be supported in next version */
17707 +static inline int au_test_fs_remote(struct super_block *sb)
17709 + return !au_test_tmpfs(sb)
17710 +#ifdef CONFIG_AUFS_BR_RAMFS
17711 + && !au_test_ramfs(sb)
17713 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
17716 +/* ---------------------------------------------------------------------- */
17719 + * Note: these functions (below) are created after reading ->getattr() in all
17720 + * filesystems under linux/fs. it means we have to do so in every update...
17724 + * some filesystems require getattr to refresh the inode attributes before
17726 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
17727 + * and leave the work for d_revalidate()
17729 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
17731 + return au_test_nfs(sb)
17732 + || au_test_fuse(sb)
17733 + /* || au_test_btrfs(sb) */ /* untested */
17738 + * filesystems which don't maintain i_size or i_blocks.
17740 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
17742 + return au_test_xfs(sb)
17743 + || au_test_btrfs(sb)
17744 + || au_test_ubifs(sb)
17745 + || au_test_hfsplus(sb) /* maintained, but incorrect */
17746 + /* || au_test_minix(sb) */ /* untested */
17751 + * filesystems which don't store the correct value in some of their inode
17754 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
17756 + return au_test_fs_bad_iattr_size(sb)
17757 + || au_test_fat(sb)
17758 + || au_test_msdos(sb)
17759 + || au_test_vfat(sb);
17762 +/* they don't check i_nlink in link(2) */
17763 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
17765 + return au_test_tmpfs(sb)
17766 +#ifdef CONFIG_AUFS_BR_RAMFS
17767 + || au_test_ramfs(sb)
17769 + || au_test_ubifs(sb)
17770 + || au_test_hfsplus(sb);
17774 + * filesystems which sets S_NOATIME and S_NOCMTIME.
17776 +static inline int au_test_fs_notime(struct super_block *sb)
17778 + return au_test_nfs(sb)
17779 + || au_test_fuse(sb)
17780 + || au_test_ubifs(sb)
17784 +/* temporary support for i#1 in cramfs */
17785 +static inline int au_test_fs_unique_ino(struct inode *inode)
17787 + if (au_test_cramfs(inode->i_sb))
17788 + return inode->i_ino != 1;
17792 +/* ---------------------------------------------------------------------- */
17795 + * the filesystem where the xino files placed must support i/o after unlink and
17796 + * maintain i_size and i_blocks.
17798 +static inline int au_test_fs_bad_xino(struct super_block *sb)
17800 + return au_test_fs_remote(sb)
17801 + || au_test_fs_bad_iattr_size(sb)
17802 + /* don't want unnecessary work for xino */
17803 + || au_test_aufs(sb)
17804 + || au_test_ecryptfs(sb)
17805 + || au_test_nilfs(sb);
17808 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
17810 + return au_test_tmpfs(sb)
17811 + || au_test_ramfs(sb);
17815 + * test if the @sb is real-readonly.
17817 +static inline int au_test_fs_rr(struct super_block *sb)
17819 + return au_test_squashfs(sb)
17820 + || au_test_iso9660(sb)
17821 + || au_test_cramfs(sb)
17822 + || au_test_romfs(sb);
17826 + * test if the @inode is nfs with 'noacl' option
17827 + * NFS always sets SB_POSIXACL regardless its mount option 'noacl.'
17829 +static inline int au_test_nfs_noacl(struct inode *inode)
17831 + return au_test_nfs(inode->i_sb)
17832 + /* && IS_POSIXACL(inode) */
17833 + && !nfs_server_capable(inode, NFS_CAP_ACLS);
17836 +#endif /* __KERNEL__ */
17837 +#endif /* __AUFS_FSTYPE_H__ */
17838 diff -urN /usr/share/empty/fs/aufs/hbl.h linux/fs/aufs/hbl.h
17839 --- /usr/share/empty/fs/aufs/hbl.h 1970-01-01 01:00:00.000000000 +0100
17840 +++ linux/fs/aufs/hbl.h 2022-11-05 23:02:18.965889284 +0100
17842 +/* SPDX-License-Identifier: GPL-2.0 */
17844 + * Copyright (C) 2017-2022 Junjiro R. Okajima
17846 + * This program is free software; you can redistribute it and/or modify
17847 + * it under the terms of the GNU General Public License as published by
17848 + * the Free Software Foundation; either version 2 of the License, or
17849 + * (at your option) any later version.
17851 + * This program is distributed in the hope that it will be useful,
17852 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17853 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17854 + * GNU General Public License for more details.
17856 + * You should have received a copy of the GNU General Public License
17857 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17861 + * helpers for hlist_bl.h
17864 +#ifndef __AUFS_HBL_H__
17865 +#define __AUFS_HBL_H__
17869 +#include <linux/list_bl.h>
17871 +static inline void au_hbl_add(struct hlist_bl_node *node,
17872 + struct hlist_bl_head *hbl)
17874 + hlist_bl_lock(hbl);
17875 + hlist_bl_add_head(node, hbl);
17876 + hlist_bl_unlock(hbl);
17879 +static inline void au_hbl_del(struct hlist_bl_node *node,
17880 + struct hlist_bl_head *hbl)
17882 + hlist_bl_lock(hbl);
17883 + hlist_bl_del(node);
17884 + hlist_bl_unlock(hbl);
17887 +#define au_hbl_for_each(pos, head) \
17888 + for (pos = hlist_bl_first(head); \
17892 +static inline unsigned long au_hbl_count(struct hlist_bl_head *hbl)
17894 + unsigned long cnt;
17895 + struct hlist_bl_node *pos;
17898 + hlist_bl_lock(hbl);
17899 + au_hbl_for_each(pos, hbl)
17901 + hlist_bl_unlock(hbl);
17905 +#endif /* __KERNEL__ */
17906 +#endif /* __AUFS_HBL_H__ */
17907 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
17908 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
17909 +++ linux/fs/aufs/hfsnotify.c 2022-11-05 23:02:18.965889284 +0100
17911 +// SPDX-License-Identifier: GPL-2.0
17913 + * Copyright (C) 2005-2022 Junjiro R. Okajima
17915 + * This program is free software; you can redistribute it and/or modify
17916 + * it under the terms of the GNU General Public License as published by
17917 + * the Free Software Foundation; either version 2 of the License, or
17918 + * (at your option) any later version.
17920 + * This program is distributed in the hope that it will be useful,
17921 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17922 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17923 + * GNU General Public License for more details.
17925 + * You should have received a copy of the GNU General Public License
17926 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17930 + * fsnotify for the lower directories
17935 +/* FS_IN_IGNORED is unnecessary */
17936 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
17937 + | FS_CREATE | FS_EVENT_ON_CHILD);
17938 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
17939 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
17941 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
17943 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
17945 + /* AuDbg("here\n"); */
17946 + au_cache_free_hnotify(hn);
17947 + smp_mb__before_atomic(); /* for atomic64_dec */
17948 + if (atomic64_dec_and_test(&au_hfsn_ifree))
17949 + wake_up(&au_hfsn_wq);
17952 +static int au_hfsn_alloc(struct au_hinode *hinode)
17955 + struct au_hnotify *hn;
17956 + struct super_block *sb;
17957 + struct au_branch *br;
17958 + struct fsnotify_mark *mark;
17959 + aufs_bindex_t bindex;
17961 + hn = hinode->hi_notify;
17962 + sb = hn->hn_aufs_inode->i_sb;
17963 + bindex = au_br_index(sb, hinode->hi_id);
17964 + br = au_sbr(sb, bindex);
17965 + AuDebugOn(!br->br_hfsn);
17967 + mark = &hn->hn_mark;
17968 + fsnotify_init_mark(mark, br->br_hfsn->hfsn_group);
17969 + mark->mask = AuHfsnMask;
17971 + * by udba rename or rmdir, aufs assign a new inode to the known
17972 + * h_inode, so specify 1 to allow dups.
17975 + err = fsnotify_add_inode_mark(mark, hinode->hi_inode, /*allow_dups*/1);
17981 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
17983 + struct fsnotify_mark *mark;
17984 + unsigned long long ull;
17985 + struct fsnotify_group *group;
17987 + ull = atomic64_inc_return(&au_hfsn_ifree);
17990 + mark = &hn->hn_mark;
17991 + spin_lock(&mark->lock);
17992 + group = mark->group;
17993 + fsnotify_get_group(group);
17994 + spin_unlock(&mark->lock);
17996 + fsnotify_destroy_mark(mark, group);
17997 + fsnotify_put_mark(mark);
17998 + fsnotify_put_group(group);
18001 + /* free hn by myself */
18005 +/* ---------------------------------------------------------------------- */
18007 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
18009 + struct fsnotify_mark *mark;
18011 + mark = &hinode->hi_notify->hn_mark;
18012 + spin_lock(&mark->lock);
18014 + AuDebugOn(mark->mask & AuHfsnMask);
18015 + mark->mask |= AuHfsnMask;
18017 + AuDebugOn(!(mark->mask & AuHfsnMask));
18018 + mark->mask &= ~AuHfsnMask;
18020 + spin_unlock(&mark->lock);
18021 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
18024 +/* ---------------------------------------------------------------------- */
18026 +/* #define AuDbgHnotify */
18027 +#ifdef AuDbgHnotify
18028 +static char *au_hfsn_name(u32 mask)
18030 +#ifdef CONFIG_AUFS_DEBUG
18031 +#define test_ret(flag) \
18033 + if (mask & flag) \
18036 + test_ret(FS_ACCESS);
18037 + test_ret(FS_MODIFY);
18038 + test_ret(FS_ATTRIB);
18039 + test_ret(FS_CLOSE_WRITE);
18040 + test_ret(FS_CLOSE_NOWRITE);
18041 + test_ret(FS_OPEN);
18042 + test_ret(FS_MOVED_FROM);
18043 + test_ret(FS_MOVED_TO);
18044 + test_ret(FS_CREATE);
18045 + test_ret(FS_DELETE);
18046 + test_ret(FS_DELETE_SELF);
18047 + test_ret(FS_MOVE_SELF);
18048 + test_ret(FS_UNMOUNT);
18049 + test_ret(FS_Q_OVERFLOW);
18050 + test_ret(FS_IN_IGNORED);
18051 + test_ret(FS_ISDIR);
18052 + test_ret(FS_IN_ONESHOT);
18053 + test_ret(FS_EVENT_ON_CHILD);
18062 +/* ---------------------------------------------------------------------- */
18064 +static void au_hfsn_free_group(struct fsnotify_group *group)
18066 + struct au_br_hfsnotify *hfsn = group->private;
18068 + /* AuDbg("here\n"); */
18069 + au_kfree_try_rcu(hfsn);
18072 +static int au_hfsn_handle_event(struct fsnotify_group *group,
18073 + u32 mask, const void *data, int data_type,
18074 + struct inode *dir,
18075 + const struct qstr *file_name, u32 cookie,
18076 + struct fsnotify_iter_info *iter_info)
18079 + struct au_hnotify *hnotify;
18080 + struct inode *h_dir, *h_inode;
18081 + struct fsnotify_mark *inode_mark;
18083 + AuDebugOn(!(data_type == FSNOTIFY_EVENT_INODE
18084 + || data_type == FSNOTIFY_EVENT_DENTRY));
18087 + /* if FS_UNMOUNT happens, there must be another bug */
18088 + AuDebugOn(mask & FS_UNMOUNT);
18089 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
18094 +#ifdef AuDbgHnotify
18096 + if (1 || file_name.len != sizeof(AUFS_XINO_FNAME) - 1
18097 + || strncmp(file_name.name, AUFS_XINO_FNAME, file_name.len)) {
18098 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
18099 + h_dir->i_ino, mask, au_hfsn_name(mask),
18100 + AuLNPair(file_name), h_inode ? h_inode->i_ino : 0);
18101 + /* WARN_ON(1); */
18106 + inode_mark = fsnotify_iter_inode_mark(iter_info);
18107 + AuDebugOn(!inode_mark);
18108 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
18109 + err = au_hnotify(h_dir, hnotify, mask, file_name, h_inode);
18115 +static struct fsnotify_ops au_hfsn_ops = {
18116 + .handle_event = au_hfsn_handle_event,
18117 + .free_group_priv = au_hfsn_free_group,
18118 + .free_mark = au_hfsn_free_mark
18121 +/* ---------------------------------------------------------------------- */
18123 +static void au_hfsn_fin_br(struct au_branch *br)
18125 + struct au_br_hfsnotify *hfsn;
18127 + hfsn = br->br_hfsn;
18130 + fsnotify_put_group(hfsn->hfsn_group);
18135 +static int au_hfsn_init_br(struct au_branch *br, int perm)
18138 + struct fsnotify_group *group;
18139 + struct au_br_hfsnotify *hfsn;
18142 + br->br_hfsn = NULL;
18143 + if (!au_br_hnotifyable(perm))
18147 + hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
18148 + if (unlikely(!hfsn))
18152 + group = fsnotify_alloc_group(&au_hfsn_ops,
18153 + /*flags - not for userspace*/0);
18154 + if (IS_ERR(group)) {
18155 + err = PTR_ERR(group);
18156 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
18160 + group->private = hfsn;
18161 + hfsn->hfsn_group = group;
18162 + br->br_hfsn = hfsn;
18163 + goto out; /* success */
18166 + au_kfree_try_rcu(hfsn);
18171 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
18176 + if (!br->br_hfsn)
18177 + err = au_hfsn_init_br(br, perm);
18182 +/* ---------------------------------------------------------------------- */
18184 +static void au_hfsn_fin(void)
18186 + AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
18187 + wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
18190 +const struct au_hnotify_op au_hnotify_op = {
18191 + .ctl = au_hfsn_ctl,
18192 + .alloc = au_hfsn_alloc,
18193 + .free = au_hfsn_free,
18195 + .fin = au_hfsn_fin,
18197 + .reset_br = au_hfsn_reset_br,
18198 + .fin_br = au_hfsn_fin_br,
18199 + .init_br = au_hfsn_init_br
18201 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
18202 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
18203 +++ linux/fs/aufs/hfsplus.c 2022-11-05 23:02:18.965889284 +0100
18205 +// SPDX-License-Identifier: GPL-2.0
18207 + * Copyright (C) 2010-2022 Junjiro R. Okajima
18209 + * This program is free software; you can redistribute it and/or modify
18210 + * it under the terms of the GNU General Public License as published by
18211 + * the Free Software Foundation; either version 2 of the License, or
18212 + * (at your option) any later version.
18214 + * This program is distributed in the hope that it will be useful,
18215 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18216 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18217 + * GNU General Public License for more details.
18219 + * You should have received a copy of the GNU General Public License
18220 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18224 + * special support for filesystems which acquires an inode mutex
18225 + * at final closing a file, eg, hfsplus.
18227 + * This trick is very simple and stupid, just to open the file before really
18228 + * necessary open to tell hfsplus that this is not the final closing.
18229 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
18230 + * and au_h_open_post() after releasing it.
18235 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
18238 + struct file *h_file;
18239 + struct dentry *h_dentry;
18241 + h_dentry = au_h_dptr(dentry, bindex);
18242 + AuDebugOn(!h_dentry);
18243 + AuDebugOn(d_is_negative(h_dentry));
18246 + if (au_test_hfsplus(h_dentry->d_sb)
18247 + && d_is_reg(h_dentry))
18248 + h_file = au_h_open(dentry, bindex,
18249 + O_RDONLY | O_NOATIME | O_LARGEFILE,
18250 + /*file*/NULL, force_wr);
18254 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
18255 + struct file *h_file)
18257 + struct au_branch *br;
18261 + br = au_sbr(dentry->d_sb, bindex);
18262 + au_lcnt_dec(&br->br_nfiles);
18265 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
18266 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
18267 +++ linux/fs/aufs/hnotify.c 2022-11-05 23:02:18.965889284 +0100
18269 +// SPDX-License-Identifier: GPL-2.0
18271 + * Copyright (C) 2005-2022 Junjiro R. Okajima
18273 + * This program is free software; you can redistribute it and/or modify
18274 + * it under the terms of the GNU General Public License as published by
18275 + * the Free Software Foundation; either version 2 of the License, or
18276 + * (at your option) any later version.
18278 + * This program is distributed in the hope that it will be useful,
18279 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18280 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18281 + * GNU General Public License for more details.
18283 + * You should have received a copy of the GNU General Public License
18284 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18288 + * abstraction to notify the direct changes on lower directories
18291 +/* #include <linux/iversion.h> */
18294 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
18297 + struct au_hnotify *hn;
18300 + hn = au_cache_alloc_hnotify();
18302 + hn->hn_aufs_inode = inode;
18303 + hinode->hi_notify = hn;
18304 + err = au_hnotify_op.alloc(hinode);
18306 + if (unlikely(err)) {
18307 + hinode->hi_notify = NULL;
18308 + au_cache_free_hnotify(hn);
18310 + * The upper dir was removed by udba, but the same named
18311 + * dir left. In this case, aufs assigns a new inode
18312 + * number and set the monitor again.
18313 + * For the lower dir, the old monitor is still left.
18315 + if (err == -EEXIST)
18324 +void au_hn_free(struct au_hinode *hinode)
18326 + struct au_hnotify *hn;
18328 + hn = hinode->hi_notify;
18330 + hinode->hi_notify = NULL;
18331 + if (au_hnotify_op.free(hinode, hn))
18332 + au_cache_free_hnotify(hn);
18336 +/* ---------------------------------------------------------------------- */
18338 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
18340 + if (hinode->hi_notify)
18341 + au_hnotify_op.ctl(hinode, do_set);
18344 +void au_hn_reset(struct inode *inode, unsigned int flags)
18346 + aufs_bindex_t bindex, bbot;
18347 + struct inode *hi;
18348 + struct dentry *iwhdentry;
18350 + bbot = au_ibbot(inode);
18351 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
18352 + hi = au_h_iptr(inode, bindex);
18356 + /* inode_lock_nested(hi, AuLsc_I_CHILD); */
18357 + iwhdentry = au_hi_wh(inode, bindex);
18361 + au_set_h_iptr(inode, bindex, NULL, 0);
18362 + au_set_h_iptr(inode, bindex, au_igrab(hi),
18363 + flags & ~AuHi_XINO);
18366 + /* inode_unlock(hi); */
18370 +/* ---------------------------------------------------------------------- */
18372 +static int hn_xino(struct inode *inode, struct inode *h_inode)
18375 + aufs_bindex_t bindex, bbot, bfound, btop;
18376 + struct inode *h_i;
18379 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18380 + pr_warn("branch root dir was changed\n");
18385 + bbot = au_ibbot(inode);
18386 + btop = au_ibtop(inode);
18387 +#if 0 /* reserved for future use */
18388 + if (bindex == bbot) {
18389 + /* keep this ino in rename case */
18393 + for (bindex = btop; bindex <= bbot; bindex++)
18394 + if (au_h_iptr(inode, bindex) == h_inode) {
18401 + for (bindex = btop; bindex <= bbot; bindex++) {
18402 + h_i = au_h_iptr(inode, bindex);
18406 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
18407 + /* ignore this error */
18408 + /* bad action? */
18411 + /* children inode number will be broken */
18418 +static int hn_gen_tree(struct dentry *dentry)
18420 + int err, i, j, ndentry;
18421 + struct au_dcsub_pages dpages;
18422 + struct au_dpage *dpage;
18423 + struct dentry **dentries;
18425 + err = au_dpages_init(&dpages, GFP_NOFS);
18426 + if (unlikely(err))
18428 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
18429 + if (unlikely(err))
18432 + for (i = 0; i < dpages.ndpage; i++) {
18433 + dpage = dpages.dpages + i;
18434 + dentries = dpage->dentries;
18435 + ndentry = dpage->ndentry;
18436 + for (j = 0; j < ndentry; j++) {
18437 + struct dentry *d;
18444 + if (d_really_is_positive(d))
18445 + /* todo: reset children xino?
18446 + cached children only? */
18447 + au_iigen_dec(d_inode(d));
18452 + au_dpages_free(&dpages);
18458 + * return 0 if processed.
18460 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
18461 + const unsigned int isdir)
18464 + struct dentry *d;
18465 + struct qstr *dname;
18468 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18469 + pr_warn("branch root dir was changed\n");
18475 + AuDebugOn(!name);
18476 + au_iigen_dec(inode);
18477 + spin_lock(&inode->i_lock);
18478 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
18479 + spin_lock(&d->d_lock);
18480 + dname = &d->d_name;
18481 + if (dname->len != nlen
18482 + && memcmp(dname->name, name, nlen)) {
18483 + spin_unlock(&d->d_lock);
18488 + spin_unlock(&d->d_lock);
18491 + spin_unlock(&inode->i_lock);
18493 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
18494 + d = d_find_any_alias(inode);
18496 + au_iigen_dec(inode);
18500 + spin_lock(&d->d_lock);
18501 + dname = &d->d_name;
18502 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
18503 + spin_unlock(&d->d_lock);
18504 + err = hn_gen_tree(d);
18505 + spin_lock(&d->d_lock);
18507 + spin_unlock(&d->d_lock);
18516 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
18520 + if (IS_ROOT(dentry)) {
18521 + pr_warn("branch root dir was changed\n");
18527 + au_digen_dec(dentry);
18528 + if (d_really_is_positive(dentry))
18529 + au_iigen_dec(d_inode(dentry));
18531 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
18532 + if (d_really_is_positive(dentry))
18533 + err = hn_gen_tree(dentry);
18540 +/* ---------------------------------------------------------------------- */
18542 +/* hnotify job flags */
18543 +#define AuHnJob_XINO0 1
18544 +#define AuHnJob_GEN (1 << 1)
18545 +#define AuHnJob_DIRENT (1 << 2)
18546 +#define AuHnJob_ISDIR (1 << 3)
18547 +#define AuHnJob_TRYXINO0 (1 << 4)
18548 +#define AuHnJob_MNTPNT (1 << 5)
18549 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
18550 +#define au_fset_hnjob(flags, name) \
18551 + do { (flags) |= AuHnJob_##name; } while (0)
18552 +#define au_fclr_hnjob(flags, name) \
18553 + do { (flags) &= ~AuHnJob_##name; } while (0)
18561 +struct au_hnotify_args {
18562 + struct inode *h_dir, *dir, *h_child_inode;
18564 + unsigned int flags[AuHnLast];
18565 + unsigned int h_child_nlen;
18566 + char h_child_name[];
18569 +struct hn_job_args {
18570 + unsigned int flags;
18571 + struct inode *inode, *h_inode, *dir, *h_dir;
18572 + struct dentry *dentry;
18577 +static int hn_job(struct hn_job_args *a)
18579 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
18583 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
18584 + hn_xino(a->inode, a->h_inode); /* ignore this error */
18586 + if (au_ftest_hnjob(a->flags, TRYXINO0)
18589 + inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
18590 + if (!a->h_inode->i_nlink
18591 + && !(a->h_inode->i_state & I_LINKABLE))
18592 + hn_xino(a->inode, a->h_inode); /* ignore this error */
18593 + inode_unlock_shared(a->h_inode);
18596 + /* make the generation obsolete */
18597 + if (au_ftest_hnjob(a->flags, GEN)) {
18600 + e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
18602 + if (e && a->dentry)
18603 + hn_gen_by_name(a->dentry, isdir);
18604 + /* ignore this error */
18607 + /* make dir entries obsolete */
18608 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
18609 + struct au_vdir *vdir;
18611 + vdir = au_ivdir(a->inode);
18613 + vdir->vd_jiffy = 0;
18614 + /* IMustLock(a->inode); */
18615 + /* inode_inc_iversion(a->inode); */
18618 + /* can do nothing but warn */
18619 + if (au_ftest_hnjob(a->flags, MNTPNT)
18621 + && d_mountpoint(a->dentry))
18622 + pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
18627 +/* ---------------------------------------------------------------------- */
18629 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
18630 + struct inode *dir)
18632 + struct dentry *dentry, *d, *parent;
18633 + struct qstr *dname;
18635 + parent = d_find_any_alias(dir);
18640 + spin_lock(&parent->d_lock);
18641 + list_for_each_entry(d, &parent->d_subdirs, d_child) {
18642 + /* AuDbg("%pd\n", d); */
18643 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
18644 + dname = &d->d_name;
18645 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
18646 + goto cont_unlock;
18650 + goto cont_unlock;
18651 + if (au_dcount(d) > 0) {
18652 + dentry = dget_dlock(d);
18653 + spin_unlock(&d->d_lock);
18658 + spin_unlock(&d->d_lock);
18660 + spin_unlock(&parent->d_lock);
18664 + di_write_lock_child(dentry);
18669 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
18670 + aufs_bindex_t bindex, ino_t h_ino)
18672 + struct inode *inode;
18677 + err = au_xino_read(sb, bindex, h_ino, &ino);
18679 + inode = ilookup(sb, ino);
18683 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
18684 + pr_warn("wrong root branch\n");
18690 + ii_write_lock_child(inode);
18696 +static void au_hn_bh(void *_args)
18698 + struct au_hnotify_args *a = _args;
18699 + struct super_block *sb;
18700 + aufs_bindex_t bindex, bbot, bfound;
18701 + unsigned char xino, try_iput;
18703 + struct inode *inode;
18705 + struct hn_job_args args;
18706 + struct dentry *dentry;
18707 + struct au_sbinfo *sbinfo;
18709 + AuDebugOn(!_args);
18710 + AuDebugOn(!a->h_dir);
18711 + AuDebugOn(!a->dir);
18712 + AuDebugOn(!a->mask);
18713 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
18714 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
18715 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
18720 + * do not lock a->dir->i_mutex here
18721 + * because of d_revalidate() may cause a deadlock.
18723 + sb = a->dir->i_sb;
18725 + sbinfo = au_sbi(sb);
18726 + AuDebugOn(!sbinfo);
18727 + si_write_lock(sb, AuLock_NOPLMW);
18729 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
18730 + switch (a->mask & FS_EVENTS_POSS_ON_CHILD) {
18731 + case FS_MOVED_FROM:
18732 + case FS_MOVED_TO:
18733 + AuWarn1("DIRREN with UDBA may not work correctly "
18734 + "for the direct rename(2)\n");
18737 + ii_read_lock_parent(a->dir);
18739 + bbot = au_ibbot(a->dir);
18740 + for (bindex = au_ibtop(a->dir); bindex <= bbot; bindex++)
18741 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
18745 + ii_read_unlock(a->dir);
18746 + if (unlikely(bfound < 0))
18749 + xino = !!au_opt_test(au_mntflags(sb), XINO);
18751 + if (a->h_child_inode)
18752 + h_ino = a->h_child_inode->i_ino;
18754 + if (a->h_child_nlen
18755 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
18756 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
18757 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
18760 + if (dentry && d_really_is_positive(dentry))
18761 + inode = d_inode(dentry);
18762 + if (xino && !inode && h_ino
18763 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
18764 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
18765 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
18766 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
18770 + args.flags = a->flags[AuHn_CHILD];
18771 + args.dentry = dentry;
18772 + args.inode = inode;
18773 + args.h_inode = a->h_child_inode;
18774 + args.dir = a->dir;
18775 + args.h_dir = a->h_dir;
18776 + args.h_name = a->h_child_name;
18777 + args.h_nlen = a->h_child_nlen;
18778 + err = hn_job(&args);
18780 + if (au_di(dentry))
18781 + di_write_unlock(dentry);
18784 + if (inode && try_iput) {
18785 + ii_write_unlock(inode);
18789 + ii_write_lock_parent(a->dir);
18790 + args.flags = a->flags[AuHn_PARENT];
18791 + args.dentry = NULL;
18792 + args.inode = a->dir;
18793 + args.h_inode = a->h_dir;
18795 + args.h_dir = NULL;
18796 + args.h_name = NULL;
18798 + err = hn_job(&args);
18799 + ii_write_unlock(a->dir);
18802 + iput(a->h_child_inode);
18805 + si_write_unlock(sb);
18806 + au_nwt_done(&sbinfo->si_nowait);
18810 +/* ---------------------------------------------------------------------- */
18812 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
18813 + const struct qstr *h_child_qstr, struct inode *h_child_inode)
18816 + unsigned int flags[AuHnLast], f;
18817 + unsigned char isdir, isroot, wh;
18818 + struct inode *dir;
18819 + struct au_hnotify_args *args;
18820 + char *p, *h_child_name;
18823 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
18824 + dir = igrab(hnotify->hn_aufs_inode);
18828 + isroot = (dir->i_ino == AUFS_ROOT_INO);
18830 + h_child_name = (void *)h_child_qstr->name;
18831 + len = h_child_qstr->len;
18832 + if (h_child_name) {
18833 + if (len > AUFS_WH_PFX_LEN
18834 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
18835 + h_child_name += AUFS_WH_PFX_LEN;
18836 + len -= AUFS_WH_PFX_LEN;
18842 + if (h_child_inode)
18843 + isdir = !!S_ISDIR(h_child_inode->i_mode);
18844 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
18845 + flags[AuHn_CHILD] = 0;
18847 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
18848 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
18849 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
18850 + switch (mask & ALL_FSNOTIFY_DIRENT_EVENTS) {
18851 + case FS_MOVED_FROM:
18852 + case FS_MOVED_TO:
18853 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
18854 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18857 + AuDebugOn(!h_child_name);
18862 + * aufs never be able to get this child inode.
18863 + * revalidation should be in d_revalidate()
18864 + * by checking i_nlink, i_generation or d_unhashed().
18866 + AuDebugOn(!h_child_name);
18867 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
18868 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
18876 + h_child_inode = NULL;
18879 + /* iput() and kfree() will be called in au_hnotify() */
18880 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
18881 + if (unlikely(!args)) {
18882 + AuErr1("no memory\n");
18886 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
18887 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
18888 + args->mask = mask;
18890 + args->h_dir = igrab(h_dir);
18891 + if (h_child_inode)
18892 + h_child_inode = igrab(h_child_inode); /* can be NULL */
18893 + args->h_child_inode = h_child_inode;
18894 + args->h_child_nlen = len;
18896 + p = (void *)args;
18897 + p += sizeof(*args);
18898 + memcpy(p, h_child_name, len);
18902 + /* NFS fires the event for silly-renamed one from kworker */
18904 + if (!dir->i_nlink
18905 + || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
18907 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
18908 + if (unlikely(err)) {
18909 + pr_err("wkq %d\n", err);
18910 + iput(args->h_child_inode);
18911 + iput(args->h_dir);
18913 + au_kfree_rcu(args);
18920 +/* ---------------------------------------------------------------------- */
18922 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
18926 + AuDebugOn(!(udba & AuOptMask_UDBA));
18929 + if (au_hnotify_op.reset_br)
18930 + err = au_hnotify_op.reset_br(udba, br, perm);
18935 +int au_hnotify_init_br(struct au_branch *br, int perm)
18940 + if (au_hnotify_op.init_br)
18941 + err = au_hnotify_op.init_br(br, perm);
18946 +void au_hnotify_fin_br(struct au_branch *br)
18948 + if (au_hnotify_op.fin_br)
18949 + au_hnotify_op.fin_br(br);
18952 +static void au_hn_destroy_cache(void)
18954 + kmem_cache_destroy(au_cache[AuCache_HNOTIFY]);
18955 + au_cache[AuCache_HNOTIFY] = NULL;
18958 +int __init au_hnotify_init(void)
18963 + au_cache[AuCache_HNOTIFY] = AuCache(au_hnotify);
18964 + if (au_cache[AuCache_HNOTIFY]) {
18966 + if (au_hnotify_op.init)
18967 + err = au_hnotify_op.init();
18968 + if (unlikely(err))
18969 + au_hn_destroy_cache();
18975 +void au_hnotify_fin(void)
18977 + if (au_hnotify_op.fin)
18978 + au_hnotify_op.fin();
18980 + /* cf. au_cache_fin() */
18981 + if (au_cache[AuCache_HNOTIFY])
18982 + au_hn_destroy_cache();
18984 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
18985 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
18986 +++ linux/fs/aufs/iinfo.c 2022-11-05 23:02:18.965889284 +0100
18988 +// SPDX-License-Identifier: GPL-2.0
18990 + * Copyright (C) 2005-2022 Junjiro R. Okajima
18992 + * This program is free software; you can redistribute it and/or modify
18993 + * it under the terms of the GNU General Public License as published by
18994 + * the Free Software Foundation; either version 2 of the License, or
18995 + * (at your option) any later version.
18997 + * This program is distributed in the hope that it will be useful,
18998 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18999 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19000 + * GNU General Public License for more details.
19002 + * You should have received a copy of the GNU General Public License
19003 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19007 + * inode private data
19012 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
19014 + struct inode *h_inode;
19015 + struct au_hinode *hinode;
19017 + IiMustAnyLock(inode);
19019 + hinode = au_hinode(au_ii(inode), bindex);
19020 + h_inode = hinode->hi_inode;
19021 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19025 +/* todo: hard/soft set? */
19026 +void au_hiput(struct au_hinode *hinode)
19028 + au_hn_free(hinode);
19029 + dput(hinode->hi_whdentry);
19030 + iput(hinode->hi_inode);
19033 +unsigned int au_hi_flags(struct inode *inode, int isdir)
19035 + unsigned int flags;
19036 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
19039 + if (au_opt_test(mnt_flags, XINO))
19040 + au_fset_hi(flags, XINO);
19041 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
19042 + au_fset_hi(flags, HNOTIFY);
19046 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
19047 + struct inode *h_inode, unsigned int flags)
19049 + struct au_hinode *hinode;
19050 + struct inode *hi;
19051 + struct au_iinfo *iinfo = au_ii(inode);
19053 + IiMustWriteLock(inode);
19055 + hinode = au_hinode(iinfo, bindex);
19056 + hi = hinode->hi_inode;
19057 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
19060 + au_hiput(hinode);
19061 + hinode->hi_inode = h_inode;
19064 + struct super_block *sb = inode->i_sb;
19065 + struct au_branch *br;
19067 + AuDebugOn(inode->i_mode
19068 + && (h_inode->i_mode & S_IFMT)
19069 + != (inode->i_mode & S_IFMT));
19070 + if (bindex == iinfo->ii_btop)
19071 + au_cpup_igen(inode, h_inode);
19072 + br = au_sbr(sb, bindex);
19073 + hinode->hi_id = br->br_id;
19074 + if (au_ftest_hi(flags, XINO)) {
19075 + err = au_xino_write(sb, bindex, h_inode->i_ino,
19077 + if (unlikely(err))
19078 + AuIOErr1("failed au_xino_write() %d\n", err);
19081 + if (au_ftest_hi(flags, HNOTIFY)
19082 + && au_br_hnotifyable(br->br_perm)) {
19083 + err = au_hn_alloc(hinode, inode);
19084 + if (unlikely(err))
19085 + AuIOErr1("au_hn_alloc() %d\n", err);
19090 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
19091 + struct dentry *h_wh)
19093 + struct au_hinode *hinode;
19095 + IiMustWriteLock(inode);
19097 + hinode = au_hinode(au_ii(inode), bindex);
19098 + AuDebugOn(hinode->hi_whdentry);
19099 + hinode->hi_whdentry = h_wh;
19102 +void au_update_iigen(struct inode *inode, int half)
19104 + struct au_iinfo *iinfo;
19105 + struct au_iigen *iigen;
19106 + unsigned int sigen;
19108 + sigen = au_sigen(inode->i_sb);
19109 + iinfo = au_ii(inode);
19110 + iigen = &iinfo->ii_generation;
19111 + spin_lock(&iigen->ig_spin);
19112 + iigen->ig_generation = sigen;
19114 + au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
19116 + au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
19117 + spin_unlock(&iigen->ig_spin);
19120 +/* it may be called at remount time, too */
19121 +void au_update_ibrange(struct inode *inode, int do_put_zero)
19123 + struct au_iinfo *iinfo;
19124 + aufs_bindex_t bindex, bbot;
19126 + AuDebugOn(au_is_bad_inode(inode));
19127 + IiMustWriteLock(inode);
19129 + iinfo = au_ii(inode);
19130 + if (do_put_zero && iinfo->ii_btop >= 0) {
19131 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19133 + struct inode *h_i;
19135 + h_i = au_hinode(iinfo, bindex)->hi_inode;
19138 + && !(h_i->i_state & I_LINKABLE))
19139 + au_set_h_iptr(inode, bindex, NULL, 0);
19143 + iinfo->ii_btop = -1;
19144 + iinfo->ii_bbot = -1;
19145 + bbot = au_sbbot(inode->i_sb);
19146 + for (bindex = 0; bindex <= bbot; bindex++)
19147 + if (au_hinode(iinfo, bindex)->hi_inode) {
19148 + iinfo->ii_btop = bindex;
19151 + if (iinfo->ii_btop >= 0)
19152 + for (bindex = bbot; bindex >= iinfo->ii_btop; bindex--)
19153 + if (au_hinode(iinfo, bindex)->hi_inode) {
19154 + iinfo->ii_bbot = bindex;
19157 + AuDebugOn(iinfo->ii_btop > iinfo->ii_bbot);
19160 +/* ---------------------------------------------------------------------- */
19162 +void au_icntnr_init_once(void *_c)
19164 + struct au_icntnr *c = _c;
19165 + struct au_iinfo *iinfo = &c->iinfo;
19167 + spin_lock_init(&iinfo->ii_generation.ig_spin);
19168 + au_rw_init(&iinfo->ii_rwsem);
19169 + inode_init_once(&c->vfs_inode);
19172 +void au_hinode_init(struct au_hinode *hinode)
19174 + hinode->hi_inode = NULL;
19175 + hinode->hi_id = -1;
19176 + au_hn_init(hinode);
19177 + hinode->hi_whdentry = NULL;
19180 +int au_iinfo_init(struct inode *inode)
19182 + struct au_iinfo *iinfo;
19183 + struct super_block *sb;
19184 + struct au_hinode *hi;
19187 + sb = inode->i_sb;
19188 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19189 + nbr = au_sbbot(sb) + 1;
19190 + if (unlikely(nbr <= 0))
19192 + hi = kmalloc_array(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
19194 + au_lcnt_inc(&au_sbi(sb)->si_ninodes);
19196 + iinfo->ii_hinode = hi;
19197 + for (i = 0; i < nbr; i++, hi++)
19198 + au_hinode_init(hi);
19200 + iinfo->ii_generation.ig_generation = au_sigen(sb);
19201 + iinfo->ii_btop = -1;
19202 + iinfo->ii_bbot = -1;
19203 + iinfo->ii_vdir = NULL;
19209 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink)
19212 + struct au_hinode *hip;
19214 + AuRwMustWriteLock(&iinfo->ii_rwsem);
19217 + hip = au_krealloc(iinfo->ii_hinode, sizeof(*hip) * nbr, GFP_NOFS,
19220 + iinfo->ii_hinode = hip;
19221 + i = iinfo->ii_bbot + 1;
19223 + for (; i < nbr; i++, hip++)
19224 + au_hinode_init(hip);
19231 +void au_iinfo_fin(struct inode *inode)
19233 + struct au_iinfo *iinfo;
19234 + struct au_hinode *hi;
19235 + struct super_block *sb;
19236 + aufs_bindex_t bindex, bbot;
19237 + const unsigned char unlinked = !inode->i_nlink;
19239 + AuDebugOn(au_is_bad_inode(inode));
19241 + sb = inode->i_sb;
19242 + au_lcnt_dec(&au_sbi(sb)->si_ninodes);
19243 + if (si_pid_test(sb))
19244 + au_xino_delete_inode(inode, unlinked);
19247 + * it is safe to hide the dependency between sbinfo and
19251 + si_noflush_read_lock(sb);
19252 + au_xino_delete_inode(inode, unlinked);
19253 + si_read_unlock(sb);
19257 + iinfo = au_ii(inode);
19258 + if (iinfo->ii_vdir)
19259 + au_vdir_free(iinfo->ii_vdir);
19261 + bindex = iinfo->ii_btop;
19262 + if (bindex >= 0) {
19263 + hi = au_hinode(iinfo, bindex);
19264 + bbot = iinfo->ii_bbot;
19265 + while (bindex++ <= bbot) {
19266 + if (hi->hi_inode)
19271 + au_kfree_rcu(iinfo->ii_hinode);
19272 + AuRwDestroy(&iinfo->ii_rwsem);
19274 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
19275 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
19276 +++ linux/fs/aufs/inode.c 2023-04-24 08:13:40.969576115 +0200
19278 +// SPDX-License-Identifier: GPL-2.0
19280 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19282 + * This program is free software; you can redistribute it and/or modify
19283 + * it under the terms of the GNU General Public License as published by
19284 + * the Free Software Foundation; either version 2 of the License, or
19285 + * (at your option) any later version.
19287 + * This program is distributed in the hope that it will be useful,
19288 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19289 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19290 + * GNU General Public License for more details.
19292 + * You should have received a copy of the GNU General Public License
19293 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19297 + * inode functions
19300 +#include <linux/iversion.h>
19303 +struct inode *au_igrab(struct inode *inode)
19306 + AuDebugOn(!atomic_read(&inode->i_count));
19312 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
19314 + au_cpup_attr_all(inode, /*force*/0);
19315 + au_update_iigen(inode, /*half*/1);
19317 + inode_inc_iversion(inode);
19320 +static int au_ii_refresh(struct inode *inode, int *update)
19324 + aufs_bindex_t bindex, new_bindex;
19325 + struct super_block *sb;
19326 + struct au_iinfo *iinfo;
19327 + struct au_hinode *p, *q, tmp;
19329 + AuDebugOn(au_is_bad_inode(inode));
19330 + IiMustWriteLock(inode);
19333 + sb = inode->i_sb;
19334 + nbr = au_sbbot(sb) + 1;
19335 + type = inode->i_mode & S_IFMT;
19336 + iinfo = au_ii(inode);
19337 + err = au_hinode_realloc(iinfo, nbr, /*may_shrink*/0);
19338 + if (unlikely(err))
19341 + AuDebugOn(iinfo->ii_btop < 0);
19342 + p = au_hinode(iinfo, iinfo->ii_btop);
19343 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
19345 + if (!p->hi_inode)
19348 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
19349 + new_bindex = au_br_index(sb, p->hi_id);
19350 + if (new_bindex == bindex)
19353 + if (new_bindex < 0) {
19356 + p->hi_inode = NULL;
19360 + if (new_bindex < iinfo->ii_btop)
19361 + iinfo->ii_btop = new_bindex;
19362 + if (iinfo->ii_bbot < new_bindex)
19363 + iinfo->ii_bbot = new_bindex;
19364 + /* swap two lower inode, and loop again */
19365 + q = au_hinode(iinfo, new_bindex);
19369 + if (tmp.hi_inode) {
19374 + au_update_ibrange(inode, /*do_put_zero*/0);
19375 + au_hinode_realloc(iinfo, nbr, /*may_shrink*/1); /* harmless if err */
19376 + e = au_dy_irefresh(inode);
19377 + if (unlikely(e && !err))
19385 +void au_refresh_iop(struct inode *inode, int force_getattr)
19388 + struct au_sbinfo *sbi = au_sbi(inode->i_sb);
19389 + const struct inode_operations *iop
19390 + = force_getattr ? aufs_iop : sbi->si_iop_array;
19392 + if (inode->i_op == iop)
19395 + switch (inode->i_mode & S_IFMT) {
19397 + type = AuIop_DIR;
19400 + type = AuIop_SYMLINK;
19403 + type = AuIop_OTHER;
19407 + inode->i_op = iop + type;
19408 + /* unnecessary smp_wmb() */
19411 +int au_refresh_hinode_self(struct inode *inode)
19415 + err = au_ii_refresh(inode, &update);
19417 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
19423 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
19425 + int err, e, update;
19426 + unsigned int flags;
19428 + aufs_bindex_t bindex, bbot;
19429 + unsigned char isdir;
19430 + struct au_hinode *p;
19431 + struct au_iinfo *iinfo;
19433 + err = au_ii_refresh(inode, &update);
19434 + if (unlikely(err))
19438 + iinfo = au_ii(inode);
19439 + p = au_hinode(iinfo, iinfo->ii_btop);
19440 + mode = (inode->i_mode & S_IFMT);
19441 + isdir = S_ISDIR(mode);
19442 + flags = au_hi_flags(inode, isdir);
19443 + bbot = au_dbbot(dentry);
19444 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
19445 + struct inode *h_i, *h_inode;
19446 + struct dentry *h_d;
19448 + h_d = au_h_dptr(dentry, bindex);
19449 + if (!h_d || d_is_negative(h_d))
19452 + h_inode = d_inode(h_d);
19453 + AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
19454 + if (iinfo->ii_btop <= bindex && bindex <= iinfo->ii_bbot) {
19455 + h_i = au_h_iptr(inode, bindex);
19457 + if (h_i == h_inode)
19463 + if (bindex < iinfo->ii_btop)
19464 + iinfo->ii_btop = bindex;
19465 + if (iinfo->ii_bbot < bindex)
19466 + iinfo->ii_bbot = bindex;
19467 + au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
19470 + au_update_ibrange(inode, /*do_put_zero*/0);
19471 + e = au_dy_irefresh(inode);
19472 + if (unlikely(e && !err))
19475 + au_refresh_hinode_attr(inode, update && isdir);
19482 +static int set_inode(struct inode *inode, struct dentry *dentry)
19485 + unsigned int flags;
19487 + aufs_bindex_t bindex, btop, btail;
19488 + unsigned char isdir;
19489 + struct dentry *h_dentry;
19490 + struct inode *h_inode;
19491 + struct au_iinfo *iinfo;
19492 + const struct inode_operations *iop;
19494 + IiMustWriteLock(inode);
19498 + iop = au_sbi(inode->i_sb)->si_iop_array;
19499 + btop = au_dbtop(dentry);
19500 + h_dentry = au_h_dptr(dentry, btop);
19501 + h_inode = d_inode(h_dentry);
19502 + mode = h_inode->i_mode;
19503 + switch (mode & S_IFMT) {
19505 + btail = au_dbtail(dentry);
19506 + inode->i_op = iop + AuIop_OTHER;
19507 + inode->i_fop = &aufs_file_fop;
19508 + err = au_dy_iaop(inode, btop, h_inode);
19509 + if (unlikely(err))
19514 + btail = au_dbtaildir(dentry);
19515 + inode->i_op = iop + AuIop_DIR;
19516 + inode->i_fop = &aufs_dir_fop;
19519 + btail = au_dbtail(dentry);
19520 + inode->i_op = iop + AuIop_SYMLINK;
19526 + btail = au_dbtail(dentry);
19527 + inode->i_op = iop + AuIop_OTHER;
19528 + init_special_inode(inode, mode, h_inode->i_rdev);
19531 + AuIOErr("Unknown file type 0%o\n", mode);
19536 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
19537 + flags = au_hi_flags(inode, isdir);
19538 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
19539 + && au_ftest_hi(flags, HNOTIFY)
19540 + && dentry->d_name.len > AUFS_WH_PFX_LEN
19541 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
19542 + au_fclr_hi(flags, HNOTIFY);
19543 + iinfo = au_ii(inode);
19544 + iinfo->ii_btop = btop;
19545 + iinfo->ii_bbot = btail;
19546 + for (bindex = btop; bindex <= btail; bindex++) {
19547 + h_dentry = au_h_dptr(dentry, bindex);
19549 + au_set_h_iptr(inode, bindex,
19550 + au_igrab(d_inode(h_dentry)), flags);
19552 + au_cpup_attr_all(inode, /*force*/1);
19554 + * to force calling aufs_get_inode_acl() every time,
19555 + * do not call cache_no_acl() for aufs inode.
19563 + * successful returns with iinfo write_locked
19565 + * zero: success, matched
19566 + * plus: no error, but unmatched
19568 +static int reval_inode(struct inode *inode, struct dentry *dentry)
19571 + unsigned int gen, igflags;
19572 + aufs_bindex_t bindex, bbot;
19573 + struct inode *h_inode, *h_dinode;
19574 + struct dentry *h_dentry;
19577 + * before this function, if aufs got any iinfo lock, it must be only
19578 + * one, the parent dir.
19579 + * it can happen by UDBA and the obsoleted inode number.
19582 + if (unlikely(inode->i_ino == parent_ino(dentry)))
19586 + ii_write_lock_new_child(inode);
19587 + h_dentry = au_h_dptr(dentry, au_dbtop(dentry));
19588 + h_dinode = d_inode(h_dentry);
19589 + bbot = au_ibbot(inode);
19590 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
19591 + h_inode = au_h_iptr(inode, bindex);
19592 + if (!h_inode || h_inode != h_dinode)
19596 + gen = au_iigen(inode, &igflags);
19597 + if (gen == au_digen(dentry)
19598 + && !au_ig_ftest(igflags, HALF_REFRESHED))
19601 + /* fully refresh inode using dentry */
19602 + err = au_refresh_hinode(inode, dentry);
19604 + au_update_iigen(inode, /*half*/0);
19608 + if (unlikely(err))
19609 + ii_write_unlock(inode);
19614 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19615 + unsigned int d_type, ino_t *ino)
19618 + const int isnondir = d_type != DT_DIR;
19620 + /* prevent hardlinked inode number from race condition */
19622 + err = au_xinondir_enter(sb, bindex, h_ino, &idx);
19623 + if (unlikely(err))
19627 + err = au_xino_read(sb, bindex, h_ino, ino);
19628 + if (unlikely(err))
19629 + goto out_xinondir;
19633 + *ino = au_xino_new_ino(sb);
19634 + if (unlikely(!*ino))
19635 + goto out_xinondir;
19636 + err = au_xino_write(sb, bindex, h_ino, *ino);
19637 + if (unlikely(err))
19638 + goto out_xinondir;
19642 + if (isnondir && idx >= 0)
19643 + au_xinondir_leave(sb, bindex, h_ino, idx);
19648 +/* successful returns with iinfo write_locked */
19649 +/* todo: return with unlocked? */
19650 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
19652 + struct inode *inode, *h_inode;
19653 + struct dentry *h_dentry;
19654 + struct super_block *sb;
19655 + ino_t h_ino, ino;
19656 + int err, idx, hlinked;
19657 + aufs_bindex_t btop;
19659 + sb = dentry->d_sb;
19660 + btop = au_dbtop(dentry);
19661 + h_dentry = au_h_dptr(dentry, btop);
19662 + h_inode = d_inode(h_dentry);
19663 + h_ino = h_inode->i_ino;
19664 + hlinked = !d_is_dir(h_dentry) && h_inode->i_nlink > 1;
19668 + * stop 'race'-ing between hardlinks under different
19672 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
19673 + inode = ERR_PTR(err);
19674 + if (unlikely(err))
19678 + err = au_xino_read(sb, btop, h_ino, &ino);
19679 + inode = ERR_PTR(err);
19680 + if (unlikely(err))
19681 + goto out_xinondir;
19684 + ino = au_xino_new_ino(sb);
19685 + if (unlikely(!ino)) {
19686 + inode = ERR_PTR(-EIO);
19687 + goto out_xinondir;
19691 + AuDbg("i%lu\n", (unsigned long)ino);
19692 + inode = au_iget_locked(sb, ino);
19693 + err = PTR_ERR(inode);
19694 + if (IS_ERR(inode))
19695 + goto out_xinondir;
19697 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
19698 + if (inode->i_state & I_NEW) {
19699 + ii_write_lock_new_child(inode);
19700 + err = set_inode(inode, dentry);
19702 + unlock_new_inode(inode);
19703 + goto out_xinondir; /* success */
19707 + * iget_failed() calls iput(), but we need to call
19708 + * ii_write_unlock() after iget_failed(). so dirty hack for
19711 + atomic_inc(&inode->i_count);
19712 + iget_failed(inode);
19713 + ii_write_unlock(inode);
19714 + au_xino_write(sb, btop, h_ino, /*ino*/0);
19715 + /* ignore this error */
19717 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
19719 + * horrible race condition between lookup, readdir and copyup
19720 + * (or something).
19722 + if (hlinked && idx >= 0)
19723 + au_xinondir_leave(sb, btop, h_ino, idx);
19724 + err = reval_inode(inode, dentry);
19725 + if (unlikely(err < 0)) {
19730 + goto out; /* success */
19731 + else if (hlinked && idx >= 0) {
19732 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
19733 + if (unlikely(err)) {
19735 + inode = ERR_PTR(err);
19741 + if (unlikely(au_test_fs_unique_ino(h_inode)))
19742 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
19743 + " b%d, %s, %pd, hi%lu, i%lu.\n",
19744 + btop, au_sbtype(h_dentry->d_sb), dentry,
19745 + (unsigned long)h_ino, (unsigned long)ino);
19747 + err = au_xino_write(sb, btop, h_ino, /*ino*/0);
19750 + if (hlinked && idx >= 0)
19751 + au_xinondir_leave(sb, btop, h_ino, idx);
19757 + inode = ERR_PTR(err);
19759 + if (hlinked && idx >= 0)
19760 + au_xinondir_leave(sb, btop, h_ino, idx);
19765 +/* ---------------------------------------------------------------------- */
19767 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19768 + struct inode *inode)
19771 + struct inode *hi;
19773 + err = au_br_rdonly(au_sbr(sb, bindex));
19775 + /* pseudo-link after flushed may happen out of bounds */
19778 + && au_ibtop(inode) <= bindex
19779 + && bindex <= au_ibbot(inode)) {
19781 + * permission check is unnecessary since vfsub routine
19782 + * will be called later
19784 + hi = au_h_iptr(inode, bindex);
19786 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
19792 +int au_test_h_perm(struct mnt_idmap *h_idmap, struct inode *h_inode,
19795 + if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
19797 + return inode_permission(h_idmap, h_inode, mask);
19800 +int au_test_h_perm_sio(struct mnt_idmap *h_idmap, struct inode *h_inode,
19803 + if (au_test_nfs(h_inode->i_sb)
19804 + && (mask & MAY_WRITE)
19805 + && S_ISDIR(h_inode->i_mode))
19806 + mask |= MAY_READ; /* force permission check */
19807 + return au_test_h_perm(h_idmap, h_inode, mask);
19809 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
19810 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
19811 +++ linux/fs/aufs/inode.h 2023-04-24 08:13:40.969576115 +0200
19813 +/* SPDX-License-Identifier: GPL-2.0 */
19815 + * Copyright (C) 2005-2022 Junjiro R. Okajima
19817 + * This program is free software; you can redistribute it and/or modify
19818 + * it under the terms of the GNU General Public License as published by
19819 + * the Free Software Foundation; either version 2 of the License, or
19820 + * (at your option) any later version.
19822 + * This program is distributed in the hope that it will be useful,
19823 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19824 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19825 + * GNU General Public License for more details.
19827 + * You should have received a copy of the GNU General Public License
19828 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19832 + * inode operations
19835 +#ifndef __AUFS_INODE_H__
19836 +#define __AUFS_INODE_H__
19840 +#include <linux/fsnotify.h>
19841 +#include "rwsem.h"
19845 +struct au_hnotify {
19846 +#ifdef CONFIG_AUFS_HNOTIFY
19847 +#ifdef CONFIG_AUFS_HFSNOTIFY
19848 + /* never use fsnotify_add_vfsmount_mark() */
19849 + struct fsnotify_mark hn_mark;
19851 + struct inode *hn_aufs_inode; /* no get/put */
19852 + struct rcu_head rcu;
19854 +} ____cacheline_aligned_in_smp;
19856 +struct au_hinode {
19857 + struct inode *hi_inode;
19858 + aufs_bindex_t hi_id;
19859 +#ifdef CONFIG_AUFS_HNOTIFY
19860 + struct au_hnotify *hi_notify;
19863 + /* reference to the copied-up whiteout with get/put */
19864 + struct dentry *hi_whdentry;
19868 +#define AuIG_HALF_REFRESHED 1
19869 +#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
19870 +#define au_ig_fset(flags, name) \
19871 + do { (flags) |= AuIG_##name; } while (0)
19872 +#define au_ig_fclr(flags, name) \
19873 + do { (flags) &= ~AuIG_##name; } while (0)
19876 + spinlock_t ig_spin;
19877 + __u32 ig_generation, ig_flags;
19882 + struct au_iigen ii_generation;
19883 + struct super_block *ii_hsb1; /* no get/put */
19885 + struct au_rwsem ii_rwsem;
19886 + aufs_bindex_t ii_btop, ii_bbot;
19888 + struct au_hinode *ii_hinode;
19889 + struct au_vdir *ii_vdir;
19892 +struct au_icntnr {
19893 + struct au_iinfo iinfo;
19894 + struct inode vfs_inode;
19895 + struct hlist_bl_node plink;
19896 + struct rcu_head rcu;
19897 +} ____cacheline_aligned_in_smp;
19899 +/* au_pin flags */
19900 +#define AuPin_DI_LOCKED 1
19901 +#define AuPin_MNT_WRITE (1 << 1)
19902 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
19903 +#define au_fset_pin(flags, name) \
19904 + do { (flags) |= AuPin_##name; } while (0)
19905 +#define au_fclr_pin(flags, name) \
19906 + do { (flags) &= ~AuPin_##name; } while (0)
19910 + struct dentry *dentry;
19911 + unsigned int udba;
19912 + unsigned char lsc_di, lsc_hi, flags;
19913 + aufs_bindex_t bindex;
19916 + struct dentry *parent;
19917 + struct au_hinode *hdir;
19918 + struct vfsmount *h_mnt;
19920 + /* temporary unlock/relock for copyup */
19921 + struct dentry *h_dentry, *h_parent;
19922 + struct au_branch *br;
19923 + struct task_struct *task;
19926 +void au_pin_hdir_unlock(struct au_pin *p);
19927 +int au_pin_hdir_lock(struct au_pin *p);
19928 +int au_pin_hdir_relock(struct au_pin *p);
19929 +void au_pin_hdir_acquire_nest(struct au_pin *p);
19930 +void au_pin_hdir_release(struct au_pin *p);
19932 +/* ---------------------------------------------------------------------- */
19934 +static inline struct au_iinfo *au_ii(struct inode *inode)
19936 + BUG_ON(is_bad_inode(inode));
19937 + return &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
19940 +/* ---------------------------------------------------------------------- */
19943 +struct inode *au_igrab(struct inode *inode);
19944 +void au_refresh_iop(struct inode *inode, int force_getattr);
19945 +int au_refresh_hinode_self(struct inode *inode);
19946 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
19947 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
19948 + unsigned int d_type, ino_t *ino);
19949 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
19950 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
19951 + struct inode *inode);
19952 +int au_test_h_perm(struct mnt_idmap *h_idmap, struct inode *h_inode,
19954 +int au_test_h_perm_sio(struct mnt_idmap *h_idmap, struct inode *h_inode,
19957 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
19958 + ino_t h_ino, unsigned int d_type, ino_t *ino)
19960 +#ifdef CONFIG_AUFS_SHWH
19961 + return au_ino(sb, bindex, h_ino, d_type, ino);
19974 +extern struct inode_operations aufs_iop[AuIop_Last], /* not const */
19975 + aufs_iop_nogetattr[AuIop_Last];
19977 +/* au_wr_dir flags */
19978 +#define AuWrDir_ADD_ENTRY 1
19979 +#define AuWrDir_ISDIR (1 << 1)
19980 +#define AuWrDir_TMPFILE (1 << 2)
19981 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
19982 +#define au_fset_wrdir(flags, name) \
19983 + do { (flags) |= AuWrDir_##name; } while (0)
19984 +#define au_fclr_wrdir(flags, name) \
19985 + do { (flags) &= ~AuWrDir_##name; } while (0)
19987 +struct au_wr_dir_args {
19988 + aufs_bindex_t force_btgt;
19989 + unsigned char flags;
19991 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
19992 + struct au_wr_dir_args *args);
19994 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
19995 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
19996 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
19997 + unsigned int udba, unsigned char flags);
19998 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
19999 + unsigned int udba, unsigned char flags) __must_check;
20000 +int au_do_pin(struct au_pin *pin) __must_check;
20001 +void au_unpin(struct au_pin *pin);
20002 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
20004 +#define AuIcpup_DID_CPUP 1
20005 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
20006 +#define au_fset_icpup(flags, name) \
20007 + do { (flags) |= AuIcpup_##name; } while (0)
20008 +#define au_fclr_icpup(flags, name) \
20009 + do { (flags) &= ~AuIcpup_##name; } while (0)
20011 +struct au_icpup_args {
20012 + unsigned char flags;
20013 + unsigned char pin_flags;
20014 + aufs_bindex_t btgt;
20015 + unsigned int udba;
20016 + struct au_pin pin;
20017 + struct path h_path;
20018 + struct inode *h_inode;
20021 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
20022 + struct au_icpup_args *a);
20024 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
20025 + struct path *h_path, int locked);
20028 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20029 + struct dentry *h_parent, int isdir);
20030 +int aufs_mknod(struct mnt_idmap *idmap, struct inode *dir,
20031 + struct dentry *dentry, umode_t mode, dev_t dev);
20032 +int aufs_symlink(struct mnt_idmap *idmap, struct inode *dir,
20033 + struct dentry *dentry, const char *symname);
20034 +int aufs_create(struct mnt_idmap *idmap, struct inode *dir,
20035 + struct dentry *dentry, umode_t mode, bool want_excl);
20036 +struct vfsub_aopen_args;
20037 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
20038 + struct vfsub_aopen_args *args);
20039 +int aufs_tmpfile(struct mnt_idmap *idmap, struct inode *dir,
20040 + struct file *file, umode_t mode);
20041 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
20042 + struct dentry *dentry);
20043 +int aufs_mkdir(struct mnt_idmap *idmap, struct inode *dir,
20044 + struct dentry *dentry, umode_t mode);
20047 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
20048 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
20049 + struct dentry *h_parent, int isdir);
20050 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
20051 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
20054 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
20055 +int aufs_rename(struct mnt_idmap *idmap,
20056 + struct inode *_src_dir, struct dentry *_src_dentry,
20057 + struct inode *_dst_dir, struct dentry *_dst_dentry,
20058 + unsigned int _flags);
20061 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
20062 +void au_hiput(struct au_hinode *hinode);
20063 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
20064 + struct dentry *h_wh);
20065 +unsigned int au_hi_flags(struct inode *inode, int isdir);
20067 +/* hinode flags */
20068 +#define AuHi_XINO 1
20069 +#define AuHi_HNOTIFY (1 << 1)
20070 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
20071 +#define au_fset_hi(flags, name) \
20072 + do { (flags) |= AuHi_##name; } while (0)
20073 +#define au_fclr_hi(flags, name) \
20074 + do { (flags) &= ~AuHi_##name; } while (0)
20076 +#ifndef CONFIG_AUFS_HNOTIFY
20077 +#undef AuHi_HNOTIFY
20078 +#define AuHi_HNOTIFY 0
20081 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
20082 + struct inode *h_inode, unsigned int flags);
20084 +void au_update_iigen(struct inode *inode, int half);
20085 +void au_update_ibrange(struct inode *inode, int do_put_zero);
20087 +void au_icntnr_init_once(void *_c);
20088 +void au_hinode_init(struct au_hinode *hinode);
20089 +int au_iinfo_init(struct inode *inode);
20090 +void au_iinfo_fin(struct inode *inode);
20091 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink);
20093 +#ifdef CONFIG_PROC_FS
20095 +int au_plink_maint(struct super_block *sb, int flags);
20097 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
20098 +int au_plink_maint_enter(struct super_block *sb);
20099 +#ifdef CONFIG_AUFS_DEBUG
20100 +void au_plink_list(struct super_block *sb);
20102 +AuStubVoid(au_plink_list, struct super_block *sb)
20104 +int au_plink_test(struct inode *inode);
20105 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
20106 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
20107 + struct dentry *h_dentry);
20108 +void au_plink_put(struct super_block *sb, int verbose);
20109 +void au_plink_clean(struct super_block *sb, int verbose);
20110 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
20112 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
20113 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
20114 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
20115 +AuStubVoid(au_plink_list, struct super_block *sb);
20116 +AuStubInt0(au_plink_test, struct inode *inode);
20117 +AuStub(struct dentry *, au_plink_lkup, return NULL,
20118 + struct inode *inode, aufs_bindex_t bindex);
20119 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
20120 + struct dentry *h_dentry);
20121 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
20122 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
20123 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
20124 +#endif /* CONFIG_PROC_FS */
20126 +#ifdef CONFIG_AUFS_XATTR
20128 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
20129 + unsigned int verbose);
20130 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
20131 +void au_xattr_init(struct super_block *sb);
20133 +AuStubInt0(au_cpup_xattr, struct path *h_dst, struct path *h_src,
20134 + int ignore_flags, unsigned int verbose);
20135 +AuStubVoid(au_xattr_init, struct super_block *sb);
20138 +#ifdef CONFIG_FS_POSIX_ACL
20139 +struct posix_acl *aufs_get_inode_acl(struct inode *inode, int type, bool rcu);
20140 +struct posix_acl *aufs_get_acl(struct mnt_idmap *idmap,
20141 + struct dentry *dentry, int type);
20142 +int aufs_set_acl(struct mnt_idmap *idmap, struct dentry *dentry,
20143 + struct posix_acl *acl, int type);
20146 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
20152 +struct au_sxattr {
20156 + const char *name;
20157 + const void *value;
20162 + struct posix_acl *acl;
20167 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
20168 + struct au_sxattr *arg);
20171 +/* ---------------------------------------------------------------------- */
20173 +/* lock subclass for iinfo */
20175 + AuLsc_II_CHILD, /* child first */
20176 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
20177 + AuLsc_II_CHILD3, /* copyup dirs */
20178 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
20179 + AuLsc_II_PARENT2,
20180 + AuLsc_II_PARENT3, /* copyup dirs */
20181 + AuLsc_II_NEW_CHILD
20185 + * ii_read_lock_child, ii_write_lock_child,
20186 + * ii_read_lock_child2, ii_write_lock_child2,
20187 + * ii_read_lock_child3, ii_write_lock_child3,
20188 + * ii_read_lock_parent, ii_write_lock_parent,
20189 + * ii_read_lock_parent2, ii_write_lock_parent2,
20190 + * ii_read_lock_parent3, ii_write_lock_parent3,
20191 + * ii_read_lock_new_child, ii_write_lock_new_child,
20193 +#define AuReadLockFunc(name, lsc) \
20194 +static inline void ii_read_lock_##name(struct inode *i) \
20196 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20199 +#define AuWriteLockFunc(name, lsc) \
20200 +static inline void ii_write_lock_##name(struct inode *i) \
20202 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
20205 +#define AuRWLockFuncs(name, lsc) \
20206 + AuReadLockFunc(name, lsc) \
20207 + AuWriteLockFunc(name, lsc)
20209 +AuRWLockFuncs(child, CHILD);
20210 +AuRWLockFuncs(child2, CHILD2);
20211 +AuRWLockFuncs(child3, CHILD3);
20212 +AuRWLockFuncs(parent, PARENT);
20213 +AuRWLockFuncs(parent2, PARENT2);
20214 +AuRWLockFuncs(parent3, PARENT3);
20215 +AuRWLockFuncs(new_child, NEW_CHILD);
20217 +#undef AuReadLockFunc
20218 +#undef AuWriteLockFunc
20219 +#undef AuRWLockFuncs
20221 +#define ii_read_unlock(i) au_rw_read_unlock(&au_ii(i)->ii_rwsem)
20222 +#define ii_write_unlock(i) au_rw_write_unlock(&au_ii(i)->ii_rwsem)
20223 +#define ii_downgrade_lock(i) au_rw_dgrade_lock(&au_ii(i)->ii_rwsem)
20225 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
20226 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
20227 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
20229 +/* ---------------------------------------------------------------------- */
20231 +static inline void au_icntnr_init(struct au_icntnr *c)
20233 +#ifdef CONFIG_AUFS_DEBUG
20234 + c->vfs_inode.i_mode = 0;
20238 +static inline unsigned int au_iigen(struct inode *inode, unsigned int *igflags)
20240 + unsigned int gen;
20241 + struct au_iinfo *iinfo;
20242 + struct au_iigen *iigen;
20244 + iinfo = au_ii(inode);
20245 + iigen = &iinfo->ii_generation;
20246 + spin_lock(&iigen->ig_spin);
20248 + *igflags = iigen->ig_flags;
20249 + gen = iigen->ig_generation;
20250 + spin_unlock(&iigen->ig_spin);
20255 +/* tiny test for inode number */
20256 +/* tmpfs generation is too rough */
20257 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
20259 + struct au_iinfo *iinfo;
20261 + iinfo = au_ii(inode);
20262 + AuRwMustAnyLock(&iinfo->ii_rwsem);
20263 + return !(iinfo->ii_hsb1 == h_inode->i_sb
20264 + && iinfo->ii_higen == h_inode->i_generation);
20267 +static inline void au_iigen_dec(struct inode *inode)
20269 + struct au_iinfo *iinfo;
20270 + struct au_iigen *iigen;
20272 + iinfo = au_ii(inode);
20273 + iigen = &iinfo->ii_generation;
20274 + spin_lock(&iigen->ig_spin);
20275 + iigen->ig_generation--;
20276 + spin_unlock(&iigen->ig_spin);
20279 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
20284 + if (unlikely(inode && au_iigen(inode, NULL) != sigen))
20290 +/* ---------------------------------------------------------------------- */
20292 +static inline struct au_hinode *au_hinode(struct au_iinfo *iinfo,
20293 + aufs_bindex_t bindex)
20295 + return iinfo->ii_hinode + bindex;
20298 +static inline int au_is_bad_inode(struct inode *inode)
20300 + return !!(is_bad_inode(inode) || !au_hinode(au_ii(inode), 0));
20303 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
20304 + aufs_bindex_t bindex)
20306 + IiMustAnyLock(inode);
20307 + return au_hinode(au_ii(inode), bindex)->hi_id;
20310 +static inline aufs_bindex_t au_ibtop(struct inode *inode)
20312 + IiMustAnyLock(inode);
20313 + return au_ii(inode)->ii_btop;
20316 +static inline aufs_bindex_t au_ibbot(struct inode *inode)
20318 + IiMustAnyLock(inode);
20319 + return au_ii(inode)->ii_bbot;
20322 +static inline struct au_vdir *au_ivdir(struct inode *inode)
20324 + IiMustAnyLock(inode);
20325 + return au_ii(inode)->ii_vdir;
20328 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
20330 + IiMustAnyLock(inode);
20331 + return au_hinode(au_ii(inode), bindex)->hi_whdentry;
20334 +static inline void au_set_ibtop(struct inode *inode, aufs_bindex_t bindex)
20336 + IiMustWriteLock(inode);
20337 + au_ii(inode)->ii_btop = bindex;
20340 +static inline void au_set_ibbot(struct inode *inode, aufs_bindex_t bindex)
20342 + IiMustWriteLock(inode);
20343 + au_ii(inode)->ii_bbot = bindex;
20346 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
20348 + IiMustWriteLock(inode);
20349 + au_ii(inode)->ii_vdir = vdir;
20352 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
20354 + IiMustAnyLock(inode);
20355 + return au_hinode(au_ii(inode), bindex);
20358 +/* ---------------------------------------------------------------------- */
20360 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
20363 + return pin->parent;
20367 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
20369 + if (pin && pin->hdir)
20370 + return pin->hdir->hi_inode;
20374 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
20377 + return pin->hdir;
20381 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
20384 + pin->dentry = dentry;
20387 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
20388 + unsigned char lflag)
20392 + au_fset_pin(pin->flags, DI_LOCKED);
20394 + au_fclr_pin(pin->flags, DI_LOCKED);
20398 +#if 0 /* reserved */
20399 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
20402 + dput(pin->parent);
20403 + pin->parent = dget(parent);
20408 +/* ---------------------------------------------------------------------- */
20411 +#ifdef CONFIG_AUFS_HNOTIFY
20412 +struct au_hnotify_op {
20413 + void (*ctl)(struct au_hinode *hinode, int do_set);
20414 + int (*alloc)(struct au_hinode *hinode);
20417 + * if it returns true, the caller should free hinode->hi_notify,
20418 + * otherwise ->free() frees it.
20420 + int (*free)(struct au_hinode *hinode,
20421 + struct au_hnotify *hn) __must_check;
20423 + void (*fin)(void);
20424 + int (*init)(void);
20426 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
20427 + void (*fin_br)(struct au_branch *br);
20428 + int (*init_br)(struct au_branch *br, int perm);
20432 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
20433 +void au_hn_free(struct au_hinode *hinode);
20434 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
20435 +void au_hn_reset(struct inode *inode, unsigned int flags);
20436 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
20437 + const struct qstr *h_child_qstr, struct inode *h_child_inode);
20438 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
20439 +int au_hnotify_init_br(struct au_branch *br, int perm);
20440 +void au_hnotify_fin_br(struct au_branch *br);
20441 +int __init au_hnotify_init(void);
20442 +void au_hnotify_fin(void);
20445 +extern const struct au_hnotify_op au_hnotify_op;
20448 +void au_hn_init(struct au_hinode *hinode)
20450 + hinode->hi_notify = NULL;
20453 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
20455 + return hinode->hi_notify;
20459 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
20460 + struct au_hinode *hinode __maybe_unused,
20461 + struct inode *inode __maybe_unused)
20462 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
20463 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
20464 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
20465 + int do_set __maybe_unused)
20466 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
20467 + unsigned int flags __maybe_unused)
20468 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
20469 + struct au_branch *br __maybe_unused,
20470 + int perm __maybe_unused)
20471 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
20472 + int perm __maybe_unused)
20473 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
20474 +AuStubInt0(__init au_hnotify_init, void)
20475 +AuStubVoid(au_hnotify_fin, void)
20476 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
20477 +#endif /* CONFIG_AUFS_HNOTIFY */
20479 +static inline void au_hn_suspend(struct au_hinode *hdir)
20481 + au_hn_ctl(hdir, /*do_set*/0);
20484 +static inline void au_hn_resume(struct au_hinode *hdir)
20486 + au_hn_ctl(hdir, /*do_set*/1);
20489 +static inline void au_hn_inode_lock(struct au_hinode *hdir)
20491 + inode_lock(hdir->hi_inode);
20492 + au_hn_suspend(hdir);
20495 +static inline void au_hn_inode_lock_nested(struct au_hinode *hdir,
20496 + unsigned int sc __maybe_unused)
20498 + inode_lock_nested(hdir->hi_inode, sc);
20499 + au_hn_suspend(hdir);
20502 +#if 0 /* unused */
20503 +#include "vfsub.h"
20504 +static inline void au_hn_inode_lock_shared_nested(struct au_hinode *hdir,
20507 + inode_lock_shared_nested(hdir->hi_inode, sc);
20508 + au_hn_suspend(hdir);
20512 +static inline void au_hn_inode_unlock(struct au_hinode *hdir)
20514 + au_hn_resume(hdir);
20515 + inode_unlock(hdir->hi_inode);
20518 +#endif /* __KERNEL__ */
20519 +#endif /* __AUFS_INODE_H__ */
20520 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
20521 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
20522 +++ linux/fs/aufs/ioctl.c 2022-11-05 23:02:18.965889284 +0100
20524 +// SPDX-License-Identifier: GPL-2.0
20526 + * Copyright (C) 2005-2022 Junjiro R. Okajima
20528 + * This program is free software; you can redistribute it and/or modify
20529 + * it under the terms of the GNU General Public License as published by
20530 + * the Free Software Foundation; either version 2 of the License, or
20531 + * (at your option) any later version.
20533 + * This program is distributed in the hope that it will be useful,
20534 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20535 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20536 + * GNU General Public License for more details.
20538 + * You should have received a copy of the GNU General Public License
20539 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20544 + * plink-management and readdir in userspace.
20545 + * assist the pathconf(3) wrapper library.
20547 + * File-based Hierarchical Storage Management.
20550 +#include <linux/compat.h>
20551 +#include <linux/file.h>
20554 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
20557 + aufs_bindex_t wbi, bindex, bbot;
20558 + struct file *h_file;
20559 + struct super_block *sb;
20560 + struct dentry *root;
20561 + struct au_branch *br;
20562 + struct aufs_wbr_fd wbrfd = {
20563 + .oflags = au_dir_roflags,
20566 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
20567 + | O_NOATIME | O_CLOEXEC;
20569 + AuDebugOn(wbrfd.oflags & ~valid);
20572 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
20573 + if (unlikely(err)) {
20579 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
20580 + wbrfd.oflags |= au_dir_roflags;
20581 + AuDbg("0%o\n", wbrfd.oflags);
20582 + if (unlikely(wbrfd.oflags & ~valid))
20586 + fd = get_unused_fd_flags(0);
20588 + if (unlikely(fd < 0))
20591 + h_file = ERR_PTR(-EINVAL);
20594 + sb = path->dentry->d_sb;
20595 + root = sb->s_root;
20596 + aufs_read_lock(root, AuLock_IR);
20597 + bbot = au_sbbot(sb);
20598 + if (wbrfd.brid >= 0) {
20599 + wbi = au_br_index(sb, wbrfd.brid);
20600 + if (unlikely(wbi < 0 || wbi > bbot))
20604 + h_file = ERR_PTR(-ENOENT);
20605 + br = au_sbr(sb, wbi);
20606 + if (!au_br_writable(br->br_perm)) {
20610 + bindex = wbi + 1;
20612 + for (; bindex <= bbot; bindex++) {
20613 + br = au_sbr(sb, bindex);
20614 + if (au_br_writable(br->br_perm)) {
20616 + br = au_sbr(sb, wbi);
20621 + AuDbg("wbi %d\n", wbi);
20623 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
20627 + aufs_read_unlock(root, AuLock_IR);
20628 + err = PTR_ERR(h_file);
20629 + if (IS_ERR(h_file))
20632 + au_lcnt_dec(&br->br_nfiles); /* cf. au_h_open() */
20633 + fd_install(fd, h_file);
20635 + goto out; /* success */
20638 + put_unused_fd(fd);
20644 +/* ---------------------------------------------------------------------- */
20646 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
20649 + struct dentry *dentry;
20652 + case AUFS_CTL_RDU:
20653 + case AUFS_CTL_RDU_INO:
20654 + err = au_rdu_ioctl(file, cmd, arg);
20657 + case AUFS_CTL_WBR_FD:
20658 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
20661 + case AUFS_CTL_IBUSY:
20662 + err = au_ibusy_ioctl(file, arg);
20665 + case AUFS_CTL_BRINFO:
20666 + err = au_brinfo_ioctl(file, arg);
20669 + case AUFS_CTL_FHSM_FD:
20670 + dentry = file->f_path.dentry;
20671 + if (IS_ROOT(dentry))
20672 + err = au_fhsm_fd(dentry->d_sb, arg);
20678 + /* do not call the lower */
20679 + AuDbg("0x%x\n", cmd);
20687 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
20692 + case AUFS_CTL_MVDOWN:
20693 + err = au_mvdown(file->f_path.dentry, (void __user *)arg);
20696 + case AUFS_CTL_WBR_FD:
20697 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
20701 + /* do not call the lower */
20702 + AuDbg("0x%x\n", cmd);
20710 +#ifdef CONFIG_COMPAT
20711 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
20712 + unsigned long arg)
20717 + case AUFS_CTL_RDU:
20718 + case AUFS_CTL_RDU_INO:
20719 + err = au_rdu_compat_ioctl(file, cmd, arg);
20722 + case AUFS_CTL_IBUSY:
20723 + err = au_ibusy_compat_ioctl(file, arg);
20726 + case AUFS_CTL_BRINFO:
20727 + err = au_brinfo_compat_ioctl(file, arg);
20731 + err = aufs_ioctl_dir(file, cmd, arg);
20738 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
20739 + unsigned long arg)
20741 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
20744 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
20745 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
20746 +++ linux/fs/aufs/i_op_add.c 2023-04-24 08:13:40.969576115 +0200
20748 +// SPDX-License-Identifier: GPL-2.0
20750 + * Copyright (C) 2005-2022 Junjiro R. Okajima
20752 + * This program is free software; you can redistribute it and/or modify
20753 + * it under the terms of the GNU General Public License as published by
20754 + * the Free Software Foundation; either version 2 of the License, or
20755 + * (at your option) any later version.
20757 + * This program is distributed in the hope that it will be useful,
20758 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20759 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20760 + * GNU General Public License for more details.
20762 + * You should have received a copy of the GNU General Public License
20763 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20767 + * inode operations (add entry)
20770 +#include <linux/iversion.h>
20774 + * final procedure of adding a new entry, except link(2).
20775 + * remove whiteout, instantiate, copyup the parent dir's times and size
20776 + * and update version.
20777 + * if it failed, re-create the removed whiteout.
20779 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
20780 + struct dentry *wh_dentry, struct dentry *dentry)
20783 + aufs_bindex_t bwh;
20784 + struct path h_path;
20785 + struct super_block *sb;
20786 + struct inode *inode, *h_dir;
20787 + struct dentry *wh;
20792 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
20793 + IMustLock(h_dir);
20794 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
20795 + bwh = au_dbwh(dentry);
20796 + h_path.dentry = wh_dentry;
20797 + h_path.mnt = au_sbr_mnt(sb, bindex);
20798 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
20800 + if (unlikely(err))
20804 + inode = au_new_inode(dentry, /*must_new*/1);
20805 + if (!IS_ERR(inode)) {
20806 + d_instantiate(dentry, inode);
20807 + dir = d_inode(dentry->d_parent); /* dir inode is locked */
20809 + au_dir_ts(dir, bindex);
20810 + inode_inc_iversion(dir);
20811 + au_fhsm_wrote(sb, bindex, /*force*/0);
20812 + return 0; /* success */
20815 + err = PTR_ERR(inode);
20820 + /* dir inode is locked */
20821 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
20822 + rerr = PTR_ERR(wh);
20823 + if (IS_ERR(wh)) {
20824 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
20825 + dentry, err, rerr);
20834 +static int au_d_may_add(struct dentry *dentry)
20839 + if (unlikely(d_unhashed(dentry)))
20841 + if (unlikely(d_really_is_positive(dentry)))
20847 + * simple tests for the adding inode operations.
20848 + * following the checks in vfs, plus the parent-child relationship.
20850 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
20851 + struct dentry *h_parent, int isdir)
20855 + struct dentry *h_dentry;
20856 + struct inode *h_inode;
20858 + err = -ENAMETOOLONG;
20859 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20862 + h_dentry = au_h_dptr(dentry, bindex);
20863 + if (d_really_is_negative(dentry)) {
20865 + if (unlikely(d_is_positive(h_dentry)))
20868 + /* rename(2) case */
20870 + if (unlikely(d_is_negative(h_dentry)))
20872 + h_inode = d_inode(h_dentry);
20873 + if (unlikely(!h_inode->i_nlink))
20876 + h_mode = h_inode->i_mode;
20879 + if (unlikely(S_ISDIR(h_mode)))
20881 + } else if (unlikely(!S_ISDIR(h_mode))) {
20888 + /* expected parent dir is locked */
20889 + if (unlikely(h_parent != h_dentry->d_parent))
20898 + * initial procedure of adding a new entry.
20899 + * prepare writable branch and the parent dir, lock it,
20900 + * and lookup whiteout for the new entry.
20902 +static struct dentry*
20903 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
20904 + struct dentry *src_dentry, struct au_pin *pin,
20905 + struct au_wr_dir_args *wr_dir_args)
20907 + struct dentry *wh_dentry, *h_parent;
20908 + struct super_block *sb;
20909 + struct au_branch *br;
20911 + unsigned int udba;
20912 + aufs_bindex_t bcpup;
20914 + AuDbg("%pd\n", dentry);
20916 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
20918 + wh_dentry = ERR_PTR(err);
20919 + if (unlikely(err < 0))
20922 + sb = dentry->d_sb;
20923 + udba = au_opt_udba(sb);
20924 + err = au_pin(pin, dentry, bcpup, udba,
20925 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20926 + wh_dentry = ERR_PTR(err);
20927 + if (unlikely(err))
20930 + h_parent = au_pinned_h_parent(pin);
20931 + if (udba != AuOpt_UDBA_NONE
20932 + && au_dbtop(dentry) == bcpup)
20933 + err = au_may_add(dentry, bcpup, h_parent,
20934 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
20935 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20936 + err = -ENAMETOOLONG;
20937 + wh_dentry = ERR_PTR(err);
20938 + if (unlikely(err))
20941 + br = au_sbr(sb, bcpup);
20943 + struct path tmp = {
20944 + .dentry = h_parent,
20945 + .mnt = au_br_mnt(br)
20947 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
20950 + wh_dentry = NULL;
20951 + if (bcpup != au_dbwh(dentry))
20952 + goto out; /* success */
20955 + * ENAMETOOLONG here means that if we allowed create such name, then it
20956 + * would not be able to removed in the future. So we don't allow such
20957 + * name here and we don't handle ENAMETOOLONG differently here.
20959 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
20962 + if (IS_ERR(wh_dentry))
20965 + return wh_dentry;
20968 +/* ---------------------------------------------------------------------- */
20970 +enum { Mknod, Symlink, Creat };
20971 +struct simple_arg {
20978 + struct vfsub_aopen_args *aopen;
20981 + const char *symname;
20990 +static int add_simple(struct inode *dir, struct dentry *dentry,
20991 + struct simple_arg *arg)
20994 + aufs_bindex_t btop;
20995 + unsigned char created;
20996 + const unsigned char try_aopen
20997 + = (arg->type == Creat && arg->u.c.try_aopen);
20998 + struct vfsub_aopen_args *aopen = arg->u.c.aopen;
20999 + struct dentry *wh_dentry, *parent;
21000 + struct inode *h_dir;
21001 + struct super_block *sb;
21002 + struct au_branch *br;
21003 + /* to reduce stack size */
21005 + struct au_dtime dt;
21006 + struct au_pin pin;
21007 + struct path h_path;
21008 + struct au_wr_dir_args wr_dir_args;
21011 + AuDbg("%pd\n", dentry);
21015 + a = kmalloc(sizeof(*a), GFP_NOFS);
21016 + if (unlikely(!a))
21018 + a->wr_dir_args.force_btgt = -1;
21019 + a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
21021 + parent = dentry->d_parent; /* dir inode is locked */
21022 + if (!try_aopen) {
21023 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21024 + if (unlikely(err))
21027 + err = au_d_may_add(dentry);
21028 + if (unlikely(err))
21031 + di_write_lock_parent(parent);
21032 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21033 + &a->pin, &a->wr_dir_args);
21034 + err = PTR_ERR(wh_dentry);
21035 + if (IS_ERR(wh_dentry))
21038 + btop = au_dbtop(dentry);
21039 + sb = dentry->d_sb;
21040 + br = au_sbr(sb, btop);
21041 + a->h_path.dentry = au_h_dptr(dentry, btop);
21042 + a->h_path.mnt = au_br_mnt(br);
21043 + h_dir = au_pinned_h_dir(&a->pin);
21044 + switch (arg->type) {
21046 + if (!try_aopen || !h_dir->i_op->atomic_open) {
21047 + err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
21048 + arg->u.c.want_excl);
21050 + if (!err && try_aopen)
21051 + aopen->file->f_mode |= FMODE_CREATED;
21054 + err = vfsub_atomic_open(h_dir, a->h_path.dentry, aopen);
21055 + AuDbg("err %d\n", err);
21056 + AuDbgFile(aopen->file);
21057 + created = err >= 0
21058 + && !!(aopen->file->f_mode & FMODE_CREATED);
21062 + err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
21066 + err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
21073 + if (unlikely(err < 0))
21076 + err = epilog(dir, btop, wh_dentry, dentry);
21078 + goto out_unpin; /* success */
21081 + if (created /* && d_is_positive(a->h_path.dentry) */) {
21082 + /* no delegation since it is just created */
21083 + rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
21086 + AuIOErr("%pd revert failure(%d, %d)\n",
21087 + dentry, err, rerr);
21090 + au_dtime_revert(&a->dt);
21092 + if (try_aopen && h_dir->i_op->atomic_open
21093 + && (aopen->file->f_mode & FMODE_OPENED))
21094 + /* aopen->file is still opened */
21095 + au_lcnt_dec(&aopen->br->br_nfiles);
21098 + au_unpin(&a->pin);
21102 + di_write_unlock(parent);
21104 + if (unlikely(err)) {
21105 + au_update_dbtop(dentry);
21109 + aufs_read_unlock(dentry, AuLock_DW);
21116 +int aufs_mknod(struct mnt_idmap *idmap, struct inode *dir,
21117 + struct dentry *dentry, umode_t mode, dev_t dev)
21119 + struct simple_arg arg = {
21126 + return add_simple(dir, dentry, &arg);
21129 +int aufs_symlink(struct mnt_idmap *idmap, struct inode *dir,
21130 + struct dentry *dentry, const char *symname)
21132 + struct simple_arg arg = {
21134 + .u.s.symname = symname
21136 + return add_simple(dir, dentry, &arg);
21139 +int aufs_create(struct mnt_idmap *idmap, struct inode *dir,
21140 + struct dentry *dentry, umode_t mode, bool want_excl)
21142 + struct simple_arg arg = {
21146 + .want_excl = want_excl
21149 + return add_simple(dir, dentry, &arg);
21152 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
21153 + struct vfsub_aopen_args *aopen_args)
21155 + struct simple_arg arg = {
21158 + .mode = aopen_args->create_mode,
21159 + .want_excl = aopen_args->open_flag & O_EXCL,
21160 + .try_aopen = true,
21161 + .aopen = aopen_args
21164 + return add_simple(dir, dentry, &arg);
21167 +int aufs_tmpfile(struct mnt_idmap *idmap, struct inode *dir,
21168 + struct file *file, umode_t mode)
21171 + aufs_bindex_t bindex;
21172 + struct path h_ppath;
21173 + struct super_block *sb;
21174 + struct au_branch *br;
21175 + struct dentry *dentry, *parent, *h_parent, *h_dentry;
21176 + struct inode *h_dir, *inode;
21177 + struct vfsmount *h_mnt;
21178 + struct mnt_idmap *h_idmap;
21179 + struct file *h_file;
21180 + struct au_wr_dir_args wr_dir_args = {
21181 + .force_btgt = -1,
21182 + .flags = AuWrDir_TMPFILE
21185 + /* copy-up may happen */
21190 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21191 + if (unlikely(err))
21194 + dentry = file->f_path.dentry;
21195 + err = au_di_init(dentry);
21196 + if (unlikely(err))
21200 + parent = d_find_any_alias(dir);
21201 + AuDebugOn(!parent);
21202 + di_write_lock_parent(parent);
21203 + if (unlikely(d_inode(parent) != dir))
21206 + err = au_digen_test(parent, au_sigen(sb));
21207 + if (unlikely(err))
21210 + bindex = au_dbtop(parent);
21211 + au_set_dbtop(dentry, bindex);
21212 + au_set_dbbot(dentry, bindex);
21213 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
21215 + if (unlikely(err < 0))
21218 + err = -EOPNOTSUPP;
21219 + h_dir = au_h_iptr(dir, bindex);
21220 + if (unlikely(!h_dir->i_op->tmpfile))
21223 + br = au_sbr(sb, bindex);
21224 + h_mnt = au_br_mnt(br);
21225 + err = vfsub_mnt_want_write(h_mnt);
21226 + if (unlikely(err))
21229 + h_idmap = mnt_idmap(h_mnt);
21230 + h_parent = au_h_dptr(parent, bindex);
21231 + h_ppath.mnt = h_mnt;
21232 + h_ppath.dentry = h_parent;
21233 + h_file = vfs_tmpfile_open(h_idmap, &h_ppath, mode, /*open_flag*/0,
21235 + if (IS_ERR(h_file)) {
21236 + err = PTR_ERR(h_file);
21241 + h_dentry = h_file->f_path.dentry;
21242 + au_set_dbtop(dentry, bindex);
21243 + au_set_dbbot(dentry, bindex);
21244 + au_set_h_dptr(dentry, bindex, dget(h_dentry));
21245 + inode = au_new_inode(dentry, /*must_new*/1);
21246 + if (IS_ERR(inode)) {
21247 + err = PTR_ERR(inode);
21248 + au_set_h_dptr(dentry, bindex, NULL);
21249 + au_set_dbtop(dentry, -1);
21250 + au_set_dbbot(dentry, -1);
21254 + if (!inode->i_nlink)
21255 + set_nlink(inode, 1);
21256 + d_tmpfile(file, inode);
21257 + au_di(dentry)->di_tmpfile = 1;
21258 + get_file(h_file);
21259 + au_di(dentry)->di_htmpfile = h_file;
21261 + /* update without i_mutex */
21262 + if (au_ibtop(dir) == au_dbtop(dentry))
21263 + au_cpup_attr_timesizes(dir);
21268 + vfsub_mnt_drop_write(h_mnt);
21270 + di_write_unlock(parent);
21272 + di_write_unlock(dentry);
21277 + au_di(dentry)->di_htmpfile = NULL;
21278 + au_di_fin(dentry);
21279 + dentry->d_fsdata = NULL;
21281 + si_read_unlock(sb);
21282 + if (!err && h_file) {
21284 + err = finish_open_simple(file, err);
21286 + au_lcnt_inc(&br->br_nfiles);
21289 + au_di(dentry)->di_htmpfile = NULL;
21290 + au_di_fin(dentry);
21291 + dentry->d_fsdata = NULL;
21295 + inode_unlock(dir);
21300 +/* ---------------------------------------------------------------------- */
21302 +struct au_link_args {
21303 + aufs_bindex_t bdst, bsrc;
21304 + struct au_pin pin;
21305 + struct path h_path;
21306 + struct dentry *src_parent, *parent;
21309 +static int au_cpup_before_link(struct dentry *src_dentry,
21310 + struct au_link_args *a)
21313 + struct dentry *h_src_dentry;
21314 + struct au_cp_generic cpg = {
21315 + .dentry = src_dentry,
21320 + .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
21323 + di_read_lock_parent(a->src_parent, AuLock_IR);
21324 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
21325 + if (unlikely(err))
21328 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
21329 + err = au_pin(&a->pin, src_dentry, a->bdst,
21330 + au_opt_udba(src_dentry->d_sb),
21331 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21332 + if (unlikely(err))
21335 + err = au_sio_cpup_simple(&cpg);
21336 + au_unpin(&a->pin);
21339 + di_read_unlock(a->src_parent, AuLock_IR);
21343 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
21344 + struct au_link_args *a)
21347 + unsigned char plink;
21348 + aufs_bindex_t bbot;
21349 + struct dentry *h_src_dentry;
21350 + struct inode *h_inode, *inode, *delegated;
21351 + struct super_block *sb;
21352 + struct file *h_file;
21356 + sb = src_dentry->d_sb;
21357 + inode = d_inode(src_dentry);
21358 + if (au_ibtop(inode) <= a->bdst)
21359 + h_inode = au_h_iptr(inode, a->bdst);
21360 + if (!h_inode || !h_inode->i_nlink) {
21361 + /* copyup src_dentry as the name of dentry. */
21362 + bbot = au_dbbot(dentry);
21363 + if (bbot < a->bsrc)
21364 + au_set_dbbot(dentry, a->bsrc);
21365 + au_set_h_dptr(dentry, a->bsrc,
21366 + dget(au_h_dptr(src_dentry, a->bsrc)));
21367 + dget(a->h_path.dentry);
21368 + au_set_h_dptr(dentry, a->bdst, NULL);
21369 + AuDbg("temporary d_inode...\n");
21370 + spin_lock(&dentry->d_lock);
21371 + dentry->d_inode = d_inode(src_dentry); /* tmp */
21372 + spin_unlock(&dentry->d_lock);
21373 + h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
21374 + if (IS_ERR(h_file))
21375 + err = PTR_ERR(h_file);
21377 + struct au_cp_generic cpg = {
21378 + .dentry = dentry,
21383 + .flags = AuCpup_KEEPLINO
21385 + err = au_sio_cpup_simple(&cpg);
21386 + au_h_open_post(dentry, a->bsrc, h_file);
21388 + dput(a->h_path.dentry);
21389 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21391 + au_set_h_dptr(dentry, a->bdst,
21392 + a->h_path.dentry);
21394 + spin_lock(&dentry->d_lock);
21395 + dentry->d_inode = NULL; /* restore */
21396 + spin_unlock(&dentry->d_lock);
21397 + AuDbg("temporary d_inode...done\n");
21398 + au_set_h_dptr(dentry, a->bsrc, NULL);
21399 + au_set_dbbot(dentry, bbot);
21401 + /* the inode of src_dentry already exists on a.bdst branch */
21402 + h_src_dentry = d_find_alias(h_inode);
21403 + if (!h_src_dentry && au_plink_test(inode)) {
21405 + h_src_dentry = au_plink_lkup(inode, a->bdst);
21406 + err = PTR_ERR(h_src_dentry);
21407 + if (IS_ERR(h_src_dentry))
21410 + if (unlikely(d_is_negative(h_src_dentry))) {
21411 + dput(h_src_dentry);
21412 + h_src_dentry = NULL;
21416 + if (h_src_dentry) {
21417 + delegated = NULL;
21418 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21419 + &a->h_path, &delegated);
21420 + if (unlikely(err == -EWOULDBLOCK)) {
21421 + pr_warn("cannot retry for NFSv4 delegation"
21422 + " for an internal link\n");
21425 + dput(h_src_dentry);
21427 + AuIOErr("no dentry found for hi%lu on b%d\n",
21428 + h_inode->i_ino, a->bdst);
21433 + if (!err && !plink)
21434 + au_plink_append(inode, a->bdst, a->h_path.dentry);
21441 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
21442 + struct dentry *dentry)
21445 + struct au_dtime dt;
21446 + struct au_link_args *a;
21447 + struct dentry *wh_dentry, *h_src_dentry;
21448 + struct inode *inode, *delegated;
21449 + struct super_block *sb;
21450 + struct au_wr_dir_args wr_dir_args = {
21451 + /* .force_btgt = -1, */
21452 + .flags = AuWrDir_ADD_ENTRY
21456 + inode = d_inode(src_dentry);
21457 + IMustLock(inode);
21460 + a = kzalloc(sizeof(*a), GFP_NOFS);
21461 + if (unlikely(!a))
21464 + a->parent = dentry->d_parent; /* dir inode is locked */
21465 + err = aufs_read_and_write_lock2(dentry, src_dentry,
21466 + AuLock_NOPLM | AuLock_GEN);
21467 + if (unlikely(err))
21469 + err = au_d_linkable(src_dentry);
21470 + if (unlikely(err))
21472 + err = au_d_may_add(dentry);
21473 + if (unlikely(err))
21476 + a->src_parent = dget_parent(src_dentry);
21477 + wr_dir_args.force_btgt = au_ibtop(inode);
21479 + di_write_lock_parent(a->parent);
21480 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
21481 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
21483 + err = PTR_ERR(wh_dentry);
21484 + if (IS_ERR(wh_dentry))
21488 + sb = dentry->d_sb;
21489 + a->bdst = au_dbtop(dentry);
21490 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
21491 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
21492 + a->bsrc = au_ibtop(inode);
21493 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21494 + if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
21495 + h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
21496 + if (!h_src_dentry) {
21497 + a->bsrc = au_dbtop(src_dentry);
21498 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
21499 + AuDebugOn(!h_src_dentry);
21500 + } else if (IS_ERR(h_src_dentry)) {
21501 + err = PTR_ERR(h_src_dentry);
21506 + * aufs doesn't touch the credential so
21507 + * security_dentry_create_files_as() is unnecessary.
21509 + if (au_opt_test(au_mntflags(sb), PLINK)) {
21510 + if (a->bdst < a->bsrc
21511 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
21512 + err = au_cpup_or_link(src_dentry, dentry, a);
21514 + delegated = NULL;
21515 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
21516 + &a->h_path, &delegated);
21517 + if (unlikely(err == -EWOULDBLOCK)) {
21518 + pr_warn("cannot retry for NFSv4 delegation"
21519 + " for an internal link\n");
21523 + dput(h_src_dentry);
21526 + * copyup src_dentry to the branch we process,
21527 + * and then link(2) to it.
21529 + dput(h_src_dentry);
21530 + if (a->bdst < a->bsrc
21531 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
21532 + au_unpin(&a->pin);
21533 + di_write_unlock(a->parent);
21534 + err = au_cpup_before_link(src_dentry, a);
21535 + di_write_lock_parent(a->parent);
21537 + err = au_pin(&a->pin, dentry, a->bdst,
21539 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21540 + if (unlikely(err))
21544 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
21546 + if (h_src_dentry && d_is_positive(h_src_dentry)) {
21547 + delegated = NULL;
21548 + err = vfsub_link(h_src_dentry,
21549 + au_pinned_h_dir(&a->pin),
21550 + &a->h_path, &delegated);
21551 + if (unlikely(err == -EWOULDBLOCK)) {
21552 + pr_warn("cannot retry"
21553 + " for NFSv4 delegation"
21554 + " for an internal link\n");
21560 + if (unlikely(err))
21564 + a->h_path.dentry = wh_dentry;
21565 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
21567 + if (unlikely(err))
21571 + au_dir_ts(dir, a->bdst);
21572 + inode_inc_iversion(dir);
21573 + inc_nlink(inode);
21574 + inode->i_ctime = dir->i_ctime;
21575 + d_instantiate(dentry, au_igrab(inode));
21576 + if (d_unhashed(a->h_path.dentry))
21577 + /* some filesystem calls d_drop() */
21579 + /* some filesystems consume an inode even hardlink */
21580 + au_fhsm_wrote(sb, a->bdst, /*force*/0);
21581 + goto out_unpin; /* success */
21584 + /* no delegation since it is just created */
21585 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
21586 + /*delegated*/NULL, /*force*/0);
21587 + if (unlikely(rerr)) {
21588 + AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
21591 + au_dtime_revert(&dt);
21593 + au_unpin(&a->pin);
21597 + di_write_unlock(a->parent);
21598 + dput(a->src_parent);
21600 + if (unlikely(err)) {
21601 + au_update_dbtop(dentry);
21604 + aufs_read_and_write_unlock2(dentry, src_dentry);
21612 +int aufs_mkdir(struct mnt_idmap *idmap, struct inode *dir,
21613 + struct dentry *dentry, umode_t mode)
21616 + aufs_bindex_t bindex;
21617 + unsigned char diropq;
21618 + struct path h_path;
21619 + struct dentry *wh_dentry, *parent, *opq_dentry;
21620 + struct inode *h_inode;
21621 + struct super_block *sb;
21623 + struct au_pin pin;
21624 + struct au_dtime dt;
21625 + } *a; /* reduce the stack usage */
21626 + struct au_wr_dir_args wr_dir_args = {
21627 + .force_btgt = -1,
21628 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
21634 + a = kmalloc(sizeof(*a), GFP_NOFS);
21635 + if (unlikely(!a))
21638 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
21639 + if (unlikely(err))
21641 + err = au_d_may_add(dentry);
21642 + if (unlikely(err))
21645 + parent = dentry->d_parent; /* dir inode is locked */
21646 + di_write_lock_parent(parent);
21647 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
21648 + &a->pin, &wr_dir_args);
21649 + err = PTR_ERR(wh_dentry);
21650 + if (IS_ERR(wh_dentry))
21653 + sb = dentry->d_sb;
21654 + bindex = au_dbtop(dentry);
21655 + h_path.dentry = au_h_dptr(dentry, bindex);
21656 + h_path.mnt = au_sbr_mnt(sb, bindex);
21657 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
21658 + if (unlikely(err))
21661 + /* make the dir opaque */
21663 + h_inode = d_inode(h_path.dentry);
21665 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
21666 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
21667 + opq_dentry = au_diropq_create(dentry, bindex);
21668 + inode_unlock(h_inode);
21669 + err = PTR_ERR(opq_dentry);
21670 + if (IS_ERR(opq_dentry))
21672 + dput(opq_dentry);
21676 + err = epilog(dir, bindex, wh_dentry, dentry);
21679 + goto out_unpin; /* success */
21684 + AuLabel(revert opq);
21685 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
21686 + rerr = au_diropq_remove(dentry, bindex);
21687 + inode_unlock(h_inode);
21689 + AuIOErr("%pd reverting diropq failed(%d, %d)\n",
21690 + dentry, err, rerr);
21696 + AuLabel(revert dir);
21697 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
21699 + AuIOErr("%pd reverting dir failed(%d, %d)\n",
21700 + dentry, err, rerr);
21703 + au_dtime_revert(&a->dt);
21705 + au_unpin(&a->pin);
21708 + di_write_unlock(parent);
21710 + if (unlikely(err)) {
21711 + au_update_dbtop(dentry);
21714 + aufs_read_unlock(dentry, AuLock_DW);
21720 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
21721 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
21722 +++ linux/fs/aufs/i_op.c 2023-04-24 08:13:40.969576115 +0200
21724 +// SPDX-License-Identifier: GPL-2.0
21726 + * Copyright (C) 2005-2022 Junjiro R. Okajima
21728 + * This program is free software; you can redistribute it and/or modify
21729 + * it under the terms of the GNU General Public License as published by
21730 + * the Free Software Foundation; either version 2 of the License, or
21731 + * (at your option) any later version.
21733 + * This program is distributed in the hope that it will be useful,
21734 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21735 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21736 + * GNU General Public License for more details.
21738 + * You should have received a copy of the GNU General Public License
21739 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21743 + * inode operations (except add/del/rename)
21746 +#include <linux/device_cgroup.h>
21747 +#include <linux/filelock.h>
21748 +#include <linux/fs_stack.h>
21749 +#include <linux/iversion.h>
21750 +#include <linux/security.h>
21753 +static int h_permission(struct inode *h_inode, int mask,
21754 + struct path *h_path, int brperm)
21757 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21758 + struct mnt_idmap *h_idmap;
21761 + if (write_mask && IS_IMMUTABLE(h_inode))
21765 + if (((mask & MAY_EXEC)
21766 + && S_ISREG(h_inode->i_mode)
21767 + && (path_noexec(h_path)
21768 + || !(h_inode->i_mode & 0111))))
21772 + * - skip the lower fs test in the case of write to ro branch.
21773 + * - nfs dir permission write check is optimized, but a policy for
21774 + * link/rename requires a real check.
21775 + * - nfs always sets SB_POSIXACL regardless its mount option 'noacl.'
21776 + * in this case, generic_permission() returns -EOPNOTSUPP.
21778 + h_idmap = mnt_idmap(h_path->mnt);
21779 + if ((write_mask && !au_br_writable(brperm))
21780 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
21781 + && write_mask && !(mask & MAY_READ))
21782 + || !h_inode->i_op->permission) {
21783 + /* AuLabel(generic_permission); */
21784 + /* AuDbg("get_inode_acl %ps\n",
21785 + h_inode->i_op->get_inode_acl); */
21786 + err = generic_permission(h_idmap, h_inode, mask);
21787 + if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
21788 + err = h_inode->i_op->permission(h_idmap, h_inode,
21792 + /* AuLabel(h_inode->permission); */
21793 + err = h_inode->i_op->permission(h_idmap, h_inode, mask);
21798 + err = devcgroup_inode_permission(h_inode, mask);
21800 + err = security_inode_permission(h_inode, mask);
21806 +static int aufs_permission(struct mnt_idmap *idmap, struct inode *inode,
21810 + aufs_bindex_t bindex, bbot;
21811 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
21812 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
21813 + struct inode *h_inode;
21814 + struct super_block *sb;
21815 + struct au_branch *br;
21817 + /* todo: support rcu-walk? */
21818 + if (mask & MAY_NOT_BLOCK)
21821 + sb = inode->i_sb;
21822 + si_read_lock(sb, AuLock_FLUSH);
21823 + ii_read_lock_child(inode);
21824 +#if 0 /* reserved for future use */
21826 + * This test may be rather 'too much' since the test is essentially done
21827 + * in the aufs_lookup(). Theoretically it is possible that the inode
21828 + * generation doesn't match to the superblock's here. But it isn't a
21829 + * big deal I suppose.
21831 + err = au_iigen_test(inode, au_sigen(sb));
21832 + if (unlikely(err))
21838 + || au_opt_test(au_mntflags(sb), DIRPERM1)) {
21839 + err = au_busy_or_stale();
21840 + h_inode = au_h_iptr(inode, au_ibtop(inode));
21841 + if (unlikely(!h_inode
21842 + || (h_inode->i_mode & S_IFMT)
21843 + != (inode->i_mode & S_IFMT)))
21847 + bindex = au_ibtop(inode);
21848 + br = au_sbr(sb, bindex);
21849 + err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
21852 + && !special_file(h_inode->i_mode)) {
21853 + /* test whether the upper writable branch exists */
21855 + for (; bindex >= 0; bindex--)
21856 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
21864 + /* non-write to dir */
21866 + bbot = au_ibbot(inode);
21867 + for (bindex = au_ibtop(inode); !err && bindex <= bbot; bindex++) {
21868 + h_inode = au_h_iptr(inode, bindex);
21870 + err = au_busy_or_stale();
21871 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
21874 + br = au_sbr(sb, bindex);
21875 + err = h_permission(h_inode, mask, &br->br_path,
21881 + ii_read_unlock(inode);
21882 + si_read_unlock(sb);
21886 +/* ---------------------------------------------------------------------- */
21888 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
21889 + unsigned int flags)
21891 + struct dentry *ret, *parent;
21892 + struct inode *inode;
21893 + struct super_block *sb;
21894 + int err, npositive;
21898 + /* todo: support rcu-walk? */
21899 + ret = ERR_PTR(-ECHILD);
21900 + if (flags & LOOKUP_RCU)
21903 + ret = ERR_PTR(-ENAMETOOLONG);
21904 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
21908 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21909 + ret = ERR_PTR(err);
21910 + if (unlikely(err))
21913 + err = au_di_init(dentry);
21914 + ret = ERR_PTR(err);
21915 + if (unlikely(err))
21919 + npositive = 0; /* suppress a warning */
21920 + parent = dentry->d_parent; /* dir inode is locked */
21921 + di_read_lock_parent(parent, AuLock_IR);
21922 + err = au_alive_dir(parent);
21924 + err = au_digen_test(parent, au_sigen(sb));
21926 + /* regardless LOOKUP_CREATE, always ALLOW_NEG */
21927 + npositive = au_lkup_dentry(dentry, au_dbtop(parent),
21928 + AuLkup_ALLOW_NEG);
21931 + di_read_unlock(parent, AuLock_IR);
21932 + ret = ERR_PTR(err);
21933 + if (unlikely(err < 0))
21937 + inode = au_new_inode(dentry, /*must_new*/0);
21938 + if (IS_ERR(inode)) {
21939 + ret = (void *)inode;
21946 + atomic_inc(&inode->i_count);
21947 + ret = d_splice_alias(inode, dentry);
21948 +#if 0 /* reserved for future use */
21949 + if (unlikely(d_need_lookup(dentry))) {
21950 + spin_lock(&dentry->d_lock);
21951 + dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
21952 + spin_unlock(&dentry->d_lock);
21956 + if (!IS_ERR(ret)) {
21958 + if (ret && ret != dentry)
21959 + ii_write_unlock(inode);
21961 + ii_write_unlock(inode);
21968 + di_write_unlock(dentry);
21970 + si_read_unlock(sb);
21975 +/* ---------------------------------------------------------------------- */
21978 + * very dirty and complicated aufs ->atomic_open().
21979 + * aufs_atomic_open()
21980 + * + au_aopen_or_create()
21982 + * + vfsub_atomic_open()
21983 + * + branch fs ->atomic_open()
21984 + * may call the actual 'open' for h_file
21985 + * + inc br_nfiles only if opened
21986 + * + au_aopen_no_open() or au_aopen_do_open()
21988 + * au_aopen_do_open()
21989 + * + finish_open()
21990 + * + au_do_aopen()
21991 + * + au_do_open() the body of all 'open'
21992 + * + au_do_open_nondir()
21993 + * set the passed h_file
21995 + * au_aopen_no_open()
21996 + * + finish_no_open()
21999 +struct aopen_node {
22000 + struct hlist_bl_node hblist;
22001 + struct file *file, *h_file;
22004 +static int au_do_aopen(struct inode *inode, struct file *file)
22006 + struct hlist_bl_head *aopen;
22007 + struct hlist_bl_node *pos;
22008 + struct aopen_node *node;
22009 + struct au_do_open_args args = {
22011 + .open = au_do_open_nondir
22014 + aopen = &au_sbi(inode->i_sb)->si_aopen;
22015 + hlist_bl_lock(aopen);
22016 + hlist_bl_for_each_entry(node, pos, aopen, hblist)
22017 + if (node->file == file) {
22018 + args.h_file = node->h_file;
22021 + hlist_bl_unlock(aopen);
22022 + /* AuDebugOn(!args.h_file); */
22024 + return au_do_open(file, &args);
22027 +static int au_aopen_do_open(struct file *file, struct dentry *dentry,
22028 + struct aopen_node *aopen_node)
22031 + struct hlist_bl_head *aopen;
22034 + aopen = &au_sbi(dentry->d_sb)->si_aopen;
22035 + au_hbl_add(&aopen_node->hblist, aopen);
22036 + err = finish_open(file, dentry, au_do_aopen);
22037 + au_hbl_del(&aopen_node->hblist, aopen);
22038 + /* AuDbgFile(file); */
22039 + AuDbg("%pd%s%s\n", dentry,
22040 + (file->f_mode & FMODE_CREATED) ? " created" : "",
22041 + (file->f_mode & FMODE_OPENED) ? " opened" : "");
22047 +static int au_aopen_no_open(struct file *file, struct dentry *dentry)
22053 + err = finish_no_open(file, dentry);
22059 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
22060 + struct file *file, unsigned int open_flag,
22061 + umode_t create_mode)
22063 + int err, did_open;
22064 + unsigned int lkup_flags;
22065 + aufs_bindex_t bindex;
22066 + struct super_block *sb;
22067 + struct dentry *parent, *d;
22068 + struct vfsub_aopen_args args = {
22069 + .open_flag = open_flag,
22070 + .create_mode = create_mode
22072 + struct aopen_node aopen_node = {
22077 + AuDbg("open_flag 0%o\n", open_flag);
22078 + AuDbgDentry(dentry);
22081 + if (!au_di(dentry)) {
22082 + lkup_flags = LOOKUP_OPEN;
22083 + if (open_flag & O_CREAT)
22084 + lkup_flags |= LOOKUP_CREATE;
22085 + d = aufs_lookup(dir, dentry, lkup_flags);
22087 + err = PTR_ERR(d);
22092 + * obsoleted dentry found.
22093 + * another error will be returned later.
22099 + AuDbgDentry(dentry);
22102 + if (d_is_positive(dentry)
22103 + || d_unhashed(dentry)
22104 + || d_unlinked(dentry)
22105 + || !(open_flag & O_CREAT)) {
22106 + err = au_aopen_no_open(file, dentry);
22107 + goto out; /* success */
22110 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
22111 + if (unlikely(err))
22114 + sb = dentry->d_sb;
22115 + parent = dentry->d_parent; /* dir is locked */
22116 + di_write_lock_parent(parent);
22117 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
22118 + if (unlikely(err < 0))
22121 + AuDbgDentry(dentry);
22122 + if (d_is_positive(dentry)) {
22123 + err = au_aopen_no_open(file, dentry);
22124 + goto out_parent; /* success */
22127 + args.file = alloc_empty_file(file->f_flags, current_cred());
22128 + err = PTR_ERR(args.file);
22129 + if (IS_ERR(args.file))
22132 + bindex = au_dbtop(dentry);
22133 + err = au_aopen_or_create(dir, dentry, &args);
22135 + AuDbgFile(args.file);
22136 + file->f_mode = args.file->f_mode & ~FMODE_OPENED;
22137 + did_open = !!(args.file->f_mode & FMODE_OPENED);
22140 + args.file = NULL;
22142 + di_write_unlock(parent);
22143 + di_write_unlock(dentry);
22144 + if (unlikely(err < 0)) {
22151 + err = au_aopen_no_open(file, dentry);
22153 + aopen_node.h_file = args.file;
22154 + err = au_aopen_do_open(file, dentry, &aopen_node);
22156 + if (unlikely(err < 0)) {
22160 + au_lcnt_dec(&args.br->br_nfiles);
22162 + goto out_sb; /* success */
22165 + di_write_unlock(parent);
22166 + di_write_unlock(dentry);
22168 + si_read_unlock(sb);
22176 +/* ---------------------------------------------------------------------- */
22178 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
22179 + const unsigned char add_entry, aufs_bindex_t bcpup,
22180 + aufs_bindex_t btop)
22183 + struct dentry *h_parent;
22184 + struct inode *h_dir;
22187 + IMustLock(d_inode(parent));
22189 + di_write_lock_parent(parent);
22192 + if (!au_h_dptr(parent, bcpup)) {
22193 + if (btop > bcpup)
22194 + err = au_cpup_dirs(dentry, bcpup);
22195 + else if (btop < bcpup)
22196 + err = au_cpdown_dirs(dentry, bcpup);
22200 + if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
22201 + h_parent = au_h_dptr(parent, bcpup);
22202 + h_dir = d_inode(h_parent);
22203 + inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
22204 + err = au_lkup_neg(dentry, bcpup, /*wh*/0);
22205 + /* todo: no unlock here */
22206 + inode_unlock_shared(h_dir);
22208 + AuDbg("bcpup %d\n", bcpup);
22210 + if (d_really_is_negative(dentry))
22211 + au_set_h_dptr(dentry, btop, NULL);
22212 + au_update_dbrange(dentry, /*do_put_zero*/0);
22217 + di_write_unlock(parent);
22219 + err = bcpup; /* success */
22226 + * decide the branch and the parent dir where we will create a new entry.
22227 + * returns new bindex or an error.
22228 + * copyup the parent dir if needed.
22230 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
22231 + struct au_wr_dir_args *args)
22234 + unsigned int flags;
22235 + aufs_bindex_t bcpup, btop, src_btop;
22236 + const unsigned char add_entry
22237 + = au_ftest_wrdir(args->flags, ADD_ENTRY)
22238 + | au_ftest_wrdir(args->flags, TMPFILE);
22239 + struct super_block *sb;
22240 + struct dentry *parent;
22241 + struct au_sbinfo *sbinfo;
22243 + sb = dentry->d_sb;
22244 + sbinfo = au_sbi(sb);
22245 + parent = dget_parent(dentry);
22246 + btop = au_dbtop(dentry);
22248 + if (args->force_btgt < 0) {
22249 + if (src_dentry) {
22250 + src_btop = au_dbtop(src_dentry);
22251 + if (src_btop < btop)
22252 + bcpup = src_btop;
22253 + } else if (add_entry) {
22255 + if (au_ftest_wrdir(args->flags, ISDIR))
22256 + au_fset_wbr(flags, DIR);
22257 + err = AuWbrCreate(sbinfo, dentry, flags);
22261 + if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
22263 + err = AuWbrCopyup(sbinfo, dentry);
22265 + if (!IS_ROOT(dentry)) {
22266 + di_read_lock_parent(parent, !AuLock_IR);
22267 + err = AuWbrCopyup(sbinfo, dentry);
22268 + di_read_unlock(parent, !AuLock_IR);
22270 + err = AuWbrCopyup(sbinfo, dentry);
22273 + if (unlikely(err < 0))
22277 + bcpup = args->force_btgt;
22278 + AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
22281 + AuDbg("btop %d, bcpup %d\n", btop, bcpup);
22283 + if (bcpup == btop)
22284 + goto out; /* success */
22286 + /* copyup the new parent into the branch we process */
22287 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, btop);
22289 + if (d_really_is_negative(dentry)) {
22290 + au_set_h_dptr(dentry, btop, NULL);
22291 + au_set_dbtop(dentry, bcpup);
22292 + au_set_dbbot(dentry, bcpup);
22294 + AuDebugOn(add_entry
22295 + && !au_ftest_wrdir(args->flags, TMPFILE)
22296 + && !au_h_dptr(dentry, bcpup));
22304 +/* ---------------------------------------------------------------------- */
22306 +void au_pin_hdir_unlock(struct au_pin *p)
22309 + au_hn_inode_unlock(p->hdir);
22312 +int au_pin_hdir_lock(struct au_pin *p)
22320 + /* even if an error happens later, keep this lock */
22321 + au_hn_inode_lock_nested(p->hdir, p->lsc_hi);
22324 + if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
22329 + err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
22330 + p->h_parent, p->br);
22336 +int au_pin_hdir_relock(struct au_pin *p)
22339 + struct inode *h_i;
22340 + struct dentry *h_d[] = {
22345 + err = au_pin_hdir_lock(p);
22346 + if (unlikely(err))
22349 + for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
22352 + if (d_is_positive(h_d[i])) {
22353 + h_i = d_inode(h_d[i]);
22354 + err = !h_i->i_nlink;
22362 +static void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
22364 + atomic_long_set(&p->hdir->hi_inode->i_rwsem.owner, (long)task);
22367 +void au_pin_hdir_acquire_nest(struct au_pin *p)
22370 + rwsem_acquire_nest(&p->hdir->hi_inode->i_rwsem.dep_map,
22371 + p->lsc_hi, 0, NULL, _RET_IP_);
22372 + au_pin_hdir_set_owner(p, current);
22376 +void au_pin_hdir_release(struct au_pin *p)
22379 + au_pin_hdir_set_owner(p, p->task);
22380 + rwsem_release(&p->hdir->hi_inode->i_rwsem.dep_map, _RET_IP_);
22384 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
22386 + if (pin && pin->parent)
22387 + return au_h_dptr(pin->parent, pin->bindex);
22391 +void au_unpin(struct au_pin *p)
22394 + au_pin_hdir_unlock(p);
22395 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
22396 + vfsub_mnt_drop_write(p->h_mnt);
22400 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22401 + di_read_unlock(p->parent, AuLock_IR);
22402 + iput(p->hdir->hi_inode);
22404 + p->parent = NULL;
22407 + /* do not clear p->task */
22410 +int au_do_pin(struct au_pin *p)
22413 + struct super_block *sb;
22414 + struct inode *h_dir;
22417 + sb = p->dentry->d_sb;
22418 + p->br = au_sbr(sb, p->bindex);
22419 + if (IS_ROOT(p->dentry)) {
22420 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
22421 + p->h_mnt = au_br_mnt(p->br);
22422 + err = vfsub_mnt_want_write(p->h_mnt);
22423 + if (unlikely(err)) {
22424 + au_fclr_pin(p->flags, MNT_WRITE);
22431 + p->h_dentry = NULL;
22432 + if (p->bindex <= au_dbbot(p->dentry))
22433 + p->h_dentry = au_h_dptr(p->dentry, p->bindex);
22435 + p->parent = dget_parent(p->dentry);
22436 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22437 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
22440 + p->h_parent = au_h_dptr(p->parent, p->bindex);
22441 + p->hdir = au_hi(d_inode(p->parent), p->bindex);
22443 + h_dir = p->hdir->hi_inode;
22447 + * if DI_LOCKED is not set, then p->parent may be different
22448 + * and h_parent can be NULL.
22450 + if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
22452 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22453 + di_read_unlock(p->parent, AuLock_IR);
22455 + p->parent = NULL;
22459 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
22460 + p->h_mnt = au_br_mnt(p->br);
22461 + err = vfsub_mnt_want_write(p->h_mnt);
22462 + if (unlikely(err)) {
22463 + au_fclr_pin(p->flags, MNT_WRITE);
22464 + if (!au_ftest_pin(p->flags, DI_LOCKED))
22465 + di_read_unlock(p->parent, AuLock_IR);
22467 + p->parent = NULL;
22473 + err = au_pin_hdir_lock(p);
22475 + goto out; /* success */
22480 + pr_err("err %d\n", err);
22481 + err = au_busy_or_stale();
22486 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
22487 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
22488 + unsigned int udba, unsigned char flags)
22490 + p->dentry = dentry;
22492 + p->lsc_di = lsc_di;
22493 + p->lsc_hi = lsc_hi;
22494 + p->flags = flags;
22495 + p->bindex = bindex;
22497 + p->parent = NULL;
22501 + p->h_dentry = NULL;
22502 + p->h_parent = NULL;
22504 + p->task = current;
22507 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
22508 + unsigned int udba, unsigned char flags)
22510 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
22512 + return au_do_pin(pin);
22515 +/* ---------------------------------------------------------------------- */
22518 + * ->setattr() and ->getattr() are called in various cases.
22519 + * chmod, stat: dentry is revalidated.
22520 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
22522 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
22524 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
22525 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
22528 + struct dentry *parent;
22531 + if (au_digen_test(dentry, sigen)) {
22532 + parent = dget_parent(dentry);
22533 + di_read_lock_parent(parent, AuLock_IR);
22534 + err = au_refresh_dentry(dentry, parent);
22535 + di_read_unlock(parent, AuLock_IR);
22543 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
22544 + struct au_icpup_args *a)
22548 + aufs_bindex_t btop, ibtop;
22549 + struct dentry *hi_wh, *parent;
22550 + struct inode *inode;
22551 + struct au_wr_dir_args wr_dir_args = {
22552 + .force_btgt = -1,
22556 + if (d_is_dir(dentry))
22557 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
22558 + /* plink or hi_wh() case */
22559 + btop = au_dbtop(dentry);
22560 + inode = d_inode(dentry);
22561 + ibtop = au_ibtop(inode);
22562 + if (btop != ibtop && !au_test_ro(inode->i_sb, ibtop, inode))
22563 + wr_dir_args.force_btgt = ibtop;
22564 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
22565 + if (unlikely(err < 0))
22569 + au_fset_icpup(a->flags, DID_CPUP);
22572 + a->pin_flags = AuPin_MNT_WRITE;
22574 + if (!IS_ROOT(dentry)) {
22575 + au_fset_pin(a->pin_flags, DI_LOCKED);
22576 + parent = dget_parent(dentry);
22577 + di_write_lock_parent(parent);
22580 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
22581 + if (unlikely(err))
22585 + a->h_path.dentry = au_h_dptr(dentry, btop);
22586 + a->h_inode = d_inode(a->h_path.dentry);
22587 + if (ia && (ia->ia_valid & ATTR_SIZE)) {
22588 + inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
22589 + if (ia->ia_size < i_size_read(a->h_inode))
22590 + sz = ia->ia_size;
22591 + inode_unlock_shared(a->h_inode);
22595 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
22596 + hi_wh = au_hi_wh(inode, a->btgt);
22598 + struct au_cp_generic cpg = {
22599 + .dentry = dentry,
22605 + err = au_sio_cpup_wh(&cpg, /*file*/NULL);
22606 + if (unlikely(err))
22608 + hi_wh = au_hi_wh(inode, a->btgt);
22609 + /* todo: revalidate hi_wh? */
22614 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
22615 + di_downgrade_lock(parent, AuLock_IR);
22619 + if (!au_ftest_icpup(a->flags, DID_CPUP))
22620 + goto out; /* success */
22622 + if (!d_unhashed(dentry)) {
22623 + struct au_cp_generic cpg = {
22624 + .dentry = dentry,
22629 + .flags = AuCpup_DTIME | AuCpup_HOPEN
22631 + err = au_sio_cpup_simple(&cpg);
22633 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22634 + } else if (!hi_wh)
22635 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
22637 + a->h_path.dentry = hi_wh; /* do not dget here */
22640 + a->h_inode = d_inode(a->h_path.dentry);
22642 + goto out; /* success */
22643 + au_unpin(&a->pin);
22646 + di_write_unlock(parent);
22651 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22655 +static int aufs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
22656 + struct iattr *ia)
22659 + struct inode *inode, *delegated;
22660 + struct super_block *sb;
22661 + struct file *file;
22662 + struct au_icpup_args *a;
22663 + struct mnt_idmap *h_idmap;
22665 + inode = d_inode(dentry);
22666 + IMustLock(inode);
22668 + err = setattr_prepare(idmap, dentry, ia);
22669 + if (unlikely(err))
22673 + a = kzalloc(sizeof(*a), GFP_NOFS);
22674 + if (unlikely(!a))
22677 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
22678 + ia->ia_valid &= ~ATTR_MODE;
22681 + sb = dentry->d_sb;
22682 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22683 + if (unlikely(err))
22686 + if (ia->ia_valid & ATTR_FILE) {
22687 + /* currently ftruncate(2) only */
22688 + AuDebugOn(!d_is_reg(dentry));
22689 + file = ia->ia_file;
22690 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1,
22692 + if (unlikely(err))
22694 + ia->ia_file = au_hf_top(file);
22695 + a->udba = AuOpt_UDBA_NONE;
22697 + /* fchmod() doesn't pass ia_file */
22698 + a->udba = au_opt_udba(sb);
22699 + di_write_lock_child(dentry);
22700 + /* no d_unlinked(), to set UDBA_NONE for root */
22701 + if (d_unhashed(dentry))
22702 + a->udba = AuOpt_UDBA_NONE;
22703 + if (a->udba != AuOpt_UDBA_NONE) {
22704 + AuDebugOn(IS_ROOT(dentry));
22705 + err = au_reval_for_attr(dentry, au_sigen(sb));
22706 + if (unlikely(err))
22711 + err = au_pin_and_icpup(dentry, ia, a);
22712 + if (unlikely(err < 0))
22714 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
22715 + ia->ia_file = NULL;
22716 + ia->ia_valid &= ~ATTR_FILE;
22719 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
22720 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
22721 + == (ATTR_MODE | ATTR_CTIME)) {
22722 + err = security_path_chmod(&a->h_path, ia->ia_mode);
22723 + if (unlikely(err))
22725 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
22726 + && (ia->ia_valid & ATTR_CTIME)) {
22727 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
22728 + if (unlikely(err))
22732 + if (ia->ia_valid & ATTR_SIZE) {
22735 + if (ia->ia_size < i_size_read(inode))
22737 + truncate_setsize(inode, ia->ia_size);
22740 + if (ia->ia_valid & ATTR_FILE)
22742 + inode_unlock(a->h_inode);
22743 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
22744 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
22746 + delegated = NULL;
22748 + err = vfsub_notify_change(&a->h_path, ia, &delegated);
22750 + err = break_deleg_wait(&delegated);
22758 + * regardless aufs 'acl' option setting.
22759 + * why don't all acl-aware fs call this func from their ->setattr()?
22761 + if (!err && (ia->ia_valid & ATTR_MODE)) {
22762 + h_idmap = mnt_idmap(a->h_path.mnt);
22763 + err = vfsub_acl_chmod(h_idmap, a->h_path.dentry, ia->ia_mode);
22766 + au_cpup_attr_changeable(inode);
22769 + inode_unlock(a->h_inode);
22770 + au_unpin(&a->pin);
22771 + if (unlikely(err))
22772 + au_update_dbtop(dentry);
22774 + di_write_unlock(dentry);
22776 + fi_write_unlock(file);
22777 + ia->ia_file = file;
22778 + ia->ia_valid |= ATTR_FILE;
22781 + si_read_unlock(sb);
22789 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
22790 +static int au_h_path_to_set_attr(struct dentry *dentry,
22791 + struct au_icpup_args *a, struct path *h_path)
22794 + struct super_block *sb;
22796 + sb = dentry->d_sb;
22797 + a->udba = au_opt_udba(sb);
22798 + /* no d_unlinked(), to set UDBA_NONE for root */
22799 + if (d_unhashed(dentry))
22800 + a->udba = AuOpt_UDBA_NONE;
22801 + if (a->udba != AuOpt_UDBA_NONE) {
22802 + AuDebugOn(IS_ROOT(dentry));
22803 + err = au_reval_for_attr(dentry, au_sigen(sb));
22804 + if (unlikely(err))
22807 + err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
22808 + if (unlikely(err < 0))
22811 + h_path->dentry = a->h_path.dentry;
22812 + h_path->mnt = au_sbr_mnt(sb, a->btgt);
22818 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
22819 + struct au_sxattr *arg)
22822 + struct path h_path;
22823 + struct super_block *sb;
22824 + struct au_icpup_args *a;
22825 + struct inode *h_inode;
22826 + struct mnt_idmap *h_idmap;
22828 + IMustLock(inode);
22831 + a = kzalloc(sizeof(*a), GFP_NOFS);
22832 + if (unlikely(!a))
22835 + sb = dentry->d_sb;
22836 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
22837 + if (unlikely(err))
22840 + h_path.dentry = NULL; /* silence gcc */
22841 + di_write_lock_child(dentry);
22842 + err = au_h_path_to_set_attr(dentry, a, &h_path);
22843 + if (unlikely(err))
22845 + h_idmap = mnt_idmap(h_path.mnt);
22847 + inode_unlock(a->h_inode);
22848 + switch (arg->type) {
22849 + case AU_XATTR_SET:
22850 + AuDebugOn(d_is_negative(h_path.dentry));
22851 + err = vfsub_setxattr(h_idmap, h_path.dentry,
22852 + arg->u.set.name, arg->u.set.value,
22853 + arg->u.set.size, arg->u.set.flags);
22856 + err = -EOPNOTSUPP;
22857 + h_inode = d_inode(h_path.dentry);
22858 + if (h_inode->i_op->set_acl) {
22859 + /* this will call posix_acl_update_mode */
22860 + err = h_inode->i_op->set_acl(h_idmap, h_path.dentry,
22861 + arg->u.acl_set.acl,
22862 + arg->u.acl_set.type);
22867 + au_cpup_attr_timesizes(inode);
22869 + au_unpin(&a->pin);
22870 + if (unlikely(err))
22871 + au_update_dbtop(dentry);
22874 + di_write_unlock(dentry);
22875 + si_read_unlock(sb);
22884 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
22885 + unsigned int nlink)
22889 + inode->i_mode = st->mode;
22890 + /* don't i_[ug]id_write() here */
22891 + inode->i_uid = st->uid;
22892 + inode->i_gid = st->gid;
22893 + inode->i_atime = st->atime;
22894 + inode->i_mtime = st->mtime;
22895 + inode->i_ctime = st->ctime;
22897 + au_cpup_attr_nlink(inode, /*force*/0);
22898 + if (S_ISDIR(inode->i_mode)) {
22899 + n = inode->i_nlink;
22902 + smp_mb(); /* for i_nlink */
22903 + /* 0 can happen */
22904 + set_nlink(inode, n);
22907 + spin_lock(&inode->i_lock);
22908 + inode->i_blocks = st->blocks;
22909 + i_size_write(inode, st->size);
22910 + spin_unlock(&inode->i_lock);
22914 + * common routine for aufs_getattr() and au_getxattr().
22915 + * returns zero or negative (an error).
22916 + * @dentry will be read-locked in success.
22918 +int au_h_path_getattr(struct dentry *dentry, struct inode *inode, int force,
22919 + struct path *h_path, int locked)
22922 + unsigned int mnt_flags, sigen;
22923 + unsigned char udba_none;
22924 + aufs_bindex_t bindex;
22925 + struct super_block *sb, *h_sb;
22927 + h_path->mnt = NULL;
22928 + h_path->dentry = NULL;
22931 + sb = dentry->d_sb;
22932 + mnt_flags = au_mntflags(sb);
22933 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
22935 + if (unlikely(locked))
22936 + goto body; /* skip locking dinfo */
22938 + /* support fstat(2) */
22939 + if (!d_unlinked(dentry) && !udba_none) {
22940 + sigen = au_sigen(sb);
22941 + err = au_digen_test(dentry, sigen);
22943 + di_read_lock_child(dentry, AuLock_IR);
22944 + err = au_dbrange_test(dentry);
22945 + if (unlikely(err)) {
22946 + di_read_unlock(dentry, AuLock_IR);
22950 + AuDebugOn(IS_ROOT(dentry));
22951 + di_write_lock_child(dentry);
22952 + err = au_dbrange_test(dentry);
22954 + err = au_reval_for_attr(dentry, sigen);
22956 + di_downgrade_lock(dentry, AuLock_IR);
22958 + di_write_unlock(dentry);
22963 + di_read_lock_child(dentry, AuLock_IR);
22967 + inode = d_inode(dentry);
22968 + if (unlikely(!inode))
22971 + bindex = au_ibtop(inode);
22972 + h_path->mnt = au_sbr_mnt(sb, bindex);
22973 + h_sb = h_path->mnt->mnt_sb;
22975 + && !au_test_fs_bad_iattr(h_sb)
22977 + goto out; /* success */
22979 + if (au_dbtop(dentry) == bindex)
22980 + h_path->dentry = au_h_dptr(dentry, bindex);
22981 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
22982 + h_path->dentry = au_plink_lkup(inode, bindex);
22983 + if (IS_ERR(h_path->dentry))
22984 + /* pretending success */
22985 + h_path->dentry = NULL;
22987 + dput(h_path->dentry);
22994 +static int aufs_getattr(struct mnt_idmap *idmap, const struct path *path,
22995 + struct kstat *st, u32 request, unsigned int query)
22998 + unsigned char positive;
22999 + struct path h_path;
23000 + struct dentry *dentry;
23001 + struct inode *inode;
23002 + struct super_block *sb;
23004 + dentry = path->dentry;
23005 + inode = d_inode(dentry);
23006 + sb = dentry->d_sb;
23007 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
23008 + if (unlikely(err))
23010 + err = au_h_path_getattr(dentry, /*inode*/NULL, /*force*/0, &h_path,
23012 + if (unlikely(err))
23014 + if (unlikely(!h_path.dentry))
23015 + /* illegally overlapped or something */
23016 + goto out_fill; /* pretending success */
23018 + positive = d_is_positive(h_path.dentry);
23020 + /* no vfsub version */
23021 + err = vfs_getattr(&h_path, st, request, query);
23024 + au_refresh_iattr(inode, st,
23025 + d_inode(h_path.dentry)->i_nlink);
23026 + goto out_fill; /* success */
23032 + generic_fillattr(idmap, inode, st);
23034 + di_read_unlock(dentry, AuLock_IR);
23036 + si_read_unlock(sb);
23042 +/* ---------------------------------------------------------------------- */
23044 +static const char *aufs_get_link(struct dentry *dentry, struct inode *inode,
23045 + struct delayed_call *done)
23048 + struct dentry *h_dentry;
23049 + struct inode *h_inode;
23051 + aufs_bindex_t bindex;
23053 + ret = NULL; /* suppress a warning */
23058 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
23059 + if (unlikely(err))
23062 + err = au_d_hashed_positive(dentry);
23063 + if (unlikely(err))
23067 + inode = d_inode(dentry);
23068 + bindex = au_ibtop(inode);
23069 + h_inode = au_h_iptr(inode, bindex);
23070 + if (unlikely(!h_inode->i_op->get_link))
23075 + if (au_dbtop(dentry) <= bindex) {
23076 + h_dentry = au_h_dptr(dentry, bindex);
23081 + h_dentry = d_find_any_alias(h_inode);
23082 + if (IS_ERR(h_dentry)) {
23083 + err = PTR_ERR(h_dentry);
23087 + if (unlikely(!h_dentry))
23091 + AuDbg("%ps\n", h_inode->i_op->get_link);
23092 + AuDbgDentry(h_dentry);
23093 + ret = vfs_get_link(h_dentry, done);
23096 + err = PTR_ERR(ret);
23099 + aufs_read_unlock(dentry, AuLock_IR);
23101 + if (unlikely(err))
23102 + ret = ERR_PTR(err);
23103 + AuTraceErrPtr(ret);
23107 +/* ---------------------------------------------------------------------- */
23109 +static int au_is_special(struct inode *inode)
23111 + return (inode->i_mode & (S_IFBLK | S_IFCHR | S_IFIFO | S_IFSOCK));
23114 +static int aufs_update_time(struct inode *inode, struct timespec64 *ts,
23118 + aufs_bindex_t bindex;
23119 + struct super_block *sb;
23120 + struct inode *h_inode;
23121 + struct vfsmount *h_mnt;
23123 + sb = inode->i_sb;
23124 + WARN_ONCE((flags & S_ATIME) && !IS_NOATIME(inode),
23125 + "unexpected s_flags 0x%lx", sb->s_flags);
23127 + /* mmap_sem might be acquired already, cf. aufs_mmap() */
23129 + si_read_lock(sb, AuLock_FLUSH);
23130 + ii_write_lock_child(inode);
23133 + bindex = au_ibtop(inode);
23134 + h_inode = au_h_iptr(inode, bindex);
23135 + if (!au_test_ro(sb, bindex, inode)) {
23136 + h_mnt = au_sbr_mnt(sb, bindex);
23137 + err = vfsub_mnt_want_write(h_mnt);
23139 + err = vfsub_update_time(h_inode, ts, flags);
23140 + vfsub_mnt_drop_write(h_mnt);
23142 + } else if (au_is_special(h_inode)) {
23144 + * Never copy-up here.
23145 + * These special files may already be opened and used for
23146 + * communicating. If we copied it up, then the communication
23147 + * would be corrupted.
23149 + AuWarn1("timestamps for i%lu are ignored "
23150 + "since it is on readonly branch (hi%lu).\n",
23151 + inode->i_ino, h_inode->i_ino);
23152 + } else if (flags & ~S_ATIME) {
23154 + AuIOErr1("unexpected flags 0x%x\n", flags);
23159 + au_cpup_attr_timesizes(inode);
23160 + ii_write_unlock(inode);
23161 + si_read_unlock(sb);
23164 + if (!err && (flags & S_VERSION))
23165 + inode_inc_iversion(inode);
23170 +/* ---------------------------------------------------------------------- */
23172 +/* no getattr version will be set by module.c:aufs_init() */
23173 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
23175 + [AuIop_SYMLINK] = {
23176 + .permission = aufs_permission,
23177 +#ifdef CONFIG_FS_POSIX_ACL
23178 + .get_inode_acl = aufs_get_inode_acl,
23179 + .get_acl = aufs_get_acl,
23180 + .set_acl = aufs_set_acl, /* unsupport for symlink? */
23183 + .setattr = aufs_setattr,
23184 + .getattr = aufs_getattr,
23186 +#ifdef CONFIG_AUFS_XATTR
23187 + .listxattr = aufs_listxattr,
23190 + .get_link = aufs_get_link
23192 + /* .update_time = aufs_update_time */
23195 + .create = aufs_create,
23196 + .lookup = aufs_lookup,
23197 + .link = aufs_link,
23198 + .unlink = aufs_unlink,
23199 + .symlink = aufs_symlink,
23200 + .mkdir = aufs_mkdir,
23201 + .rmdir = aufs_rmdir,
23202 + .mknod = aufs_mknod,
23203 + .rename = aufs_rename,
23205 + .permission = aufs_permission,
23206 +#ifdef CONFIG_FS_POSIX_ACL
23207 + .get_inode_acl = aufs_get_inode_acl,
23208 + .get_acl = aufs_get_acl,
23209 + .set_acl = aufs_set_acl,
23212 + .setattr = aufs_setattr,
23213 + .getattr = aufs_getattr,
23215 +#ifdef CONFIG_AUFS_XATTR
23216 + .listxattr = aufs_listxattr,
23219 + .update_time = aufs_update_time,
23220 + .atomic_open = aufs_atomic_open,
23221 + .tmpfile = aufs_tmpfile
23223 + [AuIop_OTHER] = {
23224 + .permission = aufs_permission,
23225 +#ifdef CONFIG_FS_POSIX_ACL
23226 + .get_inode_acl = aufs_get_inode_acl,
23227 + .get_acl = aufs_get_acl,
23228 + .set_acl = aufs_set_acl,
23231 + .setattr = aufs_setattr,
23232 + .getattr = aufs_getattr,
23234 +#ifdef CONFIG_AUFS_XATTR
23235 + .listxattr = aufs_listxattr,
23238 + .update_time = aufs_update_time
23241 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
23242 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
23243 +++ linux/fs/aufs/i_op_del.c 2023-04-24 08:13:40.969576115 +0200
23245 +// SPDX-License-Identifier: GPL-2.0
23247 + * Copyright (C) 2005-2022 Junjiro R. Okajima
23249 + * This program is free software; you can redistribute it and/or modify
23250 + * it under the terms of the GNU General Public License as published by
23251 + * the Free Software Foundation; either version 2 of the License, or
23252 + * (at your option) any later version.
23254 + * This program is distributed in the hope that it will be useful,
23255 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23256 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23257 + * GNU General Public License for more details.
23259 + * You should have received a copy of the GNU General Public License
23260 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23264 + * inode operations (del entry)
23267 +#include <linux/iversion.h>
23271 + * decide if a new whiteout for @dentry is necessary or not.
23272 + * when it is necessary, prepare the parent dir for the upper branch whose
23273 + * branch index is @bcpup for creation. the actual creation of the whiteout will
23274 + * be done by caller.
23276 + * 0: wh is unnecessary
23277 + * plus: wh is necessary
23280 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
23282 + int need_wh, err;
23283 + aufs_bindex_t btop;
23284 + struct super_block *sb;
23286 + sb = dentry->d_sb;
23287 + btop = au_dbtop(dentry);
23288 + if (*bcpup < 0) {
23290 + if (au_test_ro(sb, btop, d_inode(dentry))) {
23291 + err = AuWbrCopyup(au_sbi(sb), dentry);
23293 + if (unlikely(err < 0))
23297 + AuDebugOn(btop < *bcpup
23298 + || au_test_ro(sb, *bcpup, d_inode(dentry)));
23299 + AuDbg("bcpup %d, btop %d\n", *bcpup, btop);
23301 + if (*bcpup != btop) {
23302 + err = au_cpup_dirs(dentry, *bcpup);
23303 + if (unlikely(err))
23307 + struct au_dinfo *dinfo, *tmp;
23309 + need_wh = -ENOMEM;
23310 + dinfo = au_di(dentry);
23311 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
23313 + au_di_cp(tmp, dinfo);
23314 + au_di_swap(tmp, dinfo);
23315 + /* returns the number of positive dentries */
23316 + need_wh = au_lkup_dentry(dentry, btop + 1,
23317 + /* AuLkup_IGNORE_PERM */ 0);
23318 + au_di_swap(tmp, dinfo);
23319 + au_rw_write_unlock(&tmp->di_rwsem);
23323 + AuDbg("need_wh %d\n", need_wh);
23331 + * simple tests for the del-entry operations.
23332 + * following the checks in vfs, plus the parent-child relationship.
23334 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
23335 + struct dentry *h_parent, int isdir)
23339 + struct dentry *h_dentry, *h_latest;
23340 + struct inode *h_inode;
23341 + struct path h_ppath;
23342 + struct super_block *sb;
23343 + struct au_branch *br;
23344 + struct mnt_idmap *h_idmap;
23346 + h_dentry = au_h_dptr(dentry, bindex);
23347 + if (d_really_is_positive(dentry)) {
23349 + if (unlikely(d_is_negative(h_dentry)))
23351 + h_inode = d_inode(h_dentry);
23352 + if (unlikely(!h_inode->i_nlink))
23355 + h_mode = h_inode->i_mode;
23358 + if (unlikely(S_ISDIR(h_mode)))
23360 + } else if (unlikely(!S_ISDIR(h_mode))) {
23365 + /* rename(2) case */
23367 + if (unlikely(d_is_positive(h_dentry)))
23372 + /* expected parent dir is locked */
23373 + if (unlikely(h_parent != h_dentry->d_parent))
23378 + * rmdir a dir may break the consistency on some filesystem.
23379 + * let's try heavy test.
23382 + sb = dentry->d_sb;
23383 + br = au_sbr(sb, bindex);
23384 + h_idmap = au_br_idmap(br);
23385 + if (unlikely(!au_opt_test(au_mntflags(sb), DIRPERM1)
23386 + && au_test_h_perm(h_idmap, d_inode(h_parent),
23387 + MAY_EXEC | MAY_WRITE)))
23390 + h_ppath.dentry = h_parent;
23391 + h_ppath.mnt = au_br_mnt(br);
23392 + h_latest = au_sio_lkup_one(h_idmap, &dentry->d_name, &h_ppath);
23394 + if (IS_ERR(h_latest))
23396 + if (h_latest == h_dentry)
23405 + * decide the branch where we operate for @dentry. the branch index will be set
23406 + * @rbcpup. after deciding it, 'pin' it and store the timestamps of the parent
23407 + * dir for reverting.
23408 + * when a new whiteout is necessary, create it.
23410 +static struct dentry*
23411 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
23412 + struct au_dtime *dt, struct au_pin *pin)
23414 + struct dentry *wh_dentry;
23415 + struct super_block *sb;
23416 + struct path h_path;
23417 + int err, need_wh;
23418 + unsigned int udba;
23419 + aufs_bindex_t bcpup;
23421 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
23422 + wh_dentry = ERR_PTR(need_wh);
23423 + if (unlikely(need_wh < 0))
23426 + sb = dentry->d_sb;
23427 + udba = au_opt_udba(sb);
23429 + err = au_pin(pin, dentry, bcpup, udba,
23430 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23431 + wh_dentry = ERR_PTR(err);
23432 + if (unlikely(err))
23435 + h_path.dentry = au_pinned_h_parent(pin);
23436 + if (udba != AuOpt_UDBA_NONE
23437 + && au_dbtop(dentry) == bcpup) {
23438 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
23439 + wh_dentry = ERR_PTR(err);
23440 + if (unlikely(err))
23444 + h_path.mnt = au_sbr_mnt(sb, bcpup);
23445 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
23446 + wh_dentry = NULL;
23448 + goto out; /* success, no need to create whiteout */
23450 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
23451 + if (IS_ERR(wh_dentry))
23454 + /* returns with the parent is locked and wh_dentry is dget-ed */
23455 + goto out; /* success */
23460 + return wh_dentry;
23464 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
23465 + * in order to be revertible and save time for removing many child whiteouts
23467 + * returns 1 when there are too many child whiteout and caller should remove
23468 + * them asynchronously. returns 0 when the number of children is enough small to
23469 + * remove now or the branch fs is a remote fs.
23470 + * otherwise return an error.
23472 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
23473 + struct au_nhash *whlist, struct inode *dir)
23475 + int rmdir_later, err, dirwh;
23476 + struct dentry *h_dentry;
23477 + struct super_block *sb;
23478 + struct inode *inode;
23480 + sb = dentry->d_sb;
23481 + SiMustAnyLock(sb);
23482 + h_dentry = au_h_dptr(dentry, bindex);
23483 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
23484 + if (unlikely(err))
23487 + /* stop monitoring */
23488 + inode = d_inode(dentry);
23489 + au_hn_free(au_hi(inode, bindex));
23491 + if (!au_test_fs_remote(h_dentry->d_sb)) {
23492 + dirwh = au_sbi(sb)->si_dirwh;
23493 + rmdir_later = (dirwh <= 1);
23494 + if (!rmdir_later)
23495 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
23498 + return rmdir_later;
23501 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
23502 + if (unlikely(err)) {
23503 + AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
23504 + h_dentry, bindex, err);
23514 + * final procedure for deleting a entry.
23515 + * maintain dentry and iattr.
23517 +static void epilog(struct inode *dir, struct dentry *dentry,
23518 + aufs_bindex_t bindex)
23520 + struct inode *inode;
23522 + inode = d_inode(dentry);
23524 + inode->i_ctime = dir->i_ctime;
23526 + au_dir_ts(dir, bindex);
23527 + inode_inc_iversion(dir);
23531 + * when an error happened, remove the created whiteout and revert everything.
23533 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
23534 + aufs_bindex_t bwh, struct dentry *wh_dentry,
23535 + struct dentry *dentry, struct au_dtime *dt)
23538 + struct path h_path = {
23539 + .dentry = wh_dentry,
23540 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
23543 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
23545 + au_set_dbwh(dentry, bwh);
23546 + au_dtime_revert(dt);
23550 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
23554 +/* ---------------------------------------------------------------------- */
23556 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
23559 + aufs_bindex_t bwh, bindex, btop;
23560 + struct inode *inode, *h_dir, *delegated;
23561 + struct dentry *parent, *wh_dentry;
23562 + /* to reduce stack size */
23564 + struct au_dtime dt;
23565 + struct au_pin pin;
23566 + struct path h_path;
23572 + a = kmalloc(sizeof(*a), GFP_NOFS);
23573 + if (unlikely(!a))
23576 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
23577 + if (unlikely(err))
23579 + err = au_d_hashed_positive(dentry);
23580 + if (unlikely(err))
23582 + inode = d_inode(dentry);
23583 + IMustLock(inode);
23585 + if (unlikely(d_is_dir(dentry)))
23586 + goto out_unlock; /* possible? */
23588 + btop = au_dbtop(dentry);
23589 + bwh = au_dbwh(dentry);
23591 + parent = dentry->d_parent; /* dir inode is locked */
23592 + di_write_lock_parent(parent);
23593 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
23595 + err = PTR_ERR(wh_dentry);
23596 + if (IS_ERR(wh_dentry))
23599 + a->h_path.mnt = au_sbr_mnt(dentry->d_sb, btop);
23600 + a->h_path.dentry = au_h_dptr(dentry, btop);
23601 + dget(a->h_path.dentry);
23602 + if (bindex == btop) {
23603 + h_dir = au_pinned_h_dir(&a->pin);
23604 + delegated = NULL;
23605 + err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
23606 + if (unlikely(err == -EWOULDBLOCK)) {
23607 + pr_warn("cannot retry for NFSv4 delegation"
23608 + " for an internal unlink\n");
23612 + /* dir inode is locked */
23613 + h_dir = d_inode(wh_dentry->d_parent);
23614 + IMustLock(h_dir);
23619 + vfsub_drop_nlink(inode);
23620 + epilog(dir, dentry, bindex);
23622 + /* update target timestamps */
23623 + if (bindex == btop) {
23624 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
23626 + inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
23628 + /* todo: this timestamp may be reverted later */
23629 + inode->i_ctime = h_dir->i_ctime;
23630 + goto out_unpin; /* success */
23637 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23644 + au_unpin(&a->pin);
23646 + dput(a->h_path.dentry);
23648 + di_write_unlock(parent);
23650 + aufs_read_unlock(dentry, AuLock_DW);
23657 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
23659 + int err, rmdir_later;
23660 + aufs_bindex_t bwh, bindex, btop;
23661 + struct inode *inode;
23662 + struct dentry *parent, *wh_dentry, *h_dentry;
23663 + struct au_whtmp_rmdir *args;
23664 + /* to reduce stack size */
23666 + struct au_dtime dt;
23667 + struct au_pin pin;
23673 + a = kmalloc(sizeof(*a), GFP_NOFS);
23674 + if (unlikely(!a))
23677 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
23678 + if (unlikely(err))
23680 + err = au_alive_dir(dentry);
23681 + if (unlikely(err))
23683 + inode = d_inode(dentry);
23684 + IMustLock(inode);
23686 + if (unlikely(!d_is_dir(dentry)))
23687 + goto out_unlock; /* possible? */
23690 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
23691 + if (unlikely(!args))
23694 + parent = dentry->d_parent; /* dir inode is locked */
23695 + di_write_lock_parent(parent);
23696 + err = au_test_empty(dentry, &args->whlist);
23697 + if (unlikely(err))
23700 + btop = au_dbtop(dentry);
23701 + bwh = au_dbwh(dentry);
23703 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
23705 + err = PTR_ERR(wh_dentry);
23706 + if (IS_ERR(wh_dentry))
23709 + h_dentry = au_h_dptr(dentry, btop);
23712 + if (bindex == btop) {
23713 + err = renwh_and_rmdir(dentry, btop, &args->whlist, dir);
23715 + rmdir_later = err;
23719 + /* stop monitoring */
23720 + au_hn_free(au_hi(inode, btop));
23722 + /* dir inode is locked */
23723 + IMustLock(d_inode(wh_dentry->d_parent));
23728 + vfsub_dead_dir(inode);
23729 + au_set_dbdiropq(dentry, -1);
23730 + epilog(dir, dentry, bindex);
23732 + if (rmdir_later) {
23733 + au_whtmp_kick_rmdir(dir, btop, h_dentry, args);
23737 + goto out_unpin; /* success */
23745 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
23752 + au_unpin(&a->pin);
23756 + di_write_unlock(parent);
23758 + au_whtmp_rmdir_free(args);
23760 + aufs_read_unlock(dentry, AuLock_DW);
23767 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
23768 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
23769 +++ linux/fs/aufs/i_op_ren.c 2023-04-24 08:13:40.969576115 +0200
23771 +// SPDX-License-Identifier: GPL-2.0
23773 + * Copyright (C) 2005-2022 Junjiro R. Okajima
23775 + * This program is free software; you can redistribute it and/or modify
23776 + * it under the terms of the GNU General Public License as published by
23777 + * the Free Software Foundation; either version 2 of the License, or
23778 + * (at your option) any later version.
23780 + * This program is distributed in the hope that it will be useful,
23781 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23782 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23783 + * GNU General Public License for more details.
23785 + * You should have received a copy of the GNU General Public License
23786 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23790 + * inode operation (rename entry)
23791 + * todo: this is crazy monster
23794 +#include <linux/iversion.h>
23797 +enum { AuSRC, AuDST, AuSrcDst };
23798 +enum { AuPARENT, AuCHILD, AuParentChild };
23800 +#define AuRen_ISDIR_SRC 1
23801 +#define AuRen_ISDIR_DST (1 << 1)
23802 +#define AuRen_ISSAMEDIR (1 << 2)
23803 +#define AuRen_WHSRC (1 << 3)
23804 +#define AuRen_WHDST (1 << 4)
23805 +#define AuRen_MNT_WRITE (1 << 5)
23806 +#define AuRen_DT_DSTDIR (1 << 6)
23807 +#define AuRen_DIROPQ_SRC (1 << 7)
23808 +#define AuRen_DIROPQ_DST (1 << 8)
23809 +#define AuRen_DIRREN (1 << 9)
23810 +#define AuRen_DROPPED_SRC (1 << 10)
23811 +#define AuRen_DROPPED_DST (1 << 11)
23812 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
23813 +#define au_fset_ren(flags, name) \
23814 + do { (flags) |= AuRen_##name; } while (0)
23815 +#define au_fclr_ren(flags, name) \
23816 + do { (flags) &= ~AuRen_##name; } while (0)
23818 +#ifndef CONFIG_AUFS_DIRREN
23819 +#undef AuRen_DIRREN
23820 +#define AuRen_DIRREN 0
23823 +struct au_ren_args {
23825 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
23827 + struct inode *dir, *inode;
23828 + struct au_hinode *hdir, *hinode;
23829 + struct au_dtime dt[AuParentChild];
23830 + aufs_bindex_t btop, bdiropq;
23833 +#define src_dentry sd[AuSRC].dentry
23834 +#define src_dir sd[AuSRC].dir
23835 +#define src_inode sd[AuSRC].inode
23836 +#define src_h_dentry sd[AuSRC].h_dentry
23837 +#define src_parent sd[AuSRC].parent
23838 +#define src_h_parent sd[AuSRC].h_parent
23839 +#define src_wh_dentry sd[AuSRC].wh_dentry
23840 +#define src_hdir sd[AuSRC].hdir
23841 +#define src_hinode sd[AuSRC].hinode
23842 +#define src_h_dir sd[AuSRC].hdir->hi_inode
23843 +#define src_dt sd[AuSRC].dt
23844 +#define src_btop sd[AuSRC].btop
23845 +#define src_bdiropq sd[AuSRC].bdiropq
23847 +#define dst_dentry sd[AuDST].dentry
23848 +#define dst_dir sd[AuDST].dir
23849 +#define dst_inode sd[AuDST].inode
23850 +#define dst_h_dentry sd[AuDST].h_dentry
23851 +#define dst_parent sd[AuDST].parent
23852 +#define dst_h_parent sd[AuDST].h_parent
23853 +#define dst_wh_dentry sd[AuDST].wh_dentry
23854 +#define dst_hdir sd[AuDST].hdir
23855 +#define dst_hinode sd[AuDST].hinode
23856 +#define dst_h_dir sd[AuDST].hdir->hi_inode
23857 +#define dst_dt sd[AuDST].dt
23858 +#define dst_btop sd[AuDST].btop
23859 +#define dst_bdiropq sd[AuDST].bdiropq
23861 + struct dentry *h_trap;
23862 + struct au_branch *br;
23863 + struct path h_path;
23864 + struct au_nhash whlist;
23865 + aufs_bindex_t btgt, src_bwh;
23868 + unsigned short auren_flags;
23869 + unsigned char flags; /* syscall parameter */
23870 + unsigned char exchange;
23873 + struct au_whtmp_rmdir *thargs;
23874 + struct dentry *h_dst;
23875 + struct au_hinode *h_root;
23878 +/* ---------------------------------------------------------------------- */
23881 + * functions for reverting.
23882 + * when an error happened in a single rename systemcall, we should revert
23883 + * everything as if nothing happened.
23884 + * we don't need to revert the copied-up/down the parent dir since they are
23888 +#define RevertFailure(fmt, ...) do { \
23889 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
23890 + ##__VA_ARGS__, err, rerr); \
23894 +static void au_ren_do_rev_diropq(int err, struct au_ren_args *a, int idx)
23897 + struct dentry *d;
23898 +#define src_or_dst(member) a->sd[idx].member
23900 + d = src_or_dst(dentry); /* {src,dst}_dentry */
23901 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
23902 + rerr = au_diropq_remove(d, a->btgt);
23903 + au_hn_inode_unlock(src_or_dst(hinode));
23904 + au_set_dbdiropq(d, src_or_dst(bdiropq));
23906 + RevertFailure("remove diropq %pd", d);
23908 +#undef src_or_dst_
23911 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
23913 + if (au_ftest_ren(a->auren_flags, DIROPQ_SRC))
23914 + au_ren_do_rev_diropq(err, a, AuSRC);
23915 + if (au_ftest_ren(a->auren_flags, DIROPQ_DST))
23916 + au_ren_do_rev_diropq(err, a, AuDST);
23919 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
23922 + struct inode *delegated;
23923 + struct path h_ppath = {
23924 + .dentry = a->src_h_parent,
23925 + .mnt = a->h_path.mnt
23928 + a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name, &h_ppath);
23929 + rerr = PTR_ERR(a->h_path.dentry);
23930 + if (IS_ERR(a->h_path.dentry)) {
23931 + RevertFailure("lkup one %pd", a->src_dentry);
23935 + delegated = NULL;
23936 + rerr = vfsub_rename(a->dst_h_dir,
23937 + au_h_dptr(a->src_dentry, a->btgt),
23938 + a->src_h_dir, &a->h_path, &delegated, a->flags);
23939 + if (unlikely(rerr == -EWOULDBLOCK)) {
23940 + pr_warn("cannot retry for NFSv4 delegation"
23941 + " for an internal rename\n");
23944 + d_drop(a->h_path.dentry);
23945 + dput(a->h_path.dentry);
23946 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
23948 + RevertFailure("rename %pd", a->src_dentry);
23951 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
23954 + struct inode *delegated;
23955 + struct path h_ppath = {
23956 + .dentry = a->dst_h_parent,
23957 + .mnt = a->h_path.mnt
23960 + a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name, &h_ppath);
23961 + rerr = PTR_ERR(a->h_path.dentry);
23962 + if (IS_ERR(a->h_path.dentry)) {
23963 + RevertFailure("lkup one %pd", a->dst_dentry);
23966 + if (d_is_positive(a->h_path.dentry)) {
23967 + d_drop(a->h_path.dentry);
23968 + dput(a->h_path.dentry);
23972 + delegated = NULL;
23973 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
23974 + &delegated, a->flags);
23975 + if (unlikely(rerr == -EWOULDBLOCK)) {
23976 + pr_warn("cannot retry for NFSv4 delegation"
23977 + " for an internal rename\n");
23980 + d_drop(a->h_path.dentry);
23981 + dput(a->h_path.dentry);
23983 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
23985 + RevertFailure("rename %pd", a->h_dst);
23988 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
23992 + a->h_path.dentry = a->src_wh_dentry;
23993 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
23994 + au_set_dbwh(a->src_dentry, a->src_bwh);
23996 + RevertFailure("unlink %pd", a->src_wh_dentry);
23998 +#undef RevertFailure
24000 +/* ---------------------------------------------------------------------- */
24003 + * when we have to copyup the renaming entry, do it with the rename-target name
24004 + * in order to minimize the cost (the later actual rename is unnecessary).
24005 + * otherwise rename it on the target branch.
24007 +static int au_ren_or_cpup(struct au_ren_args *a)
24010 + struct dentry *d;
24011 + struct inode *delegated;
24013 + d = a->src_dentry;
24014 + if (au_dbtop(d) == a->btgt) {
24015 + a->h_path.dentry = a->dst_h_dentry;
24016 + AuDebugOn(au_dbtop(d) != a->btgt);
24017 + delegated = NULL;
24018 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
24019 + a->dst_h_dir, &a->h_path, &delegated,
24021 + if (unlikely(err == -EWOULDBLOCK)) {
24022 + pr_warn("cannot retry for NFSv4 delegation"
24023 + " for an internal rename\n");
24029 + if (!err && a->h_dst)
24030 + /* it will be set to dinfo later */
24036 +/* cf. aufs_rmdir() */
24037 +static int au_ren_del_whtmp(struct au_ren_args *a)
24040 + struct inode *dir;
24042 + dir = a->dst_dir;
24043 + SiMustAnyLock(dir->i_sb);
24044 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
24045 + au_sbi(dir->i_sb)->si_dirwh)
24046 + || au_test_fs_remote(a->h_dst->d_sb)) {
24047 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
24048 + if (unlikely(err))
24049 + pr_warn("failed removing whtmp dir %pd (%d), "
24050 + "ignored.\n", a->h_dst, err);
24052 + au_nhash_wh_free(&a->thargs->whlist);
24053 + a->thargs->whlist = a->whlist;
24054 + a->whlist.nh_num = 0;
24055 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
24057 + a->thargs = NULL;
24063 +/* make it 'opaque' dir. */
24064 +static int au_ren_do_diropq(struct au_ren_args *a, int idx)
24067 + struct dentry *d, *diropq;
24068 +#define src_or_dst(member) a->sd[idx].member
24071 + d = src_or_dst(dentry); /* {src,dst}_dentry */
24072 + src_or_dst(bdiropq) = au_dbdiropq(d);
24073 + src_or_dst(hinode) = au_hi(src_or_dst(inode), a->btgt);
24074 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
24075 + diropq = au_diropq_create(d, a->btgt);
24076 + au_hn_inode_unlock(src_or_dst(hinode));
24077 + if (IS_ERR(diropq))
24078 + err = PTR_ERR(diropq);
24082 +#undef src_or_dst_
24086 +static int au_ren_diropq(struct au_ren_args *a)
24089 + unsigned char always;
24090 + struct dentry *d;
24093 + d = a->dst_dentry; /* already renamed on the branch */
24094 + always = !!au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ);
24095 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24096 + && !au_ftest_ren(a->auren_flags, DIRREN)
24097 + && a->btgt != au_dbdiropq(a->src_dentry)
24098 + && (a->dst_wh_dentry
24099 + || a->btgt <= au_dbdiropq(d)
24100 + /* hide the lower to keep xino */
24101 + /* the lowers may not be a dir, but we hide them anyway */
24102 + || a->btgt < au_dbbot(d)
24105 + err = au_ren_do_diropq(a, AuSRC);
24106 + if (unlikely(err))
24108 + au_fset_ren(a->auren_flags, DIROPQ_SRC);
24110 + if (!a->exchange)
24111 + goto out; /* success */
24113 + d = a->src_dentry; /* already renamed on the branch */
24114 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24115 + && a->btgt != au_dbdiropq(a->dst_dentry)
24116 + && (a->btgt < au_dbdiropq(d)
24117 + || a->btgt < au_dbbot(d)
24119 + AuDbgDentry(a->src_dentry);
24120 + AuDbgDentry(a->dst_dentry);
24121 + err = au_ren_do_diropq(a, AuDST);
24122 + if (unlikely(err))
24123 + goto out_rev_src;
24124 + au_fset_ren(a->auren_flags, DIROPQ_DST);
24126 + goto out; /* success */
24129 + AuDbg("err %d, reverting src\n", err);
24130 + au_ren_rev_diropq(err, a);
24135 +static int do_rename(struct au_ren_args *a)
24138 + struct dentry *d, *h_d;
24140 + if (!a->exchange) {
24141 + /* prepare workqueue args for asynchronous rmdir */
24142 + h_d = a->dst_h_dentry;
24143 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
24144 + /* && !au_ftest_ren(a->auren_flags, DIRREN) */
24145 + && d_is_positive(h_d)) {
24147 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb,
24149 + if (unlikely(!a->thargs))
24151 + a->h_dst = dget(h_d);
24154 + /* create whiteout for src_dentry */
24155 + if (au_ftest_ren(a->auren_flags, WHSRC)) {
24156 + a->src_bwh = au_dbwh(a->src_dentry);
24157 + AuDebugOn(a->src_bwh >= 0);
24158 + a->src_wh_dentry = au_wh_create(a->src_dentry, a->btgt,
24159 + a->src_h_parent);
24160 + err = PTR_ERR(a->src_wh_dentry);
24161 + if (IS_ERR(a->src_wh_dentry))
24165 + /* lookup whiteout for dentry */
24166 + if (au_ftest_ren(a->auren_flags, WHDST)) {
24167 + h_d = au_wh_lkup(a->dst_h_parent,
24168 + &a->dst_dentry->d_name, a->br);
24169 + err = PTR_ERR(h_d);
24172 + if (d_is_negative(h_d))
24175 + a->dst_wh_dentry = h_d;
24178 + /* rename dentry to tmpwh */
24180 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
24181 + if (unlikely(err))
24184 + d = a->dst_dentry;
24185 + au_set_h_dptr(d, a->btgt, NULL);
24186 + err = au_lkup_neg(d, a->btgt, /*wh*/0);
24187 + if (unlikely(err))
24189 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
24193 + BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_btop != a->btgt);
24194 +#if 0 /* debugging */
24195 + BUG_ON(!au_ftest_ren(a->auren_flags, DIRREN)
24196 + && d_is_positive(a->dst_h_dentry)
24197 + && a->src_btop != a->btgt);
24200 + /* rename by vfs_rename or cpup */
24201 + err = au_ren_or_cpup(a);
24202 + if (unlikely(err))
24203 + /* leave the copied-up one */
24206 + /* make dir opaque */
24207 + err = au_ren_diropq(a);
24208 + if (unlikely(err))
24211 + /* update target timestamps */
24212 + if (a->exchange) {
24213 + AuDebugOn(au_dbtop(a->dst_dentry) != a->btgt);
24214 + a->h_path.dentry = au_h_dptr(a->dst_dentry, a->btgt);
24215 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24216 + a->dst_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24218 + AuDebugOn(au_dbtop(a->src_dentry) != a->btgt);
24219 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
24220 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
24221 + a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
24223 + if (!a->exchange) {
24224 + /* remove whiteout for dentry */
24225 + if (a->dst_wh_dentry) {
24226 + a->h_path.dentry = a->dst_wh_dentry;
24227 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
24229 + if (unlikely(err))
24233 + /* remove whtmp */
24235 + au_ren_del_whtmp(a); /* ignore this error */
24237 + au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
24240 + goto out_success;
24243 + au_ren_rev_diropq(err, a);
24245 + au_ren_rev_rename(err, a);
24249 + au_ren_rev_whtmp(err, a);
24251 + dput(a->dst_wh_dentry);
24252 + a->dst_wh_dentry = NULL;
24254 + if (a->src_wh_dentry)
24255 + au_ren_rev_whsrc(err, a);
24257 + dput(a->src_wh_dentry);
24258 + dput(a->dst_wh_dentry);
24262 + au_whtmp_rmdir_free(a->thargs);
24263 + a->thargs = NULL;
24269 +/* ---------------------------------------------------------------------- */
24272 + * test if @dentry dir can be rename destination or not.
24273 + * success means, it is a logically empty dir.
24275 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
24277 + return au_test_empty(dentry, whlist);
24281 + * test if @a->src_dentry dir can be rename source or not.
24282 + * if it can, return 0.
24284 + * - it is a logically empty dir.
24285 + * - or, it exists on writable branch and has no children including whiteouts
24286 + * on the lower branch unless DIRREN is on.
24288 +static int may_rename_srcdir(struct au_ren_args *a)
24291 + unsigned int rdhash;
24292 + aufs_bindex_t btop, btgt;
24293 + struct dentry *dentry;
24294 + struct super_block *sb;
24295 + struct au_sbinfo *sbinfo;
24297 + dentry = a->src_dentry;
24298 + sb = dentry->d_sb;
24299 + sbinfo = au_sbi(sb);
24300 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
24301 + au_fset_ren(a->auren_flags, DIRREN);
24304 + btop = au_dbtop(dentry);
24305 + if (btop != btgt) {
24306 + struct au_nhash whlist;
24308 + SiMustAnyLock(sb);
24309 + rdhash = sbinfo->si_rdhash;
24311 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
24313 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
24314 + if (unlikely(err))
24316 + err = au_test_empty(dentry, &whlist);
24317 + au_nhash_wh_free(&whlist);
24321 + if (btop == au_dbtaildir(dentry))
24322 + return 0; /* success */
24324 + err = au_test_empty_lower(dentry);
24327 + if (err == -ENOTEMPTY) {
24328 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24331 + AuWarn1("renaming dir who has child(ren) on multiple "
24332 + "branches, is not supported\n");
24339 +/* side effect: sets whlist and h_dentry */
24340 +static int au_ren_may_dir(struct au_ren_args *a)
24343 + unsigned int rdhash;
24344 + struct dentry *d;
24346 + d = a->dst_dentry;
24347 + SiMustAnyLock(d->d_sb);
24350 + if (au_ftest_ren(a->auren_flags, ISDIR_DST) && a->dst_inode) {
24351 + rdhash = au_sbi(d->d_sb)->si_rdhash;
24353 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
24354 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
24355 + if (unlikely(err))
24358 + if (!a->exchange) {
24359 + au_set_dbtop(d, a->dst_btop);
24360 + err = may_rename_dstdir(d, &a->whlist);
24361 + au_set_dbtop(d, a->btgt);
24363 + err = may_rename_srcdir(a);
24365 + a->dst_h_dentry = au_h_dptr(d, au_dbtop(d));
24366 + if (unlikely(err))
24369 + d = a->src_dentry;
24370 + a->src_h_dentry = au_h_dptr(d, au_dbtop(d));
24371 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24372 + err = may_rename_srcdir(a);
24373 + if (unlikely(err)) {
24374 + au_nhash_wh_free(&a->whlist);
24375 + a->whlist.nh_num = 0;
24382 +/* ---------------------------------------------------------------------- */
24385 + * simple tests for rename.
24386 + * following the checks in vfs, plus the parent-child relationship.
24388 +static int au_may_ren(struct au_ren_args *a)
24391 + struct inode *h_inode;
24393 + if (a->src_btop == a->btgt) {
24394 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
24395 + au_ftest_ren(a->auren_flags, ISDIR_SRC));
24396 + if (unlikely(err))
24399 + if (unlikely(a->src_h_dentry == a->h_trap))
24404 + if (a->dst_btop != a->btgt)
24407 + err = -ENOTEMPTY;
24408 + if (unlikely(a->dst_h_dentry == a->h_trap))
24412 + isdir = !!au_ftest_ren(a->auren_flags, ISDIR_DST);
24413 + if (d_really_is_negative(a->dst_dentry)) {
24414 + if (d_is_negative(a->dst_h_dentry))
24415 + err = au_may_add(a->dst_dentry, a->btgt,
24416 + a->dst_h_parent, isdir);
24418 + if (unlikely(d_is_negative(a->dst_h_dentry)))
24420 + h_inode = d_inode(a->dst_h_dentry);
24421 + if (h_inode->i_nlink)
24422 + err = au_may_del(a->dst_dentry, a->btgt,
24423 + a->dst_h_parent, isdir);
24427 + if (unlikely(err == -ENOENT || err == -EEXIST))
24433 +/* ---------------------------------------------------------------------- */
24438 + * - src_dir and dir by lock_rename()
24439 + * - inode if exists
24442 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
24444 + * + di_write_lock2_child()
24445 + * + di_write_lock_child()
24446 + * + ii_write_lock_child()
24447 + * + di_write_lock_child2()
24448 + * + ii_write_lock_child2()
24449 + * + src_parent and parent
24450 + * + di_write_lock_parent()
24451 + * + ii_write_lock_parent()
24452 + * + di_write_lock_parent2()
24453 + * + ii_write_lock_parent2()
24454 + * + lower src_dir and dir by vfsub_lock_rename()
24455 + * + verify the every relationships between child and parent. if any
24456 + * of them failed, unlock all and return -EBUSY.
24458 +static void au_ren_unlock(struct au_ren_args *a)
24460 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
24461 + a->dst_h_parent, a->dst_hdir);
24462 + if (au_ftest_ren(a->auren_flags, DIRREN)
24464 + au_hn_inode_unlock(a->h_root);
24465 + if (au_ftest_ren(a->auren_flags, MNT_WRITE))
24466 + vfsub_mnt_drop_write(au_br_mnt(a->br));
24469 +static int au_ren_lock(struct au_ren_args *a)
24472 + unsigned int udba;
24475 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
24476 + a->src_hdir = au_hi(a->src_dir, a->btgt);
24477 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
24478 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
24480 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
24481 + if (unlikely(err))
24483 + au_fset_ren(a->auren_flags, MNT_WRITE);
24484 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24485 + struct dentry *root;
24486 + struct inode *dir;
24489 + * sbinfo is already locked, so this ii_read_lock is
24490 + * unnecessary. but our debugging feature checks it.
24492 + root = a->src_inode->i_sb->s_root;
24493 + if (root != a->src_parent && root != a->dst_parent) {
24494 + dir = d_inode(root);
24495 + ii_read_lock_parent3(dir);
24496 + a->h_root = au_hi(dir, a->btgt);
24497 + ii_read_unlock(dir);
24498 + au_hn_inode_lock_nested(a->h_root, AuLsc_I_PARENT3);
24501 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
24502 + a->dst_h_parent, a->dst_hdir);
24503 + udba = au_opt_udba(a->src_dentry->d_sb);
24504 + if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
24505 + || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
24506 + err = au_busy_or_stale();
24507 + if (!err && au_dbtop(a->src_dentry) == a->btgt)
24508 + err = au_h_verify(a->src_h_dentry, udba,
24509 + d_inode(a->src_h_parent), a->src_h_parent,
24511 + if (!err && au_dbtop(a->dst_dentry) == a->btgt)
24512 + err = au_h_verify(a->dst_h_dentry, udba,
24513 + d_inode(a->dst_h_parent), a->dst_h_parent,
24516 + goto out; /* success */
24518 + err = au_busy_or_stale();
24519 + au_ren_unlock(a);
24525 +/* ---------------------------------------------------------------------- */
24527 +static void au_ren_refresh_dir(struct au_ren_args *a)
24529 + struct inode *dir;
24531 + dir = a->dst_dir;
24532 + inode_inc_iversion(dir);
24533 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
24534 + /* is this updating defined in POSIX? */
24535 + au_cpup_attr_timesizes(a->src_inode);
24536 + au_cpup_attr_nlink(dir, /*force*/1);
24538 + au_dir_ts(dir, a->btgt);
24540 + if (a->exchange) {
24541 + dir = a->src_dir;
24542 + inode_inc_iversion(dir);
24543 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24544 + /* is this updating defined in POSIX? */
24545 + au_cpup_attr_timesizes(a->dst_inode);
24546 + au_cpup_attr_nlink(dir, /*force*/1);
24548 + au_dir_ts(dir, a->btgt);
24551 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
24554 + dir = a->src_dir;
24555 + inode_inc_iversion(dir);
24556 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC))
24557 + au_cpup_attr_nlink(dir, /*force*/1);
24558 + au_dir_ts(dir, a->btgt);
24561 +static void au_ren_refresh(struct au_ren_args *a)
24563 + aufs_bindex_t bbot, bindex;
24564 + struct dentry *d, *h_d;
24565 + struct inode *i, *h_i;
24566 + struct super_block *sb;
24568 + d = a->dst_dentry;
24571 + /* already dget-ed by au_ren_or_cpup() */
24572 + au_set_h_dptr(d, a->btgt, a->h_dst);
24574 + i = a->dst_inode;
24576 + if (!a->exchange) {
24577 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST))
24578 + vfsub_drop_nlink(i);
24580 + vfsub_dead_dir(i);
24581 + au_cpup_attr_timesizes(i);
24583 + au_update_dbrange(d, /*do_put_zero*/1);
24585 + au_cpup_attr_nlink(i, /*force*/1);
24588 + for (bindex = au_dbtop(d); bindex < bbot; bindex++)
24589 + au_set_h_dptr(d, bindex, NULL);
24590 + bbot = au_dbbot(d);
24591 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++)
24592 + au_set_h_dptr(d, bindex, NULL);
24593 + au_update_dbrange(d, /*do_put_zero*/0);
24597 + || au_ftest_ren(a->auren_flags, DIRREN)) {
24598 + d_drop(a->src_dentry);
24599 + if (au_ftest_ren(a->auren_flags, DIRREN))
24600 + au_set_dbwh(a->src_dentry, -1);
24604 + d = a->src_dentry;
24605 + au_set_dbwh(d, -1);
24606 + bbot = au_dbbot(d);
24607 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24608 + h_d = au_h_dptr(d, bindex);
24610 + au_set_h_dptr(d, bindex, NULL);
24612 + au_set_dbbot(d, a->btgt);
24615 + i = a->src_inode;
24616 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
24617 + return; /* success */
24619 + bbot = au_ibbot(i);
24620 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
24621 + h_i = au_h_iptr(i, bindex);
24623 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
24624 + /* ignore this error */
24625 + au_set_h_iptr(i, bindex, NULL, 0);
24628 + au_set_ibbot(i, a->btgt);
24631 +/* ---------------------------------------------------------------------- */
24633 +/* mainly for link(2) and rename(2) */
24634 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
24636 + aufs_bindex_t bdiropq, bwh;
24637 + struct dentry *parent;
24638 + struct au_branch *br;
24640 + parent = dentry->d_parent;
24641 + IMustLock(d_inode(parent)); /* dir is locked */
24643 + bdiropq = au_dbdiropq(parent);
24644 + bwh = au_dbwh(dentry);
24645 + br = au_sbr(dentry->d_sb, btgt);
24646 + if (au_br_rdonly(br)
24647 + || (0 <= bdiropq && bdiropq < btgt)
24648 + || (0 <= bwh && bwh < btgt))
24651 + AuDbg("btgt %d\n", btgt);
24655 +/* sets src_btop, dst_btop and btgt */
24656 +static int au_ren_wbr(struct au_ren_args *a)
24659 + struct au_wr_dir_args wr_dir_args = {
24660 + /* .force_btgt = -1, */
24661 + .flags = AuWrDir_ADD_ENTRY
24664 + a->src_btop = au_dbtop(a->src_dentry);
24665 + a->dst_btop = au_dbtop(a->dst_dentry);
24666 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
24667 + || au_ftest_ren(a->auren_flags, ISDIR_DST))
24668 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
24669 + wr_dir_args.force_btgt = a->src_btop;
24670 + if (a->dst_inode && a->dst_btop < a->src_btop)
24671 + wr_dir_args.force_btgt = a->dst_btop;
24672 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
24673 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
24676 + au_update_dbtop(a->dst_dentry);
24681 +static void au_ren_dt(struct au_ren_args *a)
24683 + a->h_path.dentry = a->src_h_parent;
24684 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
24685 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR)) {
24686 + a->h_path.dentry = a->dst_h_parent;
24687 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
24690 + au_fclr_ren(a->auren_flags, DT_DSTDIR);
24691 + if (!au_ftest_ren(a->auren_flags, ISDIR_SRC)
24695 + a->h_path.dentry = a->src_h_dentry;
24696 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
24697 + if (d_is_positive(a->dst_h_dentry)) {
24698 + au_fset_ren(a->auren_flags, DT_DSTDIR);
24699 + a->h_path.dentry = a->dst_h_dentry;
24700 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
24704 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
24706 + struct dentry *h_d;
24707 + struct inode *h_inode;
24709 + au_dtime_revert(a->src_dt + AuPARENT);
24710 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR))
24711 + au_dtime_revert(a->dst_dt + AuPARENT);
24713 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC) && err != -EIO) {
24714 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
24715 + h_inode = d_inode(h_d);
24716 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
24717 + au_dtime_revert(a->src_dt + AuCHILD);
24718 + inode_unlock(h_inode);
24720 + if (au_ftest_ren(a->auren_flags, DT_DSTDIR)) {
24721 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
24722 + h_inode = d_inode(h_d);
24723 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
24724 + au_dtime_revert(a->dst_dt + AuCHILD);
24725 + inode_unlock(h_inode);
24730 +/* ---------------------------------------------------------------------- */
24732 +int aufs_rename(struct mnt_idmap *idmap,
24733 + struct inode *_src_dir, struct dentry *_src_dentry,
24734 + struct inode *_dst_dir, struct dentry *_dst_dentry,
24735 + unsigned int _flags)
24737 + int err, lock_flags;
24739 + /* reduce stack space */
24740 + struct au_ren_args *a;
24741 + struct au_pin pin;
24743 + AuDbg("%pd, %pd, 0x%x\n", _src_dentry, _dst_dentry, _flags);
24744 + IMustLock(_src_dir);
24745 + IMustLock(_dst_dir);
24748 + if (unlikely(_flags & RENAME_WHITEOUT))
24752 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
24753 + a = kzalloc(sizeof(*a), GFP_NOFS);
24754 + if (unlikely(!a))
24757 + a->flags = _flags;
24758 + BUILD_BUG_ON(sizeof(a->exchange) == sizeof(u8)
24759 + && RENAME_EXCHANGE > U8_MAX);
24760 + a->exchange = _flags & RENAME_EXCHANGE;
24761 + a->src_dir = _src_dir;
24762 + a->src_dentry = _src_dentry;
24763 + a->src_inode = NULL;
24764 + if (d_really_is_positive(a->src_dentry))
24765 + a->src_inode = d_inode(a->src_dentry);
24766 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
24767 + a->dst_dir = _dst_dir;
24768 + a->dst_dentry = _dst_dentry;
24769 + a->dst_inode = NULL;
24770 + if (d_really_is_positive(a->dst_dentry))
24771 + a->dst_inode = d_inode(a->dst_dentry);
24772 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
24773 + if (a->dst_inode) {
24775 + * if EXCHANGE && src is non-dir && dst is dir,
24776 + * dst is not locked.
24778 + /* IMustLock(a->dst_inode); */
24779 + au_igrab(a->dst_inode);
24783 + lock_flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
24784 + if (d_is_dir(a->src_dentry)) {
24785 + au_fset_ren(a->auren_flags, ISDIR_SRC);
24786 + if (unlikely(!a->exchange
24787 + && d_really_is_positive(a->dst_dentry)
24788 + && !d_is_dir(a->dst_dentry)))
24790 + lock_flags |= AuLock_DIRS;
24792 + if (a->dst_inode && d_is_dir(a->dst_dentry)) {
24793 + au_fset_ren(a->auren_flags, ISDIR_DST);
24794 + if (unlikely(!a->exchange
24795 + && d_really_is_positive(a->src_dentry)
24796 + && !d_is_dir(a->src_dentry)))
24798 + lock_flags |= AuLock_DIRS;
24800 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
24802 + if (unlikely(err))
24805 + err = au_d_hashed_positive(a->src_dentry);
24806 + if (unlikely(err))
24809 + if (a->dst_inode) {
24811 + * If it is a dir, VFS unhash it before this
24812 + * function. It means we cannot rely upon d_unhashed().
24814 + if (unlikely(!a->dst_inode->i_nlink))
24816 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST)) {
24817 + err = au_d_hashed_positive(a->dst_dentry);
24818 + if (unlikely(err && !a->exchange))
24820 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
24822 + } else if (unlikely(d_unhashed(a->dst_dentry)))
24826 + * is it possible?
24827 + * yes, it happened (in linux-3.3-rcN) but I don't know why.
24828 + * there may exist a problem somewhere else.
24831 + if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
24834 + au_fset_ren(a->auren_flags, ISSAMEDIR); /* temporary */
24835 + di_write_lock_parent(a->dst_parent);
24837 + /* which branch we process */
24838 + err = au_ren_wbr(a);
24839 + if (unlikely(err < 0))
24841 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
24842 + a->h_path.mnt = au_br_mnt(a->br);
24844 + /* are they available to be renamed */
24845 + err = au_ren_may_dir(a);
24846 + if (unlikely(err))
24847 + goto out_children;
24849 + /* prepare the writable parent dir on the same branch */
24850 + if (a->dst_btop == a->btgt) {
24851 + au_fset_ren(a->auren_flags, WHDST);
24853 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
24854 + if (unlikely(err))
24855 + goto out_children;
24859 + if (!a->exchange) {
24860 + if (a->src_dir != a->dst_dir) {
24862 + * this temporary unlock is safe,
24863 + * because both dir->i_mutex are locked.
24865 + di_write_unlock(a->dst_parent);
24866 + di_write_lock_parent(a->src_parent);
24867 + err = au_wr_dir_need_wh(a->src_dentry,
24868 + au_ftest_ren(a->auren_flags,
24871 + di_write_unlock(a->src_parent);
24872 + di_write_lock2_parent(a->src_parent, a->dst_parent,
24874 + au_fclr_ren(a->auren_flags, ISSAMEDIR);
24876 + err = au_wr_dir_need_wh(a->src_dentry,
24877 + au_ftest_ren(a->auren_flags,
24881 + if (unlikely(err < 0))
24882 + goto out_children;
24884 + au_fset_ren(a->auren_flags, WHSRC);
24887 + if (a->src_btop != a->btgt) {
24888 + err = au_pin(&pin, a->src_dentry, a->btgt,
24889 + au_opt_udba(a->src_dentry->d_sb),
24890 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24892 + struct au_cp_generic cpg = {
24893 + .dentry = a->src_dentry,
24895 + .bsrc = a->src_btop,
24898 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24900 + AuDebugOn(au_dbtop(a->src_dentry) != a->src_btop);
24901 + err = au_sio_cpup_simple(&cpg);
24904 + if (unlikely(err))
24905 + goto out_children;
24906 + a->src_btop = a->btgt;
24907 + a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
24908 + if (!a->exchange)
24909 + au_fset_ren(a->auren_flags, WHSRC);
24913 + if (a->exchange && a->dst_inode
24914 + && a->dst_btop != a->btgt) {
24915 + err = au_pin(&pin, a->dst_dentry, a->btgt,
24916 + au_opt_udba(a->dst_dentry->d_sb),
24917 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
24919 + struct au_cp_generic cpg = {
24920 + .dentry = a->dst_dentry,
24922 + .bsrc = a->dst_btop,
24925 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24927 + err = au_sio_cpup_simple(&cpg);
24930 + if (unlikely(err))
24931 + goto out_children;
24932 + a->dst_btop = a->btgt;
24933 + a->dst_h_dentry = au_h_dptr(a->dst_dentry, a->btgt);
24936 + /* lock them all */
24937 + err = au_ren_lock(a);
24938 + if (unlikely(err))
24939 + /* leave the copied-up one */
24940 + goto out_children;
24942 + if (!a->exchange) {
24943 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
24944 + err = au_may_ren(a);
24945 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
24946 + err = -ENAMETOOLONG;
24947 + if (unlikely(err))
24951 + /* store timestamps to be revertible */
24954 + /* store dirren info */
24955 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
24956 + err = au_dr_rename(a->src_dentry, a->btgt,
24957 + &a->dst_dentry->d_name, &rev);
24959 + if (unlikely(err))
24964 + err = do_rename(a);
24965 + if (unlikely(err))
24968 + if (au_ftest_ren(a->auren_flags, DIRREN))
24969 + au_dr_rename_fin(a->src_dentry, a->btgt, rev);
24971 + /* update dir attributes */
24972 + au_ren_refresh_dir(a);
24974 + /* dput/iput all lower dentries */
24975 + au_ren_refresh(a);
24977 + goto out_hdir; /* success */
24980 + if (au_ftest_ren(a->auren_flags, DIRREN))
24981 + au_dr_rename_rev(a->src_dentry, a->btgt, rev);
24983 + au_ren_rev_dt(err, a);
24985 + au_ren_unlock(a);
24987 + au_nhash_wh_free(&a->whlist);
24988 + if (err && a->dst_inode && a->dst_btop != a->btgt) {
24989 + AuDbg("btop %d, btgt %d\n", a->dst_btop, a->btgt);
24990 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
24991 + au_set_dbtop(a->dst_dentry, a->dst_btop);
24995 + if (d_unhashed(a->src_dentry))
24996 + au_fset_ren(a->auren_flags, DROPPED_SRC);
24997 + if (d_unhashed(a->dst_dentry))
24998 + au_fset_ren(a->auren_flags, DROPPED_DST);
24999 + if (!a->exchange)
25000 + d_move(a->src_dentry, a->dst_dentry);
25002 + d_exchange(a->src_dentry, a->dst_dentry);
25003 + if (au_ftest_ren(a->auren_flags, DROPPED_DST))
25004 + d_drop(a->dst_dentry);
25006 + if (au_ftest_ren(a->auren_flags, DROPPED_SRC))
25007 + d_drop(a->src_dentry);
25009 + au_update_dbtop(a->dst_dentry);
25010 + if (!a->dst_inode)
25011 + d_drop(a->dst_dentry);
25013 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
25014 + di_write_unlock(a->dst_parent);
25016 + di_write_unlock2(a->src_parent, a->dst_parent);
25018 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
25020 + iput(a->dst_inode);
25022 + au_whtmp_rmdir_free(a->thargs);
25028 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
25029 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
25030 +++ linux/fs/aufs/Kconfig 2022-11-05 23:02:18.959222617 +0100
25032 +# SPDX-License-Identifier: GPL-2.0
25034 + tristate "Aufs (Advanced multi layered unification filesystem) support"
25036 + Aufs is a stackable unification filesystem such as Unionfs,
25037 + which unifies several directories and provides a merged single
25039 + In the early days, aufs was entirely re-designed and
25040 + re-implemented Unionfs Version 1.x series. Introducing many
25041 + original ideas, approaches and improvements, it becomes totally
25042 + different from Unionfs while keeping the basic features.
25046 + prompt "Maximum number of branches"
25047 + default AUFS_BRANCH_MAX_127
25049 + Specifies the maximum number of branches (or member directories)
25050 + in a single aufs. The larger value consumes more system
25051 + resources and has a minor impact to performance.
25052 +config AUFS_BRANCH_MAX_127
25055 + Specifies the maximum number of branches (or member directories)
25056 + in a single aufs. The larger value consumes more system
25057 + resources and has a minor impact to performance.
25058 +config AUFS_BRANCH_MAX_511
25061 + Specifies the maximum number of branches (or member directories)
25062 + in a single aufs. The larger value consumes more system
25063 + resources and has a minor impact to performance.
25064 +config AUFS_BRANCH_MAX_1023
25067 + Specifies the maximum number of branches (or member directories)
25068 + in a single aufs. The larger value consumes more system
25069 + resources and has a minor impact to performance.
25070 +config AUFS_BRANCH_MAX_32767
25073 + Specifies the maximum number of branches (or member directories)
25074 + in a single aufs. The larger value consumes more system
25075 + resources and has a minor impact to performance.
25078 +config AUFS_SBILIST
25080 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
25083 + Automatic configuration for internal use.
25084 + When aufs supports Magic SysRq or /proc, enabled automatically.
25086 +config AUFS_HNOTIFY
25087 + bool "Detect direct branch access (bypassing aufs)"
25089 + If you want to modify files on branches directly, eg. bypassing aufs,
25090 + and want aufs to detect the changes of them fully, then enable this
25091 + option and use 'udba=notify' mount option.
25092 + Currently there is only one available configuration, "fsnotify".
25093 + It will have a negative impact to the performance.
25094 + See detail in aufs.5.
25097 + prompt "method" if AUFS_HNOTIFY
25098 + default AUFS_HFSNOTIFY
25099 +config AUFS_HFSNOTIFY
25104 +config AUFS_EXPORT
25105 + bool "NFS-exportable aufs"
25106 + depends on EXPORTFS
25108 + If you want to export your mounted aufs via NFS, then enable this
25109 + option. There are several requirements for this configuration.
25110 + See detail in aufs.5.
25112 +config AUFS_INO_T_64
25114 + depends on AUFS_EXPORT
25115 + depends on 64BIT && !(ALPHA || S390)
25118 + Automatic configuration for internal use.
25119 + /* typedef unsigned long/int __kernel_ino_t */
25120 + /* alpha and s390x are int */
25123 + bool "support for XATTR/EA (including Security Labels)"
25125 + If your branch fs supports XATTR/EA and you want to make them
25126 + available in aufs too, then enable this opsion and specify the
25127 + branch attributes for EA.
25128 + See detail in aufs.5.
25131 + bool "File-based Hierarchical Storage Management"
25133 + Hierarchical Storage Management (or HSM) is a well-known feature
25134 + in the storage world. Aufs provides this feature as file-based.
25135 + with multiple branches.
25136 + These multiple branches are prioritized, ie. the topmost one
25137 + should be the fastest drive and be used heavily.
25140 + bool "Readdir in userspace"
25142 + Aufs has two methods to provide a merged view for a directory,
25143 + by a user-space library and by kernel-space natively. The latter
25144 + is always enabled but sometimes large and slow.
25145 + If you enable this option, install the library in aufs2-util
25146 + package, and set some environment variables for your readdir(3),
25147 + then the work will be handled in user-space which generally
25148 + shows better performance in most cases.
25149 + See detail in aufs.5.
25151 +config AUFS_DIRREN
25152 + bool "Workaround for rename(2)-ing a directory"
25154 + By default, aufs returns EXDEV error in renameing a dir who has
25155 + his child on the lower branch, since it is a bad idea to issue
25156 + rename(2) internally for every lower branch. But user may not
25157 + accept this behaviour. So here is a workaround to allow such
25158 + rename(2) and store some extra information on the writable
25159 + branch. Obviously this costs high (and I don't like it).
25160 + To use this feature, you need to enable this configuration AND
25161 + to specify the mount option `dirren.'
25162 + See details in aufs.5 and the design documents.
25165 + bool "Show whiteouts"
25167 + If you want to make the whiteouts in aufs visible, then enable
25168 + this option and specify 'shwh' mount option. Although it may
25169 + sounds like philosophy or something, but in technically it
25170 + simply shows the name of whiteout with keeping its behaviour.
25172 +config AUFS_BR_RAMFS
25173 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
25175 + If you want to use ramfs as an aufs branch fs, then enable this
25176 + option. Generally tmpfs is recommended.
25177 + Aufs prohibited them to be a branch fs by default, because
25178 + initramfs becomes unusable after switch_root or something
25179 + generally. If you sets initramfs as an aufs branch and boot your
25180 + system by switch_root, you will meet a problem easily since the
25181 + files in initramfs may be inaccessible.
25182 + Unless you are going to use ramfs as an aufs branch fs without
25183 + switch_root or something, leave it N.
25185 +config AUFS_BR_FUSE
25186 + bool "Fuse fs as an aufs branch"
25187 + depends on FUSE_FS
25190 + If you want to use fuse-based userspace filesystem as an aufs
25191 + branch fs, then enable this option.
25192 + It implements the internal poll(2) operation which is
25193 + implemented by fuse only (curretnly).
25198 + Automatic configuration for internal use.
25200 +config AUFS_BR_HFSPLUS
25201 + bool "Hfsplus as an aufs branch"
25202 + depends on HFSPLUS_FS
25205 + If you want to use hfsplus fs as an aufs branch fs, then enable
25206 + this option. This option introduces a small overhead at
25207 + copying-up a file on hfsplus.
25209 +config AUFS_BDEV_LOOP
25211 + depends on BLK_DEV_LOOP
25214 + Automatic configuration for internal use.
25215 + Convert =[ym] into =y.
25218 + bool "Debug aufs"
25220 + Enable this to compile aufs internal debug code.
25221 + It will have a negative impact to the performance.
25223 +config AUFS_MAGIC_SYSRQ
25225 + depends on AUFS_DEBUG && MAGIC_SYSRQ
25228 + Automatic configuration for internal use.
25229 + When aufs supports Magic SysRq, enabled automatically.
25231 diff -urN /usr/share/empty/fs/aufs/lcnt.h linux/fs/aufs/lcnt.h
25232 --- /usr/share/empty/fs/aufs/lcnt.h 1970-01-01 01:00:00.000000000 +0100
25233 +++ linux/fs/aufs/lcnt.h 2022-11-05 23:02:18.965889284 +0100
25235 +/* SPDX-License-Identifier: GPL-2.0 */
25237 + * Copyright (C) 2018-2022 Junjiro R. Okajima
25239 + * This program is free software; you can redistribute it and/or modify
25240 + * it under the terms of the GNU General Public License as published by
25241 + * the Free Software Foundation; either version 2 of the License, or
25242 + * (at your option) any later version.
25244 + * This program is distributed in the hope that it will be useful,
25245 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25246 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25247 + * GNU General Public License for more details.
25249 + * You should have received a copy of the GNU General Public License
25250 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25254 + * simple long counter wrapper
25257 +#ifndef __AUFS_LCNT_H__
25258 +#define __AUFS_LCNT_H__
25262 +#include "debug.h"
25264 +#define AuLCntATOMIC 1
25265 +#define AuLCntPCPUCNT 2
25267 + * why does percpu_refcount require extra synchronize_rcu()s in
25268 + * au_br_do_free()
25270 +#define AuLCntPCPUREF 3
25272 +/* #define AuLCntChosen AuLCntATOMIC */
25273 +#define AuLCntChosen AuLCntPCPUCNT
25274 +/* #define AuLCntChosen AuLCntPCPUREF */
25276 +#if AuLCntChosen == AuLCntATOMIC
25277 +#include <linux/atomic.h>
25279 +typedef atomic_long_t au_lcnt_t;
25281 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25283 + atomic_long_set(cnt, 0);
25287 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25292 +static inline void au_lcnt_fin(au_lcnt_t *cnt __maybe_unused,
25293 + int do_sync __maybe_unused)
25298 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25300 + atomic_long_inc(cnt);
25303 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25305 + atomic_long_dec(cnt);
25308 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25310 + return atomic_long_read(cnt);
25314 +#if AuLCntChosen == AuLCntPCPUCNT
25315 +#include <linux/percpu_counter.h>
25317 +typedef struct percpu_counter au_lcnt_t;
25319 +static inline int au_lcnt_init(au_lcnt_t *cnt, void *release __maybe_unused)
25321 + return percpu_counter_init(cnt, 0, GFP_NOFS);
25324 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25329 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync __maybe_unused)
25331 + percpu_counter_destroy(cnt);
25334 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25336 + percpu_counter_inc(cnt);
25339 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25341 + percpu_counter_dec(cnt);
25344 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev __maybe_unused)
25348 + n = percpu_counter_sum(cnt);
25350 + if (LONG_MAX != LLONG_MAX
25352 + AuWarn1("%s\n", "wrap-around");
25358 +#if AuLCntChosen == AuLCntPCPUREF
25359 +#include <linux/percpu-refcount.h>
25361 +typedef struct percpu_ref au_lcnt_t;
25363 +static inline int au_lcnt_init(au_lcnt_t *cnt, percpu_ref_func_t *release)
25366 + release = percpu_ref_exit;
25367 + return percpu_ref_init(cnt, release, /*percpu mode*/0, GFP_NOFS);
25370 +static inline void au_lcnt_wait_for_fin(au_lcnt_t *cnt __maybe_unused)
25372 + synchronize_rcu();
25375 +static inline void au_lcnt_fin(au_lcnt_t *cnt, int do_sync)
25377 + percpu_ref_kill(cnt);
25379 + au_lcnt_wait_for_fin(cnt);
25382 +static inline void au_lcnt_inc(au_lcnt_t *cnt)
25384 + percpu_ref_get(cnt);
25387 +static inline void au_lcnt_dec(au_lcnt_t *cnt)
25389 + percpu_ref_put(cnt);
25393 + * avoid calling this func as possible.
25395 +static inline long au_lcnt_read(au_lcnt_t *cnt, int do_rev)
25399 + percpu_ref_switch_to_atomic_sync(cnt);
25400 + l = atomic_long_read(&cnt->count);
25402 + percpu_ref_switch_to_percpu(cnt);
25404 + /* percpu_ref is initialized by 1 instead of 0 */
25409 +#ifdef CONFIG_AUFS_DEBUG
25410 +#define AuLCntZero(val) do { \
25413 + AuDbg("%s = %ld\n", #val, l); \
25416 +#define AuLCntZero(val) do {} while (0)
25419 +#endif /* __KERNEL__ */
25420 +#endif /* __AUFS_LCNT_H__ */
25421 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
25422 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
25423 +++ linux/fs/aufs/loop.c 2022-11-05 23:02:18.965889284 +0100
25425 +// SPDX-License-Identifier: GPL-2.0
25427 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25429 + * This program is free software; you can redistribute it and/or modify
25430 + * it under the terms of the GNU General Public License as published by
25431 + * the Free Software Foundation; either version 2 of the License, or
25432 + * (at your option) any later version.
25434 + * This program is distributed in the hope that it will be useful,
25435 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25436 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25437 + * GNU General Public License for more details.
25439 + * You should have received a copy of the GNU General Public License
25440 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25444 + * support for loopback block device as a branch
25449 +/* added into drivers/block/loop.c */
25450 +static struct file *(*backing_file_func)(struct super_block *sb);
25453 + * test if two lower dentries have overlapping branches.
25455 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
25457 + struct super_block *h_sb;
25458 + struct file *backing_file;
25460 + if (unlikely(!backing_file_func)) {
25461 + /* don't load "loop" module here */
25462 + backing_file_func = symbol_get(loop_backing_file);
25463 + if (unlikely(!backing_file_func))
25464 + /* "loop" module is not loaded */
25468 + h_sb = h_adding->d_sb;
25469 + backing_file = backing_file_func(h_sb);
25470 + if (!backing_file)
25473 + h_adding = backing_file->f_path.dentry;
25475 + * h_adding can be local NFS.
25476 + * in this case aufs cannot detect the loop.
25478 + if (unlikely(h_adding->d_sb == sb))
25480 + return !!au_test_subdir(h_adding, sb->s_root);
25483 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
25484 +int au_test_loopback_kthread(void)
25487 + struct task_struct *tsk = current;
25488 + char c, comm[sizeof(tsk->comm)];
25491 + if (tsk->flags & PF_KTHREAD) {
25492 + get_task_comm(comm, tsk);
25494 + ret = ('0' <= c && c <= '9'
25495 + && !strncmp(comm, "loop", 4));
25501 +/* ---------------------------------------------------------------------- */
25503 +#define au_warn_loopback_step 16
25504 +static int au_warn_loopback_nelem = au_warn_loopback_step;
25505 +static unsigned long *au_warn_loopback_array;
25507 +void au_warn_loopback(struct super_block *h_sb)
25509 + int i, new_nelem;
25510 + unsigned long *a, magic;
25511 + static DEFINE_SPINLOCK(spin);
25513 + magic = h_sb->s_magic;
25514 + spin_lock(&spin);
25515 + a = au_warn_loopback_array;
25516 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
25517 + if (a[i] == magic) {
25518 + spin_unlock(&spin);
25522 + /* h_sb is new to us, print it */
25523 + if (i < au_warn_loopback_nelem) {
25528 + /* expand the array */
25529 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
25530 + a = au_kzrealloc(au_warn_loopback_array,
25531 + au_warn_loopback_nelem * sizeof(unsigned long),
25532 + new_nelem * sizeof(unsigned long), GFP_ATOMIC,
25533 + /*may_shrink*/0);
25535 + au_warn_loopback_nelem = new_nelem;
25536 + au_warn_loopback_array = a;
25541 + spin_unlock(&spin);
25542 + AuWarn1("realloc failed, ignored\n");
25546 + spin_unlock(&spin);
25547 + pr_warn("you may want to try another patch for loopback file "
25548 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
25551 +int au_loopback_init(void)
25554 + struct super_block *sb __maybe_unused;
25556 + BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(*au_warn_loopback_array));
25559 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
25560 + sizeof(unsigned long), GFP_NOFS);
25561 + if (unlikely(!au_warn_loopback_array))
25567 +void au_loopback_fin(void)
25569 + if (backing_file_func)
25570 + symbol_put(loop_backing_file);
25571 + au_kfree_try_rcu(au_warn_loopback_array);
25573 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
25574 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
25575 +++ linux/fs/aufs/loop.h 2022-11-05 23:02:18.965889284 +0100
25577 +/* SPDX-License-Identifier: GPL-2.0 */
25579 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25581 + * This program is free software; you can redistribute it and/or modify
25582 + * it under the terms of the GNU General Public License as published by
25583 + * the Free Software Foundation; either version 2 of the License, or
25584 + * (at your option) any later version.
25586 + * This program is distributed in the hope that it will be useful,
25587 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25588 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25589 + * GNU General Public License for more details.
25591 + * You should have received a copy of the GNU General Public License
25592 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25596 + * support for loopback mount as a branch
25599 +#ifndef __AUFS_LOOP_H__
25600 +#define __AUFS_LOOP_H__
25605 +struct super_block;
25607 +#ifdef CONFIG_AUFS_BDEV_LOOP
25608 +/* drivers/block/loop.c */
25609 +struct file *loop_backing_file(struct super_block *sb);
25612 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
25613 +int au_test_loopback_kthread(void);
25614 +void au_warn_loopback(struct super_block *h_sb);
25616 +int au_loopback_init(void);
25617 +void au_loopback_fin(void);
25619 +AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
25621 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
25622 + struct dentry *h_adding)
25623 +AuStubInt0(au_test_loopback_kthread, void)
25624 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
25626 +AuStubInt0(au_loopback_init, void)
25627 +AuStubVoid(au_loopback_fin, void)
25628 +#endif /* BLK_DEV_LOOP */
25630 +#endif /* __KERNEL__ */
25631 +#endif /* __AUFS_LOOP_H__ */
25632 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
25633 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
25634 +++ linux/fs/aufs/magic.mk 2022-11-05 23:02:18.965889284 +0100
25636 +# SPDX-License-Identifier: GPL-2.0
25638 +# defined in ${srctree}/fs/fuse/inode.c
25640 +ifdef CONFIG_FUSE_FS
25641 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
25644 +# defined in ${srctree}/fs/xfs/xfs_sb.h
25646 +ifdef CONFIG_XFS_FS
25647 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
25650 +# defined in ${srctree}/fs/configfs/mount.c
25652 +ifdef CONFIG_CONFIGFS_FS
25653 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
25656 +# defined in ${srctree}/fs/ubifs/ubifs.h
25658 +ifdef CONFIG_UBIFS_FS
25659 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
25662 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
25664 +ifdef CONFIG_HFSPLUS_FS
25665 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
25667 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
25668 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
25669 +++ linux/fs/aufs/Makefile 2022-11-05 23:02:18.959222617 +0100
25671 +# SPDX-License-Identifier: GPL-2.0
25673 +include ${src}/magic.mk
25674 +ifeq (${CONFIG_AUFS_FS},m)
25675 +include ${src}/conf.mk
25677 +-include ${src}/priv_def.mk
25679 +# cf. include/linux/kernel.h
25681 +ccflags-y += -DDEBUG
25682 +# sparse requires the full pathname
25684 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
25686 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
25689 +obj-$(CONFIG_AUFS_FS) += aufs.o
25690 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o fsctx.o \
25691 + wkq.o vfsub.o dcsub.o \
25692 + cpup.o whout.o wbr_policy.o \
25693 + dinfo.o dentry.o \
25695 + finfo.o file.o f_op.o \
25697 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
25701 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
25702 +aufs-$(CONFIG_SYSFS) += sysfs.o
25703 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
25704 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
25705 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
25706 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
25707 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
25708 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
25709 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
25710 +aufs-$(CONFIG_AUFS_DIRREN) += dirren.o
25711 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
25712 +aufs-$(CONFIG_AUFS_POLL) += poll.o
25713 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
25714 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
25715 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
25716 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
25717 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
25718 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
25719 +++ linux/fs/aufs/module.c 2022-11-05 23:02:18.965889284 +0100
25721 +// SPDX-License-Identifier: GPL-2.0
25723 + * Copyright (C) 2005-2022 Junjiro R. Okajima
25725 + * This program is free software; you can redistribute it and/or modify
25726 + * it under the terms of the GNU General Public License as published by
25727 + * the Free Software Foundation; either version 2 of the License, or
25728 + * (at your option) any later version.
25730 + * This program is distributed in the hope that it will be useful,
25731 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25732 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25733 + * GNU General Public License for more details.
25735 + * You should have received a copy of the GNU General Public License
25736 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25740 + * module global variables and operations
25743 +#include <linux/module.h>
25744 +#include <linux/seq_file.h>
25747 +/* shrinkable realloc */
25748 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink)
25756 +#if 0 /* unused */
25763 + AuDebugOn(!new_sz);
25766 + diff = au_kmidx_sub(sz, new_sz);
25772 + /* expand or SLOB */
25773 + p = krealloc(p, new_sz, gfp);
25774 + else if (new_sz < sz && may_shrink) {
25778 + q = kmalloc(new_sz, gfp);
25781 + memcpy(q, p, new_sz);
25782 + au_kfree_try_rcu(p);
25793 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
25796 + p = au_krealloc(p, new_sz, gfp, may_shrink);
25797 + if (p && new_sz > nused)
25798 + memset(p + nused, 0, new_sz - nused);
25802 +/* ---------------------------------------------------------------------- */
25806 +struct kmem_cache *au_cache[AuCache_Last];
25808 +static void au_cache_fin(void)
25813 + * Make sure all delayed rcu free inodes are flushed before we
25818 + /* excluding AuCache_HNOTIFY */
25819 + BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
25820 + for (i = 0; i < AuCache_HNOTIFY; i++) {
25821 + kmem_cache_destroy(au_cache[i]);
25822 + au_cache[i] = NULL;
25826 +static int __init au_cache_init(void)
25828 + au_cache[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
25829 + if (au_cache[AuCache_DINFO])
25830 + /* SLAB_DESTROY_BY_RCU */
25831 + au_cache[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
25832 + au_icntnr_init_once);
25833 + if (au_cache[AuCache_ICNTNR])
25834 + au_cache[AuCache_FINFO] = AuCacheCtor(au_finfo,
25835 + au_fi_init_once);
25836 + if (au_cache[AuCache_FINFO])
25837 + au_cache[AuCache_VDIR] = AuCache(au_vdir);
25838 + if (au_cache[AuCache_VDIR])
25839 + au_cache[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
25840 + if (au_cache[AuCache_DEHSTR])
25847 +/* ---------------------------------------------------------------------- */
25849 +int au_dir_roflags;
25851 +#ifdef CONFIG_AUFS_SBILIST
25853 + * iterate_supers_type() doesn't protect us from
25854 + * remounting (branch management)
25856 +struct hlist_bl_head au_sbilist;
25860 + * functions for module interface.
25862 +MODULE_LICENSE("GPL");
25863 +/* MODULE_LICENSE("GPL v2"); */
25864 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
25865 +MODULE_DESCRIPTION(AUFS_NAME
25866 + " -- Advanced multi layered unification filesystem");
25867 +MODULE_VERSION(AUFS_VERSION);
25868 +MODULE_ALIAS_FS(AUFS_NAME);
25870 +/* this module parameter has no meaning when SYSFS is disabled */
25871 +int sysaufs_brs = 1;
25872 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
25873 +module_param_named(brs, sysaufs_brs, int, 0444);
25875 +/* this module parameter has no meaning when USER_NS is disabled */
25877 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
25878 +module_param_named(allow_userns, au_userns, bool, 0444);
25880 +/* ---------------------------------------------------------------------- */
25882 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
25884 +int au_seq_path(struct seq_file *seq, struct path *path)
25888 + err = seq_path(seq, path, au_esc_chars);
25897 +/* ---------------------------------------------------------------------- */
25899 +static int __init aufs_init(void)
25904 + p = au_esc_chars;
25905 + for (i = 1; i <= ' '; i++)
25911 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
25913 + memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
25914 + for (i = 0; i < AuIop_Last; i++)
25915 + aufs_iop_nogetattr[i].getattr = NULL;
25917 + memset(au_cache, 0, sizeof(au_cache)); /* including hnotify */
25919 + au_sbilist_init();
25920 + sysaufs_brs_init();
25923 + err = sysaufs_init();
25924 + if (unlikely(err))
25926 + err = dbgaufs_init();
25927 + if (unlikely(err))
25928 + goto out_sysaufs;
25929 + err = au_procfs_init();
25930 + if (unlikely(err))
25931 + goto out_dbgaufs;
25932 + err = au_wkq_init();
25933 + if (unlikely(err))
25935 + err = au_loopback_init();
25936 + if (unlikely(err))
25938 + err = au_hnotify_init();
25939 + if (unlikely(err))
25940 + goto out_loopback;
25941 + err = au_sysrq_init();
25942 + if (unlikely(err))
25944 + err = au_cache_init();
25945 + if (unlikely(err))
25948 + aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
25949 + err = register_filesystem(&aufs_fs_type);
25950 + if (unlikely(err))
25953 + /* since we define pr_fmt, call printk directly */
25954 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
25955 + goto out; /* success */
25962 + au_hnotify_fin();
25964 + au_loopback_fin();
25978 +static void __exit aufs_exit(void)
25980 + unregister_filesystem(&aufs_fs_type);
25983 + au_hnotify_fin();
25984 + au_loopback_fin();
25992 +module_init(aufs_init);
25993 +module_exit(aufs_exit);
25994 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
25995 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
25996 +++ linux/fs/aufs/module.h 2022-11-05 23:02:18.969222617 +0100
25998 +/* SPDX-License-Identifier: GPL-2.0 */
26000 + * Copyright (C) 2005-2022 Junjiro R. Okajima
26002 + * This program is free software; you can redistribute it and/or modify
26003 + * it under the terms of the GNU General Public License as published by
26004 + * the Free Software Foundation; either version 2 of the License, or
26005 + * (at your option) any later version.
26007 + * This program is distributed in the hope that it will be useful,
26008 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26009 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26010 + * GNU General Public License for more details.
26012 + * You should have received a copy of the GNU General Public License
26013 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26017 + * module initialization and module-global
26020 +#ifndef __AUFS_MODULE_H__
26021 +#define __AUFS_MODULE_H__
26025 +#include <linux/slab.h>
26026 +#include "debug.h"
26027 +#include "dentry.h"
26030 +#include "inode.h"
26035 +/* module parameters */
26036 +extern int sysaufs_brs;
26037 +extern bool au_userns;
26039 +/* ---------------------------------------------------------------------- */
26041 +extern int au_dir_roflags;
26043 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink);
26044 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
26048 + * Comparing the size of the object with sizeof(struct rcu_head)
26049 + * case 1: object is always larger
26050 + * --> au_kfree_rcu() or au_kfree_do_rcu()
26051 + * case 2: object is always smaller
26052 + * --> au_kfree_small()
26053 + * case 3: object can be any size
26054 + * --> au_kfree_try_rcu()
26057 +static inline void au_kfree_do_rcu(const void *p)
26060 + struct rcu_head rcu;
26061 + } *a = (void *)p;
26063 + kfree_rcu(a, rcu);
26066 +#define au_kfree_rcu(_p) do { \
26067 + typeof(_p) p = (_p); \
26068 + BUILD_BUG_ON(sizeof(*p) < sizeof(struct rcu_head)); \
26070 + au_kfree_do_rcu(p); \
26073 +#define au_kfree_do_sz_test(sz) (sz >= sizeof(struct rcu_head))
26074 +#define au_kfree_sz_test(p) (p && au_kfree_do_sz_test(ksize(p)))
26076 +static inline void au_kfree_try_rcu(const void *p)
26080 + if (au_kfree_sz_test(p))
26081 + au_kfree_do_rcu(p);
26086 +static inline void au_kfree_small(const void *p)
26090 + AuDebugOn(au_kfree_sz_test(p));
26094 +static inline int au_kmidx_sub(size_t sz, size_t new_sz)
26096 +#ifndef CONFIG_SLOB
26097 + return __kmalloc_index(sz, false) - __kmalloc_index(new_sz, false);
26099 + return -1; /* SLOB is untested */
26103 +int au_seq_path(struct seq_file *seq, struct path *path);
26105 +#ifdef CONFIG_PROC_FS
26107 +int __init au_procfs_init(void);
26108 +void au_procfs_fin(void);
26110 +AuStubInt0(au_procfs_init, void);
26111 +AuStubVoid(au_procfs_fin, void);
26114 +/* ---------------------------------------------------------------------- */
26123 + AuCache_HNOTIFY, /* must be last */
26127 +extern struct kmem_cache *au_cache[AuCache_Last];
26129 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
26130 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
26131 +#define AuCacheCtor(type, ctor) \
26132 + kmem_cache_create(#type, sizeof(struct type), \
26133 + __alignof__(struct type), AuCacheFlags, ctor)
26135 +#define AuCacheFuncAlloc(name, index) \
26136 + static inline struct au_##name *au_cache_alloc_##name(void) \
26137 + { return kmem_cache_alloc(au_cache[AuCache_##index], GFP_NOFS); }
26139 +#define AuCacheFuncs(name, index) \
26140 + static inline void au_cache_free_##name##_norcu(struct au_##name *p) \
26141 + { kmem_cache_free(au_cache[AuCache_##index], p); } \
26143 + static inline void au_cache_free_##name##_rcu_cb(struct rcu_head *rcu) \
26144 + { void *p = rcu; \
26145 + p -= offsetof(struct au_##name, rcu); \
26146 + kmem_cache_free(au_cache[AuCache_##index], p); } \
26147 + static inline void au_cache_free_##name##_rcu(struct au_##name *p) \
26148 + { BUILD_BUG_ON(sizeof(struct au_##name) < sizeof(struct rcu_head)); \
26149 + call_rcu(&p->rcu, au_cache_free_##name##_rcu_cb); } \
26151 + static inline void au_cache_free_##name(struct au_##name *p) \
26152 + { /* au_cache_free_##name##_norcu(p); */ \
26153 + au_cache_free_##name##_rcu(p); }
26155 +AuCacheFuncs(dinfo, DINFO);
26156 +AuCacheFuncAlloc(dinfo, DINFO);
26158 +AuCacheFuncs(icntnr, ICNTNR);
26159 +static inline struct au_icntnr *au_cache_alloc_icntnr(struct super_block *sb)
26160 +{ return alloc_inode_sb(sb, au_cache[AuCache_ICNTNR], GFP_NOFS); }
26162 +AuCacheFuncs(finfo, FINFO);
26163 +AuCacheFuncAlloc(finfo, FINFO);
26165 +AuCacheFuncs(vdir, VDIR);
26166 +AuCacheFuncAlloc(vdir, VDIR);
26168 +AuCacheFuncs(vdir_dehstr, DEHSTR);
26169 +AuCacheFuncAlloc(vdir_dehstr, DEHSTR);
26171 +#ifdef CONFIG_AUFS_HNOTIFY
26172 +AuCacheFuncs(hnotify, HNOTIFY);
26173 +AuCacheFuncAlloc(hnotify, HNOTIFY);
26176 +#endif /* __KERNEL__ */
26177 +#endif /* __AUFS_MODULE_H__ */
26178 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
26179 --- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
26180 +++ linux/fs/aufs/mvdown.c 2022-11-05 23:02:18.969222617 +0100
26182 +// SPDX-License-Identifier: GPL-2.0
26184 + * Copyright (C) 2011-2022 Junjiro R. Okajima
26186 + * This program is free software; you can redistribute it and/or modify
26187 + * it under the terms of the GNU General Public License as published by
26188 + * the Free Software Foundation; either version 2 of the License, or
26189 + * (at your option) any later version.
26191 + * This program is distributed in the hope that it will be useful,
26192 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26193 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26194 + * GNU General Public License for more details.
26196 + * You should have received a copy of the GNU General Public License
26197 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26201 + * move-down, opposite of copy-up
26206 +struct au_mvd_args {
26208 + struct super_block *h_sb;
26209 + struct dentry *h_parent;
26210 + struct au_hinode *hdir;
26211 + struct inode *h_dir, *h_inode;
26212 + struct au_pin pin;
26213 + } info[AUFS_MVDOWN_NARRAY];
26215 + struct aufs_mvdown mvdown;
26216 + struct dentry *dentry, *parent;
26217 + struct inode *inode, *dir;
26218 + struct super_block *sb;
26219 + aufs_bindex_t bopq, bwh, bfound;
26220 + unsigned char rename_lock;
26223 +#define mvd_errno mvdown.au_errno
26224 +#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
26225 +#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
26226 +#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
26227 +#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
26229 +#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
26230 +#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
26231 +#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
26232 +#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
26233 +#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
26234 +#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
26236 +#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
26237 +#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
26238 +#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
26239 +#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
26240 +#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
26241 +#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
26243 +#define AU_MVD_PR(flag, ...) do { \
26245 + pr_err(__VA_ARGS__); \
26248 +static int find_lower_writable(struct au_mvd_args *a)
26250 + struct super_block *sb;
26251 + aufs_bindex_t bindex, bbot;
26252 + struct au_branch *br;
26255 + bindex = a->mvd_bsrc;
26256 + bbot = au_sbbot(sb);
26257 + if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
26258 + for (bindex++; bindex <= bbot; bindex++) {
26259 + br = au_sbr(sb, bindex);
26260 + if (au_br_fhsm(br->br_perm)
26261 + && !sb_rdonly(au_br_sb(br)))
26264 + else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
26265 + for (bindex++; bindex <= bbot; bindex++) {
26266 + br = au_sbr(sb, bindex);
26267 + if (!au_br_rdonly(br))
26271 + for (bindex++; bindex <= bbot; bindex++) {
26272 + br = au_sbr(sb, bindex);
26273 + if (!sb_rdonly(au_br_sb(br))) {
26274 + if (au_br_rdonly(br))
26276 + |= AUFS_MVDOWN_ROLOWER_R;
26284 +/* make the parent dir on bdst */
26285 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
26290 + a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
26291 + a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
26292 + a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
26293 + a->mvd_h_dst_parent = NULL;
26294 + if (au_dbbot(a->parent) >= a->mvd_bdst)
26295 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26296 + if (!a->mvd_h_dst_parent) {
26297 + err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
26298 + if (unlikely(err)) {
26299 + AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
26302 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
26310 +/* lock them all */
26311 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
26314 + struct dentry *h_trap;
26316 + a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
26317 + a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
26318 + err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
26319 + au_opt_udba(a->sb),
26320 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26322 + if (unlikely(err)) {
26323 + AU_MVD_PR(dmsg, "pin_dst failed\n");
26327 + if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
26328 + a->rename_lock = 0;
26329 + au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26330 + AuLsc_DI_PARENT, AuLsc_I_PARENT3,
26331 + au_opt_udba(a->sb),
26332 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26333 + err = au_do_pin(&a->mvd_pin_src);
26335 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26336 + if (unlikely(err)) {
26337 + AU_MVD_PR(dmsg, "pin_src failed\n");
26340 + goto out; /* success */
26343 + a->rename_lock = 1;
26344 + au_pin_hdir_unlock(&a->mvd_pin_dst);
26345 + err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
26346 + au_opt_udba(a->sb),
26347 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
26349 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
26350 + if (unlikely(err)) {
26351 + AU_MVD_PR(dmsg, "pin_src failed\n");
26352 + au_pin_hdir_lock(&a->mvd_pin_dst);
26355 + au_pin_hdir_unlock(&a->mvd_pin_src);
26356 + h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26357 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26359 + err = (h_trap != a->mvd_h_src_parent);
26361 + err = (h_trap != a->mvd_h_dst_parent);
26363 + BUG_ON(err); /* it should never happen */
26364 + if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
26367 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26368 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26369 + au_pin_hdir_lock(&a->mvd_pin_src);
26370 + au_unpin(&a->mvd_pin_src);
26371 + au_pin_hdir_lock(&a->mvd_pin_dst);
26374 + goto out; /* success */
26377 + au_unpin(&a->mvd_pin_dst);
26383 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
26385 + if (!a->rename_lock)
26386 + au_unpin(&a->mvd_pin_src);
26388 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
26389 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
26390 + au_pin_hdir_lock(&a->mvd_pin_src);
26391 + au_unpin(&a->mvd_pin_src);
26392 + au_pin_hdir_lock(&a->mvd_pin_dst);
26394 + au_unpin(&a->mvd_pin_dst);
26397 +/* copy-down the file */
26398 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
26401 + struct au_cp_generic cpg = {
26402 + .dentry = a->dentry,
26403 + .bdst = a->mvd_bdst,
26404 + .bsrc = a->mvd_bsrc,
26406 + .pin = &a->mvd_pin_dst,
26407 + .flags = AuCpup_DTIME | AuCpup_HOPEN
26410 + AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
26411 + if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26412 + au_fset_cpup(cpg.flags, OVERWRITE);
26413 + if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
26414 + au_fset_cpup(cpg.flags, RWDST);
26415 + err = au_sio_cpdown_simple(&cpg);
26416 + if (unlikely(err))
26417 + AU_MVD_PR(dmsg, "cpdown failed\n");
26424 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
26427 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
26430 + struct path h_path;
26431 + struct au_branch *br;
26432 + struct inode *delegated;
26434 + br = au_sbr(a->sb, a->mvd_bdst);
26435 + h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
26436 + err = PTR_ERR(h_path.dentry);
26437 + if (IS_ERR(h_path.dentry)) {
26438 + AU_MVD_PR(dmsg, "wh_lkup failed\n");
26443 + if (d_is_positive(h_path.dentry)) {
26444 + h_path.mnt = au_br_mnt(br);
26445 + delegated = NULL;
26446 + err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
26447 + &delegated, /*force*/0);
26448 + if (unlikely(err == -EWOULDBLOCK)) {
26449 + pr_warn("cannot retry for NFSv4 delegation"
26450 + " for an internal unlink\n");
26453 + if (unlikely(err))
26454 + AU_MVD_PR(dmsg, "wh_unlink failed\n");
26456 + dput(h_path.dentry);
26464 + * unlink the topmost h_dentry
26466 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
26469 + struct path h_path;
26470 + struct inode *delegated;
26472 + h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
26473 + h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
26474 + delegated = NULL;
26475 + err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
26476 + if (unlikely(err == -EWOULDBLOCK)) {
26477 + pr_warn("cannot retry for NFSv4 delegation"
26478 + " for an internal unlink\n");
26481 + if (unlikely(err))
26482 + AU_MVD_PR(dmsg, "unlink failed\n");
26488 +/* Since mvdown succeeded, we ignore an error of this function */
26489 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
26492 + struct au_branch *br;
26494 + a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
26495 + br = au_sbr(a->sb, a->mvd_bsrc);
26496 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
26498 + br = au_sbr(a->sb, a->mvd_bdst);
26499 + a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
26500 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
26503 + a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
26505 + AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
26509 + * copy-down the file and unlink the bsrc file.
26510 + * - unlink the bdst whout if exist
26511 + * - copy-down the file (with whtmp name and rename)
26512 + * - unlink the bsrc file
26514 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
26518 + err = au_do_mkdir(dmsg, a);
26520 + err = au_do_lock(dmsg, a);
26521 + if (unlikely(err))
26525 + * do not revert the activities we made on bdst since they should be
26526 + * harmless in aufs.
26529 + err = au_do_cpdown(dmsg, a);
26531 + err = au_do_unlink_wh(dmsg, a);
26532 + if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
26533 + err = au_do_unlink(dmsg, a);
26534 + if (unlikely(err))
26537 + AuDbg("%pd2, 0x%x, %d --> %d\n",
26538 + a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
26539 + if (find_lower_writable(a) < 0)
26540 + a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
26542 + if (a->mvdown.flags & AUFS_MVDOWN_STFS)
26543 + au_do_stfs(dmsg, a);
26545 + /* maintain internal array */
26546 + if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
26547 + au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
26548 + au_set_dbtop(a->dentry, a->mvd_bdst);
26549 + au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
26550 + au_set_ibtop(a->inode, a->mvd_bdst);
26552 + /* hide the lower */
26553 + au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
26554 + au_set_dbbot(a->dentry, a->mvd_bsrc);
26555 + au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
26556 + au_set_ibbot(a->inode, a->mvd_bsrc);
26558 + if (au_dbbot(a->dentry) < a->mvd_bdst)
26559 + au_set_dbbot(a->dentry, a->mvd_bdst);
26560 + if (au_ibbot(a->inode) < a->mvd_bdst)
26561 + au_set_ibbot(a->inode, a->mvd_bdst);
26564 + au_do_unlock(dmsg, a);
26570 +/* ---------------------------------------------------------------------- */
26572 +/* make sure the file is idle */
26573 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
26575 + int err, plinked;
26578 + plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
26579 + if (au_dbtop(a->dentry) == a->mvd_bsrc
26580 + && au_dcount(a->dentry) == 1
26581 + && atomic_read(&a->inode->i_count) == 1
26582 + /* && a->mvd_h_src_inode->i_nlink == 1 */
26583 + && (!plinked || !au_plink_test(a->inode))
26584 + && a->inode->i_nlink == 1)
26589 + "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
26590 + a->mvd_bsrc, au_dbtop(a->dentry), au_dcount(a->dentry),
26591 + atomic_read(&a->inode->i_count), a->inode->i_nlink,
26592 + a->mvd_h_src_inode->i_nlink,
26593 + plinked, plinked ? au_plink_test(a->inode) : 0);
26600 +/* make sure the parent dir is fine */
26601 +static int au_mvd_args_parent(const unsigned char dmsg,
26602 + struct au_mvd_args *a)
26605 + aufs_bindex_t bindex;
26608 + if (unlikely(au_alive_dir(a->parent))) {
26610 + AU_MVD_PR(dmsg, "parent dir is dead\n");
26614 + a->bopq = au_dbdiropq(a->parent);
26615 + bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
26616 + AuDbg("b%d\n", bindex);
26617 + if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
26618 + || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
26620 + a->mvd_errno = EAU_MVDOWN_OPAQUE;
26621 + AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
26622 + a->bopq, a->mvd_bdst);
26630 +static int au_mvd_args_intermediate(const unsigned char dmsg,
26631 + struct au_mvd_args *a)
26634 + struct au_dinfo *dinfo, *tmp;
26636 + /* lookup the next lower positive entry */
26638 + tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
26639 + if (unlikely(!tmp))
26644 + dinfo = au_di(a->dentry);
26645 + au_di_cp(tmp, dinfo);
26646 + au_di_swap(tmp, dinfo);
26648 + /* returns the number of positive dentries */
26649 + err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1,
26650 + /* AuLkup_IGNORE_PERM */ 0);
26652 + a->bwh = au_dbwh(a->dentry);
26653 + else if (err > 0)
26654 + a->bfound = au_dbtop(a->dentry);
26656 + au_di_swap(tmp, dinfo);
26657 + au_rw_write_unlock(&tmp->di_rwsem);
26659 + if (unlikely(err < 0))
26660 + AU_MVD_PR(dmsg, "failed look-up lower\n");
26663 + * here, we have these cases.
26665 + * no positive dentry under bsrc. there are more sub-cases.
26667 + * there no whiteout, we can safely move-down.
26670 + * bsrc < bwh && bwh < bdst
26671 + * there is a whiteout on RO branch. cannot proceed.
26673 + * there is a whiteout on the RW target branch. it should
26676 + * there is a whiteout somewhere unrelated branch.
26677 + * -1 < bfound && bfound <= bsrc
26680 + * found, but it is on RO branch between bsrc and bdst. cannot
26683 + * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
26686 + * found, after we create the file on bdst, it will be hidden.
26689 + AuDebugOn(a->bfound == -1
26691 + && a->bwh <= a->mvd_bsrc);
26692 + AuDebugOn(-1 < a->bfound
26693 + && a->bfound <= a->mvd_bsrc);
26696 + if (a->bfound == -1
26697 + && a->mvd_bsrc < a->bwh
26699 + && a->bwh < a->mvd_bdst) {
26700 + a->mvd_errno = EAU_MVDOWN_WHITEOUT;
26701 + AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
26702 + a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
26704 + } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
26705 + a->mvd_errno = EAU_MVDOWN_UPPER;
26706 + AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
26707 + a->mvd_bdst, a->bfound);
26711 + err = 0; /* success */
26718 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
26723 + if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
26724 + && a->bfound == a->mvd_bdst)
26730 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
26733 + struct au_branch *br;
26736 + if (unlikely(S_ISDIR(a->inode->i_mode)))
26740 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
26741 + a->mvd_bsrc = au_ibtop(a->inode);
26743 + a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
26744 + if (unlikely(a->mvd_bsrc < 0
26745 + || (a->mvd_bsrc < au_dbtop(a->dentry)
26746 + || au_dbbot(a->dentry) < a->mvd_bsrc
26747 + || !au_h_dptr(a->dentry, a->mvd_bsrc))
26748 + || (a->mvd_bsrc < au_ibtop(a->inode)
26749 + || au_ibbot(a->inode) < a->mvd_bsrc
26750 + || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
26751 + a->mvd_errno = EAU_MVDOWN_NOUPPER;
26752 + AU_MVD_PR(dmsg, "no upper\n");
26756 + if (unlikely(a->mvd_bsrc == au_sbbot(a->sb))) {
26757 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
26758 + AU_MVD_PR(dmsg, "on the bottom\n");
26761 + a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
26762 + br = au_sbr(a->sb, a->mvd_bsrc);
26763 + err = au_br_rdonly(br);
26764 + if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
26765 + if (unlikely(err))
26767 + } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
26768 + || IS_APPEND(a->mvd_h_src_inode))) {
26770 + a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
26776 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
26777 + a->mvd_bdst = find_lower_writable(a);
26778 + if (unlikely(a->mvd_bdst < 0)) {
26779 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
26780 + AU_MVD_PR(dmsg, "no writable lower branch\n");
26784 + a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
26785 + if (unlikely(a->mvd_bdst < 0
26786 + || au_sbbot(a->sb) < a->mvd_bdst)) {
26787 + a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
26788 + AU_MVD_PR(dmsg, "no lower brid\n");
26793 + err = au_mvd_args_busy(dmsg, a);
26795 + err = au_mvd_args_parent(dmsg, a);
26797 + err = au_mvd_args_intermediate(dmsg, a);
26799 + err = au_mvd_args_exist(dmsg, a);
26801 + AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
26808 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
26811 + unsigned char dmsg;
26812 + struct au_mvd_args *args;
26813 + struct inode *inode;
26815 + inode = d_inode(dentry);
26817 + if (unlikely(!capable(CAP_SYS_ADMIN)))
26821 + args = kmalloc(sizeof(*args), GFP_NOFS);
26822 + if (unlikely(!args))
26825 + err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
26827 + /* VERIFY_WRITE */
26828 + err = !access_ok(uarg, sizeof(*uarg));
26829 + if (unlikely(err)) {
26834 + AuDbg("flags 0x%x\n", args->mvdown.flags);
26835 + args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
26836 + args->mvdown.au_errno = 0;
26837 + args->dentry = dentry;
26838 + args->inode = inode;
26839 + args->sb = dentry->d_sb;
26842 + dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
26843 + args->parent = dget_parent(dentry);
26844 + args->dir = d_inode(args->parent);
26845 + inode_lock_nested(args->dir, I_MUTEX_PARENT);
26846 + dput(args->parent);
26847 + if (unlikely(args->parent != dentry->d_parent)) {
26848 + AU_MVD_PR(dmsg, "parent dir is moved\n");
26852 + inode_lock_nested(inode, I_MUTEX_CHILD);
26853 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
26854 + if (unlikely(err))
26857 + di_write_lock_parent(args->parent);
26858 + err = au_mvd_args(dmsg, args);
26859 + if (unlikely(err))
26862 + err = au_do_mvdown(dmsg, args);
26863 + if (unlikely(err))
26866 + au_cpup_attr_timesizes(args->dir);
26867 + au_cpup_attr_timesizes(inode);
26868 + if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
26869 + au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
26870 + /* au_digen_dec(dentry); */
26873 + di_write_unlock(args->parent);
26874 + aufs_read_unlock(dentry, AuLock_DW);
26876 + inode_unlock(inode);
26878 + inode_unlock(args->dir);
26880 + e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
26883 + au_kfree_rcu(args);
26888 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
26889 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
26890 +++ linux/fs/aufs/opts.c 2022-11-05 23:02:18.969222617 +0100
26892 +// SPDX-License-Identifier: GPL-2.0
26894 + * Copyright (C) 2005-2022 Junjiro R. Okajima
26896 + * This program is free software; you can redistribute it and/or modify
26897 + * it under the terms of the GNU General Public License as published by
26898 + * the Free Software Foundation; either version 2 of the License, or
26899 + * (at your option) any later version.
26901 + * This program is distributed in the hope that it will be useful,
26902 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26903 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26904 + * GNU General Public License for more details.
26906 + * You should have received a copy of the GNU General Public License
26907 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26911 + * mount options/flags
26914 +#include <linux/types.h> /* a distribution requires */
26915 +#include <linux/parser.h>
26918 +/* ---------------------------------------------------------------------- */
26920 +static const char *au_parser_pattern(int val, match_table_t tbl)
26922 + struct match_token *p;
26925 + while (p->pattern) {
26926 + if (p->token == val)
26927 + return p->pattern;
26934 +static const char *au_optstr(int *val, match_table_t tbl)
26936 + struct match_token *p;
26943 + while (p->pattern) {
26945 + && (v & p->token) == p->token) {
26946 + *val &= ~p->token;
26947 + return p->pattern;
26956 +/* ---------------------------------------------------------------------- */
26958 +static match_table_t brperm = {
26959 + {AuBrPerm_RO, AUFS_BRPERM_RO},
26960 + {AuBrPerm_RR, AUFS_BRPERM_RR},
26961 + {AuBrPerm_RW, AUFS_BRPERM_RW},
26965 +static match_table_t brattr = {
26967 + {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
26968 + {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
26969 + /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
26970 + {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
26971 +#ifdef CONFIG_AUFS_FHSM
26972 + {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
26974 +#ifdef CONFIG_AUFS_XATTR
26975 + {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
26976 + {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
26977 + {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
26978 + {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
26979 + {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
26980 + {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
26983 + /* ro/rr branch */
26984 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
26987 + {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
26988 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
26993 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
27000 + p = strchr(str, '+');
27003 + v = match_token(str, table, args);
27005 + if (v & AuBrAttr_CMOO_Mask)
27006 + attr &= ~AuBrAttr_CMOO_Mask;
27011 + pr_warn("ignored branch attribute %s\n", str);
27021 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
27029 + p = au_optstr(&perm, brattr);
27032 + memcpy(q, p, sz + 1);
27038 + p = au_optstr(&perm, brattr);
27042 + memcpy(q, p, sz + 1);
27048 + return q - str->a;
27051 +int au_br_perm_val(char *perm)
27053 + int val, bad, sz;
27055 + substring_t args[MAX_OPT_ARGS];
27056 + au_br_perm_str_t attr;
27058 + p = strchr(perm, '+');
27061 + val = match_token(perm, brperm, args);
27065 + pr_warn("ignored branch permission %s\n", perm);
27066 + val = AuBrPerm_RO;
27072 + val |= br_attr_val(p + 1, brattr, args);
27075 + switch (val & AuBrPerm_Mask) {
27076 + case AuBrPerm_RO:
27077 + case AuBrPerm_RR:
27078 + bad = val & AuBrWAttr_Mask;
27079 + val &= ~AuBrWAttr_Mask;
27081 + case AuBrPerm_RW:
27082 + bad = val & AuBrRAttr_Mask;
27083 + val &= ~AuBrRAttr_Mask;
27088 + * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
27089 + * does not treat it as an error, just warning.
27090 + * this is a tiny guard for the user operation.
27092 + if (val & AuBrAttr_UNPIN) {
27093 + bad |= AuBrAttr_UNPIN;
27094 + val &= ~AuBrAttr_UNPIN;
27097 + if (unlikely(bad)) {
27098 + sz = au_do_optstr_br_attr(&attr, bad);
27100 + pr_warn("ignored branch attribute %s\n", attr.a);
27107 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
27109 + au_br_perm_str_t attr;
27115 + p = au_optstr(&perm, brperm);
27116 + AuDebugOn(!p || !*p);
27118 + memcpy(q, p, sz + 1);
27121 + sz = au_do_optstr_br_attr(&attr, perm);
27124 + memcpy(q, attr.a, sz + 1);
27127 + AuDebugOn(strlen(str->a) >= sizeof(str->a));
27130 +/* ---------------------------------------------------------------------- */
27132 +static match_table_t udbalevel = {
27133 + {AuOpt_UDBA_REVAL, "reval"},
27134 + {AuOpt_UDBA_NONE, "none"},
27135 +#ifdef CONFIG_AUFS_HNOTIFY
27136 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
27137 +#ifdef CONFIG_AUFS_HFSNOTIFY
27138 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
27144 +int au_udba_val(char *str)
27146 + substring_t args[MAX_OPT_ARGS];
27148 + return match_token(str, udbalevel, args);
27151 +const char *au_optstr_udba(int udba)
27153 + return au_parser_pattern(udba, udbalevel);
27156 +/* ---------------------------------------------------------------------- */
27158 +static match_table_t au_wbr_create_policy = {
27159 + {AuWbrCreate_TDP, "tdp"},
27160 + {AuWbrCreate_TDP, "top-down-parent"},
27161 + {AuWbrCreate_RR, "rr"},
27162 + {AuWbrCreate_RR, "round-robin"},
27163 + {AuWbrCreate_MFS, "mfs"},
27164 + {AuWbrCreate_MFS, "most-free-space"},
27165 + {AuWbrCreate_MFSV, "mfs:%d"},
27166 + {AuWbrCreate_MFSV, "most-free-space:%d"},
27168 + /* top-down regardless the parent, and then mfs */
27169 + {AuWbrCreate_TDMFS, "tdmfs:%d"},
27170 + {AuWbrCreate_TDMFSV, "tdmfs:%d:%d"},
27172 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
27173 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
27174 + {AuWbrCreate_PMFS, "pmfs"},
27175 + {AuWbrCreate_PMFSV, "pmfs:%d"},
27176 + {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
27177 + {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
27182 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
27183 + struct au_opt_wbr_create *create)
27186 + unsigned long long ull;
27189 + if (!match_u64(arg, &ull))
27190 + create->mfsrr_watermark = ull;
27192 + pr_err("bad integer in %s\n", str);
27199 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
27200 + struct au_opt_wbr_create *create)
27205 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
27206 + create->mfs_second = n;
27208 + pr_err("bad integer in %s\n", str);
27215 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
27218 + substring_t args[MAX_OPT_ARGS];
27220 + err = match_token(str, au_wbr_create_policy, args);
27221 + create->wbr_create = err;
27223 + case AuWbrCreate_MFSRRV:
27224 + case AuWbrCreate_TDMFSV:
27225 + case AuWbrCreate_PMFSRRV:
27226 + e = au_wbr_mfs_wmark(&args[0], str, create);
27228 + e = au_wbr_mfs_sec(&args[1], str, create);
27232 + case AuWbrCreate_MFSRR:
27233 + case AuWbrCreate_TDMFS:
27234 + case AuWbrCreate_PMFSRR:
27235 + e = au_wbr_mfs_wmark(&args[0], str, create);
27236 + if (unlikely(e)) {
27241 + case AuWbrCreate_MFS:
27242 + case AuWbrCreate_PMFS:
27243 + create->mfs_second = AUFS_MFS_DEF_SEC;
27245 + case AuWbrCreate_MFSV:
27246 + case AuWbrCreate_PMFSV:
27247 + e = au_wbr_mfs_sec(&args[0], str, create);
27256 +const char *au_optstr_wbr_create(int wbr_create)
27258 + return au_parser_pattern(wbr_create, au_wbr_create_policy);
27261 +static match_table_t au_wbr_copyup_policy = {
27262 + {AuWbrCopyup_TDP, "tdp"},
27263 + {AuWbrCopyup_TDP, "top-down-parent"},
27264 + {AuWbrCopyup_BUP, "bup"},
27265 + {AuWbrCopyup_BUP, "bottom-up-parent"},
27266 + {AuWbrCopyup_BU, "bu"},
27267 + {AuWbrCopyup_BU, "bottom-up"},
27271 +int au_wbr_copyup_val(char *str)
27273 + substring_t args[MAX_OPT_ARGS];
27275 + return match_token(str, au_wbr_copyup_policy, args);
27278 +const char *au_optstr_wbr_copyup(int wbr_copyup)
27280 + return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
27283 +/* ---------------------------------------------------------------------- */
27285 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
27286 + aufs_bindex_t bindex)
27289 + struct au_opt_add *add = &opt->add;
27292 + add->bindex = bindex;
27293 + add->perm = AuBrPerm_RO;
27294 + add->pathname = opt_str;
27295 + p = strchr(opt_str, '=');
27299 + add->perm = au_br_perm_val(p);
27302 + err = vfsub_kern_path(add->pathname, AuOpt_LkupDirFlags, &add->path);
27305 + add->perm = AuBrPerm_RO;
27306 + if (au_test_fs_rr(add->path.dentry->d_sb))
27307 + add->perm = AuBrPerm_RR;
27308 + else if (!bindex && !(sb_flags & SB_RDONLY))
27309 + add->perm = AuBrPerm_RW;
27311 + opt->type = Opt_add;
27314 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
27321 +static int au_opt_wbr_create(struct super_block *sb,
27322 + struct au_opt_wbr_create *create)
27325 + struct au_sbinfo *sbinfo;
27327 + SiMustWriteLock(sb);
27329 + err = 1; /* handled */
27330 + sbinfo = au_sbi(sb);
27331 + if (sbinfo->si_wbr_create_ops->fin) {
27332 + err = sbinfo->si_wbr_create_ops->fin(sb);
27337 + sbinfo->si_wbr_create = create->wbr_create;
27338 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
27339 + switch (create->wbr_create) {
27340 + case AuWbrCreate_MFSRRV:
27341 + case AuWbrCreate_MFSRR:
27342 + case AuWbrCreate_TDMFS:
27343 + case AuWbrCreate_TDMFSV:
27344 + case AuWbrCreate_PMFSRR:
27345 + case AuWbrCreate_PMFSRRV:
27346 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
27348 + case AuWbrCreate_MFS:
27349 + case AuWbrCreate_MFSV:
27350 + case AuWbrCreate_PMFS:
27351 + case AuWbrCreate_PMFSV:
27352 + sbinfo->si_wbr_mfs.mfs_expire
27353 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
27357 + if (sbinfo->si_wbr_create_ops->init)
27358 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
27365 + * plus: processed without an error
27366 + * zero: unprocessed
27368 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
27369 + struct au_opts *opts)
27372 + struct au_sbinfo *sbinfo;
27374 + SiMustWriteLock(sb);
27376 + err = 1; /* handled */
27377 + sbinfo = au_sbi(sb);
27378 + switch (opt->type) {
27380 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27381 + sbinfo->si_mntflags |= opt->udba;
27382 + opts->given_udba |= opt->udba;
27387 + au_opt_set(sbinfo->si_mntflags, PLINK);
27389 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27390 + au_plink_put(sb, /*verbose*/1);
27391 + au_opt_clr(sbinfo->si_mntflags, PLINK);
27394 + case Opt_list_plink:
27395 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27396 + au_plink_list(sb);
27401 + au_opt_set(sbinfo->si_mntflags, DIO);
27402 + au_fset_opts(opts->flags, REFRESH_DYAOP);
27404 + au_opt_clr(sbinfo->si_mntflags, DIO);
27405 + au_fset_opts(opts->flags, REFRESH_DYAOP);
27409 + case Opt_fhsm_sec:
27410 + au_fhsm_set(sbinfo, opt->fhsm_second);
27413 + case Opt_diropq_a:
27414 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27416 + case Opt_diropq_w:
27417 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
27420 + case Opt_warn_perm:
27422 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
27424 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
27427 + case Opt_verbose:
27429 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
27431 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
27436 + au_opt_set(sbinfo->si_mntflags, SUM);
27438 + au_opt_clr(sbinfo->si_mntflags, SUM);
27439 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
27443 + au_opt_clr(sbinfo->si_mntflags, SUM);
27444 + au_opt_set(sbinfo->si_mntflags, SUM_W);
27447 + case Opt_wbr_create:
27448 + err = au_opt_wbr_create(sb, &opt->wbr_create);
27450 + case Opt_wbr_copyup:
27451 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
27452 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
27456 + sbinfo->si_dirwh = opt->dirwh;
27459 + case Opt_rdcache:
27460 + sbinfo->si_rdcache
27461 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
27464 + sbinfo->si_rdblk = opt->rdblk;
27467 + sbinfo->si_rdhash = opt->rdhash;
27472 + au_opt_set(sbinfo->si_mntflags, SHWH);
27474 + au_opt_clr(sbinfo->si_mntflags, SHWH);
27477 + case Opt_dirperm1:
27479 + au_opt_set(sbinfo->si_mntflags, DIRPERM1);
27481 + au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
27484 + case Opt_trunc_xino:
27486 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
27488 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
27491 + case Opt_trunc_xino_path:
27492 + case Opt_itrunc_xino:
27493 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex,
27499 + case Opt_trunc_xib:
27501 + au_fset_opts(opts->flags, TRUNC_XIB);
27503 + au_fclr_opts(opts->flags, TRUNC_XIB);
27509 + if (!au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27510 + err = au_dr_opt_set(sb);
27515 + au_opt_set(sbinfo->si_mntflags, DIRREN);
27517 + if (au_opt_test(sbinfo->si_mntflags, DIRREN)) {
27518 + err = au_dr_opt_clr(sb, au_ftest_opts(opts->flags,
27524 + au_opt_clr(sbinfo->si_mntflags, DIRREN);
27530 + sb->s_flags |= SB_POSIXACL;
27532 + sb->s_flags &= ~SB_POSIXACL;
27544 + * returns tri-state.
27545 + * plus: processed without an error
27546 + * zero: unprocessed
27549 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
27550 + struct au_opts *opts)
27552 + int err, do_refresh;
27555 + switch (opt->type) {
27557 + opt->add.bindex = au_sbbot(sb) + 1;
27558 + if (opt->add.bindex < 0)
27559 + opt->add.bindex = 0;
27561 + /* Always goto add, not fallthrough */
27562 + case Opt_prepend:
27563 + opt->add.bindex = 0;
27565 + add: /* indented label */
27567 + err = au_br_add(sb, &opt->add,
27568 + au_ftest_opts(opts->flags, REMOUNT));
27571 + au_fset_opts(opts->flags, REFRESH);
27577 + err = au_br_del(sb, &opt->del,
27578 + au_ftest_opts(opts->flags, REMOUNT));
27581 + au_fset_opts(opts->flags, TRUNC_XIB);
27582 + au_fset_opts(opts->flags, REFRESH);
27588 + err = au_br_mod(sb, &opt->mod,
27589 + au_ftest_opts(opts->flags, REMOUNT),
27594 + au_fset_opts(opts->flags, REFRESH);
27601 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
27602 + struct au_opt_xino **opt_xino,
27603 + struct au_opts *opts)
27608 + switch (opt->type) {
27610 + err = au_xino_set(sb, &opt->xino,
27611 + !!au_ftest_opts(opts->flags, REMOUNT));
27613 + *opt_xino = &opt->xino;
27617 + *opt_xino = (void *)-1;
27624 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
27625 + unsigned int pending)
27628 + aufs_bindex_t bindex, bbot;
27629 + unsigned char do_plink, skip, do_free, can_no_dreval;
27630 + struct au_branch *br;
27631 + struct au_wbr *wbr;
27632 + struct dentry *root, *dentry;
27633 + struct inode *dir, *h_dir;
27634 + struct au_sbinfo *sbinfo;
27635 + struct au_hinode *hdir;
27637 + SiMustAnyLock(sb);
27639 + sbinfo = au_sbi(sb);
27640 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
27642 + if (!(sb_flags & SB_RDONLY)) {
27643 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
27644 + pr_warn("first branch should be rw\n");
27645 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
27646 + pr_warn_once("shwh should be used with ro\n");
27649 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
27650 + && !au_opt_test(sbinfo->si_mntflags, XINO))
27651 + pr_warn_once("udba=*notify requires xino\n");
27653 + if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
27654 + pr_warn_once("dirperm1 breaks the protection"
27655 + " by the permission bits on the lower branch\n");
27659 + root = sb->s_root;
27660 + dir = d_inode(root);
27661 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
27662 + can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
27664 + bbot = au_sbbot(sb);
27665 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
27667 + h_dir = au_h_iptr(dir, bindex);
27668 + br = au_sbr(sb, bindex);
27670 + if ((br->br_perm & AuBrAttr_ICEX)
27671 + && !h_dir->i_op->listxattr)
27672 + br->br_perm &= ~AuBrAttr_ICEX;
27673 +#if 0 /* untested */
27674 + if ((br->br_perm & AuBrAttr_ICEX_SEC)
27675 + && (au_br_sb(br)->s_flags & SB_NOSEC))
27676 + br->br_perm &= ~AuBrAttr_ICEX_SEC;
27680 + wbr = br->br_wbr;
27682 + wbr_wh_read_lock(wbr);
27684 + if (!au_br_writable(br->br_perm)) {
27687 + || (!wbr->wbr_whbase
27688 + && !wbr->wbr_plink
27689 + && !wbr->wbr_orph));
27690 + } else if (!au_br_wh_linkable(br->br_perm)) {
27691 + /* skip = (!br->br_whbase && !br->br_orph); */
27692 + skip = (!wbr || !wbr->wbr_whbase);
27693 + if (skip && wbr) {
27695 + skip = !!wbr->wbr_plink;
27697 + skip = !wbr->wbr_plink;
27700 + /* skip = (br->br_whbase && br->br_ohph); */
27701 + skip = (wbr && wbr->wbr_whbase);
27704 + skip = !!wbr->wbr_plink;
27706 + skip = !wbr->wbr_plink;
27710 + wbr_wh_read_unlock(wbr);
27712 + if (can_no_dreval) {
27713 + dentry = br->br_path.dentry;
27714 + spin_lock(&dentry->d_lock);
27715 + if (dentry->d_flags &
27716 + (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
27717 + can_no_dreval = 0;
27718 + spin_unlock(&dentry->d_lock);
27721 + if (au_br_fhsm(br->br_perm)) {
27723 + AuDebugOn(!br->br_fhsm);
27729 + hdir = au_hi(dir, bindex);
27730 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
27732 + wbr_wh_write_lock(wbr);
27733 + err = au_wh_init(br, sb);
27735 + wbr_wh_write_unlock(wbr);
27736 + au_hn_inode_unlock(hdir);
27738 + if (!err && do_free) {
27739 + au_kfree_rcu(wbr);
27740 + br->br_wbr = NULL;
27744 + if (can_no_dreval)
27745 + au_fset_si(sbinfo, NO_DREVAL);
27747 + au_fclr_si(sbinfo, NO_DREVAL);
27750 + au_fset_si(sbinfo, FHSM);
27751 + for (bindex = bbot; bindex >= 0; bindex--) {
27752 + br = au_sbr(sb, bindex);
27753 + if (au_br_fhsm(br->br_perm)) {
27754 + au_fhsm_set_bottom(sb, bindex);
27759 + au_fclr_si(sbinfo, FHSM);
27760 + au_fhsm_set_bottom(sb, -1);
27766 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
27769 + unsigned int tmp;
27770 + aufs_bindex_t bindex, bbot;
27771 + struct au_opt *opt;
27772 + struct au_opt_xino *opt_xino, xino;
27773 + struct au_sbinfo *sbinfo;
27774 + struct au_branch *br;
27775 + struct inode *dir;
27777 + SiMustWriteLock(sb);
27782 + while (err >= 0 && opt->type != Opt_tail)
27783 + err = au_opt_simple(sb, opt++, opts);
27786 + else if (unlikely(err < 0))
27789 + /* disable xino and udba temporary */
27790 + sbinfo = au_sbi(sb);
27791 + tmp = sbinfo->si_mntflags;
27792 + au_opt_clr(sbinfo->si_mntflags, XINO);
27793 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
27796 + while (err >= 0 && opt->type != Opt_tail)
27797 + err = au_opt_br(sb, opt++, opts);
27800 + else if (unlikely(err < 0))
27803 + bbot = au_sbbot(sb);
27804 + if (unlikely(bbot < 0)) {
27806 + pr_err("no branches\n");
27810 + if (au_opt_test(tmp, XINO))
27811 + au_opt_set(sbinfo->si_mntflags, XINO);
27813 + while (!err && opt->type != Opt_tail)
27814 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
27815 + if (unlikely(err))
27818 + err = au_opts_verify(sb, sb->s_flags, tmp);
27819 + if (unlikely(err))
27822 + /* restore xino */
27823 + if (au_opt_test(tmp, XINO) && !opt_xino) {
27824 + xino.file = au_xino_def(sb);
27825 + err = PTR_ERR(xino.file);
27826 + if (IS_ERR(xino.file))
27829 + err = au_xino_set(sb, &xino, /*remount*/0);
27831 + if (unlikely(err))
27835 + /* restore udba */
27836 + tmp &= AuOptMask_UDBA;
27837 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27838 + sbinfo->si_mntflags |= tmp;
27839 + bbot = au_sbbot(sb);
27840 + for (bindex = 0; bindex <= bbot; bindex++) {
27841 + br = au_sbr(sb, bindex);
27842 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
27843 + if (unlikely(err))
27844 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
27846 + /* go on even if err */
27848 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
27849 + dir = d_inode(sb->s_root);
27850 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
27857 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
27860 + unsigned char no_dreval;
27861 + struct inode *dir;
27862 + struct au_opt_xino *opt_xino;
27863 + struct au_opt *opt;
27864 + struct au_sbinfo *sbinfo;
27866 + SiMustWriteLock(sb);
27868 + err = au_dr_opt_flush(sb);
27869 + if (unlikely(err))
27871 + au_fset_opts(opts->flags, DR_FLUSHED);
27873 + dir = d_inode(sb->s_root);
27874 + sbinfo = au_sbi(sb);
27877 + while (err >= 0 && opt->type != Opt_tail) {
27878 + err = au_opt_simple(sb, opt, opts);
27880 + err = au_opt_br(sb, opt, opts);
27882 + err = au_opt_xino(sb, opt, &opt_xino, opts);
27888 + /* go on even err */
27890 + no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
27891 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
27892 + if (unlikely(rerr && !err))
27895 + if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
27896 + au_fset_opts(opts->flags, REFRESH_IDOP);
27898 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
27899 + rerr = au_xib_trunc(sb);
27900 + if (unlikely(rerr && !err))
27904 + /* will be handled by the caller */
27905 + if (!au_ftest_opts(opts->flags, REFRESH)
27906 + && (opts->given_udba
27907 + || au_opt_test(sbinfo->si_mntflags, XINO)
27908 + || au_ftest_opts(opts->flags, REFRESH_IDOP)
27910 + au_fset_opts(opts->flags, REFRESH);
27912 + AuDbg("status 0x%x\n", opts->flags);
27918 +/* ---------------------------------------------------------------------- */
27920 +unsigned int au_opt_udba(struct super_block *sb)
27922 + return au_mntflags(sb) & AuOptMask_UDBA;
27924 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
27925 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
27926 +++ linux/fs/aufs/opts.h 2022-11-05 23:02:18.969222617 +0100
27928 +/* SPDX-License-Identifier: GPL-2.0 */
27930 + * Copyright (C) 2005-2022 Junjiro R. Okajima
27932 + * This program is free software; you can redistribute it and/or modify
27933 + * it under the terms of the GNU General Public License as published by
27934 + * the Free Software Foundation; either version 2 of the License, or
27935 + * (at your option) any later version.
27937 + * This program is distributed in the hope that it will be useful,
27938 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27939 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27940 + * GNU General Public License for more details.
27942 + * You should have received a copy of the GNU General Public License
27943 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27947 + * mount options/flags
27950 +#ifndef __AUFS_OPTS_H__
27951 +#define __AUFS_OPTS_H__
27955 +#include <linux/fs_parser.h>
27956 +#include <linux/namei.h>
27957 +#include <linux/path.h>
27961 + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
27962 + Opt_idel, Opt_imod,
27963 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
27964 + Opt_xino, Opt_noxino,
27965 + Opt_trunc_xino, Opt_trunc_xino_v,
27966 + Opt_trunc_xino_path, Opt_itrunc_xino,
27969 + Opt_plink, Opt_list_plink,
27972 + Opt_diropq, Opt_diropq_a, Opt_diropq_w,
27974 + Opt_wbr_copyup, Opt_wbr_create,
27976 + Opt_verbose, Opt_noverbose,
27977 + Opt_sum, Opt_wsum,
27981 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
27984 +/* ---------------------------------------------------------------------- */
27987 +#define AuOpt_XINO 1 /* external inode number bitmap
27988 + and translation table */
27989 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
27990 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
27991 +#define AuOpt_UDBA_REVAL (1 << 3)
27992 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
27993 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
27994 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
27995 +#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
27997 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
27998 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
27999 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
28000 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
28001 +#define AuOpt_VERBOSE (1 << 13) /* print the cause of error */
28002 +#define AuOpt_DIO (1 << 14) /* direct io */
28003 +#define AuOpt_DIRREN (1 << 15) /* directory rename */
28005 +#ifndef CONFIG_AUFS_HNOTIFY
28006 +#undef AuOpt_UDBA_HNOTIFY
28007 +#define AuOpt_UDBA_HNOTIFY 0
28009 +#ifndef CONFIG_AUFS_DIRREN
28010 +#undef AuOpt_DIRREN
28011 +#define AuOpt_DIRREN 0
28013 +#ifndef CONFIG_AUFS_SHWH
28015 +#define AuOpt_SHWH 0
28018 +#define AuOpt_Def (AuOpt_XINO \
28019 + | AuOpt_UDBA_REVAL \
28021 + /* | AuOpt_DIRPERM1 */ \
28022 + | AuOpt_WARN_PERM)
28023 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
28024 + | AuOpt_UDBA_REVAL \
28025 + | AuOpt_UDBA_HNOTIFY)
28027 +#define AuOpt_LkupDirFlags (LOOKUP_FOLLOW | LOOKUP_DIRECTORY)
28029 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
28030 +#define au_opt_set(flags, name) do { \
28031 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
28032 + ((flags) |= AuOpt_##name); \
28034 +#define au_opt_set_udba(flags, name) do { \
28035 + (flags) &= ~AuOptMask_UDBA; \
28036 + ((flags) |= AuOpt_##name); \
28038 +#define au_opt_clr(flags, name) do { \
28039 + ((flags) &= ~AuOpt_##name); \
28042 +static inline unsigned int au_opts_plink(unsigned int mntflags)
28044 +#ifdef CONFIG_PROC_FS
28047 + return mntflags & ~AuOpt_PLINK;
28051 +/* ---------------------------------------------------------------------- */
28053 +/* policies to select one among multiple writable branches */
28055 + AuWbrCreate_TDP, /* top down parent */
28056 + AuWbrCreate_RR, /* round robin */
28057 + AuWbrCreate_MFS, /* most free space */
28058 + AuWbrCreate_MFSV, /* mfs with seconds */
28059 + AuWbrCreate_MFSRR, /* mfs then rr */
28060 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
28061 + AuWbrCreate_TDMFS, /* top down regardless parent and mfs */
28062 + AuWbrCreate_TDMFSV, /* top down regardless parent and mfs */
28063 + AuWbrCreate_PMFS, /* parent and mfs */
28064 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
28065 + AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
28066 + AuWbrCreate_PMFSRRV, /* plus seconds */
28068 + AuWbrCreate_Def = AuWbrCreate_TDP
28072 + AuWbrCopyup_TDP, /* top down parent */
28073 + AuWbrCopyup_BUP, /* bottom up parent */
28074 + AuWbrCopyup_BU, /* bottom up */
28076 + AuWbrCopyup_Def = AuWbrCopyup_TDP
28079 +/* ---------------------------------------------------------------------- */
28083 +struct au_opt_add {
28084 + aufs_bindex_t bindex;
28087 + struct path path;
28090 +struct au_opt_del {
28092 + struct path h_path;
28095 +struct au_opt_mod {
28098 + struct dentry *h_root;
28101 +struct au_opt_xino {
28103 + struct file *file;
28106 +struct au_opt_xino_itrunc {
28107 + aufs_bindex_t bindex;
28110 +struct au_opt_wbr_create {
28113 + unsigned long long mfsrr_watermark;
28119 + struct au_opt_xino xino;
28120 + struct au_opt_xino_itrunc xino_itrunc;
28121 + struct au_opt_add add;
28122 + struct au_opt_del del;
28123 + struct au_opt_mod mod;
28126 + unsigned int rdblk;
28127 + unsigned int rdhash;
28129 + struct au_opt_wbr_create wbr_create;
28131 + unsigned int fhsm_second;
28132 + bool tf; /* generic flag, true or false */
28137 +#define AuOpts_REMOUNT 1
28138 +#define AuOpts_REFRESH (1 << 1)
28139 +#define AuOpts_TRUNC_XIB (1 << 2)
28140 +#define AuOpts_REFRESH_DYAOP (1 << 3)
28141 +#define AuOpts_REFRESH_IDOP (1 << 4)
28142 +#define AuOpts_DR_FLUSHED (1 << 5)
28143 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
28144 +#define au_fset_opts(flags, name) \
28145 + do { (flags) |= AuOpts_##name; } while (0)
28146 +#define au_fclr_opts(flags, name) \
28147 + do { (flags) &= ~AuOpts_##name; } while (0)
28149 +#ifndef CONFIG_AUFS_DIRREN
28150 +#undef AuOpts_DR_FLUSHED
28151 +#define AuOpts_DR_FLUSHED 0
28155 + struct au_opt *opt;
28158 + unsigned int given_udba;
28159 + unsigned int flags;
28160 + unsigned long sb_flags;
28163 +/* ---------------------------------------------------------------------- */
28166 +int au_br_perm_val(char *perm);
28167 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
28168 +int au_udba_val(char *str);
28169 +const char *au_optstr_udba(int udba);
28170 +int au_wbr_create_val(char *str, struct au_opt_wbr_create *create);
28171 +const char *au_optstr_wbr_create(int wbr_create);
28172 +int au_wbr_copyup_val(char *str);
28173 +const char *au_optstr_wbr_copyup(int wbr_copyup);
28175 +int au_opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
28176 + aufs_bindex_t bindex);
28177 +struct super_block;
28178 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
28179 + unsigned int pending);
28180 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
28181 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
28183 +unsigned int au_opt_udba(struct super_block *sb);
28186 +int aufs_fsctx_init(struct fs_context *fc);
28187 +extern const struct fs_parameter_spec aufs_fsctx_paramspec[];
28189 +#endif /* __KERNEL__ */
28190 +#endif /* __AUFS_OPTS_H__ */
28191 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
28192 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
28193 +++ linux/fs/aufs/plink.c 2022-11-05 23:02:18.969222617 +0100
28195 +// SPDX-License-Identifier: GPL-2.0
28197 + * Copyright (C) 2005-2022 Junjiro R. Okajima
28199 + * This program is free software; you can redistribute it and/or modify
28200 + * it under the terms of the GNU General Public License as published by
28201 + * the Free Software Foundation; either version 2 of the License, or
28202 + * (at your option) any later version.
28204 + * This program is distributed in the hope that it will be useful,
28205 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28206 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28207 + * GNU General Public License for more details.
28209 + * You should have received a copy of the GNU General Public License
28210 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28220 + * the pseudo-link maintenance mode.
28221 + * during a user process maintains the pseudo-links,
28222 + * prohibit adding a new plink and branch manipulation.
28226 + * For entry functions which will handle plink, and i_mutex is already held
28228 + * They cannot wait and should return an error at once.
28229 + * Callers has to check the error.
28231 + * For entry functions which will handle plink, but i_mutex is not held
28233 + * They can wait the plink maintenance mode to finish.
28235 + * They behave like F_SETLK and F_SETLKW.
28236 + * If the caller never handle plink, then both flags are unnecessary.
28239 +int au_plink_maint(struct super_block *sb, int flags)
28243 + struct task_struct *parent, *prev;
28244 + struct au_sbinfo *sbi;
28246 + SiMustAnyLock(sb);
28249 + if (!au_opt_test(au_mntflags(sb), PLINK))
28252 + sbi = au_sbi(sb);
28253 + pid = sbi->si_plink_maint_pid;
28254 + if (!pid || pid == current->pid)
28257 + /* todo: it highly depends upon /sbin/mount.aufs */
28259 + parent = current;
28263 + parent = rcu_dereference(parent->real_parent);
28264 + if (parent == prev)
28266 + ppid = task_pid_vnr(parent);
28267 + if (pid == ppid) {
28268 + rcu_read_unlock();
28273 + rcu_read_unlock();
28275 + if (au_ftest_lock(flags, NOPLMW)) {
28276 + /* if there is no i_mutex lock in VFS, we don't need to wait */
28277 + /* AuDebugOn(!lockdep_depth(current)); */
28278 + while (sbi->si_plink_maint_pid) {
28279 + si_read_unlock(sb);
28280 + /* gave up wake_up_bit() */
28281 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
28283 + if (au_ftest_lock(flags, FLUSH))
28284 + au_nwt_flush(&sbi->si_nowait);
28285 + si_noflush_read_lock(sb);
28287 + } else if (au_ftest_lock(flags, NOPLM)) {
28288 + AuDbg("ppid %d, pid %d\n", ppid, pid);
28296 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
28298 + spin_lock(&sbinfo->si_plink_maint_lock);
28299 + sbinfo->si_plink_maint_pid = 0;
28300 + spin_unlock(&sbinfo->si_plink_maint_lock);
28301 + wake_up_all(&sbinfo->si_plink_wq);
28304 +int au_plink_maint_enter(struct super_block *sb)
28307 + struct au_sbinfo *sbinfo;
28310 + sbinfo = au_sbi(sb);
28311 + /* make sure i am the only one in this fs */
28312 + si_write_lock(sb, AuLock_FLUSH);
28313 + if (au_opt_test(au_mntflags(sb), PLINK)) {
28314 + spin_lock(&sbinfo->si_plink_maint_lock);
28315 + if (!sbinfo->si_plink_maint_pid)
28316 + sbinfo->si_plink_maint_pid = current->pid;
28319 + spin_unlock(&sbinfo->si_plink_maint_lock);
28321 + si_write_unlock(sb);
28326 +/* ---------------------------------------------------------------------- */
28328 +#ifdef CONFIG_AUFS_DEBUG
28329 +void au_plink_list(struct super_block *sb)
28332 + struct au_sbinfo *sbinfo;
28333 + struct hlist_bl_head *hbl;
28334 + struct hlist_bl_node *pos;
28335 + struct au_icntnr *icntnr;
28337 + SiMustAnyLock(sb);
28339 + sbinfo = au_sbi(sb);
28340 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28341 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28343 + for (i = 0; i < AuPlink_NHASH; i++) {
28344 + hbl = sbinfo->si_plink + i;
28345 + hlist_bl_lock(hbl);
28346 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28347 + AuDbg("%lu\n", icntnr->vfs_inode.i_ino);
28348 + hlist_bl_unlock(hbl);
28353 +/* is the inode pseudo-linked? */
28354 +int au_plink_test(struct inode *inode)
28357 + struct au_sbinfo *sbinfo;
28358 + struct hlist_bl_head *hbl;
28359 + struct hlist_bl_node *pos;
28360 + struct au_icntnr *icntnr;
28362 + sbinfo = au_sbi(inode->i_sb);
28363 + AuRwMustAnyLock(&sbinfo->si_rwsem);
28364 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
28365 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28368 + i = au_plink_hash(inode->i_ino);
28369 + hbl = sbinfo->si_plink + i;
28370 + hlist_bl_lock(hbl);
28371 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
28372 + if (&icntnr->vfs_inode == inode) {
28376 + hlist_bl_unlock(hbl);
28380 +/* ---------------------------------------------------------------------- */
28383 + * generate a name for plink.
28384 + * the file will be stored under AUFS_WH_PLINKDIR.
28386 +/* 20 is max digits length of ulong 64 */
28387 +#define PLINK_NAME_LEN ((20 + 1) * 2)
28389 +static int plink_name(char *name, int len, struct inode *inode,
28390 + aufs_bindex_t bindex)
28393 + struct inode *h_inode;
28395 + h_inode = au_h_iptr(inode, bindex);
28396 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
28400 +struct au_do_plink_lkup_args {
28401 + struct dentry **errp;
28402 + struct qstr *tgtname;
28403 + struct path *h_ppath;
28406 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
28407 + struct path *h_ppath)
28409 + struct dentry *h_dentry;
28410 + struct inode *h_inode;
28412 + h_inode = d_inode(h_ppath->dentry);
28413 + inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2);
28414 + h_dentry = vfsub_lkup_one(tgtname, h_ppath);
28415 + inode_unlock_shared(h_inode);
28420 +static void au_call_do_plink_lkup(void *args)
28422 + struct au_do_plink_lkup_args *a = args;
28423 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_ppath);
28426 +/* lookup the plink-ed @inode under the branch at @bindex */
28427 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
28429 + struct dentry *h_dentry;
28430 + struct au_branch *br;
28431 + struct path h_ppath;
28433 + char a[PLINK_NAME_LEN];
28434 + struct qstr tgtname = QSTR_INIT(a, 0);
28436 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
28438 + br = au_sbr(inode->i_sb, bindex);
28439 + h_ppath.dentry = br->br_wbr->wbr_plink;
28440 + h_ppath.mnt = au_br_mnt(br);
28441 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28443 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28444 + struct au_do_plink_lkup_args args = {
28445 + .errp = &h_dentry,
28446 + .tgtname = &tgtname,
28447 + .h_ppath = &h_ppath
28450 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
28451 + if (unlikely(wkq_err))
28452 + h_dentry = ERR_PTR(wkq_err);
28454 + h_dentry = au_do_plink_lkup(&tgtname, &h_ppath);
28459 +/* create a pseudo-link */
28460 +static int do_whplink(struct qstr *tgt, struct path *h_ppath,
28461 + struct dentry *h_dentry)
28464 + struct path h_path;
28465 + struct inode *h_dir, *delegated;
28467 + h_dir = d_inode(h_ppath->dentry);
28468 + inode_lock_nested(h_dir, AuLsc_I_CHILD2);
28469 + h_path.mnt = h_ppath->mnt;
28471 + h_path.dentry = vfsub_lkup_one(tgt, h_ppath);
28472 + err = PTR_ERR(h_path.dentry);
28473 + if (IS_ERR(h_path.dentry))
28477 + /* wh.plink dir is not monitored */
28478 + /* todo: is it really safe? */
28479 + if (d_is_positive(h_path.dentry)
28480 + && d_inode(h_path.dentry) != d_inode(h_dentry)) {
28481 + delegated = NULL;
28482 + err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
28483 + if (unlikely(err == -EWOULDBLOCK)) {
28484 + pr_warn("cannot retry for NFSv4 delegation"
28485 + " for an internal unlink\n");
28488 + dput(h_path.dentry);
28489 + h_path.dentry = NULL;
28493 + if (!err && d_is_negative(h_path.dentry)) {
28494 + delegated = NULL;
28495 + err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
28496 + if (unlikely(err == -EWOULDBLOCK)) {
28497 + pr_warn("cannot retry for NFSv4 delegation"
28498 + " for an internal link\n");
28502 + dput(h_path.dentry);
28505 + inode_unlock(h_dir);
28509 +struct do_whplink_args {
28511 + struct qstr *tgt;
28512 + struct path *h_ppath;
28513 + struct dentry *h_dentry;
28516 +static void call_do_whplink(void *args)
28518 + struct do_whplink_args *a = args;
28519 + *a->errp = do_whplink(a->tgt, a->h_ppath, a->h_dentry);
28522 +static int whplink(struct dentry *h_dentry, struct inode *inode,
28523 + aufs_bindex_t bindex)
28525 + int err, wkq_err;
28526 + struct au_branch *br;
28527 + struct au_wbr *wbr;
28528 + struct path h_ppath;
28529 + char a[PLINK_NAME_LEN];
28530 + struct qstr tgtname = QSTR_INIT(a, 0);
28532 + br = au_sbr(inode->i_sb, bindex);
28533 + wbr = br->br_wbr;
28534 + h_ppath.dentry = wbr->wbr_plink;
28535 + h_ppath.mnt = au_br_mnt(br);
28536 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
28538 + /* always superio. */
28539 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
28540 + struct do_whplink_args args = {
28543 + .h_ppath = &h_ppath,
28544 + .h_dentry = h_dentry
28546 + wkq_err = au_wkq_wait(call_do_whplink, &args);
28547 + if (unlikely(wkq_err))
28550 + err = do_whplink(&tgtname, &h_ppath, h_dentry);
28556 + * create a new pseudo-link for @h_dentry on @bindex.
28557 + * the linked inode is held in aufs @inode.
28559 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
28560 + struct dentry *h_dentry)
28562 + struct super_block *sb;
28563 + struct au_sbinfo *sbinfo;
28564 + struct hlist_bl_head *hbl;
28565 + struct hlist_bl_node *pos;
28566 + struct au_icntnr *icntnr;
28567 + int found, err, cnt, i;
28569 + sb = inode->i_sb;
28570 + sbinfo = au_sbi(sb);
28571 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28572 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28574 + found = au_plink_test(inode);
28578 + i = au_plink_hash(inode->i_ino);
28579 + hbl = sbinfo->si_plink + i;
28582 + hlist_bl_lock(hbl);
28583 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink) {
28584 + if (&icntnr->vfs_inode == inode) {
28590 + icntnr = container_of(inode, struct au_icntnr, vfs_inode);
28591 + hlist_bl_add_head(&icntnr->plink, hbl);
28593 + hlist_bl_unlock(hbl);
28595 + cnt = au_hbl_count(hbl);
28596 +#define msg "unexpectedly unbalanced or too many pseudo-links"
28597 + if (cnt > AUFS_PLINK_WARN)
28598 + AuWarn1(msg ", %d\n", cnt);
28600 + err = whplink(h_dentry, inode, bindex);
28601 + if (unlikely(err)) {
28602 + pr_warn("err %d, damaged pseudo link.\n", err);
28603 + au_hbl_del(&icntnr->plink, hbl);
28604 + iput(&icntnr->vfs_inode);
28607 + iput(&icntnr->vfs_inode);
28610 +/* free all plinks */
28611 +void au_plink_put(struct super_block *sb, int verbose)
28614 + struct au_sbinfo *sbinfo;
28615 + struct hlist_bl_head *hbl;
28616 + struct hlist_bl_node *pos, *tmp;
28617 + struct au_icntnr *icntnr;
28619 + SiMustWriteLock(sb);
28621 + sbinfo = au_sbi(sb);
28622 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28623 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28625 + /* no spin_lock since sbinfo is write-locked */
28627 + for (i = 0; i < AuPlink_NHASH; i++) {
28628 + hbl = sbinfo->si_plink + i;
28629 + if (!warned && verbose && !hlist_bl_empty(hbl)) {
28630 + pr_warn("pseudo-link is not flushed");
28633 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink)
28634 + iput(&icntnr->vfs_inode);
28635 + INIT_HLIST_BL_HEAD(hbl);
28639 +void au_plink_clean(struct super_block *sb, int verbose)
28641 + struct dentry *root;
28643 + root = sb->s_root;
28644 + aufs_write_lock(root);
28645 + if (au_opt_test(au_mntflags(sb), PLINK))
28646 + au_plink_put(sb, verbose);
28647 + aufs_write_unlock(root);
28650 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
28653 + aufs_bindex_t btop, bbot, bindex;
28656 + btop = au_ibtop(inode);
28657 + bbot = au_ibbot(inode);
28659 + for (bindex = btop; bindex <= bbot; bindex++) {
28660 + if (!au_h_iptr(inode, bindex)
28661 + || au_ii_br_id(inode, bindex) != br_id)
28663 + au_set_h_iptr(inode, bindex, NULL, 0);
28668 + for (bindex = btop; bindex <= bbot; bindex++)
28669 + if (au_h_iptr(inode, bindex)) {
28679 +/* free the plinks on a branch specified by @br_id */
28680 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
28682 + struct au_sbinfo *sbinfo;
28683 + struct hlist_bl_head *hbl;
28684 + struct hlist_bl_node *pos, *tmp;
28685 + struct au_icntnr *icntnr;
28686 + struct inode *inode;
28689 + SiMustWriteLock(sb);
28691 + sbinfo = au_sbi(sb);
28692 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
28693 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
28695 + /* no bit_lock since sbinfo is write-locked */
28696 + for (i = 0; i < AuPlink_NHASH; i++) {
28697 + hbl = sbinfo->si_plink + i;
28698 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) {
28699 + inode = au_igrab(&icntnr->vfs_inode);
28700 + ii_write_lock_child(inode);
28701 + do_put = au_plink_do_half_refresh(inode, br_id);
28703 + hlist_bl_del(&icntnr->plink);
28706 + ii_write_unlock(inode);
28711 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
28712 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
28713 +++ linux/fs/aufs/poll.c 2022-11-05 23:02:18.969222617 +0100
28715 +// SPDX-License-Identifier: GPL-2.0
28717 + * Copyright (C) 2005-2022 Junjiro R. Okajima
28719 + * This program is free software; you can redistribute it and/or modify
28720 + * it under the terms of the GNU General Public License as published by
28721 + * the Free Software Foundation; either version 2 of the License, or
28722 + * (at your option) any later version.
28724 + * This program is distributed in the hope that it will be useful,
28725 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28726 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28727 + * GNU General Public License for more details.
28729 + * You should have received a copy of the GNU General Public License
28730 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28735 + * There is only one filesystem which implements ->poll operation, currently.
28740 +__poll_t aufs_poll(struct file *file, struct poll_table_struct *pt)
28743 + struct file *h_file;
28744 + struct super_block *sb;
28746 + /* We should pretend an error happened. */
28747 + mask = EPOLLERR /* | EPOLLIN | EPOLLOUT */;
28748 + sb = file->f_path.dentry->d_sb;
28749 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
28751 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
28752 + if (IS_ERR(h_file)) {
28753 + AuDbg("h_file %ld\n", PTR_ERR(h_file));
28757 + mask = vfs_poll(h_file, pt);
28758 + fput(h_file); /* instead of au_read_post() */
28761 + si_read_unlock(sb);
28762 + if (mask & EPOLLERR)
28763 + AuDbg("mask 0x%x\n", mask);
28766 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
28767 --- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
28768 +++ linux/fs/aufs/posix_acl.c 2023-04-24 08:13:40.969576115 +0200
28770 +// SPDX-License-Identifier: GPL-2.0
28772 + * Copyright (C) 2014-2022 Junjiro R. Okajima
28774 + * This program is free software; you can redistribute it and/or modify
28775 + * it under the terms of the GNU General Public License as published by
28776 + * the Free Software Foundation; either version 2 of the License, or
28777 + * (at your option) any later version.
28779 + * This program is distributed in the hope that it will be useful,
28780 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28781 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28782 + * GNU General Public License for more details.
28784 + * You should have received a copy of the GNU General Public License
28785 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28789 + * posix acl operations
28792 +#include <linux/fs.h>
28795 +struct posix_acl *aufs_get_inode_acl(struct inode *inode, int type, bool rcu)
28797 + struct posix_acl *acl;
28799 + aufs_bindex_t bindex;
28800 + struct inode *h_inode;
28801 + struct super_block *sb;
28803 + acl = ERR_PTR(-ECHILD);
28808 + sb = inode->i_sb;
28809 + si_read_lock(sb, AuLock_FLUSH);
28810 + ii_read_lock_child(inode);
28811 + if (!(sb->s_flags & SB_POSIXACL))
28814 + bindex = au_ibtop(inode);
28815 + h_inode = au_h_iptr(inode, bindex);
28816 + if (unlikely(!h_inode
28817 + || ((h_inode->i_mode & S_IFMT)
28818 + != (inode->i_mode & S_IFMT)))) {
28819 + err = au_busy_or_stale();
28820 + acl = ERR_PTR(err);
28824 + /* always topmost only */
28825 + acl = get_inode_acl(h_inode, type);
28827 + forget_cached_acl(inode, type);
28829 + set_cached_acl(inode, type, acl);
28832 + ii_read_unlock(inode);
28833 + si_read_unlock(sb);
28836 + AuTraceErrPtr(acl);
28840 +struct posix_acl *aufs_get_acl(struct mnt_idmap *idmap,
28841 + struct dentry *dentry, int type)
28843 + struct posix_acl *acl;
28844 + struct inode *inode;
28846 + inode = d_inode(dentry);
28847 + acl = aufs_get_inode_acl(inode, type, /*rcu*/false);
28852 +int aufs_set_acl(struct mnt_idmap *idmap, struct dentry *dentry,
28853 + struct posix_acl *acl, int type)
28857 + struct inode *inode;
28858 + struct au_sxattr arg = {
28859 + .type = AU_ACL_SET,
28866 + inode = d_inode(dentry);
28867 + IMustLock(inode);
28869 + ssz = au_sxattr(dentry, inode, &arg);
28870 + /* forget even it if succeeds since the branch might set differently */
28871 + forget_cached_acl(inode, type);
28878 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
28879 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
28880 +++ linux/fs/aufs/procfs.c 2022-11-05 23:02:18.969222617 +0100
28882 +// SPDX-License-Identifier: GPL-2.0
28884 + * Copyright (C) 2010-2022 Junjiro R. Okajima
28886 + * This program is free software; you can redistribute it and/or modify
28887 + * it under the terms of the GNU General Public License as published by
28888 + * the Free Software Foundation; either version 2 of the License, or
28889 + * (at your option) any later version.
28891 + * This program is distributed in the hope that it will be useful,
28892 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28893 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28894 + * GNU General Public License for more details.
28896 + * You should have received a copy of the GNU General Public License
28897 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28901 + * procfs interfaces
28904 +#include <linux/proc_fs.h>
28907 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
28909 + struct au_sbinfo *sbinfo;
28911 + sbinfo = file->private_data;
28913 + au_plink_maint_leave(sbinfo);
28914 + kobject_put(&sbinfo->si_kobj);
28920 +static void au_procfs_plm_write_clean(struct file *file)
28922 + struct au_sbinfo *sbinfo;
28924 + sbinfo = file->private_data;
28926 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
28929 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
28932 + struct super_block *sb;
28933 + struct au_sbinfo *sbinfo;
28934 + struct hlist_bl_node *pos;
28937 + if (unlikely(file->private_data))
28941 + /* don't use au_sbilist_lock() here */
28942 + hlist_bl_lock(&au_sbilist);
28943 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
28944 + if (id == sysaufs_si_id(sbinfo)) {
28945 + if (kobject_get_unless_zero(&sbinfo->si_kobj))
28946 + sb = sbinfo->si_sb;
28949 + hlist_bl_unlock(&au_sbilist);
28952 + if (unlikely(!sb))
28955 + err = au_plink_maint_enter(sb);
28957 + /* keep kobject_get() */
28958 + file->private_data = sbinfo;
28960 + kobject_put(&sbinfo->si_kobj);
28966 + * Accept a valid "si=xxxx" only.
28967 + * Once it is accepted successfully, accept "clean" too.
28969 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
28970 + size_t count, loff_t *ppos)
28973 + unsigned long id;
28974 + /* last newline is allowed */
28975 + char buf[3 + sizeof(unsigned long) * 2 + 1];
28978 + if (unlikely(!capable(CAP_SYS_ADMIN)))
28982 + if (unlikely(count > sizeof(buf)))
28985 + err = copy_from_user(buf, ubuf, count);
28986 + if (unlikely(err)) {
28993 + if (!strcmp("clean", buf)) {
28994 + au_procfs_plm_write_clean(file);
28995 + goto out_success;
28996 + } else if (unlikely(strncmp("si=", buf, 3)))
28999 + err = kstrtoul(buf + 3, 16, &id);
29000 + if (unlikely(err))
29003 + err = au_procfs_plm_write_si(file, id);
29004 + if (unlikely(err))
29008 + err = count; /* success */
29013 +static const struct proc_ops au_procfs_plm_op = {
29014 + .proc_write = au_procfs_plm_write,
29015 + .proc_release = au_procfs_plm_release
29018 +/* ---------------------------------------------------------------------- */
29020 +static struct proc_dir_entry *au_procfs_dir;
29022 +void au_procfs_fin(void)
29024 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
29025 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29028 +int __init au_procfs_init(void)
29031 + struct proc_dir_entry *entry;
29034 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
29035 + if (unlikely(!au_procfs_dir))
29038 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | 0200,
29039 + au_procfs_dir, &au_procfs_plm_op);
29040 + if (unlikely(!entry))
29044 + goto out; /* success */
29048 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
29052 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
29053 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
29054 +++ linux/fs/aufs/rdu.c 2022-12-17 09:21:34.799855195 +0100
29056 +// SPDX-License-Identifier: GPL-2.0
29058 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29060 + * This program is free software; you can redistribute it and/or modify
29061 + * it under the terms of the GNU General Public License as published by
29062 + * the Free Software Foundation; either version 2 of the License, or
29063 + * (at your option) any later version.
29065 + * This program is distributed in the hope that it will be useful,
29066 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29067 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29068 + * GNU General Public License for more details.
29070 + * You should have received a copy of the GNU General Public License
29071 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29075 + * readdir in userspace.
29078 +#include <linux/compat.h>
29079 +#include <linux/fs_stack.h>
29080 +#include <linux/security.h>
29083 +/* bits for struct aufs_rdu.flags */
29084 +#define AuRdu_CALLED 1
29085 +#define AuRdu_CONT (1 << 1)
29086 +#define AuRdu_FULL (1 << 2)
29087 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
29088 +#define au_fset_rdu(flags, name) \
29089 + do { (flags) |= AuRdu_##name; } while (0)
29090 +#define au_fclr_rdu(flags, name) \
29091 + do { (flags) &= ~AuRdu_##name; } while (0)
29093 +struct au_rdu_arg {
29094 + struct dir_context ctx;
29095 + struct aufs_rdu *rdu;
29096 + union au_rdu_ent_ul ent;
29097 + unsigned long end;
29099 + struct super_block *sb;
29103 +static bool au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
29104 + loff_t offset, u64 h_ino, unsigned int d_type)
29107 + struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
29108 + struct aufs_rdu *rdu = arg->rdu;
29109 + struct au_rdu_ent ent;
29113 + au_fset_rdu(rdu->cookie.flags, CALLED);
29114 + len = au_rdu_len(nlen);
29115 + if (arg->ent.ul + len < arg->end) {
29117 + ent.bindex = rdu->cookie.bindex;
29118 + ent.type = d_type;
29120 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
29121 + ent.type = DT_UNKNOWN;
29123 + /* unnecessary to support mmap_sem since this is a dir */
29125 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
29127 + if (copy_to_user(arg->ent.e->name, name, nlen))
29129 + /* the terminating NULL */
29130 + if (__put_user(0, arg->ent.e->name + nlen))
29133 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
29134 + arg->ent.ul += len;
29138 + au_fset_rdu(rdu->cookie.flags, FULL);
29140 + rdu->tail = arg->ent;
29144 + /* AuTraceErr(err); */
29148 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
29152 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
29154 + /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
29155 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
29157 + if (unlikely(offset != cookie->h_pos))
29163 + au_fclr_rdu(cookie->flags, CALLED);
29165 + err = vfsub_iterate_dir(h_file, &arg->ctx);
29169 + && au_ftest_rdu(cookie->flags, CALLED)
29170 + && !au_ftest_rdu(cookie->flags, FULL));
29171 + cookie->h_pos = h_file->f_pos;
29178 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
29181 + aufs_bindex_t bbot;
29182 + struct au_rdu_arg arg = {
29184 + .actor = au_rdu_fill
29187 + struct dentry *dentry;
29188 + struct inode *inode;
29189 + struct file *h_file;
29190 + struct au_rdu_cookie *cookie = &rdu->cookie;
29192 + /* VERIFY_WRITE */
29193 + err = !access_ok(rdu->ent.e, rdu->sz);
29194 + if (unlikely(err)) {
29200 + rdu->tail = rdu->ent;
29203 + arg.ent = rdu->ent;
29204 + arg.end = arg.ent.ul;
29205 + arg.end += rdu->sz;
29208 + if (unlikely(!file->f_op->iterate && !file->f_op->iterate_shared))
29211 + err = security_file_permission(file, MAY_READ);
29213 + if (unlikely(err))
29216 + dentry = file->f_path.dentry;
29217 + inode = d_inode(dentry);
29218 + inode_lock_shared(inode);
29220 + arg.sb = inode->i_sb;
29221 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
29222 + if (unlikely(err))
29224 + err = au_alive_dir(dentry);
29225 + if (unlikely(err))
29227 + /* todo: reval? */
29228 + fi_read_lock(file);
29231 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
29232 + && cookie->generation != au_figen(file)))
29237 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
29239 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
29241 + bbot = au_fbtop(file);
29242 + if (cookie->bindex < bbot)
29243 + cookie->bindex = bbot;
29244 + bbot = au_fbbot_dir(file);
29245 + /* AuDbg("b%d, b%d\n", cookie->bindex, bbot); */
29246 + for (; !err && cookie->bindex <= bbot;
29247 + cookie->bindex++, cookie->h_pos = 0) {
29248 + h_file = au_hf_dir(file, cookie->bindex);
29252 + au_fclr_rdu(cookie->flags, FULL);
29253 + err = au_rdu_do(h_file, &arg);
29255 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
29258 + AuDbg("rent %llu\n", rdu->rent);
29260 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
29261 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
29262 + au_fset_rdu(cookie->flags, CONT);
29263 + cookie->generation = au_figen(file);
29266 + ii_read_lock_child(inode);
29267 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibtop(inode)));
29268 + ii_read_unlock(inode);
29271 + fi_read_unlock(file);
29273 + si_read_unlock(arg.sb);
29275 + inode_unlock_shared(inode);
29281 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
29285 + unsigned long long nent;
29286 + union au_rdu_ent_ul *u;
29287 + struct au_rdu_ent ent;
29288 + struct super_block *sb;
29291 + nent = rdu->nent;
29293 + sb = file->f_path.dentry->d_sb;
29294 + si_read_lock(sb, AuLock_FLUSH);
29295 + while (nent-- > 0) {
29296 + /* unnecessary to support mmap_sem since this is a dir */
29297 + err = copy_from_user(&ent, u->e, sizeof(ent));
29299 + /* VERIFY_WRITE */
29300 + err = !access_ok(&u->e->ino, sizeof(ino));
29301 + if (unlikely(err)) {
29307 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
29309 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
29311 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
29313 + if (unlikely(err)) {
29318 + err = __put_user(ino, &u->e->ino);
29319 + if (unlikely(err)) {
29324 + u->ul += au_rdu_len(ent.nlen);
29326 + si_read_unlock(sb);
29331 +/* ---------------------------------------------------------------------- */
29333 +static int au_rdu_verify(struct aufs_rdu *rdu)
29335 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
29336 + "%llu, b%d, 0x%x, g%u}\n",
29337 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
29339 + rdu->rent, rdu->shwh, rdu->full,
29340 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
29341 + rdu->cookie.generation);
29343 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
29347 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
29351 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29354 + struct aufs_rdu rdu;
29355 + void __user *p = (void __user *)arg;
29357 + err = copy_from_user(&rdu, p, sizeof(rdu));
29358 + if (unlikely(err)) {
29363 + err = au_rdu_verify(&rdu);
29364 + if (unlikely(err))
29368 + case AUFS_CTL_RDU:
29369 + err = au_rdu(file, &rdu);
29370 + if (unlikely(err))
29373 + e = copy_to_user(p, &rdu, sizeof(rdu));
29374 + if (unlikely(e)) {
29379 + case AUFS_CTL_RDU_INO:
29380 + err = au_rdu_ino(file, &rdu);
29384 + /* err = -ENOTTY; */
29393 +#ifdef CONFIG_COMPAT
29394 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
29397 + struct aufs_rdu rdu;
29398 + void __user *p = compat_ptr(arg);
29400 + /* todo: get_user()? */
29401 + err = copy_from_user(&rdu, p, sizeof(rdu));
29402 + if (unlikely(err)) {
29407 + rdu.ent.e = compat_ptr(rdu.ent.ul);
29408 + err = au_rdu_verify(&rdu);
29409 + if (unlikely(err))
29413 + case AUFS_CTL_RDU:
29414 + err = au_rdu(file, &rdu);
29415 + if (unlikely(err))
29418 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
29419 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
29420 + e = copy_to_user(p, &rdu, sizeof(rdu));
29421 + if (unlikely(e)) {
29426 + case AUFS_CTL_RDU_INO:
29427 + err = au_rdu_ino(file, &rdu);
29431 + /* err = -ENOTTY; */
29440 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
29441 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
29442 +++ linux/fs/aufs/rwsem.h 2022-11-05 23:02:18.969222617 +0100
29444 +/* SPDX-License-Identifier: GPL-2.0 */
29446 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29448 + * This program is free software; you can redistribute it and/or modify
29449 + * it under the terms of the GNU General Public License as published by
29450 + * the Free Software Foundation; either version 2 of the License, or
29451 + * (at your option) any later version.
29453 + * This program is distributed in the hope that it will be useful,
29454 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29455 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29456 + * GNU General Public License for more details.
29458 + * You should have received a copy of the GNU General Public License
29459 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29463 + * simple read-write semaphore wrappers
29466 +#ifndef __AUFS_RWSEM_H__
29467 +#define __AUFS_RWSEM_H__
29471 +#include "debug.h"
29473 +/* in the future, the name 'au_rwsem' will be totally gone */
29474 +#define au_rwsem rw_semaphore
29476 +/* to debug easier, do not make them inlined functions */
29477 +#define AuRwMustNoWaiters(rw) AuDebugOn(rwsem_is_contended(rw))
29479 +#ifdef CONFIG_LOCKDEP
29480 +/* rwsem_is_locked() is unusable */
29481 +#define AuRwMustReadLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29482 + && !lockdep_recursing(current) \
29484 + && !lockdep_is_held_type(rw, 1))
29485 +#define AuRwMustWriteLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29486 + && !lockdep_recursing(current) \
29488 + && !lockdep_is_held_type(rw, 0))
29489 +#define AuRwMustAnyLock(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29490 + && !lockdep_recursing(current) \
29492 + && !lockdep_is_held(rw))
29493 +#define AuRwDestroy(rw) AuDebugOn(IS_ENABLED(CONFIG_LOCKDEP) \
29494 + && !lockdep_recursing(current) \
29496 + && lockdep_is_held(rw))
29498 +#define AuRwMustReadLock(rw) do {} while (0)
29499 +#define AuRwMustWriteLock(rw) do {} while (0)
29500 +#define AuRwMustAnyLock(rw) do {} while (0)
29501 +#define AuRwDestroy(rw) do {} while (0)
29504 +#define au_rw_init(rw) init_rwsem(rw)
29506 +#define au_rw_init_wlock(rw) do { \
29507 + au_rw_init(rw); \
29508 + down_write(rw); \
29511 +#define au_rw_init_wlock_nested(rw, lsc) do { \
29512 + au_rw_init(rw); \
29513 + down_write_nested(rw, lsc); \
29516 +#define au_rw_read_lock(rw) down_read(rw)
29517 +#define au_rw_read_lock_nested(rw, lsc) down_read_nested(rw, lsc)
29518 +#define au_rw_read_unlock(rw) up_read(rw)
29519 +#define au_rw_dgrade_lock(rw) downgrade_write(rw)
29520 +#define au_rw_write_lock(rw) down_write(rw)
29521 +#define au_rw_write_lock_nested(rw, lsc) down_write_nested(rw, lsc)
29522 +#define au_rw_write_unlock(rw) up_write(rw)
29523 +/* why is not _nested version defined? */
29524 +#define au_rw_read_trylock(rw) down_read_trylock(rw)
29525 +#define au_rw_write_trylock(rw) down_write_trylock(rw)
29527 +#endif /* __KERNEL__ */
29528 +#endif /* __AUFS_RWSEM_H__ */
29529 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
29530 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
29531 +++ linux/fs/aufs/sbinfo.c 2022-11-05 23:02:18.969222617 +0100
29533 +// SPDX-License-Identifier: GPL-2.0
29535 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29537 + * This program is free software; you can redistribute it and/or modify
29538 + * it under the terms of the GNU General Public License as published by
29539 + * the Free Software Foundation; either version 2 of the License, or
29540 + * (at your option) any later version.
29542 + * This program is distributed in the hope that it will be useful,
29543 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29544 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29545 + * GNU General Public License for more details.
29547 + * You should have received a copy of the GNU General Public License
29548 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29552 + * superblock private data
29555 +#include <linux/iversion.h>
29559 + * they are necessary regardless sysfs is disabled.
29561 +void au_si_free(struct kobject *kobj)
29564 + struct au_sbinfo *sbinfo;
29565 + char *locked __maybe_unused; /* debug only */
29567 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
29568 + for (i = 0; i < AuPlink_NHASH; i++)
29569 + AuDebugOn(!hlist_bl_empty(sbinfo->si_plink + i));
29570 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
29572 + AuLCntZero(au_lcnt_read(&sbinfo->si_ninodes, /*do_rev*/0));
29573 + au_lcnt_fin(&sbinfo->si_ninodes, /*do_sync*/0);
29574 + AuLCntZero(au_lcnt_read(&sbinfo->si_nfiles, /*do_rev*/0));
29575 + au_lcnt_fin(&sbinfo->si_nfiles, /*do_sync*/0);
29577 + dbgaufs_si_fin(sbinfo);
29578 + au_rw_write_lock(&sbinfo->si_rwsem);
29579 + au_br_free(sbinfo);
29580 + au_rw_write_unlock(&sbinfo->si_rwsem);
29582 + au_kfree_try_rcu(sbinfo->si_branch);
29583 + mutex_destroy(&sbinfo->si_xib_mtx);
29584 + AuRwDestroy(&sbinfo->si_rwsem);
29586 + au_lcnt_wait_for_fin(&sbinfo->si_ninodes);
29587 + /* si_nfiles is waited too */
29588 + au_kfree_rcu(sbinfo);
29591 +struct au_sbinfo *au_si_alloc(struct super_block *sb)
29593 + struct au_sbinfo *sbinfo;
29597 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
29598 + if (unlikely(!sbinfo))
29601 + /* will be reallocated separately */
29602 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
29603 + if (unlikely(!sbinfo->si_branch))
29606 + err = sysaufs_si_init(sbinfo);
29608 + dbgaufs_si_null(sbinfo);
29609 + err = dbgaufs_si_init(sbinfo);
29610 + if (unlikely(err))
29611 + kobject_put(&sbinfo->si_kobj);
29613 + if (unlikely(err))
29616 + au_nwt_init(&sbinfo->si_nowait);
29617 + au_rw_init_wlock(&sbinfo->si_rwsem);
29619 + au_lcnt_init(&sbinfo->si_ninodes, /*release*/NULL);
29620 + au_lcnt_init(&sbinfo->si_nfiles, /*release*/NULL);
29622 + sbinfo->si_bbot = -1;
29623 + sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
29625 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
29626 + sbinfo->si_wbr_create = AuWbrCreate_Def;
29627 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
29628 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
29630 + au_fhsm_init(sbinfo);
29632 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
29634 + sbinfo->si_xino_jiffy = jiffies;
29635 + sbinfo->si_xino_expire
29636 + = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
29637 + mutex_init(&sbinfo->si_xib_mtx);
29638 + /* leave si_xib_last_pindex and si_xib_next_bit */
29640 + INIT_HLIST_BL_HEAD(&sbinfo->si_aopen);
29642 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
29643 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
29644 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
29645 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
29647 + for (i = 0; i < AuPlink_NHASH; i++)
29648 + INIT_HLIST_BL_HEAD(sbinfo->si_plink + i);
29649 + init_waitqueue_head(&sbinfo->si_plink_wq);
29650 + spin_lock_init(&sbinfo->si_plink_maint_lock);
29652 + INIT_HLIST_BL_HEAD(&sbinfo->si_files);
29654 + /* with getattr by default */
29655 + sbinfo->si_iop_array = aufs_iop;
29657 + /* leave other members for sysaufs and si_mnt. */
29658 + sbinfo->si_sb = sb;
29660 + sb->s_fs_info = sbinfo;
29663 + return sbinfo; /* success */
29666 + au_kfree_try_rcu(sbinfo->si_branch);
29668 + au_kfree_rcu(sbinfo);
29670 + return ERR_PTR(err);
29673 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink)
29676 + struct au_branch **brp;
29678 + AuRwMustWriteLock(&sbinfo->si_rwsem);
29681 + sz = sizeof(*brp) * (sbinfo->si_bbot + 1);
29682 + if (unlikely(!sz))
29683 + sz = sizeof(*brp);
29684 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS,
29687 + sbinfo->si_branch = brp;
29694 +/* ---------------------------------------------------------------------- */
29696 +unsigned int au_sigen_inc(struct super_block *sb)
29698 + unsigned int gen;
29699 + struct inode *inode;
29701 + SiMustWriteLock(sb);
29703 + gen = ++au_sbi(sb)->si_generation;
29704 + au_update_digen(sb->s_root);
29705 + inode = d_inode(sb->s_root);
29706 + au_update_iigen(inode, /*half*/0);
29707 + inode_inc_iversion(inode);
29711 +aufs_bindex_t au_new_br_id(struct super_block *sb)
29713 + aufs_bindex_t br_id;
29715 + struct au_sbinfo *sbinfo;
29717 + SiMustWriteLock(sb);
29719 + sbinfo = au_sbi(sb);
29720 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
29721 + br_id = ++sbinfo->si_last_br_id;
29722 + AuDebugOn(br_id < 0);
29723 + if (br_id && au_br_index(sb, br_id) < 0)
29730 +/* ---------------------------------------------------------------------- */
29732 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
29733 +int si_read_lock(struct super_block *sb, int flags)
29738 + if (au_ftest_lock(flags, FLUSH))
29739 + au_nwt_flush(&au_sbi(sb)->si_nowait);
29741 + si_noflush_read_lock(sb);
29742 + err = au_plink_maint(sb, flags);
29743 + if (unlikely(err))
29744 + si_read_unlock(sb);
29749 +int si_write_lock(struct super_block *sb, int flags)
29753 + if (au_ftest_lock(flags, FLUSH))
29754 + au_nwt_flush(&au_sbi(sb)->si_nowait);
29756 + si_noflush_write_lock(sb);
29757 + err = au_plink_maint(sb, flags);
29758 + if (unlikely(err))
29759 + si_write_unlock(sb);
29764 +/* dentry and super_block lock. call at entry point */
29765 +int aufs_read_lock(struct dentry *dentry, int flags)
29768 + struct super_block *sb;
29770 + sb = dentry->d_sb;
29771 + err = si_read_lock(sb, flags);
29772 + if (unlikely(err))
29775 + if (au_ftest_lock(flags, DW))
29776 + di_write_lock_child(dentry);
29778 + di_read_lock_child(dentry, flags);
29780 + if (au_ftest_lock(flags, GEN)) {
29781 + err = au_digen_test(dentry, au_sigen(sb));
29782 + if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
29783 + AuDebugOn(!err && au_dbrange_test(dentry));
29785 + err = au_dbrange_test(dentry);
29786 + if (unlikely(err))
29787 + aufs_read_unlock(dentry, flags);
29794 +void aufs_read_unlock(struct dentry *dentry, int flags)
29796 + if (au_ftest_lock(flags, DW))
29797 + di_write_unlock(dentry);
29799 + di_read_unlock(dentry, flags);
29800 + si_read_unlock(dentry->d_sb);
29803 +void aufs_write_lock(struct dentry *dentry)
29805 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
29806 + di_write_lock_child(dentry);
29809 +void aufs_write_unlock(struct dentry *dentry)
29811 + di_write_unlock(dentry);
29812 + si_write_unlock(dentry->d_sb);
29815 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
29818 + unsigned int sigen;
29819 + struct super_block *sb;
29822 + err = si_read_lock(sb, flags);
29823 + if (unlikely(err))
29826 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
29828 + if (au_ftest_lock(flags, GEN)) {
29829 + sigen = au_sigen(sb);
29830 + err = au_digen_test(d1, sigen);
29831 + AuDebugOn(!err && au_dbrange_test(d1));
29833 + err = au_digen_test(d2, sigen);
29834 + AuDebugOn(!err && au_dbrange_test(d2));
29836 + if (unlikely(err))
29837 + aufs_read_and_write_unlock2(d1, d2);
29844 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
29846 + di_write_unlock2(d1, d2);
29847 + si_read_unlock(d1->d_sb);
29849 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
29850 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
29851 +++ linux/fs/aufs/super.c 2022-11-05 23:02:18.969222617 +0100
29853 +// SPDX-License-Identifier: GPL-2.0
29855 + * Copyright (C) 2005-2022 Junjiro R. Okajima
29857 + * This program is free software; you can redistribute it and/or modify
29858 + * it under the terms of the GNU General Public License as published by
29859 + * the Free Software Foundation; either version 2 of the License, or
29860 + * (at your option) any later version.
29862 + * This program is distributed in the hope that it will be useful,
29863 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29864 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29865 + * GNU General Public License for more details.
29867 + * You should have received a copy of the GNU General Public License
29868 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29872 + * mount and super_block operations
29875 +#include <linux/iversion.h>
29876 +#include <linux/mm.h>
29877 +#include <linux/seq_file.h>
29878 +#include <linux/statfs.h>
29879 +#include <linux/vmalloc.h>
29883 + * super_operations
29885 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
29887 + struct au_icntnr *c;
29889 + c = au_cache_alloc_icntnr(sb);
29891 + au_icntnr_init(c);
29892 + inode_set_iversion(&c->vfs_inode, 1); /* sigen(sb); */
29893 + c->iinfo.ii_hinode = NULL;
29894 + return &c->vfs_inode;
29899 +static void aufs_destroy_inode(struct inode *inode)
29901 + if (!au_is_bad_inode(inode))
29902 + au_iinfo_fin(inode);
29905 +static void aufs_free_inode(struct inode *inode)
29907 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
29910 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
29912 + struct inode *inode;
29915 + inode = iget_locked(sb, ino);
29916 + if (unlikely(!inode)) {
29917 + inode = ERR_PTR(-ENOMEM);
29920 + if (!(inode->i_state & I_NEW))
29923 + err = au_xigen_new(inode);
29925 + err = au_iinfo_init(inode);
29927 + inode_inc_iversion(inode);
29929 + iget_failed(inode);
29930 + inode = ERR_PTR(err);
29934 + /* never return NULL */
29935 + AuDebugOn(!inode);
29936 + AuTraceErrPtr(inode);
29940 +/* lock free root dinfo */
29941 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
29944 + aufs_bindex_t bindex, bbot;
29945 + struct path path;
29946 + struct au_hdentry *hdp;
29947 + struct au_branch *br;
29948 + au_br_perm_str_t perm;
29951 + bbot = au_sbbot(sb);
29953 + hdp = au_hdentry(au_di(sb->s_root), bindex);
29954 + for (; !err && bindex <= bbot; bindex++, hdp++) {
29955 + br = au_sbr(sb, bindex);
29956 + path.mnt = au_br_mnt(br);
29957 + path.dentry = hdp->hd_dentry;
29958 + err = au_seq_path(seq, &path);
29960 + au_optstr_br_perm(&perm, br->br_perm);
29961 + seq_printf(seq, "=%s", perm.a);
29962 + if (bindex != bbot)
29963 + seq_putc(seq, ':');
29966 + if (unlikely(err || seq_has_overflowed(seq)))
29972 +static void au_gen_fmt(char *fmt, int len __maybe_unused, const char *pat,
29973 + const char *append)
29978 + while (*pat != ':')
29981 + strcpy(p, append);
29982 + AuDebugOn(strlen(fmt) >= len);
29985 +static void au_show_wbr_create(struct seq_file *m, int v,
29986 + struct au_sbinfo *sbinfo)
29990 + struct au_wbr_mfs *mfs;
29992 + AuRwMustAnyLock(&sbinfo->si_rwsem);
29994 + seq_puts(m, ",create=");
29995 + pat = au_optstr_wbr_create(v);
29996 + mfs = &sbinfo->si_wbr_mfs;
29998 + case AuWbrCreate_TDP:
29999 + case AuWbrCreate_RR:
30000 + case AuWbrCreate_MFS:
30001 + case AuWbrCreate_PMFS:
30002 + seq_puts(m, pat);
30004 + case AuWbrCreate_MFSRR:
30005 + case AuWbrCreate_TDMFS:
30006 + case AuWbrCreate_PMFSRR:
30007 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu");
30008 + seq_printf(m, fmt, mfs->mfsrr_watermark);
30010 + case AuWbrCreate_MFSV:
30011 + case AuWbrCreate_PMFSV:
30012 + au_gen_fmt(fmt, sizeof(fmt), pat, "%lu");
30013 + seq_printf(m, fmt,
30014 + jiffies_to_msecs(mfs->mfs_expire)
30017 + case AuWbrCreate_MFSRRV:
30018 + case AuWbrCreate_TDMFSV:
30019 + case AuWbrCreate_PMFSRRV:
30020 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu:%lu");
30021 + seq_printf(m, fmt, mfs->mfsrr_watermark,
30022 + jiffies_to_msecs(mfs->mfs_expire) / MSEC_PER_SEC);
30029 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
30031 +#ifdef CONFIG_SYSFS
30035 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
30036 + aufs_bindex_t bindex, brid;
30037 + struct qstr *name;
30039 + struct dentry *d, *h_root;
30040 + struct au_branch *br;
30042 + AuRwMustAnyLock(&sbinfo->si_rwsem);
30045 + f = au_sbi(sb)->si_xib;
30049 + /* stop printing the default xino path on the first writable branch */
30051 + bindex = au_xi_root(sb, f->f_path.dentry);
30052 + if (bindex >= 0) {
30053 + br = au_sbr_sb(sb, bindex);
30054 + h_root = au_br_dentry(br);
30057 + d = f->f_path.dentry;
30058 + name = &d->d_name;
30059 + /* safe ->d_parent because the file is unlinked */
30060 + if (d->d_parent == h_root
30061 + && name->len == len
30062 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
30065 + seq_puts(seq, ",xino=");
30066 + err = au_xino_path(seq, f);
30073 +/* seq_file will re-call me in case of too long string */
30074 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
30077 + unsigned int mnt_flags, v;
30078 + struct super_block *sb;
30079 + struct au_sbinfo *sbinfo;
30081 +#define AuBool(name, str) do { \
30082 + v = au_opt_test(mnt_flags, name); \
30083 + if (v != au_opt_test(AuOpt_Def, name)) \
30084 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
30087 +#define AuStr(name, str) do { \
30088 + v = mnt_flags & AuOptMask_##name; \
30089 + if (v != (AuOpt_Def & AuOptMask_##name)) \
30090 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
30093 +#define AuUInt(name, str, val) do { \
30094 + if (val != AUFS_##name##_DEF) \
30095 + seq_printf(m, "," #str "=%u", val); \
30098 + sb = dentry->d_sb;
30099 + if (sb->s_flags & SB_POSIXACL)
30100 + seq_puts(m, ",acl");
30101 +#if 0 /* reserved for future use */
30102 + if (sb->s_flags & SB_I_VERSION)
30103 + seq_puts(m, ",i_version");
30106 + /* lock free root dinfo */
30107 + si_noflush_read_lock(sb);
30108 + sbinfo = au_sbi(sb);
30109 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
30111 + mnt_flags = au_mntflags(sb);
30112 + if (au_opt_test(mnt_flags, XINO)) {
30113 + err = au_show_xino(m, sb);
30114 + if (unlikely(err))
30117 + seq_puts(m, ",noxino");
30119 + AuBool(TRUNC_XINO, trunc_xino);
30120 + AuStr(UDBA, udba);
30121 + AuBool(SHWH, shwh);
30122 + AuBool(PLINK, plink);
30123 + AuBool(DIO, dio);
30124 + AuBool(DIRPERM1, dirperm1);
30126 + v = sbinfo->si_wbr_create;
30127 + if (v != AuWbrCreate_Def)
30128 + au_show_wbr_create(m, v, sbinfo);
30130 + v = sbinfo->si_wbr_copyup;
30131 + if (v != AuWbrCopyup_Def)
30132 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
30134 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
30135 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
30136 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
30138 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
30140 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
30141 + AuUInt(RDCACHE, rdcache, v);
30143 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
30144 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
30146 + au_fhsm_show(m, sbinfo);
30148 + AuBool(DIRREN, dirren);
30149 + AuBool(SUM, sum);
30150 + /* AuBool(SUM_W, wsum); */
30151 + AuBool(WARN_PERM, warn_perm);
30152 + AuBool(VERBOSE, verbose);
30155 + /* be sure to print "br:" last */
30156 + if (!sysaufs_brs) {
30157 + seq_puts(m, ",br:");
30158 + au_show_brs(m, sb);
30160 + si_read_unlock(sb);
30168 +/* ---------------------------------------------------------------------- */
30170 +/* sum mode which returns the summation for statfs(2) */
30172 +static u64 au_add_till_max(u64 a, u64 b)
30180 + return ULLONG_MAX;
30183 +static u64 au_mul_till_max(u64 a, long mul)
30191 + return ULLONG_MAX;
30194 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
30197 + long bsize, factor;
30198 + u64 blocks, bfree, bavail, files, ffree;
30199 + aufs_bindex_t bbot, bindex, i;
30200 + unsigned char shared;
30201 + struct path h_path;
30202 + struct super_block *h_sb;
30205 + bsize = LONG_MAX;
30211 + bbot = au_sbbot(sb);
30212 + for (bindex = 0; bindex <= bbot; bindex++) {
30213 + h_path.mnt = au_sbr_mnt(sb, bindex);
30214 + h_sb = h_path.mnt->mnt_sb;
30216 + for (i = 0; !shared && i < bindex; i++)
30217 + shared = (au_sbr_sb(sb, i) == h_sb);
30221 + /* sb->s_root for NFS is unreliable */
30222 + h_path.dentry = h_path.mnt->mnt_root;
30223 + err = vfs_statfs(&h_path, buf);
30224 + if (unlikely(err))
30227 + if (bsize > buf->f_bsize) {
30229 + * we will reduce bsize, so we have to expand blocks
30230 + * etc. to match them again
30232 + factor = (bsize / buf->f_bsize);
30233 + blocks = au_mul_till_max(blocks, factor);
30234 + bfree = au_mul_till_max(bfree, factor);
30235 + bavail = au_mul_till_max(bavail, factor);
30236 + bsize = buf->f_bsize;
30239 + factor = (buf->f_bsize / bsize);
30240 + blocks = au_add_till_max(blocks,
30241 + au_mul_till_max(buf->f_blocks, factor));
30242 + bfree = au_add_till_max(bfree,
30243 + au_mul_till_max(buf->f_bfree, factor));
30244 + bavail = au_add_till_max(bavail,
30245 + au_mul_till_max(buf->f_bavail, factor));
30246 + files = au_add_till_max(files, buf->f_files);
30247 + ffree = au_add_till_max(ffree, buf->f_ffree);
30250 + buf->f_bsize = bsize;
30251 + buf->f_blocks = blocks;
30252 + buf->f_bfree = bfree;
30253 + buf->f_bavail = bavail;
30254 + buf->f_files = files;
30255 + buf->f_ffree = ffree;
30256 + buf->f_frsize = 0;
30262 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
30265 + struct path h_path;
30266 + struct super_block *sb;
30268 + /* lock free root dinfo */
30269 + sb = dentry->d_sb;
30270 + si_noflush_read_lock(sb);
30271 + if (!au_opt_test(au_mntflags(sb), SUM)) {
30272 + /* sb->s_root for NFS is unreliable */
30273 + h_path.mnt = au_sbr_mnt(sb, 0);
30274 + h_path.dentry = h_path.mnt->mnt_root;
30275 + err = vfs_statfs(&h_path, buf);
30277 + err = au_statfs_sum(sb, buf);
30278 + si_read_unlock(sb);
30281 + buf->f_type = AUFS_SUPER_MAGIC;
30282 + buf->f_namelen = AUFS_MAX_NAMELEN;
30283 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
30285 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
30290 +/* ---------------------------------------------------------------------- */
30292 +static int aufs_sync_fs(struct super_block *sb, int wait)
30295 + aufs_bindex_t bbot, bindex;
30296 + struct au_branch *br;
30297 + struct super_block *h_sb;
30300 + si_noflush_read_lock(sb);
30301 + bbot = au_sbbot(sb);
30302 + for (bindex = 0; bindex <= bbot; bindex++) {
30303 + br = au_sbr(sb, bindex);
30304 + if (!au_br_writable(br->br_perm))
30307 + h_sb = au_sbr_sb(sb, bindex);
30308 + e = vfsub_sync_filesystem(h_sb);
30309 + if (unlikely(e && !err))
30311 + /* go on even if an error happens */
30313 + si_read_unlock(sb);
30318 +/* ---------------------------------------------------------------------- */
30320 +/* final actions when unmounting a file system */
30321 +static void aufs_put_super(struct super_block *sb)
30323 + struct au_sbinfo *sbinfo;
30325 + sbinfo = au_sbi(sb);
30327 + kobject_put(&sbinfo->si_kobj);
30330 +/* ---------------------------------------------------------------------- */
30332 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30333 + struct super_block *sb, void *arg)
30336 + unsigned long long n, sz;
30343 + if (*hint > ULLONG_MAX / sizeof(array)) {
30344 + array = ERR_PTR(-EMFILE);
30345 + pr_err("hint %llu\n", *hint);
30349 + sz = sizeof(array) * *hint;
30350 + array = kzalloc(sz, GFP_NOFS);
30351 + if (unlikely(!array))
30352 + array = vzalloc(sz);
30353 + if (unlikely(!array)) {
30354 + array = ERR_PTR(-ENOMEM);
30358 + n = cb(sb, array, *hint, arg);
30359 + AuDebugOn(n > *hint);
30366 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
30367 + unsigned long long max __maybe_unused,
30370 + unsigned long long n;
30371 + struct inode **p, *inode;
30372 + struct list_head *head;
30377 + spin_lock(&sb->s_inode_list_lock);
30378 + list_for_each_entry(inode, head, i_sb_list) {
30379 + if (!au_is_bad_inode(inode)
30380 + && au_ii(inode)->ii_btop >= 0) {
30381 + spin_lock(&inode->i_lock);
30382 + if (atomic_read(&inode->i_count)) {
30386 + AuDebugOn(n > max);
30388 + spin_unlock(&inode->i_lock);
30391 + spin_unlock(&sb->s_inode_list_lock);
30396 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
30398 + struct au_sbinfo *sbi;
30400 + sbi = au_sbi(sb);
30401 + *max = au_lcnt_read(&sbi->si_ninodes, /*do_rev*/1);
30402 + return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
30405 +void au_iarray_free(struct inode **a, unsigned long long max)
30407 + unsigned long long ull;
30409 + for (ull = 0; ull < max; ull++)
30414 +/* ---------------------------------------------------------------------- */
30417 + * refresh dentry and inode at remount time.
30419 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
30420 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
30421 + struct dentry *parent)
30425 + di_write_lock_child(dentry);
30426 + di_read_lock_parent(parent, AuLock_IR);
30427 + err = au_refresh_dentry(dentry, parent);
30428 + if (!err && dir_flags)
30429 + au_hn_reset(d_inode(dentry), dir_flags);
30430 + di_read_unlock(parent, AuLock_IR);
30431 + di_write_unlock(dentry);
30436 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
30437 + struct au_sbinfo *sbinfo,
30438 + const unsigned int dir_flags, unsigned int do_idop)
30441 + struct dentry *parent;
30444 + parent = dget_parent(dentry);
30445 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
30446 + if (d_really_is_positive(dentry)) {
30447 + if (!d_is_dir(dentry))
30448 + err = au_do_refresh(dentry, /*dir_flags*/0,
30451 + err = au_do_refresh(dentry, dir_flags, parent);
30452 + if (unlikely(err))
30453 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
30456 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
30457 + AuDbgDentry(dentry);
30463 + au_refresh_dop(dentry, /*force_reval*/0);
30465 + au_refresh_dop(dentry, /*force_reval*/1);
30471 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
30473 + int err, i, j, ndentry, e;
30474 + unsigned int sigen;
30475 + struct au_dcsub_pages dpages;
30476 + struct au_dpage *dpage;
30477 + struct dentry **dentries, *d;
30478 + struct au_sbinfo *sbinfo;
30479 + struct dentry *root = sb->s_root;
30480 + const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
30483 + au_refresh_dop(root, /*force_reval*/0);
30485 + err = au_dpages_init(&dpages, GFP_NOFS);
30486 + if (unlikely(err))
30488 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
30489 + if (unlikely(err))
30492 + sigen = au_sigen(sb);
30493 + sbinfo = au_sbi(sb);
30494 + for (i = 0; i < dpages.ndpage; i++) {
30495 + dpage = dpages.dpages + i;
30496 + dentries = dpage->dentries;
30497 + ndentry = dpage->ndentry;
30498 + for (j = 0; j < ndentry; j++) {
30500 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
30502 + if (unlikely(e && !err))
30504 + /* go on even err */
30509 + au_dpages_free(&dpages);
30514 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
30517 + unsigned int sigen;
30518 + unsigned long long max, ull;
30519 + struct inode *inode, **array;
30521 + array = au_iarray_alloc(sb, &max);
30522 + err = PTR_ERR(array);
30523 + if (IS_ERR(array))
30527 + sigen = au_sigen(sb);
30528 + for (ull = 0; ull < max; ull++) {
30529 + inode = array[ull];
30530 + if (unlikely(!inode))
30534 + ii_write_lock_child(inode);
30535 + if (au_iigen(inode, NULL) != sigen) {
30536 + e = au_refresh_hinode_self(inode);
30537 + if (unlikely(e)) {
30538 + au_refresh_iop(inode, /*force_getattr*/1);
30539 + pr_err("error %d, i%lu\n", e, inode->i_ino);
30542 + /* go on even if err */
30545 + if (!e && do_idop)
30546 + au_refresh_iop(inode, /*force_getattr*/0);
30547 + ii_write_unlock(inode);
30550 + au_iarray_free(array, max);
30556 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
30559 + unsigned int udba;
30560 + aufs_bindex_t bindex, bbot;
30561 + struct dentry *root;
30562 + struct inode *inode;
30563 + struct au_branch *br;
30564 + struct au_sbinfo *sbi;
30566 + au_sigen_inc(sb);
30567 + sbi = au_sbi(sb);
30568 + au_fclr_si(sbi, FAILED_REFRESH_DIR);
30570 + root = sb->s_root;
30571 + DiMustNoWaiters(root);
30572 + inode = d_inode(root);
30573 + IiMustNoWaiters(inode);
30575 + udba = au_opt_udba(sb);
30576 + bbot = au_sbbot(sb);
30577 + for (bindex = 0; bindex <= bbot; bindex++) {
30578 + br = au_sbr(sb, bindex);
30579 + err = au_hnotify_reset_br(udba, br, br->br_perm);
30580 + if (unlikely(err))
30581 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
30583 + /* go on even if err */
30585 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
30588 + if (au_ftest_si(sbi, NO_DREVAL)) {
30589 + AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
30590 + sb->s_d_op = &aufs_dop_noreval;
30591 + AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
30592 + sbi->si_iop_array = aufs_iop_nogetattr;
30594 + AuDebugOn(sb->s_d_op == &aufs_dop);
30595 + sb->s_d_op = &aufs_dop;
30596 + AuDebugOn(sbi->si_iop_array == aufs_iop);
30597 + sbi->si_iop_array = aufs_iop;
30599 + pr_info("reset to %ps and %ps\n",
30600 + sb->s_d_op, sbi->si_iop_array);
30603 + di_write_unlock(root);
30604 + err = au_refresh_d(sb, do_idop);
30605 + e = au_refresh_i(sb, do_idop);
30606 + if (unlikely(e && !err))
30608 + /* aufs_write_lock() calls ..._child() */
30609 + di_write_lock_child(root);
30611 + au_cpup_attr_all(inode, /*force*/1);
30613 + if (unlikely(err))
30614 + AuIOErr("refresh failed, ignored, %d\n", err);
30617 +const struct super_operations aufs_sop = {
30618 + .alloc_inode = aufs_alloc_inode,
30619 + .destroy_inode = aufs_destroy_inode,
30620 + .free_inode = aufs_free_inode,
30621 + /* always deleting, no clearing */
30622 + .drop_inode = generic_delete_inode,
30623 + .show_options = aufs_show_options,
30624 + .statfs = aufs_statfs,
30625 + .put_super = aufs_put_super,
30626 + .sync_fs = aufs_sync_fs
30629 +/* ---------------------------------------------------------------------- */
30631 +int au_alloc_root(struct super_block *sb)
30634 + struct inode *inode;
30635 + struct dentry *root;
30638 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
30639 + err = PTR_ERR(inode);
30640 + if (IS_ERR(inode))
30643 + inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
30644 + inode->i_fop = &aufs_dir_fop;
30645 + inode->i_mode = S_IFDIR;
30646 + set_nlink(inode, 2);
30647 + unlock_new_inode(inode);
30649 + root = d_make_root(inode);
30650 + if (unlikely(!root))
30652 + err = PTR_ERR(root);
30653 + if (IS_ERR(root))
30656 + err = au_di_init(root);
30658 + sb->s_root = root;
30659 + return 0; /* success */
30667 +/* ---------------------------------------------------------------------- */
30669 +static void aufs_kill_sb(struct super_block *sb)
30671 + struct au_sbinfo *sbinfo;
30672 + struct dentry *root;
30674 + sbinfo = au_sbi(sb);
30678 + au_sbilist_del(sb);
30680 + root = sb->s_root;
30682 + aufs_write_lock(root);
30684 + __si_write_lock(sb);
30687 + if (sbinfo->si_wbr_create_ops->fin)
30688 + sbinfo->si_wbr_create_ops->fin(sb);
30689 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
30690 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
30691 + au_remount_refresh(sb, /*do_idop*/0);
30693 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
30694 + au_plink_put(sb, /*verbose*/1);
30697 + au_dr_opt_flush(sb);
30700 + aufs_write_unlock(root);
30702 + __si_write_unlock(sb);
30704 + sbinfo->si_sb = NULL;
30705 + au_nwt_flush(&sbinfo->si_nowait);
30708 + kill_anon_super(sb);
30711 +struct file_system_type aufs_fs_type = {
30712 + .name = AUFS_FSTYPE,
30713 + /* a race between rename and others */
30714 + .fs_flags = FS_RENAME_DOES_D_MOVE
30716 + /*| FS_ALLOW_IDMAP*/
30718 + .init_fs_context = aufs_fsctx_init,
30719 + .parameters = aufs_fsctx_paramspec,
30720 + .kill_sb = aufs_kill_sb,
30721 + /* no need to __module_get() and module_put(). */
30722 + .owner = THIS_MODULE,
30724 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
30725 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
30726 +++ linux/fs/aufs/super.h 2022-11-05 23:02:18.969222617 +0100
30728 +/* SPDX-License-Identifier: GPL-2.0 */
30730 + * Copyright (C) 2005-2022 Junjiro R. Okajima
30732 + * This program is free software; you can redistribute it and/or modify
30733 + * it under the terms of the GNU General Public License as published by
30734 + * the Free Software Foundation; either version 2 of the License, or
30735 + * (at your option) any later version.
30737 + * This program is distributed in the hope that it will be useful,
30738 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30739 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30740 + * GNU General Public License for more details.
30742 + * You should have received a copy of the GNU General Public License
30743 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30747 + * super_block operations
30750 +#ifndef __AUFS_SUPER_H__
30751 +#define __AUFS_SUPER_H__
30755 +#include <linux/fs.h>
30756 +#include <linux/kobject.h>
30759 +#include "rwsem.h"
30762 +/* policies to select one among multiple writable branches */
30763 +struct au_wbr_copyup_operations {
30764 + int (*copyup)(struct dentry *dentry);
30767 +#define AuWbr_DIR 1 /* target is a dir */
30768 +#define AuWbr_PARENT (1 << 1) /* always require a parent */
30770 +#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
30771 +#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
30772 +#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
30774 +struct au_wbr_create_operations {
30775 + int (*create)(struct dentry *dentry, unsigned int flags);
30776 + int (*init)(struct super_block *sb);
30777 + int (*fin)(struct super_block *sb);
30780 +struct au_wbr_mfs {
30781 + struct mutex mfs_lock; /* protect this structure */
30782 + unsigned long mfs_jiffy;
30783 + unsigned long mfs_expire;
30784 + aufs_bindex_t mfs_bindex;
30786 + unsigned long long mfsrr_bytes;
30787 + unsigned long long mfsrr_watermark;
30790 +#define AuPlink_NHASH 100
30791 +static inline int au_plink_hash(ino_t ino)
30793 + return ino % AuPlink_NHASH;
30796 +/* File-based Hierarchical Storage Management */
30798 +#ifdef CONFIG_AUFS_FHSM
30799 + /* allow only one process who can receive the notification */
30800 + spinlock_t fhsm_spin;
30802 + wait_queue_head_t fhsm_wqh;
30803 + atomic_t fhsm_readable;
30805 + /* these are protected by si_rwsem */
30806 + unsigned long fhsm_expire;
30807 + aufs_bindex_t fhsm_bottom;
30812 +struct au_sbinfo {
30813 + /* nowait tasks in the system-wide workqueue */
30814 + struct au_nowait_tasks si_nowait;
30817 + * tried sb->s_umount, but failed due to the dependency between i_mutex.
30818 + * rwsem for au_sbinfo is necessary.
30820 + struct au_rwsem si_rwsem;
30823 + * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
30826 + au_lcnt_t si_ninodes, si_nfiles;
30828 + /* branch management */
30829 + unsigned int si_generation;
30831 + /* see AuSi_ flags */
30832 + unsigned char au_si_status;
30834 + aufs_bindex_t si_bbot;
30836 + /* dirty trick to keep br_id plus */
30837 + unsigned int si_last_br_id :
30838 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
30839 + struct au_branch **si_branch;
30841 + /* policy to select a writable branch */
30842 + unsigned char si_wbr_copyup;
30843 + unsigned char si_wbr_create;
30844 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
30845 + struct au_wbr_create_operations *si_wbr_create_ops;
30847 + /* round robin */
30848 + atomic_t si_wbr_rr_next;
30850 + /* most free space */
30851 + struct au_wbr_mfs si_wbr_mfs;
30853 + /* File-based Hierarchical Storage Management */
30854 + struct au_fhsm si_fhsm;
30856 + /* mount flags */
30857 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
30858 + unsigned int si_mntflags;
30860 + /* external inode number (bitmap and translation table) */
30861 + loff_t si_ximaxent; /* max entries in a xino */
30863 + struct file *si_xib;
30864 + struct mutex si_xib_mtx; /* protect xib members */
30865 + unsigned long *si_xib_buf;
30866 + unsigned long si_xib_last_pindex;
30867 + int si_xib_next_bit;
30869 + unsigned long si_xino_jiffy;
30870 + unsigned long si_xino_expire;
30871 + /* reserved for future use */
30872 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
30874 +#ifdef CONFIG_AUFS_EXPORT
30875 + /* i_generation */
30876 + /* todo: make xigen file an array to support many inode numbers */
30877 + struct file *si_xigen;
30878 + atomic_t si_xigen_next;
30881 + /* dirty trick to support atomic_open */
30882 + struct hlist_bl_head si_aopen;
30884 + /* vdir parameters */
30885 + unsigned long si_rdcache; /* max cache time in jiffies */
30886 + unsigned int si_rdblk; /* deblk size */
30887 + unsigned int si_rdhash; /* hash size */
30890 + * If the number of whiteouts are larger than si_dirwh, leave all of
30891 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
30892 + * future fsck.aufs or kernel thread will remove them later.
30893 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
30895 + unsigned int si_dirwh;
30897 + /* pseudo_link list */
30898 + struct hlist_bl_head si_plink[AuPlink_NHASH];
30899 + wait_queue_head_t si_plink_wq;
30900 + spinlock_t si_plink_maint_lock;
30901 + pid_t si_plink_maint_pid;
30904 + struct hlist_bl_head si_files;
30906 + /* with/without getattr, brother of sb->s_d_op */
30907 + const struct inode_operations *si_iop_array;
30910 + * sysfs and lifetime management.
30911 + * this is not a small structure and it may be a waste of memory in case
30912 + * of sysfs is disabled, particularly when many aufs-es are mounted.
30913 + * but using sysfs is majority.
30915 + struct kobject si_kobj;
30916 +#ifdef CONFIG_DEBUG_FS
30917 + struct dentry *si_dbgaufs;
30918 + struct dentry *si_dbgaufs_plink;
30919 + struct dentry *si_dbgaufs_xib;
30920 +#ifdef CONFIG_AUFS_EXPORT
30921 + struct dentry *si_dbgaufs_xigen;
30925 +#ifdef CONFIG_AUFS_SBILIST
30926 + struct hlist_bl_node si_list;
30929 + /* dirty, necessary for unmounting, sysfs and sysrq */
30930 + struct super_block *si_sb;
30933 +/* sbinfo status flags */
30935 + * set true when refresh_dirs() failed at remount time.
30936 + * then try refreshing dirs at access time again.
30937 + * if it is false, refreshing dirs at access time is unnecessary
30939 +#define AuSi_FAILED_REFRESH_DIR 1
30940 +#define AuSi_FHSM (1 << 1) /* fhsm is active now */
30941 +#define AuSi_NO_DREVAL (1 << 2) /* disable all d_revalidate */
30943 +#ifndef CONFIG_AUFS_FHSM
30945 +#define AuSi_FHSM 0
30948 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
30949 + unsigned int flag)
30951 + AuRwMustAnyLock(&sbi->si_rwsem);
30952 + return sbi->au_si_status & flag;
30954 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
30955 +#define au_fset_si(sbinfo, name) do { \
30956 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30957 + (sbinfo)->au_si_status |= AuSi_##name; \
30959 +#define au_fclr_si(sbinfo, name) do { \
30960 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30961 + (sbinfo)->au_si_status &= ~AuSi_##name; \
30964 +/* ---------------------------------------------------------------------- */
30966 +/* policy to select one among writable branches */
30967 +#define AuWbrCopyup(sbinfo, ...) \
30968 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
30969 +#define AuWbrCreate(sbinfo, ...) \
30970 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
30972 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
30973 +#define AuLock_DW 1 /* write-lock dentry */
30974 +#define AuLock_IR (1 << 1) /* read-lock inode */
30975 +#define AuLock_IW (1 << 2) /* write-lock inode */
30976 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
30977 +#define AuLock_DIRS (1 << 4) /* target is a pair of dirs */
30978 + /* except RENAME_EXCHANGE */
30979 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
30980 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
30981 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
30982 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
30983 +#define au_fset_lock(flags, name) \
30984 + do { (flags) |= AuLock_##name; } while (0)
30985 +#define au_fclr_lock(flags, name) \
30986 + do { (flags) &= ~AuLock_##name; } while (0)
30988 +/* ---------------------------------------------------------------------- */
30991 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
30993 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
30994 + unsigned long long max, void *arg);
30995 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30996 + struct super_block *sb, void *arg);
30997 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
30998 +void au_iarray_free(struct inode **a, unsigned long long max);
31000 +void au_remount_refresh(struct super_block *sb, unsigned int do_idop);
31001 +extern const struct super_operations aufs_sop;
31002 +int au_alloc_root(struct super_block *sb);
31003 +extern struct file_system_type aufs_fs_type;
31006 +void au_si_free(struct kobject *kobj);
31007 +struct au_sbinfo *au_si_alloc(struct super_block *sb);
31008 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink);
31010 +unsigned int au_sigen_inc(struct super_block *sb);
31011 +aufs_bindex_t au_new_br_id(struct super_block *sb);
31013 +int si_read_lock(struct super_block *sb, int flags);
31014 +int si_write_lock(struct super_block *sb, int flags);
31015 +int aufs_read_lock(struct dentry *dentry, int flags);
31016 +void aufs_read_unlock(struct dentry *dentry, int flags);
31017 +void aufs_write_lock(struct dentry *dentry);
31018 +void aufs_write_unlock(struct dentry *dentry);
31019 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
31020 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
31022 +/* wbr_policy.c */
31023 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
31024 +extern struct au_wbr_create_operations au_wbr_create_ops[];
31025 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
31026 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
31027 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop);
31030 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
31032 +#ifdef CONFIG_AUFS_FHSM
31035 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
31039 + spin_lock(&fhsm->fhsm_spin);
31040 + pid = fhsm->fhsm_pid;
31041 + spin_unlock(&fhsm->fhsm_spin);
31046 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
31047 +void au_fhsm_wrote_all(struct super_block *sb, int force);
31048 +int au_fhsm_fd(struct super_block *sb, int oflags);
31049 +int au_fhsm_br_alloc(struct au_branch *br);
31050 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
31051 +void au_fhsm_fin(struct super_block *sb);
31052 +void au_fhsm_init(struct au_sbinfo *sbinfo);
31053 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
31054 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
31056 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
31058 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
31059 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
31060 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
31061 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
31062 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
31063 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
31064 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
31065 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
31066 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
31069 +/* ---------------------------------------------------------------------- */
31071 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
31073 + return sb->s_fs_info;
31076 +/* ---------------------------------------------------------------------- */
31078 +#ifdef CONFIG_AUFS_EXPORT
31079 +int au_test_nfsd(void);
31080 +void au_export_init(struct super_block *sb);
31081 +void au_xigen_inc(struct inode *inode);
31082 +int au_xigen_new(struct inode *inode);
31083 +int au_xigen_set(struct super_block *sb, struct path *path);
31084 +void au_xigen_clr(struct super_block *sb);
31086 +static inline int au_busy_or_stale(void)
31088 + if (!au_test_nfsd())
31093 +AuStubInt0(au_test_nfsd, void)
31094 +AuStubVoid(au_export_init, struct super_block *sb)
31095 +AuStubVoid(au_xigen_inc, struct inode *inode)
31096 +AuStubInt0(au_xigen_new, struct inode *inode)
31097 +AuStubInt0(au_xigen_set, struct super_block *sb, struct path *path)
31098 +AuStubVoid(au_xigen_clr, struct super_block *sb)
31099 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
31100 +#endif /* CONFIG_AUFS_EXPORT */
31102 +/* ---------------------------------------------------------------------- */
31104 +#ifdef CONFIG_AUFS_SBILIST
31106 +extern struct hlist_bl_head au_sbilist;
31108 +static inline void au_sbilist_init(void)
31110 + INIT_HLIST_BL_HEAD(&au_sbilist);
31113 +static inline void au_sbilist_add(struct super_block *sb)
31115 + au_hbl_add(&au_sbi(sb)->si_list, &au_sbilist);
31118 +static inline void au_sbilist_del(struct super_block *sb)
31120 + au_hbl_del(&au_sbi(sb)->si_list, &au_sbilist);
31123 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
31124 +static inline void au_sbilist_lock(void)
31126 + hlist_bl_lock(&au_sbilist);
31129 +static inline void au_sbilist_unlock(void)
31131 + hlist_bl_unlock(&au_sbilist);
31133 +#define AuGFP_SBILIST GFP_ATOMIC
31135 +AuStubVoid(au_sbilist_lock, void)
31136 +AuStubVoid(au_sbilist_unlock, void)
31137 +#define AuGFP_SBILIST GFP_NOFS
31138 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
31140 +AuStubVoid(au_sbilist_init, void)
31141 +AuStubVoid(au_sbilist_add, struct super_block *sb)
31142 +AuStubVoid(au_sbilist_del, struct super_block *sb)
31143 +AuStubVoid(au_sbilist_lock, void)
31144 +AuStubVoid(au_sbilist_unlock, void)
31145 +#define AuGFP_SBILIST GFP_NOFS
31148 +/* ---------------------------------------------------------------------- */
31150 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
31153 + * This function is a dynamic '__init' function actually,
31154 + * so the tiny check for si_rwsem is unnecessary.
31156 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
31157 +#ifdef CONFIG_DEBUG_FS
31158 + sbinfo->si_dbgaufs = NULL;
31159 + sbinfo->si_dbgaufs_plink = NULL;
31160 + sbinfo->si_dbgaufs_xib = NULL;
31161 +#ifdef CONFIG_AUFS_EXPORT
31162 + sbinfo->si_dbgaufs_xigen = NULL;
31167 +/* ---------------------------------------------------------------------- */
31169 +/* current->atomic_flags */
31170 +/* this value should never corrupt the ones defined in linux/sched.h */
31171 +#define PFA_AUFS 0x10
31173 +TASK_PFA_TEST(AUFS, test_aufs) /* task_test_aufs */
31174 +TASK_PFA_SET(AUFS, aufs) /* task_set_aufs */
31175 +TASK_PFA_CLEAR(AUFS, aufs) /* task_clear_aufs */
31177 +static inline int si_pid_test(struct super_block *sb)
31179 + return !!task_test_aufs(current);
31182 +static inline void si_pid_clr(struct super_block *sb)
31184 + AuDebugOn(!task_test_aufs(current));
31185 + task_clear_aufs(current);
31188 +static inline void si_pid_set(struct super_block *sb)
31190 + AuDebugOn(task_test_aufs(current));
31191 + task_set_aufs(current);
31194 +/* ---------------------------------------------------------------------- */
31196 +/* lock superblock. mainly for entry point functions */
31197 +#define __si_read_lock(sb) au_rw_read_lock(&au_sbi(sb)->si_rwsem)
31198 +#define __si_write_lock(sb) au_rw_write_lock(&au_sbi(sb)->si_rwsem)
31199 +#define __si_read_trylock(sb) au_rw_read_trylock(&au_sbi(sb)->si_rwsem)
31200 +#define __si_write_trylock(sb) au_rw_write_trylock(&au_sbi(sb)->si_rwsem)
31202 +#define __si_read_trylock_nested(sb) \
31203 + au_rw_read_trylock_nested(&au_sbi(sb)->si_rwsem)
31204 +#define __si_write_trylock_nested(sb) \
31205 + au_rw_write_trylock_nested(&au_sbi(sb)->si_rwsem)
31208 +#define __si_read_unlock(sb) au_rw_read_unlock(&au_sbi(sb)->si_rwsem)
31209 +#define __si_write_unlock(sb) au_rw_write_unlock(&au_sbi(sb)->si_rwsem)
31210 +#define __si_downgrade_lock(sb) au_rw_dgrade_lock(&au_sbi(sb)->si_rwsem)
31212 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
31213 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
31214 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
31216 +static inline void si_noflush_read_lock(struct super_block *sb)
31218 + __si_read_lock(sb);
31222 +static inline int si_noflush_read_trylock(struct super_block *sb)
31226 + locked = __si_read_trylock(sb);
31232 +static inline void si_noflush_write_lock(struct super_block *sb)
31234 + __si_write_lock(sb);
31238 +static inline int si_noflush_write_trylock(struct super_block *sb)
31242 + locked = __si_write_trylock(sb);
31248 +#if 0 /* reserved */
31249 +static inline int si_read_trylock(struct super_block *sb, int flags)
31251 + if (au_ftest_lock(flags, FLUSH))
31252 + au_nwt_flush(&au_sbi(sb)->si_nowait);
31253 + return si_noflush_read_trylock(sb);
31257 +static inline void si_read_unlock(struct super_block *sb)
31260 + __si_read_unlock(sb);
31263 +#if 0 /* reserved */
31264 +static inline int si_write_trylock(struct super_block *sb, int flags)
31266 + if (au_ftest_lock(flags, FLUSH))
31267 + au_nwt_flush(&au_sbi(sb)->si_nowait);
31268 + return si_noflush_write_trylock(sb);
31272 +static inline void si_write_unlock(struct super_block *sb)
31275 + __si_write_unlock(sb);
31278 +#if 0 /* reserved */
31279 +static inline void si_downgrade_lock(struct super_block *sb)
31281 + __si_downgrade_lock(sb);
31285 +/* ---------------------------------------------------------------------- */
31287 +static inline aufs_bindex_t au_sbbot(struct super_block *sb)
31289 + SiMustAnyLock(sb);
31290 + return au_sbi(sb)->si_bbot;
31293 +static inline unsigned int au_mntflags(struct super_block *sb)
31295 + SiMustAnyLock(sb);
31296 + return au_sbi(sb)->si_mntflags;
31299 +static inline unsigned int au_sigen(struct super_block *sb)
31301 + SiMustAnyLock(sb);
31302 + return au_sbi(sb)->si_generation;
31305 +static inline struct au_branch *au_sbr(struct super_block *sb,
31306 + aufs_bindex_t bindex)
31308 + SiMustAnyLock(sb);
31309 + return au_sbi(sb)->si_branch[0 + bindex];
31312 +static inline loff_t au_xi_maxent(struct super_block *sb)
31314 + SiMustAnyLock(sb);
31315 + return au_sbi(sb)->si_ximaxent;
31318 +#endif /* __KERNEL__ */
31319 +#endif /* __AUFS_SUPER_H__ */
31320 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
31321 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
31322 +++ linux/fs/aufs/sysaufs.c 2022-11-05 23:02:18.969222617 +0100
31324 +// SPDX-License-Identifier: GPL-2.0
31326 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31328 + * This program is free software; you can redistribute it and/or modify
31329 + * it under the terms of the GNU General Public License as published by
31330 + * the Free Software Foundation; either version 2 of the License, or
31331 + * (at your option) any later version.
31333 + * This program is distributed in the hope that it will be useful,
31334 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31335 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31336 + * GNU General Public License for more details.
31338 + * You should have received a copy of the GNU General Public License
31339 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31343 + * sysfs interface and lifetime management
31344 + * they are necessary regardless sysfs is disabled.
31347 +#include <linux/random.h>
31350 +unsigned long sysaufs_si_mask;
31351 +struct kset *sysaufs_kset;
31353 +#define AuSiAttr(_name) { \
31354 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
31355 + .show = sysaufs_si_##_name, \
31358 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
31359 +struct attribute *sysaufs_si_attrs[] = {
31360 + &sysaufs_si_attr_xi_path.attr,
31363 +ATTRIBUTE_GROUPS(sysaufs_si);
31365 +static const struct sysfs_ops au_sbi_ops = {
31366 + .show = sysaufs_si_show
31369 +static struct kobj_type au_sbi_ktype = {
31370 + .release = au_si_free,
31371 + .sysfs_ops = &au_sbi_ops,
31372 + .default_groups = sysaufs_si_groups
31375 +/* ---------------------------------------------------------------------- */
31377 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
31381 + sbinfo->si_kobj.kset = sysaufs_kset;
31382 + /* cf. sysaufs_name() */
31383 + err = kobject_init_and_add
31384 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
31385 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
31390 +void sysaufs_fin(void)
31392 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31393 + kset_unregister(sysaufs_kset);
31396 +int __init sysaufs_init(void)
31401 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
31402 + } while (!sysaufs_si_mask);
31405 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
31406 + if (unlikely(!sysaufs_kset))
31408 + err = PTR_ERR(sysaufs_kset);
31409 + if (IS_ERR(sysaufs_kset))
31411 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
31412 + if (unlikely(err))
31413 + kset_unregister(sysaufs_kset);
31418 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
31419 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
31420 +++ linux/fs/aufs/sysaufs.h 2022-11-05 23:02:18.969222617 +0100
31422 +/* SPDX-License-Identifier: GPL-2.0 */
31424 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31426 + * This program is free software; you can redistribute it and/or modify
31427 + * it under the terms of the GNU General Public License as published by
31428 + * the Free Software Foundation; either version 2 of the License, or
31429 + * (at your option) any later version.
31431 + * This program is distributed in the hope that it will be useful,
31432 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31433 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31434 + * GNU General Public License for more details.
31436 + * You should have received a copy of the GNU General Public License
31437 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31441 + * sysfs interface and mount lifetime management
31444 +#ifndef __SYSAUFS_H__
31445 +#define __SYSAUFS_H__
31449 +#include <linux/sysfs.h>
31450 +#include "module.h"
31452 +struct super_block;
31455 +struct sysaufs_si_attr {
31456 + struct attribute attr;
31457 + int (*show)(struct seq_file *seq, struct super_block *sb);
31460 +/* ---------------------------------------------------------------------- */
31463 +extern unsigned long sysaufs_si_mask;
31464 +extern struct kset *sysaufs_kset;
31465 +extern struct attribute *sysaufs_si_attrs[];
31466 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
31467 +int __init sysaufs_init(void);
31468 +void sysaufs_fin(void);
31470 +/* ---------------------------------------------------------------------- */
31472 +/* some people doesn't like to show a pointer in kernel */
31473 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
31475 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
31478 +#define SysaufsSiNamePrefix "si_"
31479 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
31480 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
31482 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
31483 + sysaufs_si_id(sbinfo));
31487 +#ifdef CONFIG_SYSFS
31489 +extern struct attribute_group *sysaufs_attr_group;
31491 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
31492 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31494 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
31495 +#ifdef CONFIG_COMPAT
31496 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
31499 +void sysaufs_br_init(struct au_branch *br);
31500 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
31501 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
31503 +#define sysaufs_brs_init() do {} while (0)
31506 +#define sysaufs_attr_group NULL
31508 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
31509 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
31510 + struct attribute *attr, char *buf)
31511 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
31512 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
31513 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
31515 +static inline void sysaufs_brs_init(void)
31520 +#endif /* CONFIG_SYSFS */
31522 +#endif /* __KERNEL__ */
31523 +#endif /* __SYSAUFS_H__ */
31524 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
31525 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
31526 +++ linux/fs/aufs/sysfs.c 2022-11-05 23:02:18.969222617 +0100
31528 +// SPDX-License-Identifier: GPL-2.0
31530 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31532 + * This program is free software; you can redistribute it and/or modify
31533 + * it under the terms of the GNU General Public License as published by
31534 + * the Free Software Foundation; either version 2 of the License, or
31535 + * (at your option) any later version.
31537 + * This program is distributed in the hope that it will be useful,
31538 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31539 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31540 + * GNU General Public License for more details.
31542 + * You should have received a copy of the GNU General Public License
31543 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31547 + * sysfs interface
31550 +#include <linux/compat.h>
31551 +#include <linux/seq_file.h>
31554 +#ifdef CONFIG_AUFS_FS_MODULE
31555 +/* this entry violates the "one line per file" policy of sysfs */
31556 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
31560 + static char *conf =
31561 +/* this file is generated at compiling */
31562 +#include "conf.str"
31565 + err = snprintf(buf, PAGE_SIZE, conf);
31566 + if (unlikely(err >= PAGE_SIZE))
31571 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
31574 +static struct attribute *au_attr[] = {
31575 +#ifdef CONFIG_AUFS_FS_MODULE
31576 + &au_config_attr.attr,
31578 + NULL, /* need to NULL terminate the list of attributes */
31581 +static struct attribute_group sysaufs_attr_group_body = {
31585 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
31587 +/* ---------------------------------------------------------------------- */
31589 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
31593 + SiMustAnyLock(sb);
31596 + if (au_opt_test(au_mntflags(sb), XINO)) {
31597 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
31598 + seq_putc(seq, '\n');
31604 + * the lifetime of branch is independent from the entry under sysfs.
31605 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
31608 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
31609 + aufs_bindex_t bindex, int idx)
31612 + struct path path;
31613 + struct dentry *root;
31614 + struct au_branch *br;
31615 + au_br_perm_str_t perm;
31617 + AuDbg("b%d\n", bindex);
31620 + root = sb->s_root;
31621 + di_read_lock_parent(root, !AuLock_IR);
31622 + br = au_sbr(sb, bindex);
31625 + case AuBrSysfs_BR:
31626 + path.mnt = au_br_mnt(br);
31627 + path.dentry = au_h_dptr(root, bindex);
31628 + err = au_seq_path(seq, &path);
31630 + au_optstr_br_perm(&perm, br->br_perm);
31631 + seq_printf(seq, "=%s\n", perm.a);
31634 + case AuBrSysfs_BRID:
31635 + seq_printf(seq, "%d\n", br->br_id);
31638 + di_read_unlock(root, !AuLock_IR);
31639 + if (unlikely(err || seq_has_overflowed(seq)))
31645 +/* ---------------------------------------------------------------------- */
31647 +static struct seq_file *au_seq(char *p, ssize_t len)
31649 + struct seq_file *seq;
31651 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
31653 + /* mutex_init(&seq.lock); */
31656 + return seq; /* success */
31659 + seq = ERR_PTR(-ENOMEM);
31663 +#define SysaufsBr_PREFIX "br"
31664 +#define SysaufsBrid_PREFIX "brid"
31666 +/* todo: file size may exceed PAGE_SIZE */
31667 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31673 + aufs_bindex_t bbot;
31674 + struct au_sbinfo *sbinfo;
31675 + struct super_block *sb;
31676 + struct seq_file *seq;
31678 + struct attribute **cattr;
31680 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
31681 + sb = sbinfo->si_sb;
31684 + * prevent a race condition between sysfs and aufs.
31685 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
31686 + * prohibits maintaining the sysfs entries.
31687 + * hew we acquire read lock after sysfs_get_active_two().
31688 + * on the other hand, the remount process may maintain the sysfs/aufs
31689 + * entries after acquiring write lock.
31690 + * it can cause a deadlock.
31691 + * simply we gave up processing read here.
31694 + if (unlikely(!si_noflush_read_trylock(sb)))
31697 + seq = au_seq(buf, PAGE_SIZE);
31698 + err = PTR_ERR(seq);
31702 + name = (void *)attr->name;
31703 + cattr = sysaufs_si_attrs;
31705 + if (!strcmp(name, (*cattr)->name)) {
31706 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
31713 + if (!strncmp(name, SysaufsBrid_PREFIX,
31714 + sizeof(SysaufsBrid_PREFIX) - 1)) {
31715 + idx = AuBrSysfs_BRID;
31716 + name += sizeof(SysaufsBrid_PREFIX) - 1;
31717 + } else if (!strncmp(name, SysaufsBr_PREFIX,
31718 + sizeof(SysaufsBr_PREFIX) - 1)) {
31719 + idx = AuBrSysfs_BR;
31720 + name += sizeof(SysaufsBr_PREFIX) - 1;
31724 + err = kstrtol(name, 10, &l);
31726 + bbot = au_sbbot(sb);
31728 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
31735 + err = seq->count;
31736 + /* sysfs limit */
31737 + if (unlikely(err == PAGE_SIZE))
31740 + au_kfree_rcu(seq);
31742 + si_read_unlock(sb);
31747 +/* ---------------------------------------------------------------------- */
31749 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
31753 + aufs_bindex_t bindex, bbot;
31756 + struct seq_file *seq;
31757 + struct au_branch *br;
31759 + si_read_lock(sb, AuLock_FLUSH);
31760 + bbot = au_sbbot(sb);
31766 + buf = (void *)__get_free_page(GFP_NOFS);
31767 + if (unlikely(!buf))
31770 + seq = au_seq(buf, PAGE_SIZE);
31771 + err = PTR_ERR(seq);
31775 + sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
31776 + for (bindex = 0; bindex <= bbot; bindex++, arg++) {
31777 + /* VERIFY_WRITE */
31778 + err = !access_ok(arg, sizeof(*arg));
31779 + if (unlikely(err))
31782 + br = au_sbr(sb, bindex);
31783 + brid = br->br_id;
31784 + BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
31785 + err = __put_user(brid, &arg->id);
31786 + if (unlikely(err))
31789 + BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
31790 + err = __put_user(br->br_perm, &arg->perm);
31791 + if (unlikely(err))
31794 + err = au_seq_path(seq, &br->br_path);
31795 + if (unlikely(err))
31797 + seq_putc(seq, '\0');
31798 + if (!seq_has_overflowed(seq)) {
31799 + err = copy_to_user(arg->path, seq->buf, seq->count);
31801 + if (unlikely(err))
31808 + if (unlikely(err))
31812 + au_kfree_rcu(seq);
31814 + free_page((unsigned long)buf);
31816 + si_read_unlock(sb);
31820 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
31822 + return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
31825 +#ifdef CONFIG_COMPAT
31826 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
31828 + return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
31832 +/* ---------------------------------------------------------------------- */
31834 +void sysaufs_br_init(struct au_branch *br)
31837 + struct au_brsysfs *br_sysfs;
31838 + struct attribute *attr;
31840 + br_sysfs = br->br_sysfs;
31841 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31842 + attr = &br_sysfs->attr;
31843 + sysfs_attr_init(attr);
31844 + attr->name = br_sysfs->name;
31845 + attr->mode = 0444;
31850 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
31852 + struct au_branch *br;
31853 + struct kobject *kobj;
31854 + struct au_brsysfs *br_sysfs;
31856 + aufs_bindex_t bbot;
31858 + if (!sysaufs_brs)
31861 + kobj = &au_sbi(sb)->si_kobj;
31862 + bbot = au_sbbot(sb);
31863 + for (; bindex <= bbot; bindex++) {
31864 + br = au_sbr(sb, bindex);
31865 + br_sysfs = br->br_sysfs;
31866 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31867 + sysfs_remove_file(kobj, &br_sysfs->attr);
31873 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
31876 + aufs_bindex_t bbot;
31877 + struct kobject *kobj;
31878 + struct au_branch *br;
31879 + struct au_brsysfs *br_sysfs;
31881 + if (!sysaufs_brs)
31884 + kobj = &au_sbi(sb)->si_kobj;
31885 + bbot = au_sbbot(sb);
31886 + for (; bindex <= bbot; bindex++) {
31887 + br = au_sbr(sb, bindex);
31888 + br_sysfs = br->br_sysfs;
31889 + snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
31890 + SysaufsBr_PREFIX "%d", bindex);
31891 + snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
31892 + SysaufsBrid_PREFIX "%d", bindex);
31893 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31894 + err = sysfs_create_file(kobj, &br_sysfs->attr);
31895 + if (unlikely(err))
31896 + pr_warn("failed %s under sysfs(%d)\n",
31897 + br_sysfs->name, err);
31902 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
31903 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
31904 +++ linux/fs/aufs/sysrq.c 2022-11-05 23:02:18.969222617 +0100
31906 +// SPDX-License-Identifier: GPL-2.0
31908 + * Copyright (C) 2005-2022 Junjiro R. Okajima
31910 + * This program is free software; you can redistribute it and/or modify
31911 + * it under the terms of the GNU General Public License as published by
31912 + * the Free Software Foundation; either version 2 of the License, or
31913 + * (at your option) any later version.
31915 + * This program is distributed in the hope that it will be useful,
31916 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31917 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31918 + * GNU General Public License for more details.
31920 + * You should have received a copy of the GNU General Public License
31921 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31925 + * magic sysrq handler
31928 +/* #include <linux/sysrq.h> */
31929 +#include <linux/writeback.h>
31932 +/* ---------------------------------------------------------------------- */
31934 +static void sysrq_sb(struct super_block *sb)
31937 + struct au_sbinfo *sbinfo;
31938 + struct file *file;
31939 + struct hlist_bl_head *files;
31940 + struct hlist_bl_node *pos;
31941 + struct au_finfo *finfo;
31944 + plevel = au_plevel;
31945 + au_plevel = KERN_WARNING;
31947 + /* since we define pr_fmt, call printk directly */
31948 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
31950 + sbinfo = au_sbi(sb);
31951 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
31952 + pr("superblock\n");
31955 +#if 0 /* reserved */
31957 + int err, i, j, ndentry;
31958 + struct au_dcsub_pages dpages;
31959 + struct au_dpage *dpage;
31961 + err = au_dpages_init(&dpages, GFP_ATOMIC);
31962 + if (unlikely(err))
31964 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
31966 + for (i = 0; i < dpages.ndpage; i++) {
31967 + dpage = dpages.dpages + i;
31968 + ndentry = dpage->ndentry;
31969 + for (j = 0; j < ndentry; j++)
31970 + au_dpri_dentry(dpage->dentries[j]);
31972 + au_dpages_free(&dpages);
31976 + pr("isolated inode\n");
31977 + spin_lock(&sb->s_inode_list_lock);
31978 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
31979 + spin_lock(&i->i_lock);
31980 + if (hlist_empty(&i->i_dentry))
31981 + au_dpri_inode(i);
31982 + spin_unlock(&i->i_lock);
31984 + spin_unlock(&sb->s_inode_list_lock);
31987 + files = &au_sbi(sb)->si_files;
31988 + hlist_bl_lock(files);
31989 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
31992 + file = finfo->fi_file;
31993 + mode = file_inode(file)->i_mode;
31994 + if (!special_file(mode))
31995 + au_dpri_file(file);
31997 + hlist_bl_unlock(files);
32001 + au_plevel = plevel;
32004 +/* ---------------------------------------------------------------------- */
32006 +/* module parameter */
32007 +static char *aufs_sysrq_key = "a";
32008 +module_param_named(sysrq, aufs_sysrq_key, charp, 0444);
32009 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
32011 +static void au_sysrq(int key __maybe_unused)
32013 + struct au_sbinfo *sbinfo;
32014 + struct hlist_bl_node *pos;
32017 + au_sbilist_lock();
32018 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
32019 + sysrq_sb(sbinfo->si_sb);
32020 + au_sbilist_unlock();
32024 +static struct sysrq_key_op au_sysrq_op = {
32025 + .handler = au_sysrq,
32026 + .help_msg = "Aufs",
32027 + .action_msg = "Aufs",
32028 + .enable_mask = SYSRQ_ENABLE_DUMP
32031 +/* ---------------------------------------------------------------------- */
32033 +int __init au_sysrq_init(void)
32039 + key = *aufs_sysrq_key;
32040 + if ('a' <= key && key <= 'z')
32041 + err = register_sysrq_key(key, &au_sysrq_op);
32042 + if (unlikely(err))
32043 + pr_err("err %d, sysrq=%c\n", err, key);
32047 +void au_sysrq_fin(void)
32051 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
32052 + if (unlikely(err))
32053 + pr_err("err %d (ignored)\n", err);
32055 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
32056 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
32057 +++ linux/fs/aufs/vdir.c 2022-12-17 09:21:34.799855195 +0100
32059 +// SPDX-License-Identifier: GPL-2.0
32061 + * Copyright (C) 2005-2022 Junjiro R. Okajima
32063 + * This program is free software; you can redistribute it and/or modify
32064 + * it under the terms of the GNU General Public License as published by
32065 + * the Free Software Foundation; either version 2 of the License, or
32066 + * (at your option) any later version.
32068 + * This program is distributed in the hope that it will be useful,
32069 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32070 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32071 + * GNU General Public License for more details.
32073 + * You should have received a copy of the GNU General Public License
32074 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32078 + * virtual or vertical directory
32081 +#include <linux/iversion.h>
32084 +static unsigned int calc_size(int nlen)
32086 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
32089 +static int set_deblk_end(union au_vdir_deblk_p *p,
32090 + union au_vdir_deblk_p *deblk_end)
32092 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
32093 + p->de->de_str.len = 0;
32097 + return -1; /* error */
32100 +/* returns true or false */
32101 +static int is_deblk_end(union au_vdir_deblk_p *p,
32102 + union au_vdir_deblk_p *deblk_end)
32104 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
32105 + return !p->de->de_str.len;
32109 +static unsigned char *last_deblk(struct au_vdir *vdir)
32111 + return vdir->vd_deblk[vdir->vd_nblk - 1];
32114 +/* ---------------------------------------------------------------------- */
32116 +/* estimate the appropriate size for name hash table */
32117 +unsigned int au_rdhash_est(loff_t sz)
32125 + if (sz < AUFS_RDHASH_DEF)
32126 + n = AUFS_RDHASH_DEF;
32127 + /* pr_info("n %u\n", n); */
32132 + * the allocated memory has to be freed by
32133 + * au_nhash_wh_free() or au_nhash_de_free().
32135 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
32137 + struct hlist_head *head;
32141 + sz = sizeof(*nhash->nh_head) * num_hash;
32142 + head = kmalloc(sz, gfp);
32144 + nhash->nh_num = num_hash;
32145 + nhash->nh_head = head;
32146 + for (u = 0; u < num_hash; u++)
32147 + INIT_HLIST_HEAD(head++);
32148 + return 0; /* success */
32154 +static void nhash_count(struct hlist_head *head)
32156 +#if 0 /* debugging */
32158 + struct hlist_node *pos;
32161 + hlist_for_each(pos, head)
32163 + pr_info("%lu\n", n);
32167 +static void au_nhash_wh_do_free(struct hlist_head *head)
32169 + struct au_vdir_wh *pos;
32170 + struct hlist_node *node;
32172 + hlist_for_each_entry_safe(pos, node, head, wh_hash)
32173 + au_kfree_rcu(pos);
32176 +static void au_nhash_de_do_free(struct hlist_head *head)
32178 + struct au_vdir_dehstr *pos;
32179 + struct hlist_node *node;
32181 + hlist_for_each_entry_safe(pos, node, head, hash)
32182 + au_cache_free_vdir_dehstr(pos);
32185 +static void au_nhash_do_free(struct au_nhash *nhash,
32186 + void (*free)(struct hlist_head *head))
32189 + struct hlist_head *head;
32191 + n = nhash->nh_num;
32195 + head = nhash->nh_head;
32196 + while (n-- > 0) {
32197 + nhash_count(head);
32200 + au_kfree_try_rcu(nhash->nh_head);
32203 +void au_nhash_wh_free(struct au_nhash *whlist)
32205 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
32208 +static void au_nhash_de_free(struct au_nhash *delist)
32210 + au_nhash_do_free(delist, au_nhash_de_do_free);
32213 +/* ---------------------------------------------------------------------- */
32215 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
32219 + unsigned int u, n;
32220 + struct hlist_head *head;
32221 + struct au_vdir_wh *pos;
32224 + n = whlist->nh_num;
32225 + head = whlist->nh_head;
32226 + for (u = 0; u < n; u++, head++)
32227 + hlist_for_each_entry(pos, head, wh_hash)
32228 + if (pos->wh_bindex == btgt && ++num > limit)
32233 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
32234 + unsigned char *name,
32235 + unsigned int len)
32238 + /* const unsigned int magic_bit = 12; */
32240 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
32247 + /* v = hash_long(v, magic_bit); */
32248 + v %= nhash->nh_num;
32249 + return nhash->nh_head + v;
32252 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
32255 + return str->len == nlen && !memcmp(str->name, name, nlen);
32258 +/* returns found or not */
32259 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
32261 + struct hlist_head *head;
32262 + struct au_vdir_wh *pos;
32263 + struct au_vdir_destr *str;
32265 + head = au_name_hash(whlist, name, nlen);
32266 + hlist_for_each_entry(pos, head, wh_hash) {
32267 + str = &pos->wh_str;
32268 + AuDbg("%.*s\n", str->len, str->name);
32269 + if (au_nhash_test_name(str, name, nlen))
32275 +/* returns found(true) or not */
32276 +static int test_known(struct au_nhash *delist, char *name, int nlen)
32278 + struct hlist_head *head;
32279 + struct au_vdir_dehstr *pos;
32280 + struct au_vdir_destr *str;
32282 + head = au_name_hash(delist, name, nlen);
32283 + hlist_for_each_entry(pos, head, hash) {
32285 + AuDbg("%.*s\n", str->len, str->name);
32286 + if (au_nhash_test_name(str, name, nlen))
32292 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
32293 + unsigned char d_type)
32295 +#ifdef CONFIG_AUFS_SHWH
32296 + wh->wh_ino = ino;
32297 + wh->wh_type = d_type;
32301 +/* ---------------------------------------------------------------------- */
32303 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
32304 + unsigned int d_type, aufs_bindex_t bindex,
32305 + unsigned char shwh)
32308 + struct au_vdir_destr *str;
32309 + struct au_vdir_wh *wh;
32311 + AuDbg("%.*s\n", nlen, name);
32312 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
32315 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
32316 + if (unlikely(!wh))
32320 + wh->wh_bindex = bindex;
32322 + au_shwh_init_wh(wh, ino, d_type);
32323 + str = &wh->wh_str;
32325 + memcpy(str->name, name, nlen);
32326 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
32333 +static int append_deblk(struct au_vdir *vdir)
32336 + unsigned long ul;
32337 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
32338 + union au_vdir_deblk_p p, deblk_end;
32339 + unsigned char **o;
32342 + o = au_krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
32343 + GFP_NOFS, /*may_shrink*/0);
32344 + if (unlikely(!o))
32347 + vdir->vd_deblk = o;
32348 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
32350 + ul = vdir->vd_nblk++;
32351 + vdir->vd_deblk[ul] = p.deblk;
32352 + vdir->vd_last.ul = ul;
32353 + vdir->vd_last.p.deblk = p.deblk;
32354 + deblk_end.deblk = p.deblk + deblk_sz;
32355 + err = set_deblk_end(&p, &deblk_end);
32362 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
32363 + unsigned int d_type, struct au_nhash *delist)
32367 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
32368 + union au_vdir_deblk_p p, *room, deblk_end;
32369 + struct au_vdir_dehstr *dehstr;
32371 + p.deblk = last_deblk(vdir);
32372 + deblk_end.deblk = p.deblk + deblk_sz;
32373 + room = &vdir->vd_last.p;
32374 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
32375 + || !is_deblk_end(room, &deblk_end));
32377 + sz = calc_size(nlen);
32378 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
32379 + err = append_deblk(vdir);
32380 + if (unlikely(err))
32383 + p.deblk = last_deblk(vdir);
32384 + deblk_end.deblk = p.deblk + deblk_sz;
32386 + AuDebugOn(room->deblk != p.deblk);
32390 + dehstr = au_cache_alloc_vdir_dehstr();
32391 + if (unlikely(!dehstr))
32394 + dehstr->str = &room->de->de_str;
32395 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
32396 + room->de->de_ino = ino;
32397 + room->de->de_type = d_type;
32398 + room->de->de_str.len = nlen;
32399 + memcpy(room->de->de_str.name, name, nlen);
32402 + room->deblk += sz;
32403 + if (unlikely(set_deblk_end(room, &deblk_end)))
32404 + err = append_deblk(vdir);
32411 +/* ---------------------------------------------------------------------- */
32413 +void au_vdir_free(struct au_vdir *vdir)
32415 + unsigned char **deblk;
32417 + deblk = vdir->vd_deblk;
32418 + while (vdir->vd_nblk--)
32419 + au_kfree_try_rcu(*deblk++);
32420 + au_kfree_try_rcu(vdir->vd_deblk);
32421 + au_cache_free_vdir(vdir);
32424 +static struct au_vdir *alloc_vdir(struct file *file)
32426 + struct au_vdir *vdir;
32427 + struct super_block *sb;
32430 + sb = file->f_path.dentry->d_sb;
32431 + SiMustAnyLock(sb);
32434 + vdir = au_cache_alloc_vdir();
32435 + if (unlikely(!vdir))
32438 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
32439 + if (unlikely(!vdir->vd_deblk))
32442 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
32443 + if (!vdir->vd_deblk_sz) {
32444 + /* estimate the appropriate size for deblk */
32445 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
32446 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
32448 + vdir->vd_nblk = 0;
32449 + vdir->vd_version = 0;
32450 + vdir->vd_jiffy = 0;
32451 + err = append_deblk(vdir);
32453 + return vdir; /* success */
32455 + au_kfree_try_rcu(vdir->vd_deblk);
32458 + au_cache_free_vdir(vdir);
32460 + vdir = ERR_PTR(err);
32464 +static int reinit_vdir(struct au_vdir *vdir)
32467 + union au_vdir_deblk_p p, deblk_end;
32469 + while (vdir->vd_nblk > 1) {
32470 + au_kfree_try_rcu(vdir->vd_deblk[vdir->vd_nblk - 1]);
32471 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
32474 + p.deblk = vdir->vd_deblk[0];
32475 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
32476 + err = set_deblk_end(&p, &deblk_end);
32477 + /* keep vd_dblk_sz */
32478 + vdir->vd_last.ul = 0;
32479 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32480 + vdir->vd_version = 0;
32481 + vdir->vd_jiffy = 0;
32486 +/* ---------------------------------------------------------------------- */
32488 +#define AuFillVdir_CALLED 1
32489 +#define AuFillVdir_WHABLE (1 << 1)
32490 +#define AuFillVdir_SHWH (1 << 2)
32491 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
32492 +#define au_fset_fillvdir(flags, name) \
32493 + do { (flags) |= AuFillVdir_##name; } while (0)
32494 +#define au_fclr_fillvdir(flags, name) \
32495 + do { (flags) &= ~AuFillVdir_##name; } while (0)
32497 +#ifndef CONFIG_AUFS_SHWH
32498 +#undef AuFillVdir_SHWH
32499 +#define AuFillVdir_SHWH 0
32502 +struct fillvdir_arg {
32503 + struct dir_context ctx;
32504 + struct file *file;
32505 + struct au_vdir *vdir;
32506 + struct au_nhash delist;
32507 + struct au_nhash whlist;
32508 + aufs_bindex_t bindex;
32509 + unsigned int flags;
32513 +static bool fillvdir(struct dir_context *ctx, const char *__name, int nlen,
32514 + loff_t offset __maybe_unused, u64 h_ino,
32515 + unsigned int d_type)
32517 + struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
32518 + char *name = (void *)__name;
32519 + struct super_block *sb;
32521 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
32524 + sb = arg->file->f_path.dentry->d_sb;
32525 + au_fset_fillvdir(arg->flags, CALLED);
32527 + if (nlen <= AUFS_WH_PFX_LEN
32528 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
32529 + if (test_known(&arg->delist, name, nlen)
32530 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
32531 + goto out; /* already exists or whiteouted */
32533 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
32535 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
32536 + d_type = DT_UNKNOWN;
32537 + arg->err = append_de(arg->vdir, name, nlen, ino,
32538 + d_type, &arg->delist);
32540 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
32541 + name += AUFS_WH_PFX_LEN;
32542 + nlen -= AUFS_WH_PFX_LEN;
32543 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
32544 + goto out; /* already whiteouted */
32546 + ino = 0; /* just to suppress a warning */
32548 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
32551 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
32552 + d_type = DT_UNKNOWN;
32553 + arg->err = au_nhash_append_wh
32554 + (&arg->whlist, name, nlen, ino, d_type,
32555 + arg->bindex, shwh);
32561 + arg->vdir->vd_jiffy = jiffies;
32563 + AuTraceErr(arg->err);
32564 + return !arg->err;
32567 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
32568 + struct au_nhash *whlist, struct au_nhash *delist)
32570 +#ifdef CONFIG_AUFS_SHWH
32572 + unsigned int nh, u;
32573 + struct hlist_head *head;
32574 + struct au_vdir_wh *pos;
32575 + struct hlist_node *n;
32577 + struct au_vdir_destr *destr;
32579 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
32582 + o = p = (void *)__get_free_page(GFP_NOFS);
32583 + if (unlikely(!p))
32587 + nh = whlist->nh_num;
32588 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32589 + p += AUFS_WH_PFX_LEN;
32590 + for (u = 0; u < nh; u++) {
32591 + head = whlist->nh_head + u;
32592 + hlist_for_each_entry_safe(pos, n, head, wh_hash) {
32593 + destr = &pos->wh_str;
32594 + memcpy(p, destr->name, destr->len);
32595 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
32596 + pos->wh_ino, pos->wh_type, delist);
32597 + if (unlikely(err))
32602 + free_page((unsigned long)o);
32612 +static int au_do_read_vdir(struct fillvdir_arg *arg)
32615 + unsigned int rdhash;
32617 + aufs_bindex_t bbot, bindex, btop;
32618 + unsigned char shwh;
32619 + struct file *hf, *file;
32620 + struct super_block *sb;
32622 + file = arg->file;
32623 + sb = file->f_path.dentry->d_sb;
32624 + SiMustAnyLock(sb);
32626 + rdhash = au_sbi(sb)->si_rdhash;
32628 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
32629 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
32630 + if (unlikely(err))
32632 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
32633 + if (unlikely(err))
32639 + if (au_opt_test(au_mntflags(sb), SHWH)) {
32641 + au_fset_fillvdir(arg->flags, SHWH);
32643 + btop = au_fbtop(file);
32644 + bbot = au_fbbot_dir(file);
32645 + for (bindex = btop; !err && bindex <= bbot; bindex++) {
32646 + hf = au_hf_dir(file, bindex);
32650 + offset = vfsub_llseek(hf, 0, SEEK_SET);
32652 + if (unlikely(offset))
32655 + arg->bindex = bindex;
32656 + au_fclr_fillvdir(arg->flags, WHABLE);
32658 + || (bindex != bbot
32659 + && au_br_whable(au_sbr_perm(sb, bindex))))
32660 + au_fset_fillvdir(arg->flags, WHABLE);
32663 + au_fclr_fillvdir(arg->flags, CALLED);
32665 + err = vfsub_iterate_dir(hf, &arg->ctx);
32668 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
32671 + * dir_relax() may be good for concurrency, but aufs should not
32672 + * use it since it will cause a lockdep problem.
32676 + if (!err && shwh)
32677 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
32679 + au_nhash_wh_free(&arg->whlist);
32682 + au_nhash_de_free(&arg->delist);
32687 +static int read_vdir(struct file *file, int may_read)
32690 + unsigned long expire;
32691 + unsigned char do_read;
32692 + struct fillvdir_arg arg = {
32694 + .actor = fillvdir
32697 + struct inode *inode;
32698 + struct au_vdir *vdir, *allocated;
32701 + inode = file_inode(file);
32702 + IMustLock(inode);
32703 + IiMustWriteLock(inode);
32704 + SiMustAnyLock(inode->i_sb);
32706 + allocated = NULL;
32708 + expire = au_sbi(inode->i_sb)->si_rdcache;
32709 + vdir = au_ivdir(inode);
32712 + vdir = alloc_vdir(file);
32713 + err = PTR_ERR(vdir);
32714 + if (IS_ERR(vdir))
32717 + allocated = vdir;
32718 + } else if (may_read
32719 + && (!inode_eq_iversion(inode, vdir->vd_version)
32720 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
32722 + err = reinit_vdir(vdir);
32723 + if (unlikely(err))
32728 + return 0; /* success */
32732 + err = au_do_read_vdir(&arg);
32734 + /* file->f_pos = 0; */ /* todo: ctx->pos? */
32735 + vdir->vd_version = inode_query_iversion(inode);
32736 + vdir->vd_last.ul = 0;
32737 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32739 + au_set_ivdir(inode, allocated);
32740 + } else if (allocated)
32741 + au_vdir_free(allocated);
32747 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
32750 + unsigned long ul, n;
32751 + const unsigned int deblk_sz = src->vd_deblk_sz;
32753 + AuDebugOn(tgt->vd_nblk != 1);
32756 + if (tgt->vd_nblk < src->vd_nblk) {
32757 + unsigned char **p;
32759 + p = au_krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
32760 + GFP_NOFS, /*may_shrink*/0);
32761 + if (unlikely(!p))
32763 + tgt->vd_deblk = p;
32766 + if (tgt->vd_deblk_sz != deblk_sz) {
32767 + unsigned char *p;
32769 + tgt->vd_deblk_sz = deblk_sz;
32770 + p = au_krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS,
32771 + /*may_shrink*/1);
32772 + if (unlikely(!p))
32774 + tgt->vd_deblk[0] = p;
32776 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
32777 + tgt->vd_version = src->vd_version;
32778 + tgt->vd_jiffy = src->vd_jiffy;
32780 + n = src->vd_nblk;
32781 + for (ul = 1; ul < n; ul++) {
32782 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
32784 + if (unlikely(!tgt->vd_deblk[ul]))
32788 + tgt->vd_nblk = n;
32789 + tgt->vd_last.ul = tgt->vd_last.ul;
32790 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
32791 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
32792 + - src->vd_deblk[src->vd_last.ul];
32794 + return 0; /* success */
32797 + rerr = reinit_vdir(tgt);
32802 +int au_vdir_init(struct file *file)
32805 + struct inode *inode;
32806 + struct au_vdir *vdir_cache, *allocated;
32808 + /* test file->f_pos here instead of ctx->pos */
32809 + err = read_vdir(file, !file->f_pos);
32810 + if (unlikely(err))
32813 + allocated = NULL;
32814 + vdir_cache = au_fvdir_cache(file);
32815 + if (!vdir_cache) {
32816 + vdir_cache = alloc_vdir(file);
32817 + err = PTR_ERR(vdir_cache);
32818 + if (IS_ERR(vdir_cache))
32820 + allocated = vdir_cache;
32821 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
32822 + /* test file->f_pos here instead of ctx->pos */
32823 + err = reinit_vdir(vdir_cache);
32824 + if (unlikely(err))
32827 + return 0; /* success */
32829 + inode = file_inode(file);
32830 + err = copy_vdir(vdir_cache, au_ivdir(inode));
32832 + file->f_version = inode_query_iversion(inode);
32834 + au_set_fvdir_cache(file, allocated);
32835 + } else if (allocated)
32836 + au_vdir_free(allocated);
32842 +static loff_t calc_offset(struct au_vdir *vdir)
32845 + union au_vdir_deblk_p p;
32847 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
32848 + offset = vdir->vd_last.p.deblk - p.deblk;
32849 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
32853 +/* returns true or false */
32854 +static int seek_vdir(struct file *file, struct dir_context *ctx)
32857 + unsigned int deblk_sz;
32858 + unsigned long ul, n;
32860 + union au_vdir_deblk_p p, deblk_end;
32861 + struct au_vdir *vdir_cache;
32864 + vdir_cache = au_fvdir_cache(file);
32865 + offset = calc_offset(vdir_cache);
32866 + AuDbg("offset %lld\n", offset);
32867 + if (ctx->pos == offset)
32870 + vdir_cache->vd_last.ul = 0;
32871 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
32876 + deblk_sz = vdir_cache->vd_deblk_sz;
32877 + ul = div64_u64(ctx->pos, deblk_sz);
32878 + AuDbg("ul %lu\n", ul);
32879 + if (ul >= vdir_cache->vd_nblk)
32882 + n = vdir_cache->vd_nblk;
32883 + for (; ul < n; ul++) {
32884 + p.deblk = vdir_cache->vd_deblk[ul];
32885 + deblk_end.deblk = p.deblk + deblk_sz;
32887 + offset *= deblk_sz;
32888 + while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
32891 + l = calc_size(p.de->de_str.len);
32895 + if (!is_deblk_end(&p, &deblk_end)) {
32897 + vdir_cache->vd_last.ul = ul;
32898 + vdir_cache->vd_last.p = p;
32906 + AuDbg("valid %d\n", !valid);
32910 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
32912 + unsigned int l, deblk_sz;
32913 + union au_vdir_deblk_p deblk_end;
32914 + struct au_vdir *vdir_cache;
32915 + struct au_vdir_de *de;
32917 + if (!seek_vdir(file, ctx))
32920 + vdir_cache = au_fvdir_cache(file);
32921 + deblk_sz = vdir_cache->vd_deblk_sz;
32923 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32924 + deblk_end.deblk += deblk_sz;
32925 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
32926 + de = vdir_cache->vd_last.p.de;
32927 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
32928 + de->de_str.len, de->de_str.name, ctx->pos,
32929 + (unsigned long)de->de_ino, de->de_type);
32930 + if (unlikely(!dir_emit(ctx, de->de_str.name,
32931 + de->de_str.len, de->de_ino,
32933 + /* todo: ignore the error caused by udba? */
32934 + /* return err; */
32938 + l = calc_size(de->de_str.len);
32939 + vdir_cache->vd_last.p.deblk += l;
32942 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
32943 + vdir_cache->vd_last.ul++;
32944 + vdir_cache->vd_last.p.deblk
32945 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32946 + ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
32955 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
32956 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
32957 +++ linux/fs/aufs/vfsub.c 2023-04-24 08:13:40.969576115 +0200
32959 +// SPDX-License-Identifier: GPL-2.0
32961 + * Copyright (C) 2005-2022 Junjiro R. Okajima
32963 + * This program is free software; you can redistribute it and/or modify
32964 + * it under the terms of the GNU General Public License as published by
32965 + * the Free Software Foundation; either version 2 of the License, or
32966 + * (at your option) any later version.
32968 + * This program is distributed in the hope that it will be useful,
32969 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32970 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32971 + * GNU General Public License for more details.
32973 + * You should have received a copy of the GNU General Public License
32974 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32978 + * sub-routines for VFS
32981 +#include <linux/mnt_namespace.h>
32982 +#include <linux/nsproxy.h>
32983 +#include <linux/security.h>
32984 +#include <linux/splice.h>
32987 +#ifdef CONFIG_AUFS_BR_FUSE
32988 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb)
32990 + if (!au_test_fuse(h_sb) || !au_userns)
32993 + return is_current_mnt_ns(mnt) ? 0 : -EACCES;
32997 +int vfsub_sync_filesystem(struct super_block *h_sb)
33002 + down_read(&h_sb->s_umount);
33003 + err = sync_filesystem(h_sb);
33004 + up_read(&h_sb->s_umount);
33010 +/* ---------------------------------------------------------------------- */
33012 +int vfsub_update_h_iattr(struct path *h_path, int *did)
33016 + struct super_block *h_sb;
33019 + * Always needs h_path->mnt for LSM or FUSE branch.
33021 + AuDebugOn(!h_path->mnt);
33023 + /* for remote fs, leave work for its getattr or d_revalidate */
33024 + /* for bad i_attr fs, handle them in aufs_getattr() */
33025 + /* still some fs may acquire i_mutex. we need to skip them */
33029 + h_sb = h_path->dentry->d_sb;
33030 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
33032 + err = vfsub_getattr(h_path, &st);
33037 +/* ---------------------------------------------------------------------- */
33039 +struct file *vfsub_dentry_open(struct path *path, int flags)
33041 + return dentry_open(path, flags /* | __FMODE_NONOTIFY */,
33045 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
33047 + struct file *file;
33050 + file = filp_open(path,
33051 + oflags /* | __FMODE_NONOTIFY */,
33054 + if (IS_ERR(file))
33056 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33063 + * Ideally this function should call VFS:do_last() in order to keep all its
33064 + * checkings. But it is very hard for aufs to regenerate several VFS internal
33065 + * structure such as nameidata. This is a second (or third) best approach.
33066 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
33068 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33069 + struct vfsub_aopen_args *args)
33072 + struct au_branch *br = args->br;
33073 + struct file *file = args->file;
33074 + /* copied from linux/fs/namei.c:atomic_open() */
33075 + struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
33078 + AuDebugOn(!dir->i_op->atomic_open);
33080 + err = au_br_test_oflag(args->open_flag, br);
33081 + if (unlikely(err))
33084 + au_lcnt_inc(&br->br_nfiles);
33085 + file->f_path.dentry = DENTRY_NOT_SET;
33086 + file->f_path.mnt = au_br_mnt(br);
33087 + AuDbg("%ps\n", dir->i_op->atomic_open);
33088 + err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
33089 + args->create_mode);
33090 + if (unlikely(err < 0)) {
33091 + au_lcnt_dec(&br->br_nfiles);
33095 + /* temporary workaround for nfsv4 branch */
33096 + if (au_test_nfs(dir->i_sb))
33097 + nfs_mark_for_revalidate(dir);
33099 + if (file->f_mode & FMODE_CREATED)
33100 + fsnotify_create(dir, dentry);
33101 + if (!(file->f_mode & FMODE_OPENED)) {
33102 + au_lcnt_dec(&br->br_nfiles);
33106 + /* todo: call VFS:may_open() here */
33107 + /* todo: ima_file_check() too? */
33108 + if (!err && (args->open_flag & __FMODE_EXEC))
33109 + err = deny_write_access(file);
33111 + fsnotify_open(file);
33113 + au_lcnt_dec(&br->br_nfiles);
33114 + /* note that the file is created and still opened */
33120 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
33124 + err = kern_path(name, flags, path);
33125 + if (!err && d_is_positive(path->dentry))
33126 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33130 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33131 + struct path *ppath, int len)
33133 + struct path path;
33135 + path.dentry = lookup_one_len_unlocked(name, ppath->dentry, len);
33136 + if (IS_ERR(path.dentry))
33138 + if (d_is_positive(path.dentry)) {
33139 + path.mnt = ppath->mnt;
33140 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33144 + AuTraceErrPtr(path.dentry);
33145 + return path.dentry;
33148 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33151 + struct path path;
33153 + /* VFS checks it too, but by WARN_ON_ONCE() */
33154 + IMustLock(d_inode(ppath->dentry));
33156 + path.dentry = lookup_one_len(name, ppath->dentry, len);
33157 + if (IS_ERR(path.dentry))
33159 + if (d_is_positive(path.dentry)) {
33160 + path.mnt = ppath->mnt;
33161 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
33165 + AuTraceErrPtr(path.dentry);
33166 + return path.dentry;
33169 +void vfsub_call_lkup_one(void *args)
33171 + struct vfsub_lkup_one_args *a = args;
33172 + *a->errp = vfsub_lkup_one(a->name, a->ppath);
33175 +/* ---------------------------------------------------------------------- */
33177 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33178 + struct dentry *d2, struct au_hinode *hdir2)
33180 + struct dentry *d;
33183 + d = lock_rename(d1, d2);
33185 + au_hn_suspend(hdir1);
33186 + if (hdir1 != hdir2)
33187 + au_hn_suspend(hdir2);
33192 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33193 + struct dentry *d2, struct au_hinode *hdir2)
33195 + au_hn_resume(hdir1);
33196 + if (hdir1 != hdir2)
33197 + au_hn_resume(hdir2);
33199 + unlock_rename(d1, d2);
33203 +/* ---------------------------------------------------------------------- */
33205 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
33208 + struct dentry *d;
33209 + struct mnt_idmap *idmap;
33213 + d = path->dentry;
33214 + path->dentry = d->d_parent;
33215 + err = security_path_mknod(path, d, mode, 0);
33216 + path->dentry = d;
33217 + if (unlikely(err))
33219 + idmap = mnt_idmap(path->mnt);
33222 + err = vfs_create(idmap, dir, path->dentry, mode, want_excl);
33225 + struct path tmp = *path;
33228 + vfsub_update_h_iattr(&tmp, &did);
33230 + tmp.dentry = path->dentry->d_parent;
33231 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33240 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
33243 + struct dentry *d;
33244 + struct mnt_idmap *idmap;
33248 + d = path->dentry;
33249 + path->dentry = d->d_parent;
33250 + err = security_path_symlink(path, d, symname);
33251 + path->dentry = d;
33252 + if (unlikely(err))
33254 + idmap = mnt_idmap(path->mnt);
33257 + err = vfs_symlink(idmap, dir, path->dentry, symname);
33260 + struct path tmp = *path;
33263 + vfsub_update_h_iattr(&tmp, &did);
33265 + tmp.dentry = path->dentry->d_parent;
33266 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33275 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
33278 + struct dentry *d;
33279 + struct mnt_idmap *idmap;
33283 + d = path->dentry;
33284 + path->dentry = d->d_parent;
33285 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
33286 + path->dentry = d;
33287 + if (unlikely(err))
33289 + idmap = mnt_idmap(path->mnt);
33292 + err = vfs_mknod(idmap, dir, path->dentry, mode, dev);
33295 + struct path tmp = *path;
33298 + vfsub_update_h_iattr(&tmp, &did);
33300 + tmp.dentry = path->dentry->d_parent;
33301 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33310 +static int au_test_nlink(struct inode *inode)
33312 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
33314 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
33315 + || inode->i_nlink < link_max)
33320 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
33321 + struct inode **delegated_inode)
33324 + struct dentry *d;
33325 + struct mnt_idmap *idmap;
33329 + err = au_test_nlink(d_inode(src_dentry));
33330 + if (unlikely(err))
33333 + /* we don't call may_linkat() */
33334 + d = path->dentry;
33335 + path->dentry = d->d_parent;
33336 + err = security_path_link(src_dentry, path, d);
33337 + path->dentry = d;
33338 + if (unlikely(err))
33340 + idmap = mnt_idmap(path->mnt);
33343 + err = vfs_link(src_dentry, idmap, dir, path->dentry, delegated_inode);
33346 + struct path tmp = *path;
33349 + /* fuse has different memory inode for the same inumber */
33350 + vfsub_update_h_iattr(&tmp, &did);
33352 + tmp.dentry = path->dentry->d_parent;
33353 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33354 + tmp.dentry = src_dentry;
33355 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33364 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
33365 + struct inode *dir, struct path *path,
33366 + struct inode **delegated_inode, unsigned int flags)
33369 + struct renamedata rd;
33370 + struct path tmp = {
33373 + struct dentry *d;
33376 + IMustLock(src_dir);
33378 + d = path->dentry;
33379 + path->dentry = d->d_parent;
33380 + tmp.dentry = src_dentry->d_parent;
33381 + err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
33382 + path->dentry = d;
33383 + if (unlikely(err))
33386 + rd.old_mnt_idmap = mnt_idmap(path->mnt);
33387 + rd.old_dir = src_dir;
33388 + rd.old_dentry = src_dentry;
33389 + rd.new_mnt_idmap = rd.old_mnt_idmap;
33390 + rd.new_dir = dir;
33391 + rd.new_dentry = path->dentry;
33392 + rd.delegated_inode = delegated_inode;
33393 + rd.flags = flags;
33395 + err = vfs_rename(&rd);
33400 + tmp.dentry = d->d_parent;
33401 + vfsub_update_h_iattr(&tmp, &did);
33403 + tmp.dentry = src_dentry;
33404 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33405 + tmp.dentry = src_dentry->d_parent;
33406 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33415 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
33418 + struct dentry *d;
33419 + struct mnt_idmap *idmap;
33423 + d = path->dentry;
33424 + path->dentry = d->d_parent;
33425 + err = security_path_mkdir(path, d, mode);
33426 + path->dentry = d;
33427 + if (unlikely(err))
33429 + idmap = mnt_idmap(path->mnt);
33432 + err = vfs_mkdir(idmap, dir, path->dentry, mode);
33435 + struct path tmp = *path;
33438 + vfsub_update_h_iattr(&tmp, &did);
33440 + tmp.dentry = path->dentry->d_parent;
33441 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
33450 +int vfsub_rmdir(struct inode *dir, struct path *path)
33453 + struct dentry *d;
33454 + struct mnt_idmap *idmap;
33458 + d = path->dentry;
33459 + path->dentry = d->d_parent;
33460 + err = security_path_rmdir(path, d);
33461 + path->dentry = d;
33462 + if (unlikely(err))
33464 + idmap = mnt_idmap(path->mnt);
33467 + err = vfs_rmdir(idmap, dir, path->dentry);
33470 + struct path tmp = {
33471 + .dentry = path->dentry->d_parent,
33475 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33482 +/* ---------------------------------------------------------------------- */
33484 +/* todo: support mmap_sem? */
33485 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
33491 + err = vfs_read(file, ubuf, count, ppos);
33494 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33498 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
33504 + err = kernel_read(file, kbuf, count, ppos);
33508 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33512 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
33518 + err = vfs_write(file, ubuf, count, ppos);
33521 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33525 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
33530 + err = kernel_write(file, kbuf, count, ppos);
33533 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33537 +int vfsub_flush(struct file *file, fl_owner_t id)
33542 + if (file->f_op->flush) {
33543 + if (!au_test_nfs(file->f_path.dentry->d_sb))
33544 + err = file->f_op->flush(file, id);
33547 + err = file->f_op->flush(file, id);
33551 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
33557 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
33561 + AuDbg("%pD, ctx{%ps, %llu}\n", file, ctx->actor, ctx->pos);
33564 + err = iterate_dir(file, ctx);
33567 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
33572 +long vfsub_splice_to(struct file *in, loff_t *ppos,
33573 + struct pipe_inode_info *pipe, size_t len,
33574 + unsigned int flags)
33579 + err = do_splice_to(in, ppos, pipe, len, flags);
33581 + file_accessed(in);
33583 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
33587 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
33588 + loff_t *ppos, size_t len, unsigned int flags)
33593 + err = do_splice_from(pipe, out, ppos, len, flags);
33596 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
33600 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
33604 + /* file can be NULL */
33606 + err = vfs_fsync(file, datasync);
33610 + AuDebugOn(!file);
33611 + path = &file->f_path;
33613 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33618 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
33619 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33620 + struct file *h_file)
33623 + struct inode *h_inode;
33624 + struct super_block *h_sb;
33625 + struct mnt_idmap *h_idmap;
33628 + err = vfsub_truncate(h_path, length);
33632 + h_inode = d_inode(h_path->dentry);
33633 + h_sb = h_inode->i_sb;
33635 + sb_start_write(h_sb);
33637 + err = security_file_truncate(h_file);
33639 + h_idmap = mnt_idmap(h_path->mnt);
33641 + err = do_truncate(h_idmap, h_path->dentry, length, attr,
33646 + sb_end_write(h_sb);
33653 +/* ---------------------------------------------------------------------- */
33655 +struct au_vfsub_mkdir_args {
33657 + struct inode *dir;
33658 + struct path *path;
33662 +static void au_call_vfsub_mkdir(void *args)
33664 + struct au_vfsub_mkdir_args *a = args;
33665 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
33668 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
33670 + int err, do_sio, wkq_err;
33671 + struct mnt_idmap *idmap;
33673 + idmap = mnt_idmap(path->mnt);
33674 + do_sio = au_test_h_perm_sio(idmap, dir, MAY_EXEC | MAY_WRITE);
33677 + err = vfsub_mkdir(dir, path, mode);
33680 + struct au_vfsub_mkdir_args args = {
33686 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
33687 + if (unlikely(wkq_err))
33694 +struct au_vfsub_rmdir_args {
33696 + struct inode *dir;
33697 + struct path *path;
33700 +static void au_call_vfsub_rmdir(void *args)
33702 + struct au_vfsub_rmdir_args *a = args;
33703 + *a->errp = vfsub_rmdir(a->dir, a->path);
33706 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
33708 + int err, do_sio, wkq_err;
33709 + struct mnt_idmap *idmap;
33711 + idmap = mnt_idmap(path->mnt);
33712 + do_sio = au_test_h_perm_sio(idmap, dir, MAY_EXEC | MAY_WRITE);
33715 + err = vfsub_rmdir(dir, path);
33718 + struct au_vfsub_rmdir_args args = {
33723 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
33724 + if (unlikely(wkq_err))
33731 +/* ---------------------------------------------------------------------- */
33733 +struct notify_change_args {
33735 + struct path *path;
33736 + struct iattr *ia;
33737 + struct inode **delegated_inode;
33740 +static void call_notify_change(void *args)
33742 + struct notify_change_args *a = args;
33743 + struct inode *h_inode;
33744 + struct mnt_idmap *idmap;
33746 + h_inode = d_inode(a->path->dentry);
33747 + IMustLock(h_inode);
33749 + *a->errp = -EPERM;
33750 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
33751 + idmap = mnt_idmap(a->path->mnt);
33753 + *a->errp = notify_change(idmap, a->path->dentry, a->ia,
33754 + a->delegated_inode);
33757 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
33759 + AuTraceErr(*a->errp);
33762 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33763 + struct inode **delegated_inode)
33766 + struct notify_change_args args = {
33770 + .delegated_inode = delegated_inode
33773 + call_notify_change(&args);
33778 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33779 + struct inode **delegated_inode)
33781 + int err, wkq_err;
33782 + struct notify_change_args args = {
33786 + .delegated_inode = delegated_inode
33789 + wkq_err = au_wkq_wait(call_notify_change, &args);
33790 + if (unlikely(wkq_err))
33796 +/* ---------------------------------------------------------------------- */
33798 +struct unlink_args {
33800 + struct inode *dir;
33801 + struct path *path;
33802 + struct inode **delegated_inode;
33805 +static void call_unlink(void *args)
33807 + struct unlink_args *a = args;
33808 + struct dentry *d = a->path->dentry;
33809 + struct inode *h_inode;
33810 + struct mnt_idmap *idmap;
33811 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
33812 + && au_dcount(d) == 1);
33814 + IMustLock(a->dir);
33816 + a->path->dentry = d->d_parent;
33817 + *a->errp = security_path_unlink(a->path, d);
33818 + a->path->dentry = d;
33819 + if (unlikely(*a->errp))
33822 + if (!stop_sillyrename)
33825 + if (d_is_positive(d)) {
33826 + h_inode = d_inode(d);
33830 + idmap = mnt_idmap(a->path->mnt);
33832 + *a->errp = vfs_unlink(idmap, a->dir, d, a->delegated_inode);
33835 + struct path tmp = {
33836 + .dentry = d->d_parent,
33837 + .mnt = a->path->mnt
33839 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33842 + if (!stop_sillyrename)
33847 + AuTraceErr(*a->errp);
33851 + * @dir: must be locked.
33852 + * @dentry: target dentry.
33854 +int vfsub_unlink(struct inode *dir, struct path *path,
33855 + struct inode **delegated_inode, int force)
33858 + struct unlink_args args = {
33862 + .delegated_inode = delegated_inode
33866 + call_unlink(&args);
33870 + wkq_err = au_wkq_wait(call_unlink, &args);
33871 + if (unlikely(wkq_err))
33877 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
33878 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
33879 +++ linux/fs/aufs/vfsub.h 2023-04-24 08:13:40.969576115 +0200
33881 +/* SPDX-License-Identifier: GPL-2.0 */
33883 + * Copyright (C) 2005-2022 Junjiro R. Okajima
33885 + * This program is free software; you can redistribute it and/or modify
33886 + * it under the terms of the GNU General Public License as published by
33887 + * the Free Software Foundation; either version 2 of the License, or
33888 + * (at your option) any later version.
33890 + * This program is distributed in the hope that it will be useful,
33891 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33892 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33893 + * GNU General Public License for more details.
33895 + * You should have received a copy of the GNU General Public License
33896 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33900 + * sub-routines for VFS
33903 +#ifndef __AUFS_VFSUB_H__
33904 +#define __AUFS_VFSUB_H__
33908 +#include <linux/fs.h>
33909 +#include <linux/mount.h>
33910 +#include <linux/posix_acl.h>
33911 +#include <linux/xattr.h>
33912 +#include "debug.h"
33914 +/* copied from linux/fs/internal.h */
33915 +/* todo: BAD approach!! */
33916 +extern void __mnt_drop_write(struct vfsmount *);
33917 +extern struct file *alloc_empty_file(int, const struct cred *);
33919 +/* ---------------------------------------------------------------------- */
33921 +/* lock subclass for lower inode */
33922 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
33923 +/* reduce? gave up. */
33925 + AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
33926 + AuLsc_I_PARENT, /* lower inode, parent first */
33927 + AuLsc_I_PARENT2, /* copyup dirs */
33928 + AuLsc_I_PARENT3, /* copyup wh */
33934 +/* to debug easier, do not make them inlined functions */
33935 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
33936 +#define IMustLock(i) AuDebugOn(!inode_is_locked(i))
33938 +/* ---------------------------------------------------------------------- */
33940 +static inline void vfsub_drop_nlink(struct inode *inode)
33942 + AuDebugOn(!inode->i_nlink);
33943 + drop_nlink(inode);
33946 +static inline void vfsub_dead_dir(struct inode *inode)
33948 + AuDebugOn(!S_ISDIR(inode->i_mode));
33949 + inode->i_flags |= S_DEAD;
33950 + clear_nlink(inode);
33953 +static inline int vfsub_native_ro(struct inode *inode)
33955 + return sb_rdonly(inode->i_sb)
33956 + || IS_RDONLY(inode)
33957 + /* || IS_APPEND(inode) */
33958 + || IS_IMMUTABLE(inode);
33961 +#ifdef CONFIG_AUFS_BR_FUSE
33962 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb);
33964 +AuStubInt0(vfsub_test_mntns, struct vfsmount *mnt, struct super_block *h_sb);
33967 +int vfsub_sync_filesystem(struct super_block *h_sb);
33969 +/* ---------------------------------------------------------------------- */
33971 +int vfsub_update_h_iattr(struct path *h_path, int *did);
33972 +struct file *vfsub_dentry_open(struct path *path, int flags);
33973 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
33975 +struct vfsub_aopen_args {
33976 + struct file *file;
33977 + unsigned int open_flag;
33978 + umode_t create_mode;
33979 + struct au_branch *br;
33981 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33982 + struct vfsub_aopen_args *args);
33983 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
33985 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33986 + struct path *ppath, int len);
33987 +struct dentry *vfsub_lookup_one_len(const char *name, struct path *ppath,
33990 +struct vfsub_lkup_one_args {
33991 + struct dentry **errp;
33992 + struct qstr *name;
33993 + struct path *ppath;
33996 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
33997 + struct path *ppath)
33999 + return vfsub_lookup_one_len(name->name, ppath, name->len);
34002 +void vfsub_call_lkup_one(void *args);
34004 +/* ---------------------------------------------------------------------- */
34006 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
34011 + err = mnt_want_write(mnt);
34016 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
34019 + mnt_drop_write(mnt);
34023 +#if 0 /* reserved */
34024 +static inline void vfsub_mnt_drop_write_file(struct file *file)
34027 + mnt_drop_write_file(file);
34032 +/* ---------------------------------------------------------------------- */
34035 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
34036 + struct dentry *d2, struct au_hinode *hdir2);
34037 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
34038 + struct dentry *d2, struct au_hinode *hdir2);
34040 +int vfsub_create(struct inode *dir, struct path *path, int mode,
34042 +int vfsub_symlink(struct inode *dir, struct path *path,
34043 + const char *symname);
34044 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
34045 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
34046 + struct path *path, struct inode **delegated_inode);
34047 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
34048 + struct inode *hdir, struct path *path,
34049 + struct inode **delegated_inode, unsigned int flags);
34050 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
34051 +int vfsub_rmdir(struct inode *dir, struct path *path);
34053 +/* ---------------------------------------------------------------------- */
34055 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
34057 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
34059 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
34061 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
34063 +int vfsub_flush(struct file *file, fl_owner_t id);
34064 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
34066 +static inline loff_t vfsub_f_size_read(struct file *file)
34068 + return i_size_read(file_inode(file));
34071 +static inline unsigned int vfsub_file_flags(struct file *file)
34073 + unsigned int flags;
34075 + spin_lock(&file->f_lock);
34076 + flags = file->f_flags;
34077 + spin_unlock(&file->f_lock);
34082 +static inline int vfsub_file_execed(struct file *file)
34084 + /* todo: direct access f_flags */
34085 + return !!(vfsub_file_flags(file) & __FMODE_EXEC);
34088 +#if 0 /* reserved */
34089 +static inline void vfsub_file_accessed(struct file *h_file)
34091 + file_accessed(h_file);
34092 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
34096 +#if 0 /* reserved */
34097 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
34098 + struct dentry *h_dentry)
34100 + struct path h_path = {
34101 + .dentry = h_dentry,
34104 + touch_atime(&h_path);
34105 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
34109 +static inline int vfsub_update_time(struct inode *h_inode,
34110 + struct timespec64 *ts, int flags)
34112 + return inode_update_time(h_inode, ts, flags);
34113 + /* no vfsub_update_h_iattr() since we don't have struct path */
34116 +#ifdef CONFIG_FS_POSIX_ACL
34117 +static inline int vfsub_acl_chmod(struct mnt_idmap *h_idmap,
34118 + struct dentry *h_dentry, umode_t h_mode)
34122 + err = posix_acl_chmod(h_idmap, h_dentry, h_mode);
34123 + if (err == -EOPNOTSUPP)
34128 +AuStubInt0(vfsub_acl_chmod, struct mnt_idmap *h_idmap,
34129 + struct dentry *h_dentry, umode_t h_mode);
34132 +long vfsub_splice_to(struct file *in, loff_t *ppos,
34133 + struct pipe_inode_info *pipe, size_t len,
34134 + unsigned int flags);
34135 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
34136 + loff_t *ppos, size_t len, unsigned int flags);
34138 +static inline long vfsub_truncate(struct path *path, loff_t length)
34143 + err = vfs_truncate(path, length);
34148 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
34149 + struct file *h_file);
34150 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
34153 + * re-use branch fs's ioctl(FICLONE) while aufs itself doesn't support such
34156 +static inline loff_t vfsub_clone_file_range(struct file *src, struct file *dst,
34162 + err = vfs_clone_file_range(src, 0, dst, 0, len, /*remap_flags*/0);
34168 +/* copy_file_range(2) is a systemcall */
34169 +static inline ssize_t vfsub_copy_file_range(struct file *src, loff_t src_pos,
34170 + struct file *dst, loff_t dst_pos,
34171 + size_t len, unsigned int flags)
34176 + ssz = vfs_copy_file_range(src, src_pos, dst, dst_pos, len, flags);
34182 +/* ---------------------------------------------------------------------- */
34184 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
34189 + err = vfs_llseek(file, offset, origin);
34194 +/* ---------------------------------------------------------------------- */
34196 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
34197 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
34198 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
34199 + struct inode **delegated_inode);
34200 +int vfsub_notify_change(struct path *path, struct iattr *ia,
34201 + struct inode **delegated_inode);
34202 +int vfsub_unlink(struct inode *dir, struct path *path,
34203 + struct inode **delegated_inode, int force);
34205 +static inline int vfsub_getattr(const struct path *path, struct kstat *st)
34207 + return vfs_getattr(path, st, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
34210 +/* ---------------------------------------------------------------------- */
34212 +static inline int vfsub_setxattr(struct mnt_idmap *idmap,
34213 + struct dentry *dentry, const char *name,
34214 + const void *value, size_t size, int flags)
34219 + err = vfs_setxattr(idmap, dentry, name, value, size, flags);
34225 +static inline int vfsub_removexattr(struct mnt_idmap *idmap,
34226 + struct dentry *dentry, const char *name)
34231 + err = vfs_removexattr(idmap, dentry, name);
34237 +#ifdef CONFIG_FS_POSIX_ACL
34238 +static inline int vfsub_set_acl(struct mnt_idmap *idmap,
34239 + struct dentry *dentry, const char *name,
34240 + struct posix_acl *acl)
34245 + err = vfs_set_acl(idmap, dentry, name, acl);
34251 +static inline int vfsub_remove_acl(struct mnt_idmap *idmap,
34252 + struct dentry *dentry, const char *name)
34257 + err = vfs_remove_acl(idmap, dentry, name);
34263 +AuStubInt0(vfsub_set_acl, struct mnt_idmap *idmap, struct dentry *dentry,
34264 + const char *name, struct posix_acl *acl);
34265 +AuStubInt0(vfsub_remove_acl, struct mnt_idmap *idmap,
34266 + struct dentry *dentry, const char *name);
34269 +#endif /* __KERNEL__ */
34270 +#endif /* __AUFS_VFSUB_H__ */
34271 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
34272 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
34273 +++ linux/fs/aufs/wbr_policy.c 2022-11-05 23:02:18.969222617 +0100
34275 +// SPDX-License-Identifier: GPL-2.0
34277 + * Copyright (C) 2005-2022 Junjiro R. Okajima
34279 + * This program is free software; you can redistribute it and/or modify
34280 + * it under the terms of the GNU General Public License as published by
34281 + * the Free Software Foundation; either version 2 of the License, or
34282 + * (at your option) any later version.
34284 + * This program is distributed in the hope that it will be useful,
34285 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
34286 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
34287 + * GNU General Public License for more details.
34289 + * You should have received a copy of the GNU General Public License
34290 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
34294 + * policies for selecting one among multiple writable branches
34297 +#include <linux/statfs.h>
34300 +/* subset of cpup_attr() */
34301 +static noinline_for_stack
34302 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
34306 + struct inode *h_isrc;
34308 + h_isrc = d_inode(h_src);
34309 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
34310 + ia.ia_mode = h_isrc->i_mode;
34311 + ia.ia_uid = h_isrc->i_uid;
34312 + ia.ia_gid = h_isrc->i_gid;
34313 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
34314 + au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
34315 + /* no delegation since it is just created */
34316 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34318 + /* is this nfs only? */
34319 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
34320 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
34321 + ia.ia_mode = h_isrc->i_mode;
34322 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
34328 +#define AuCpdown_PARENT_OPQ 1
34329 +#define AuCpdown_WHED (1 << 1)
34330 +#define AuCpdown_MADE_DIR (1 << 2)
34331 +#define AuCpdown_DIROPQ (1 << 3)
34332 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
34333 +#define au_fset_cpdown(flags, name) \
34334 + do { (flags) |= AuCpdown_##name; } while (0)
34335 +#define au_fclr_cpdown(flags, name) \
34336 + do { (flags) &= ~AuCpdown_##name; } while (0)
34338 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
34339 + unsigned int *flags)
34342 + struct dentry *opq_dentry;
34344 + opq_dentry = au_diropq_create(dentry, bdst);
34345 + err = PTR_ERR(opq_dentry);
34346 + if (IS_ERR(opq_dentry))
34348 + dput(opq_dentry);
34349 + au_fset_cpdown(*flags, DIROPQ);
34355 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
34356 + struct inode *dir, aufs_bindex_t bdst)
34359 + struct path h_path;
34360 + struct au_branch *br;
34362 + br = au_sbr(dentry->d_sb, bdst);
34363 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
34364 + err = PTR_ERR(h_path.dentry);
34365 + if (IS_ERR(h_path.dentry))
34369 + if (d_is_positive(h_path.dentry)) {
34370 + h_path.mnt = au_br_mnt(br);
34371 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
34374 + dput(h_path.dentry);
34380 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
34381 + struct au_pin *pin,
34382 + struct dentry *h_parent, void *arg)
34385 + aufs_bindex_t bopq, btop;
34386 + struct path h_path;
34387 + struct dentry *parent;
34388 + struct inode *h_dir, *h_inode, *inode, *dir;
34389 + unsigned int *flags = arg;
34391 + btop = au_dbtop(dentry);
34392 + /* dentry is di-locked */
34393 + parent = dget_parent(dentry);
34394 + dir = d_inode(parent);
34395 + h_dir = d_inode(h_parent);
34396 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
34397 + IMustLock(h_dir);
34399 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
34400 + if (unlikely(err < 0))
34402 + h_path.dentry = au_h_dptr(dentry, bdst);
34403 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
34404 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path, 0755);
34405 + if (unlikely(err))
34407 + au_fset_cpdown(*flags, MADE_DIR);
34409 + bopq = au_dbdiropq(dentry);
34410 + au_fclr_cpdown(*flags, WHED);
34411 + au_fclr_cpdown(*flags, DIROPQ);
34412 + if (au_dbwh(dentry) == bdst)
34413 + au_fset_cpdown(*flags, WHED);
34414 + if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
34415 + au_fset_cpdown(*flags, PARENT_OPQ);
34416 + h_inode = d_inode(h_path.dentry);
34417 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
34418 + if (au_ftest_cpdown(*flags, WHED)) {
34419 + err = au_cpdown_dir_opq(dentry, bdst, flags);
34420 + if (unlikely(err)) {
34421 + inode_unlock(h_inode);
34426 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, btop));
34427 + inode_unlock(h_inode);
34428 + if (unlikely(err))
34431 + if (au_ftest_cpdown(*flags, WHED)) {
34432 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
34433 + if (unlikely(err))
34437 + inode = d_inode(dentry);
34438 + if (au_ibbot(inode) < bdst)
34439 + au_set_ibbot(inode, bdst);
34440 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
34441 + au_hi_flags(inode, /*isdir*/1));
34442 + au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
34443 + goto out; /* success */
34447 + if (au_ftest_cpdown(*flags, DIROPQ)) {
34448 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
34449 + rerr = au_diropq_remove(dentry, bdst);
34450 + inode_unlock(h_inode);
34451 + if (unlikely(rerr)) {
34452 + AuIOErr("failed removing diropq for %pd b%d (%d)\n",
34453 + dentry, bdst, rerr);
34459 + if (au_ftest_cpdown(*flags, MADE_DIR)) {
34460 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
34461 + if (unlikely(rerr)) {
34462 + AuIOErr("failed removing %pd b%d (%d)\n",
34463 + dentry, bdst, rerr);
34468 + au_set_h_dptr(dentry, bdst, NULL);
34469 + if (au_dbbot(dentry) == bdst)
34470 + au_update_dbbot(dentry);
34476 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
34479 + unsigned int flags;
34482 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
34487 +/* ---------------------------------------------------------------------- */
34489 +/* policies for create */
34491 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
34493 + int err, i, j, ndentry;
34494 + aufs_bindex_t bopq;
34495 + struct au_dcsub_pages dpages;
34496 + struct au_dpage *dpage;
34497 + struct dentry **dentries, *parent, *d;
34499 + err = au_dpages_init(&dpages, GFP_NOFS);
34500 + if (unlikely(err))
34502 + parent = dget_parent(dentry);
34503 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
34504 + if (unlikely(err))
34508 + for (i = 0; i < dpages.ndpage; i++) {
34509 + dpage = dpages.dpages + i;
34510 + dentries = dpage->dentries;
34511 + ndentry = dpage->ndentry;
34512 + for (j = 0; j < ndentry; j++) {
34514 + di_read_lock_parent2(d, !AuLock_IR);
34515 + bopq = au_dbdiropq(d);
34516 + di_read_unlock(d, !AuLock_IR);
34517 + if (bopq >= 0 && bopq < err)
34524 + au_dpages_free(&dpages);
34529 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
34531 + for (; bindex >= 0; bindex--)
34532 + if (!au_br_rdonly(au_sbr(sb, bindex)))
34537 +/* top down parent */
34538 +static int au_wbr_create_tdp(struct dentry *dentry,
34539 + unsigned int flags __maybe_unused)
34542 + aufs_bindex_t btop, bindex;
34543 + struct super_block *sb;
34544 + struct dentry *parent, *h_parent;
34546 + sb = dentry->d_sb;
34547 + btop = au_dbtop(dentry);
34549 + if (!au_br_rdonly(au_sbr(sb, btop)))
34553 + parent = dget_parent(dentry);
34554 + for (bindex = au_dbtop(parent); bindex < btop; bindex++) {
34555 + h_parent = au_h_dptr(parent, bindex);
34556 + if (!h_parent || d_is_negative(h_parent))
34559 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
34566 + /* bottom up here */
34567 + if (unlikely(err < 0)) {
34568 + err = au_wbr_bu(sb, btop - 1);
34570 + err = au_wbr_nonopq(dentry, err);
34574 + AuDbg("b%d\n", err);
34578 +/* ---------------------------------------------------------------------- */
34580 +/* an exception for the policy other than tdp */
34581 +static int au_wbr_create_exp(struct dentry *dentry)
34584 + aufs_bindex_t bwh, bdiropq;
34585 + struct dentry *parent;
34588 + bwh = au_dbwh(dentry);
34589 + parent = dget_parent(dentry);
34590 + bdiropq = au_dbdiropq(parent);
34592 + if (bdiropq >= 0)
34593 + err = min(bdiropq, bwh);
34596 + AuDbg("%d\n", err);
34597 + } else if (bdiropq >= 0) {
34599 + AuDbg("%d\n", err);
34604 + err = au_wbr_nonopq(dentry, err);
34606 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
34609 + AuDbg("%d\n", err);
34613 +/* ---------------------------------------------------------------------- */
34616 +static int au_wbr_create_init_rr(struct super_block *sb)
34620 + err = au_wbr_bu(sb, au_sbbot(sb));
34621 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
34624 + AuDbg("b%d\n", err);
34628 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
34632 + aufs_bindex_t bindex, bbot;
34633 + struct super_block *sb;
34636 + err = au_wbr_create_exp(dentry);
34640 + sb = dentry->d_sb;
34641 + next = &au_sbi(sb)->si_wbr_rr_next;
34642 + bbot = au_sbbot(sb);
34644 + for (bindex = 0; bindex <= bbot; bindex++) {
34645 + if (!au_ftest_wbr(flags, DIR)) {
34646 + err = atomic_dec_return(next) + 1;
34647 + /* modulo for 0 is meaningless */
34648 + if (unlikely(!err))
34649 + err = atomic_dec_return(next) + 1;
34651 + err = atomic_read(next);
34652 + AuDbg("%d\n", err);
34655 + AuDbg("%d\n", err);
34656 + if (!au_br_rdonly(au_sbr(sb, err)))
34662 + err = au_wbr_nonopq(dentry, err);
34665 + AuDbg("%d\n", err);
34669 +/* ---------------------------------------------------------------------- */
34671 +/* most free space */
34672 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
34674 + struct super_block *sb;
34675 + struct au_branch *br;
34676 + struct au_wbr_mfs *mfs;
34677 + struct dentry *h_parent;
34678 + aufs_bindex_t bindex, bbot;
34680 + unsigned long long b, bavail;
34681 + struct path h_path;
34682 + /* reduce the stack usage */
34683 + struct kstatfs *st;
34685 + st = kmalloc(sizeof(*st), GFP_NOFS);
34686 + if (unlikely(!st)) {
34687 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
34692 + sb = dentry->d_sb;
34693 + mfs = &au_sbi(sb)->si_wbr_mfs;
34694 + MtxMustLock(&mfs->mfs_lock);
34695 + mfs->mfs_bindex = -EROFS;
34696 + mfs->mfsrr_bytes = 0;
34699 + bbot = au_sbbot(sb);
34701 + bindex = au_dbtop(parent);
34702 + bbot = au_dbtaildir(parent);
34705 + for (; bindex <= bbot; bindex++) {
34707 + h_parent = au_h_dptr(parent, bindex);
34708 + if (!h_parent || d_is_negative(h_parent))
34711 + br = au_sbr(sb, bindex);
34712 + if (au_br_rdonly(br))
34715 + /* sb->s_root for NFS is unreliable */
34716 + h_path.mnt = au_br_mnt(br);
34717 + h_path.dentry = h_path.mnt->mnt_root;
34718 + err = vfs_statfs(&h_path, st);
34719 + if (unlikely(err)) {
34720 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
34724 + /* when the available size is equal, select the lower one */
34725 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
34726 + || sizeof(b) < sizeof(st->f_bsize));
34727 + b = st->f_bavail * st->f_bsize;
34728 + br->br_wbr->wbr_bytes = b;
34729 + if (b >= bavail) {
34731 + mfs->mfs_bindex = bindex;
34732 + mfs->mfs_jiffy = jiffies;
34736 + mfs->mfsrr_bytes = bavail;
34737 + AuDbg("b%d\n", mfs->mfs_bindex);
34738 + au_kfree_rcu(st);
34741 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
34744 + struct dentry *parent;
34745 + struct super_block *sb;
34746 + struct au_wbr_mfs *mfs;
34748 + err = au_wbr_create_exp(dentry);
34752 + sb = dentry->d_sb;
34754 + if (au_ftest_wbr(flags, PARENT))
34755 + parent = dget_parent(dentry);
34756 + mfs = &au_sbi(sb)->si_wbr_mfs;
34757 + mutex_lock(&mfs->mfs_lock);
34758 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34759 + || mfs->mfs_bindex < 0
34760 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
34761 + au_mfs(dentry, parent);
34762 + mutex_unlock(&mfs->mfs_lock);
34763 + err = mfs->mfs_bindex;
34767 + err = au_wbr_nonopq(dentry, err);
34770 + AuDbg("b%d\n", err);
34774 +static int au_wbr_create_init_mfs(struct super_block *sb)
34776 + struct au_wbr_mfs *mfs;
34778 + mfs = &au_sbi(sb)->si_wbr_mfs;
34779 + mutex_init(&mfs->mfs_lock);
34780 + mfs->mfs_jiffy = 0;
34781 + mfs->mfs_bindex = -EROFS;
34786 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
34788 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
34792 +/* ---------------------------------------------------------------------- */
34794 +/* top down regardless parent, and then mfs */
34795 +static int au_wbr_create_tdmfs(struct dentry *dentry,
34796 + unsigned int flags __maybe_unused)
34799 + aufs_bindex_t bwh, btail, bindex, bfound, bmfs;
34800 + unsigned long long watermark;
34801 + struct super_block *sb;
34802 + struct au_wbr_mfs *mfs;
34803 + struct au_branch *br;
34804 + struct dentry *parent;
34806 + sb = dentry->d_sb;
34807 + mfs = &au_sbi(sb)->si_wbr_mfs;
34808 + mutex_lock(&mfs->mfs_lock);
34809 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34810 + || mfs->mfs_bindex < 0)
34811 + au_mfs(dentry, /*parent*/NULL);
34812 + watermark = mfs->mfsrr_watermark;
34813 + bmfs = mfs->mfs_bindex;
34814 + mutex_unlock(&mfs->mfs_lock);
34816 + /* another style of au_wbr_create_exp() */
34817 + bwh = au_dbwh(dentry);
34818 + parent = dget_parent(dentry);
34819 + btail = au_dbtaildir(parent);
34820 + if (bwh >= 0 && bwh < btail)
34823 + err = au_wbr_nonopq(dentry, btail);
34824 + if (unlikely(err < 0))
34828 + for (bindex = 0; bindex <= btail; bindex++) {
34829 + br = au_sbr(sb, bindex);
34830 + if (au_br_rdonly(br))
34832 + if (br->br_wbr->wbr_bytes > watermark) {
34843 + AuDbg("b%d\n", err);
34847 +/* ---------------------------------------------------------------------- */
34849 +/* most free space and then round robin */
34850 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
34853 + struct au_wbr_mfs *mfs;
34855 + err = au_wbr_create_mfs(dentry, flags);
34857 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
34858 + mutex_lock(&mfs->mfs_lock);
34859 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
34860 + err = au_wbr_create_rr(dentry, flags);
34861 + mutex_unlock(&mfs->mfs_lock);
34864 + AuDbg("b%d\n", err);
34868 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
34872 + au_wbr_create_init_mfs(sb); /* ignore */
34873 + err = au_wbr_create_init_rr(sb);
34878 +/* ---------------------------------------------------------------------- */
34880 +/* top down parent and most free space */
34881 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
34884 + unsigned long long b;
34885 + aufs_bindex_t bindex, btop, bbot;
34886 + struct super_block *sb;
34887 + struct dentry *parent, *h_parent;
34888 + struct au_branch *br;
34890 + err = au_wbr_create_tdp(dentry, flags);
34891 + if (unlikely(err < 0))
34893 + parent = dget_parent(dentry);
34894 + btop = au_dbtop(parent);
34895 + bbot = au_dbtaildir(parent);
34896 + if (btop == bbot)
34897 + goto out_parent; /* success */
34899 + e2 = au_wbr_create_mfs(dentry, flags);
34901 + goto out_parent; /* success */
34903 + /* when the available size is equal, select upper one */
34904 + sb = dentry->d_sb;
34905 + br = au_sbr(sb, err);
34906 + b = br->br_wbr->wbr_bytes;
34907 + AuDbg("b%d, %llu\n", err, b);
34909 + for (bindex = btop; bindex <= bbot; bindex++) {
34910 + h_parent = au_h_dptr(parent, bindex);
34911 + if (!h_parent || d_is_negative(h_parent))
34914 + br = au_sbr(sb, bindex);
34915 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
34916 + b = br->br_wbr->wbr_bytes;
34918 + AuDbg("b%d, %llu\n", err, b);
34923 + err = au_wbr_nonopq(dentry, err);
34928 + AuDbg("b%d\n", err);
34932 +/* ---------------------------------------------------------------------- */
34935 + * - top down parent
34936 + * - most free space with parent
34937 + * - most free space round-robin regardless parent
34939 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
34942 + unsigned long long watermark;
34943 + struct super_block *sb;
34944 + struct au_branch *br;
34945 + struct au_wbr_mfs *mfs;
34947 + err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
34948 + if (unlikely(err < 0))
34951 + sb = dentry->d_sb;
34952 + br = au_sbr(sb, err);
34953 + mfs = &au_sbi(sb)->si_wbr_mfs;
34954 + mutex_lock(&mfs->mfs_lock);
34955 + watermark = mfs->mfsrr_watermark;
34956 + mutex_unlock(&mfs->mfs_lock);
34957 + if (br->br_wbr->wbr_bytes < watermark)
34958 + /* regardless the parent dir */
34959 + err = au_wbr_create_mfsrr(dentry, flags);
34962 + AuDbg("b%d\n", err);
34966 +/* ---------------------------------------------------------------------- */
34968 +/* policies for copyup */
34970 +/* top down parent */
34971 +static int au_wbr_copyup_tdp(struct dentry *dentry)
34973 + return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
34976 +/* bottom up parent */
34977 +static int au_wbr_copyup_bup(struct dentry *dentry)
34980 + aufs_bindex_t bindex, btop;
34981 + struct dentry *parent, *h_parent;
34982 + struct super_block *sb;
34985 + sb = dentry->d_sb;
34986 + parent = dget_parent(dentry);
34987 + btop = au_dbtop(parent);
34988 + for (bindex = au_dbtop(dentry); bindex >= btop; bindex--) {
34989 + h_parent = au_h_dptr(parent, bindex);
34990 + if (!h_parent || d_is_negative(h_parent))
34993 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
35000 + /* bottom up here */
35001 + if (unlikely(err < 0))
35002 + err = au_wbr_bu(sb, btop - 1);
35004 + AuDbg("b%d\n", err);
35009 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop)
35013 + err = au_wbr_bu(dentry->d_sb, btop);
35014 + AuDbg("b%d\n", err);
35016 + err = au_wbr_nonopq(dentry, err);
35018 + AuDbg("b%d\n", err);
35022 +static int au_wbr_copyup_bu(struct dentry *dentry)
35025 + aufs_bindex_t btop;
35027 + btop = au_dbtop(dentry);
35028 + err = au_wbr_do_copyup_bu(dentry, btop);
35032 +/* ---------------------------------------------------------------------- */
35034 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
35035 + [AuWbrCopyup_TDP] = {
35036 + .copyup = au_wbr_copyup_tdp
35038 + [AuWbrCopyup_BUP] = {
35039 + .copyup = au_wbr_copyup_bup
35041 + [AuWbrCopyup_BU] = {
35042 + .copyup = au_wbr_copyup_bu
35046 +struct au_wbr_create_operations au_wbr_create_ops[] = {
35047 + [AuWbrCreate_TDP] = {
35048 + .create = au_wbr_create_tdp
35050 + [AuWbrCreate_RR] = {
35051 + .create = au_wbr_create_rr,
35052 + .init = au_wbr_create_init_rr
35054 + [AuWbrCreate_MFS] = {
35055 + .create = au_wbr_create_mfs,
35056 + .init = au_wbr_create_init_mfs,
35057 + .fin = au_wbr_create_fin_mfs
35059 + [AuWbrCreate_MFSV] = {
35060 + .create = au_wbr_create_mfs,
35061 + .init = au_wbr_create_init_mfs,
35062 + .fin = au_wbr_create_fin_mfs
35064 + [AuWbrCreate_MFSRR] = {
35065 + .create = au_wbr_create_mfsrr,
35066 + .init = au_wbr_create_init_mfsrr,
35067 + .fin = au_wbr_create_fin_mfs
35069 + [AuWbrCreate_MFSRRV] = {
35070 + .create = au_wbr_create_mfsrr,
35071 + .init = au_wbr_create_init_mfsrr,
35072 + .fin = au_wbr_create_fin_mfs
35074 + [AuWbrCreate_TDMFS] = {
35075 + .create = au_wbr_create_tdmfs,
35076 + .init = au_wbr_create_init_mfs,
35077 + .fin = au_wbr_create_fin_mfs
35079 + [AuWbrCreate_TDMFSV] = {
35080 + .create = au_wbr_create_tdmfs,
35081 + .init = au_wbr_create_init_mfs,
35082 + .fin = au_wbr_create_fin_mfs
35084 + [AuWbrCreate_PMFS] = {
35085 + .create = au_wbr_create_pmfs,
35086 + .init = au_wbr_create_init_mfs,
35087 + .fin = au_wbr_create_fin_mfs
35089 + [AuWbrCreate_PMFSV] = {
35090 + .create = au_wbr_create_pmfs,
35091 + .init = au_wbr_create_init_mfs,
35092 + .fin = au_wbr_create_fin_mfs
35094 + [AuWbrCreate_PMFSRR] = {
35095 + .create = au_wbr_create_pmfsrr,
35096 + .init = au_wbr_create_init_mfsrr,
35097 + .fin = au_wbr_create_fin_mfs
35099 + [AuWbrCreate_PMFSRRV] = {
35100 + .create = au_wbr_create_pmfsrr,
35101 + .init = au_wbr_create_init_mfsrr,
35102 + .fin = au_wbr_create_fin_mfs
35105 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
35106 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
35107 +++ linux/fs/aufs/whout.c 2023-04-24 08:13:40.972909448 +0200
35109 +// SPDX-License-Identifier: GPL-2.0
35111 + * Copyright (C) 2005-2022 Junjiro R. Okajima
35113 + * This program is free software; you can redistribute it and/or modify
35114 + * it under the terms of the GNU General Public License as published by
35115 + * the Free Software Foundation; either version 2 of the License, or
35116 + * (at your option) any later version.
35118 + * This program is distributed in the hope that it will be useful,
35119 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35120 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35121 + * GNU General Public License for more details.
35123 + * You should have received a copy of the GNU General Public License
35124 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35128 + * whiteout for logical deletion and opaque directory
35133 +#define WH_MASK 0444
35136 + * If a directory contains this file, then it is opaque. We start with the
35137 + * .wh. flag so that it is blocked by lookup.
35139 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
35140 + sizeof(AUFS_WH_DIROPQ) - 1);
35143 + * generate whiteout name, which is NOT terminated by NULL.
35144 + * @name: original d_name.name
35145 + * @len: original d_name.len
35146 + * @wh: whiteout qstr
35147 + * returns zero when succeeds, otherwise error.
35148 + * succeeded value as wh->name should be freed by kfree().
35150 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
35154 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
35155 + return -ENAMETOOLONG;
35157 + wh->len = name->len + AUFS_WH_PFX_LEN;
35158 + p = kmalloc(wh->len, GFP_NOFS);
35161 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35162 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
35169 +/* ---------------------------------------------------------------------- */
35172 + * test if the @wh_name exists under @h_ppath.
35173 + * @try_sio specifies the necessary of super-io.
35175 +int au_wh_test(struct mnt_idmap *h_idmap, struct path *h_ppath,
35176 + struct qstr *wh_name, int try_sio)
35179 + struct dentry *wh_dentry;
35182 + wh_dentry = vfsub_lkup_one(wh_name, h_ppath);
35184 + wh_dentry = au_sio_lkup_one(h_idmap, wh_name, h_ppath);
35185 + err = PTR_ERR(wh_dentry);
35186 + if (IS_ERR(wh_dentry)) {
35187 + if (err == -ENAMETOOLONG)
35193 + if (d_is_negative(wh_dentry))
35194 + goto out_wh; /* success */
35197 + if (d_is_reg(wh_dentry))
35198 + goto out_wh; /* success */
35201 + AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
35202 + wh_dentry, d_inode(wh_dentry)->i_mode);
35211 + * test if the @h_path->dentry sets opaque or not.
35213 +int au_diropq_test(struct mnt_idmap *h_idmap, struct path *h_path)
35216 + struct inode *h_dir;
35218 + h_dir = d_inode(h_path->dentry);
35219 + err = au_wh_test(h_idmap, h_path, &diropq_name,
35220 + au_test_h_perm_sio(h_idmap, h_dir, MAY_EXEC));
35225 + * returns a negative dentry whose name is unique and temporary.
35227 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
35228 + struct qstr *prefix)
35230 + struct dentry *dentry;
35232 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
35234 + /* strict atomic_t is unnecessary here */
35235 + static unsigned short cnt;
35237 + struct path h_ppath;
35238 + struct mnt_idmap *h_idmap;
35240 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
35243 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
35244 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
35245 + dentry = ERR_PTR(-ENAMETOOLONG);
35246 + if (unlikely(qs.len > NAME_MAX))
35248 + dentry = ERR_PTR(-ENOMEM);
35249 + name = kmalloc(qs.len + 1, GFP_NOFS);
35250 + if (unlikely(!name))
35254 + /* doubly whiteout-ed */
35255 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
35256 + p = name + AUFS_WH_PFX_LEN * 2;
35257 + memcpy(p, prefix->name, prefix->len);
35258 + p += prefix->len;
35260 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
35262 + h_ppath.dentry = h_parent;
35263 + h_ppath.mnt = au_br_mnt(br);
35264 + h_idmap = au_br_idmap(br);
35266 + for (i = 0; i < 3; i++) {
35267 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
35268 + dentry = au_sio_lkup_one(h_idmap, &qs, &h_ppath);
35269 + if (IS_ERR(dentry) || d_is_negative(dentry))
35273 + /* pr_warn("could not get random name\n"); */
35274 + dentry = ERR_PTR(-EEXIST);
35275 + AuDbg("%.*s\n", AuLNPair(&qs));
35279 + if (name != defname)
35280 + au_kfree_try_rcu(name);
35282 + AuTraceErrPtr(dentry);
35287 + * rename the @h_dentry on @br to the whiteouted temporary name.
35289 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
35292 + struct path h_path = {
35293 + .mnt = au_br_mnt(br)
35295 + struct inode *h_dir, *delegated;
35296 + struct dentry *h_parent;
35298 + h_parent = h_dentry->d_parent; /* dir inode is locked */
35299 + h_dir = d_inode(h_parent);
35300 + IMustLock(h_dir);
35302 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
35303 + err = PTR_ERR(h_path.dentry);
35304 + if (IS_ERR(h_path.dentry))
35307 + /* under the same dir, no need to lock_rename() */
35308 + delegated = NULL;
35309 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated,
35312 + if (unlikely(err == -EWOULDBLOCK)) {
35313 + pr_warn("cannot retry for NFSv4 delegation"
35314 + " for an internal rename\n");
35317 + dput(h_path.dentry);
35324 +/* ---------------------------------------------------------------------- */
35326 + * functions for removing a whiteout
35329 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
35332 + struct inode *delegated;
35335 + * forces superio when the dir has a sticky bit.
35336 + * this may be a violation of unix fs semantics.
35338 + force = (h_dir->i_mode & S_ISVTX)
35339 + && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
35340 + delegated = NULL;
35341 + err = vfsub_unlink(h_dir, h_path, &delegated, force);
35342 + if (unlikely(err == -EWOULDBLOCK)) {
35343 + pr_warn("cannot retry for NFSv4 delegation"
35344 + " for an internal unlink\n");
35350 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
35351 + struct dentry *dentry)
35355 + err = do_unlink_wh(h_dir, h_path);
35356 + if (!err && dentry)
35357 + au_set_dbwh(dentry, -1);
35362 +static int unlink_wh_name(struct path *h_ppath, struct qstr *wh)
35365 + struct path h_path;
35368 + h_path.dentry = vfsub_lkup_one(wh, h_ppath);
35369 + if (IS_ERR(h_path.dentry))
35370 + err = PTR_ERR(h_path.dentry);
35372 + if (d_is_reg(h_path.dentry)) {
35373 + h_path.mnt = h_ppath->mnt;
35374 + err = do_unlink_wh(d_inode(h_ppath->dentry), &h_path);
35376 + dput(h_path.dentry);
35382 +/* ---------------------------------------------------------------------- */
35384 + * initialize/clean whiteout for a branch
35387 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
35391 + struct inode *delegated;
35393 + if (d_is_negative(whpath->dentry))
35397 + err = vfsub_rmdir(h_dir, whpath);
35399 + delegated = NULL;
35400 + err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
35401 + if (unlikely(err == -EWOULDBLOCK)) {
35402 + pr_warn("cannot retry for NFSv4 delegation"
35403 + " for an internal unlink\n");
35407 + if (unlikely(err))
35408 + pr_warn("failed removing %pd (%d), ignored.\n",
35409 + whpath->dentry, err);
35412 +static int test_linkable(struct dentry *h_root)
35414 + struct inode *h_dir = d_inode(h_root);
35416 + if (h_dir->i_op->link)
35419 + pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
35420 + h_root, au_sbtype(h_root->d_sb));
35421 + return -ENOSYS; /* the branch doesn't have its ->link() */
35424 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
35425 +static int au_whdir(struct inode *h_dir, struct path *path)
35430 + if (d_is_negative(path->dentry)) {
35433 + if (au_test_nfs(path->dentry->d_sb))
35435 + err = vfsub_mkdir(h_dir, path, mode);
35436 + } else if (d_is_dir(path->dentry))
35439 + pr_err("unknown %pd exists\n", path->dentry);
35444 +struct au_wh_base {
35445 + const struct qstr *name;
35446 + struct dentry *dentry;
35449 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
35450 + struct path *h_path)
35452 + h_path->dentry = base[AuBrWh_BASE].dentry;
35453 + au_wh_clean(h_dir, h_path, /*isdir*/0);
35454 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35455 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35456 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35457 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35461 + * returns tri-state,
35462 + * minus: error, caller should print the message
35464 + * plus: error, caller should NOT print the message
35466 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
35467 + int do_plink, struct au_wh_base base[],
35468 + struct path *h_path)
35471 + struct inode *h_dir;
35473 + h_dir = d_inode(h_root);
35474 + h_path->dentry = base[AuBrWh_BASE].dentry;
35475 + au_wh_clean(h_dir, h_path, /*isdir*/0);
35476 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35478 + err = test_linkable(h_root);
35479 + if (unlikely(err)) {
35484 + err = au_whdir(h_dir, h_path);
35485 + if (unlikely(err))
35487 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35489 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35490 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35491 + err = au_whdir(h_dir, h_path);
35492 + if (unlikely(err))
35494 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35501 + * for the moment, aufs supports the branch filesystem which does not support
35502 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
35503 + * copyup failed. finally, such filesystem will not be used as the writable
35506 + * returns tri-state, see above.
35508 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
35509 + int do_plink, struct au_wh_base base[],
35510 + struct path *h_path)
35513 + struct inode *h_dir;
35515 + WbrWhMustWriteLock(wbr);
35517 + err = test_linkable(h_root);
35518 + if (unlikely(err)) {
35524 + * todo: should this create be done in /sbin/mount.aufs helper?
35527 + h_dir = d_inode(h_root);
35528 + if (d_is_negative(base[AuBrWh_BASE].dentry)) {
35529 + h_path->dentry = base[AuBrWh_BASE].dentry;
35530 + err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
35531 + } else if (d_is_reg(base[AuBrWh_BASE].dentry))
35534 + pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
35535 + if (unlikely(err))
35538 + h_path->dentry = base[AuBrWh_PLINK].dentry;
35540 + err = au_whdir(h_dir, h_path);
35541 + if (unlikely(err))
35543 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
35545 + au_wh_clean(h_dir, h_path, /*isdir*/1);
35546 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
35548 + h_path->dentry = base[AuBrWh_ORPH].dentry;
35549 + err = au_whdir(h_dir, h_path);
35550 + if (unlikely(err))
35552 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
35559 + * initialize the whiteout base file/dir for @br.
35561 +int au_wh_init(struct au_branch *br, struct super_block *sb)
35564 + const unsigned char do_plink
35565 + = !!au_opt_test(au_mntflags(sb), PLINK);
35566 + struct inode *h_dir;
35567 + struct path path = br->br_path;
35568 + struct dentry *h_root = path.dentry;
35569 + struct au_wbr *wbr = br->br_wbr;
35570 + static const struct qstr base_name[] = {
35571 + [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
35572 + sizeof(AUFS_BASE_NAME) - 1),
35573 + [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
35574 + sizeof(AUFS_PLINKDIR_NAME) - 1),
35575 + [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
35576 + sizeof(AUFS_ORPHDIR_NAME) - 1)
35578 + struct au_wh_base base[] = {
35579 + [AuBrWh_BASE] = {
35580 + .name = base_name + AuBrWh_BASE,
35583 + [AuBrWh_PLINK] = {
35584 + .name = base_name + AuBrWh_PLINK,
35587 + [AuBrWh_ORPH] = {
35588 + .name = base_name + AuBrWh_ORPH,
35594 + WbrWhMustWriteLock(wbr);
35596 + for (i = 0; i < AuBrWh_Last; i++) {
35597 + /* doubly whiteouted */
35598 + struct dentry *d;
35600 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
35601 + err = PTR_ERR(d);
35605 + base[i].dentry = d;
35607 + && wbr->wbr_wh[i]
35608 + && wbr->wbr_wh[i] != base[i].dentry);
35612 + for (i = 0; i < AuBrWh_Last; i++) {
35613 + dput(wbr->wbr_wh[i]);
35614 + wbr->wbr_wh[i] = NULL;
35618 + if (!au_br_writable(br->br_perm)) {
35619 + h_dir = d_inode(h_root);
35620 + au_wh_init_ro(h_dir, base, &path);
35621 + } else if (!au_br_wh_linkable(br->br_perm)) {
35622 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
35628 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
35634 + goto out; /* success */
35637 + pr_err("an error(%d) on the writable branch %pd(%s)\n",
35638 + err, h_root, au_sbtype(h_root->d_sb));
35640 + for (i = 0; i < AuBrWh_Last; i++)
35641 + dput(base[i].dentry);
35645 +/* ---------------------------------------------------------------------- */
35647 + * whiteouts are all hard-linked usually.
35648 + * when its link count reaches a ceiling, we create a new whiteout base
35649 + * asynchronously.
35652 +struct reinit_br_wh {
35653 + struct super_block *sb;
35654 + struct au_branch *br;
35657 +static void reinit_br_wh(void *arg)
35660 + aufs_bindex_t bindex;
35661 + struct path h_path;
35662 + struct reinit_br_wh *a = arg;
35663 + struct au_wbr *wbr;
35664 + struct inode *dir, *delegated;
35665 + struct dentry *h_root;
35666 + struct au_hinode *hdir;
35669 + wbr = a->br->br_wbr;
35670 + /* big aufs lock */
35671 + si_noflush_write_lock(a->sb);
35672 + if (!au_br_writable(a->br->br_perm))
35674 + bindex = au_br_index(a->sb, a->br->br_id);
35675 + if (unlikely(bindex < 0))
35678 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
35679 + dir = d_inode(a->sb->s_root);
35680 + hdir = au_hi(dir, bindex);
35681 + h_root = au_h_dptr(a->sb->s_root, bindex);
35682 + AuDebugOn(h_root != au_br_dentry(a->br));
35684 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35685 + wbr_wh_write_lock(wbr);
35686 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
35689 + h_path.dentry = wbr->wbr_whbase;
35690 + h_path.mnt = au_br_mnt(a->br);
35691 + delegated = NULL;
35692 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
35694 + if (unlikely(err == -EWOULDBLOCK)) {
35695 + pr_warn("cannot retry for NFSv4 delegation"
35696 + " for an internal unlink\n");
35700 + pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
35703 + dput(wbr->wbr_whbase);
35704 + wbr->wbr_whbase = NULL;
35706 + err = au_wh_init(a->br, a->sb);
35707 + wbr_wh_write_unlock(wbr);
35708 + au_hn_inode_unlock(hdir);
35709 + di_read_unlock(a->sb->s_root, AuLock_IR);
35711 + au_fhsm_wrote(a->sb, bindex, /*force*/0);
35715 + atomic_dec(&wbr->wbr_wh_running);
35716 + au_lcnt_dec(&a->br->br_count);
35717 + si_write_unlock(a->sb);
35718 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
35720 + if (unlikely(err))
35721 + AuIOErr("err %d\n", err);
35724 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
35726 + int do_dec, wkq_err;
35727 + struct reinit_br_wh *arg;
35730 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
35733 + /* ignore ENOMEM */
35734 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
35737 + * dec(wh_running), kfree(arg) and dec(br_count)
35738 + * in reinit function
35742 + au_lcnt_inc(&br->br_count);
35743 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
35744 + if (unlikely(wkq_err)) {
35745 + atomic_dec(&br->br_wbr->wbr_wh_running);
35746 + au_lcnt_dec(&br->br_count);
35747 + au_kfree_rcu(arg);
35754 + atomic_dec(&br->br_wbr->wbr_wh_running);
35757 +/* ---------------------------------------------------------------------- */
35760 + * create the whiteout @wh.
35762 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
35763 + struct dentry *wh)
35766 + struct path h_path = {
35769 + struct au_branch *br;
35770 + struct au_wbr *wbr;
35771 + struct dentry *h_parent;
35772 + struct inode *h_dir, *delegated;
35774 + h_parent = wh->d_parent; /* dir inode is locked */
35775 + h_dir = d_inode(h_parent);
35776 + IMustLock(h_dir);
35778 + br = au_sbr(sb, bindex);
35779 + h_path.mnt = au_br_mnt(br);
35780 + wbr = br->br_wbr;
35781 + wbr_wh_read_lock(wbr);
35782 + if (wbr->wbr_whbase) {
35783 + delegated = NULL;
35784 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
35785 + if (unlikely(err == -EWOULDBLOCK)) {
35786 + pr_warn("cannot retry for NFSv4 delegation"
35787 + " for an internal link\n");
35790 + if (!err || err != -EMLINK)
35793 + /* link count full. re-initialize br_whbase. */
35794 + kick_reinit_br_wh(sb, br);
35797 + /* return this error in this context */
35798 + err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
35800 + au_fhsm_wrote(sb, bindex, /*force*/0);
35803 + wbr_wh_read_unlock(wbr);
35807 +/* ---------------------------------------------------------------------- */
35810 + * create or remove the diropq.
35812 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
35813 + unsigned int flags)
35815 + struct dentry *opq_dentry;
35816 + struct super_block *sb;
35817 + struct au_branch *br;
35818 + struct path h_path;
35821 + sb = dentry->d_sb;
35822 + br = au_sbr(sb, bindex);
35823 + h_path.dentry = au_h_dptr(dentry, bindex);
35824 + h_path.mnt = au_br_mnt(br);
35825 + opq_dentry = vfsub_lkup_one(&diropq_name, &h_path);
35826 + if (IS_ERR(opq_dentry))
35829 + if (au_ftest_diropq(flags, CREATE)) {
35830 + err = link_or_create_wh(sb, bindex, opq_dentry);
35832 + au_set_dbdiropq(dentry, bindex);
35833 + goto out; /* success */
35836 + h_path.dentry = opq_dentry;
35837 + err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &h_path);
35839 + au_set_dbdiropq(dentry, -1);
35841 + dput(opq_dentry);
35842 + opq_dentry = ERR_PTR(err);
35845 + return opq_dentry;
35848 +struct do_diropq_args {
35849 + struct dentry **errp;
35850 + struct dentry *dentry;
35851 + aufs_bindex_t bindex;
35852 + unsigned int flags;
35855 +static void call_do_diropq(void *args)
35857 + struct do_diropq_args *a = args;
35858 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
35861 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35862 + unsigned int flags)
35864 + struct dentry *diropq, *h_dentry;
35865 + struct mnt_idmap *h_idmap;
35867 + h_idmap = au_sbr_idmap(dentry->d_sb, bindex);
35868 + h_dentry = au_h_dptr(dentry, bindex);
35869 + if (!au_test_h_perm_sio(h_idmap, d_inode(h_dentry),
35870 + MAY_EXEC | MAY_WRITE))
35871 + diropq = do_diropq(dentry, bindex, flags);
35874 + struct do_diropq_args args = {
35876 + .dentry = dentry,
35877 + .bindex = bindex,
35881 + wkq_err = au_wkq_wait(call_do_diropq, &args);
35882 + if (unlikely(wkq_err))
35883 + diropq = ERR_PTR(wkq_err);
35889 +/* ---------------------------------------------------------------------- */
35892 + * lookup whiteout dentry.
35893 + * @h_parent: lower parent dentry which must exist and be locked
35894 + * @base_name: name of dentry which will be whiteouted
35895 + * returns dentry for whiteout.
35897 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35898 + struct au_branch *br)
35901 + struct qstr wh_name;
35902 + struct dentry *wh_dentry;
35903 + struct path h_path;
35905 + err = au_wh_name_alloc(&wh_name, base_name);
35906 + wh_dentry = ERR_PTR(err);
35908 + h_path.dentry = h_parent;
35909 + h_path.mnt = au_br_mnt(br);
35910 + wh_dentry = vfsub_lkup_one(&wh_name, &h_path);
35911 + au_kfree_try_rcu(wh_name.name);
35913 + return wh_dentry;
35917 + * link/create a whiteout for @dentry on @bindex.
35919 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35920 + struct dentry *h_parent)
35922 + struct dentry *wh_dentry;
35923 + struct super_block *sb;
35926 + sb = dentry->d_sb;
35927 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
35928 + if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
35929 + err = link_or_create_wh(sb, bindex, wh_dentry);
35931 + au_set_dbwh(dentry, bindex);
35932 + au_fhsm_wrote(sb, bindex, /*force*/0);
35935 + wh_dentry = ERR_PTR(err);
35939 + return wh_dentry;
35942 +/* ---------------------------------------------------------------------- */
35944 +/* Delete all whiteouts in this directory on branch bindex. */
35945 +static int del_wh_children(struct path *h_path, struct au_nhash *whlist,
35946 + aufs_bindex_t bindex)
35949 + unsigned long ul, n;
35950 + struct qstr wh_name;
35952 + struct hlist_head *head;
35953 + struct au_vdir_wh *pos;
35954 + struct au_vdir_destr *str;
35957 + p = (void *)__get_free_page(GFP_NOFS);
35958 + wh_name.name = p;
35959 + if (unlikely(!wh_name.name))
35963 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35964 + p += AUFS_WH_PFX_LEN;
35965 + n = whlist->nh_num;
35966 + head = whlist->nh_head;
35967 + for (ul = 0; !err && ul < n; ul++, head++) {
35968 + hlist_for_each_entry(pos, head, wh_hash) {
35969 + if (pos->wh_bindex != bindex)
35972 + str = &pos->wh_str;
35973 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
35974 + memcpy(p, str->name, str->len);
35975 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
35976 + err = unlink_wh_name(h_path, &wh_name);
35981 + AuIOErr("whiteout name too long %.*s\n",
35982 + str->len, str->name);
35987 + free_page((unsigned long)wh_name.name);
35993 +struct del_wh_children_args {
35995 + struct path *h_path;
35996 + struct au_nhash *whlist;
35997 + aufs_bindex_t bindex;
36000 +static void call_del_wh_children(void *args)
36002 + struct del_wh_children_args *a = args;
36003 + *a->errp = del_wh_children(a->h_path, a->whlist, a->bindex);
36006 +/* ---------------------------------------------------------------------- */
36008 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
36010 + struct au_whtmp_rmdir *whtmp;
36012 + unsigned int rdhash;
36014 + SiMustAnyLock(sb);
36016 + whtmp = kzalloc(sizeof(*whtmp), gfp);
36017 + if (unlikely(!whtmp)) {
36018 + whtmp = ERR_PTR(-ENOMEM);
36022 + /* no estimation for dir size */
36023 + rdhash = au_sbi(sb)->si_rdhash;
36025 + rdhash = AUFS_RDHASH_DEF;
36026 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
36027 + if (unlikely(err)) {
36028 + au_kfree_rcu(whtmp);
36029 + whtmp = ERR_PTR(err);
36036 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
36039 + au_lcnt_dec(&whtmp->br->br_count);
36040 + dput(whtmp->wh_dentry);
36041 + iput(whtmp->dir);
36042 + au_nhash_wh_free(&whtmp->whlist);
36043 + au_kfree_rcu(whtmp);
36047 + * rmdir the whiteouted temporary named dir @h_dentry.
36048 + * @whlist: whiteouted children.
36050 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36051 + struct dentry *wh_dentry, struct au_nhash *whlist)
36054 + unsigned int h_nlink;
36055 + struct path wh_path;
36056 + struct inode *wh_inode, *h_dir;
36057 + struct au_branch *br;
36058 + struct mnt_idmap *h_idmap;
36060 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
36061 + IMustLock(h_dir);
36063 + br = au_sbr(dir->i_sb, bindex);
36064 + wh_path.dentry = wh_dentry;
36065 + wh_path.mnt = au_br_mnt(br);
36066 + h_idmap = au_br_idmap(br);
36067 + wh_inode = d_inode(wh_dentry);
36068 + inode_lock_nested(wh_inode, AuLsc_I_CHILD);
36071 + * someone else might change some whiteouts while we were sleeping.
36072 + * it means this whlist may have an obsoleted entry.
36074 + if (!au_test_h_perm_sio(h_idmap, wh_inode, MAY_EXEC | MAY_WRITE))
36075 + err = del_wh_children(&wh_path, whlist, bindex);
36078 + struct del_wh_children_args args = {
36080 + .h_path = &wh_path,
36081 + .whlist = whlist,
36085 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
36086 + if (unlikely(wkq_err))
36089 + inode_unlock(wh_inode);
36092 + h_nlink = h_dir->i_nlink;
36093 + err = vfsub_rmdir(h_dir, &wh_path);
36094 + /* some fs doesn't change the parent nlink in some cases */
36095 + h_nlink -= h_dir->i_nlink;
36099 + if (au_ibtop(dir) == bindex) {
36100 + /* todo: dir->i_mutex is necessary */
36101 + au_cpup_attr_timesizes(dir);
36103 + vfsub_drop_nlink(dir);
36105 + return 0; /* success */
36108 + pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
36112 +static void call_rmdir_whtmp(void *args)
36115 + aufs_bindex_t bindex;
36116 + struct au_whtmp_rmdir *a = args;
36117 + struct super_block *sb;
36118 + struct dentry *h_parent;
36119 + struct inode *h_dir;
36120 + struct au_hinode *hdir;
36122 + /* rmdir by nfsd may cause deadlock with this i_mutex */
36123 + /* inode_lock(a->dir); */
36125 + sb = a->dir->i_sb;
36126 + si_read_lock(sb, !AuLock_FLUSH);
36127 + if (!au_br_writable(a->br->br_perm))
36129 + bindex = au_br_index(sb, a->br->br_id);
36130 + if (unlikely(bindex < 0))
36134 + ii_write_lock_parent(a->dir);
36135 + h_parent = dget_parent(a->wh_dentry);
36136 + h_dir = d_inode(h_parent);
36137 + hdir = au_hi(a->dir, bindex);
36138 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
36139 + if (unlikely(err))
36141 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
36142 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
36145 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
36146 + au_hn_inode_unlock(hdir);
36147 + vfsub_mnt_drop_write(au_br_mnt(a->br));
36151 + ii_write_unlock(a->dir);
36153 + /* inode_unlock(a->dir); */
36154 + au_whtmp_rmdir_free(a);
36155 + si_read_unlock(sb);
36156 + au_nwt_done(&au_sbi(sb)->si_nowait);
36157 + if (unlikely(err))
36158 + AuIOErr("err %d\n", err);
36161 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36162 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
36165 + struct super_block *sb;
36169 + /* all post-process will be done in do_rmdir_whtmp(). */
36171 + args->dir = au_igrab(dir);
36172 + args->br = au_sbr(sb, bindex);
36173 + au_lcnt_inc(&args->br->br_count);
36174 + args->wh_dentry = dget(wh_dentry);
36175 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
36176 + if (unlikely(wkq_err)) {
36177 + pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
36178 + au_whtmp_rmdir_free(args);
36181 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
36182 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
36183 +++ linux/fs/aufs/whout.h 2023-04-24 08:13:40.972909448 +0200
36185 +/* SPDX-License-Identifier: GPL-2.0 */
36187 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36189 + * This program is free software; you can redistribute it and/or modify
36190 + * it under the terms of the GNU General Public License as published by
36191 + * the Free Software Foundation; either version 2 of the License, or
36192 + * (at your option) any later version.
36194 + * This program is distributed in the hope that it will be useful,
36195 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36196 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36197 + * GNU General Public License for more details.
36199 + * You should have received a copy of the GNU General Public License
36200 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36204 + * whiteout for logical deletion and opaque directory
36207 +#ifndef __AUFS_WHOUT_H__
36208 +#define __AUFS_WHOUT_H__
36215 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
36216 +int au_wh_test(struct mnt_idmap *h_idmap, struct path *h_ppath,
36217 + struct qstr *wh_name, int try_sio);
36218 +int au_diropq_test(struct mnt_idmap *h_idmap, struct path *h_path);
36220 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
36221 + struct qstr *prefix);
36222 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
36223 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
36224 + struct dentry *dentry);
36225 +int au_wh_init(struct au_branch *br, struct super_block *sb);
36227 +/* diropq flags */
36228 +#define AuDiropq_CREATE 1
36229 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
36230 +#define au_fset_diropq(flags, name) \
36231 + do { (flags) |= AuDiropq_##name; } while (0)
36232 +#define au_fclr_diropq(flags, name) \
36233 + do { (flags) &= ~AuDiropq_##name; } while (0)
36235 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
36236 + unsigned int flags);
36237 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
36238 + struct au_branch *br);
36239 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
36240 + struct dentry *h_parent);
36242 +/* real rmdir for the whiteout-ed dir */
36243 +struct au_whtmp_rmdir {
36244 + struct inode *dir;
36245 + struct au_branch *br;
36246 + struct dentry *wh_dentry;
36247 + struct au_nhash whlist;
36250 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
36251 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
36252 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
36253 + struct dentry *wh_dentry, struct au_nhash *whlist);
36254 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
36255 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
36257 +/* ---------------------------------------------------------------------- */
36259 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
36260 + aufs_bindex_t bindex)
36262 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
36265 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
36267 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
36270 +#endif /* __KERNEL__ */
36271 +#endif /* __AUFS_WHOUT_H__ */
36272 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
36273 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
36274 +++ linux/fs/aufs/wkq.c 2022-11-05 23:02:18.972555950 +0100
36276 +// SPDX-License-Identifier: GPL-2.0
36278 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36280 + * This program is free software; you can redistribute it and/or modify
36281 + * it under the terms of the GNU General Public License as published by
36282 + * the Free Software Foundation; either version 2 of the License, or
36283 + * (at your option) any later version.
36285 + * This program is distributed in the hope that it will be useful,
36286 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36287 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36288 + * GNU General Public License for more details.
36290 + * You should have received a copy of the GNU General Public License
36291 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36295 + * workqueue for asynchronous/super-io operations
36296 + * todo: try new credential scheme
36299 +#include <linux/module.h>
36302 +/* internal workqueue named AUFS_WKQ_NAME */
36304 +static struct workqueue_struct *au_wkq;
36306 +struct au_wkinfo {
36307 + struct work_struct wk;
36308 + struct kobject *kobj;
36310 + unsigned int flags; /* see wkq.h */
36312 + au_wkq_func_t func;
36315 +#ifdef CONFIG_LOCKDEP
36317 + struct held_lock **hlock;
36320 + struct completion *comp;
36323 +/* ---------------------------------------------------------------------- */
36325 + * Aufs passes some operations to the workqueue such as the internal copyup.
36326 + * This scheme looks rather unnatural for LOCKDEP debugging feature, since the
36327 + * job run by workqueue depends upon the locks acquired in the other task.
36328 + * Delegating a small operation to the workqueue, aufs passes its lockdep
36329 + * information too. And the job in the workqueue restores the info in order to
36330 + * pretend as if it acquired those locks. This is just to make LOCKDEP work
36331 + * correctly and expectedly.
36334 +#ifndef CONFIG_LOCKDEP
36335 +AuStubInt0(au_wkq_lockdep_alloc, struct au_wkinfo *wkinfo);
36336 +AuStubVoid(au_wkq_lockdep_free, struct au_wkinfo *wkinfo);
36337 +AuStubVoid(au_wkq_lockdep_pre, struct au_wkinfo *wkinfo);
36338 +AuStubVoid(au_wkq_lockdep_post, struct au_wkinfo *wkinfo);
36339 +AuStubVoid(au_wkq_lockdep_init, struct au_wkinfo *wkinfo);
36341 +static void au_wkq_lockdep_init(struct au_wkinfo *wkinfo)
36343 + wkinfo->hlock = NULL;
36344 + wkinfo->dont_check = 0;
36351 +static int au_wkq_lockdep_test(struct lock_class_key *key, const char *name)
36353 + static DEFINE_SPINLOCK(spin);
36356 + struct lock_class_key *key;
36358 + { .name = "&sbinfo->si_rwsem" },
36359 + { .name = "&finfo->fi_rwsem" },
36360 + { .name = "&dinfo->di_rwsem" },
36361 + { .name = "&iinfo->ii_rwsem" }
36366 + /* lockless read from 'set.' see below */
36367 + if (set == ARRAY_SIZE(a)) {
36368 + for (i = 0; i < ARRAY_SIZE(a); i++)
36369 + if (a[i].key == key)
36374 + spin_lock(&spin);
36376 + for (i = 0; i < ARRAY_SIZE(a); i++)
36377 + if (a[i].key == key) {
36378 + spin_unlock(&spin);
36381 + for (i = 0; i < ARRAY_SIZE(a); i++) {
36383 + if (unlikely(a[i].key == key)) { /* rare but possible */
36384 + spin_unlock(&spin);
36389 + if (strstr(a[i].name, name)) {
36391 + * the order of these three lines is important for the
36392 + * lockless read above.
36395 + spin_unlock(&spin);
36397 + /* AuDbg("%d, %s\n", set, name); */
36401 + spin_unlock(&spin);
36410 +static int au_wkq_lockdep_alloc(struct au_wkinfo *wkinfo)
36413 + struct task_struct *curr;
36414 + struct held_lock **hl, *held_locks, *p;
36418 + wkinfo->dont_check = lockdep_recursing(curr);
36419 + if (wkinfo->dont_check)
36421 + n = curr->lockdep_depth;
36426 + wkinfo->hlock = kmalloc_array(n + 1, sizeof(*wkinfo->hlock), GFP_NOFS);
36427 + if (unlikely(!wkinfo->hlock))
36431 +#if 0 /* left for debugging */
36432 + if (0 && au_debug_test())
36433 + lockdep_print_held_locks(curr);
36435 + held_locks = curr->held_locks;
36436 + hl = wkinfo->hlock;
36438 + p = held_locks++;
36439 + if (au_wkq_lockdep_test(p->instance->key, p->instance->name))
36448 +static void au_wkq_lockdep_free(struct au_wkinfo *wkinfo)
36450 + au_kfree_try_rcu(wkinfo->hlock);
36453 +static void au_wkq_lockdep_pre(struct au_wkinfo *wkinfo)
36455 + struct held_lock *p, **hl = wkinfo->hlock;
36458 + if (wkinfo->dont_check)
36462 + while ((p = *hl++)) { /* assignment */
36463 + subclass = lockdep_hlock_class(p)->subclass;
36464 + /* AuDbg("%s, %d\n", p->instance->name, subclass); */
36466 + rwsem_acquire_read(p->instance, subclass, 0,
36467 + /*p->acquire_ip*/_RET_IP_);
36469 + rwsem_acquire(p->instance, subclass, 0,
36470 + /*p->acquire_ip*/_RET_IP_);
36474 +static void au_wkq_lockdep_post(struct au_wkinfo *wkinfo)
36476 + struct held_lock *p, **hl = wkinfo->hlock;
36478 + if (wkinfo->dont_check)
36482 + while ((p = *hl++)) /* assignment */
36483 + rwsem_release(p->instance, /*p->acquire_ip*/_RET_IP_);
36487 +static void wkq_func(struct work_struct *wk)
36489 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
36491 + AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
36492 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
36494 + au_wkq_lockdep_pre(wkinfo);
36495 + wkinfo->func(wkinfo->args);
36496 + au_wkq_lockdep_post(wkinfo);
36497 + if (au_ftest_wkq(wkinfo->flags, WAIT))
36498 + complete(wkinfo->comp);
36500 + kobject_put(wkinfo->kobj);
36501 + module_put(THIS_MODULE); /* todo: ?? */
36502 + au_kfree_rcu(wkinfo);
36507 + * Since struct completion is large, try allocating it dynamically.
36509 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
36511 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
36513 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
36515 + init_completion(*comp);
36516 + wkinfo->comp = *comp;
36522 +static void au_wkq_comp_free(struct completion *comp)
36524 + au_kfree_rcu(comp);
36527 +static void au_wkq_run(struct au_wkinfo *wkinfo)
36529 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
36530 + if (au_wkq_test()) {
36531 + AuWarn1("wkq from wkq, unless silly-rename on NFS,"
36532 + " due to a dead dir by UDBA,"
36533 + " or async xino write?\n");
36534 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
36537 + au_dbg_verify_kthread();
36539 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
36540 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
36541 + queue_work(au_wkq, &wkinfo->wk);
36543 + INIT_WORK(&wkinfo->wk, wkq_func);
36544 + schedule_work(&wkinfo->wk);
36549 + * Be careful. It is easy to make deadlock happen.
36550 + * processA: lock, wkq and wait
36551 + * processB: wkq and wait, lock in wkq
36554 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
36557 + AuWkqCompDeclare(comp);
36558 + struct au_wkinfo wkinfo = {
36564 + err = au_wkq_comp_alloc(&wkinfo, &comp);
36565 + if (unlikely(err))
36567 + err = au_wkq_lockdep_alloc(&wkinfo);
36568 + if (unlikely(err))
36571 + au_wkq_run(&wkinfo);
36572 + /* no timeout, no interrupt */
36573 + wait_for_completion(wkinfo.comp);
36575 + au_wkq_lockdep_free(&wkinfo);
36578 + au_wkq_comp_free(comp);
36580 + destroy_work_on_stack(&wkinfo.wk);
36585 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
36586 + * problem in a concurrent umounting.
36588 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36589 + unsigned int flags)
36592 + struct au_wkinfo *wkinfo;
36594 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
36597 + * wkq_func() must free this wkinfo.
36598 + * it highly depends upon the implementation of workqueue.
36601 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
36603 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
36604 + wkinfo->flags = flags & ~AuWkq_WAIT;
36605 + wkinfo->func = func;
36606 + wkinfo->args = args;
36607 + wkinfo->comp = NULL;
36608 + au_wkq_lockdep_init(wkinfo);
36609 + kobject_get(wkinfo->kobj);
36610 + __module_get(THIS_MODULE); /* todo: ?? */
36612 + au_wkq_run(wkinfo);
36615 + au_nwt_done(&au_sbi(sb)->si_nowait);
36621 +/* ---------------------------------------------------------------------- */
36623 +void au_nwt_init(struct au_nowait_tasks *nwt)
36625 + atomic_set(&nwt->nw_len, 0);
36626 + /* smp_mb(); */ /* atomic_set */
36627 + init_waitqueue_head(&nwt->nw_wq);
36630 +void au_wkq_fin(void)
36632 + destroy_workqueue(au_wkq);
36635 +int __init au_wkq_init(void)
36640 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
36641 + if (IS_ERR(au_wkq))
36642 + err = PTR_ERR(au_wkq);
36643 + else if (!au_wkq)
36648 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
36649 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
36650 +++ linux/fs/aufs/wkq.h 2022-11-05 23:02:18.972555950 +0100
36652 +/* SPDX-License-Identifier: GPL-2.0 */
36654 + * Copyright (C) 2005-2022 Junjiro R. Okajima
36656 + * This program is free software; you can redistribute it and/or modify
36657 + * it under the terms of the GNU General Public License as published by
36658 + * the Free Software Foundation; either version 2 of the License, or
36659 + * (at your option) any later version.
36661 + * This program is distributed in the hope that it will be useful,
36662 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36663 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36664 + * GNU General Public License for more details.
36666 + * You should have received a copy of the GNU General Public License
36667 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36671 + * workqueue for asynchronous/super-io operations
36672 + * todo: try new credentials management scheme
36675 +#ifndef __AUFS_WKQ_H__
36676 +#define __AUFS_WKQ_H__
36680 +#include <linux/wait.h>
36682 +struct super_block;
36684 +/* ---------------------------------------------------------------------- */
36687 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
36689 +struct au_nowait_tasks {
36691 + wait_queue_head_t nw_wq;
36694 +/* ---------------------------------------------------------------------- */
36696 +typedef void (*au_wkq_func_t)(void *args);
36699 +#define AuWkq_WAIT 1
36700 +#define AuWkq_NEST (1 << 1)
36701 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
36702 +#define au_fset_wkq(flags, name) \
36703 + do { (flags) |= AuWkq_##name; } while (0)
36704 +#define au_fclr_wkq(flags, name) \
36705 + do { (flags) &= ~AuWkq_##name; } while (0)
36708 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
36709 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36710 + unsigned int flags);
36711 +void au_nwt_init(struct au_nowait_tasks *nwt);
36712 +int __init au_wkq_init(void);
36713 +void au_wkq_fin(void);
36715 +/* ---------------------------------------------------------------------- */
36717 +static inline int au_wkq_test(void)
36719 + return current->flags & PF_WQ_WORKER;
36722 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
36724 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
36727 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
36729 + if (atomic_dec_and_test(&nwt->nw_len))
36730 + wake_up_all(&nwt->nw_wq);
36733 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
36735 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
36739 +#endif /* __KERNEL__ */
36740 +#endif /* __AUFS_WKQ_H__ */
36741 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
36742 --- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
36743 +++ linux/fs/aufs/xattr.c 2023-07-04 23:43:55.849649534 +0200
36745 +// SPDX-License-Identifier: GPL-2.0
36747 + * Copyright (C) 2014-2022 Junjiro R. Okajima
36749 + * This program is free software; you can redistribute it and/or modify
36750 + * it under the terms of the GNU General Public License as published by
36751 + * the Free Software Foundation; either version 2 of the License, or
36752 + * (at your option) any later version.
36754 + * This program is distributed in the hope that it will be useful,
36755 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36756 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36757 + * GNU General Public License for more details.
36759 + * You should have received a copy of the GNU General Public License
36760 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36764 + * handling xattr functions
36767 +#include <linux/fs.h>
36768 +#include <linux/xattr.h>
36771 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
36773 + if (!ignore_flags)
36781 + if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
36786 +#define cmp(brattr, prefix) do { \
36787 + if (!strncmp(name, XATTR_##prefix##_PREFIX, \
36788 + XATTR_##prefix##_PREFIX_LEN)) { \
36789 + if (ignore_flags & AuBrAttr_ICEX_##brattr) \
36795 + cmp(SEC, SECURITY);
36796 + cmp(SYS, SYSTEM);
36797 + cmp(TR, TRUSTED);
36801 + if (ignore_flags & AuBrAttr_ICEX_OTH)
36808 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
36810 +static int au_do_cpup_xattr(struct path *h_dst, struct path *h_src,
36811 + char *name, char **buf, unsigned int ignore_flags,
36812 + unsigned int verbose)
36816 + struct inode *h_idst;
36817 + struct dentry *h_dst_dentry, *h_src_dentry;
36818 + struct mnt_idmap *h_dst_idmap, *h_src_idmap;
36819 + struct posix_acl *acl;
36821 + is_acl = !!is_posix_acl_xattr(name);
36822 + h_src_idmap = mnt_idmap(h_src->mnt);
36823 + h_src_dentry = h_src->dentry;
36825 + acl = vfs_get_acl(h_src_idmap, h_src_dentry, name);
36827 + if (unlikely(IS_ERR(acl))) {
36828 + err = PTR_ERR(acl);
36829 + if (err == -ENODATA)
36831 + else if (err == -EOPNOTSUPP
36832 + && au_test_nfs_noacl(d_inode(h_src_dentry)))
36834 + else if (verbose || au_debug_test())
36835 + pr_err("%s, err %d\n", name, err);
36839 + ssz = vfs_getxattr_alloc(h_src_idmap, h_src_dentry, name, buf,
36841 + if (unlikely(ssz <= 0)) {
36843 + if (err == -ENODATA)
36845 + else if (err == -EOPNOTSUPP
36846 + && (ignore_flags & au_xattr_out_of_list))
36848 + else if (err && (verbose || au_debug_test()))
36849 + pr_err("%s, err %d\n", name, err);
36854 + /* unlock it temporary */
36855 + h_dst_idmap = mnt_idmap(h_dst->mnt);
36856 + h_dst_dentry = h_dst->dentry;
36857 + h_idst = d_inode(h_dst_dentry);
36858 + inode_unlock(h_idst);
36860 + err = vfsub_set_acl(h_dst_idmap, h_dst_dentry, name, acl);
36861 + posix_acl_release(acl);
36863 + err = vfsub_setxattr(h_dst_idmap, h_dst_dentry, name, *buf,
36864 + ssz, /*flags*/0);
36865 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36866 + if (unlikely(err)) {
36867 + if (verbose || au_debug_test())
36868 + pr_err("%s, err %d\n", name, err);
36869 + err = au_xattr_ignore(err, name, ignore_flags);
36876 +int au_cpup_xattr(struct path *h_dst, struct path *h_src, int ignore_flags,
36877 + unsigned int verbose)
36879 + int err, unlocked;
36881 + struct dentry *h_dst_dentry, *h_src_dentry;
36882 + struct inode *h_isrc, *h_idst;
36883 + char *value, *p, *o, *e;
36885 + /* try stopping to update the source inode while we are referencing */
36886 + /* there should not be the parent-child relationship between them */
36887 + h_dst_dentry = h_dst->dentry;
36888 + h_idst = d_inode(h_dst_dentry);
36889 + h_src_dentry = h_src->dentry;
36890 + h_isrc = d_inode(h_src_dentry);
36891 + inode_unlock(h_idst);
36892 + inode_lock_shared_nested(h_isrc, AuLsc_I_CHILD);
36893 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36896 + /* some filesystems don't list POSIX ACL, for example tmpfs */
36897 + ssz = vfs_listxattr(h_src_dentry, NULL, 0);
36899 + if (unlikely(err < 0)) {
36901 + if (err == -ENODATA
36902 + || err == -EOPNOTSUPP)
36903 + err = 0; /* ignore */
36912 + p = kmalloc(ssz, GFP_NOFS);
36914 + if (unlikely(!p))
36916 + err = vfs_listxattr(h_src_dentry, p, ssz);
36918 + inode_unlock_shared(h_isrc);
36920 + AuDbg("err %d, ssz %zd\n", err, ssz);
36921 + if (unlikely(err < 0))
36927 + ignore_flags |= au_xattr_out_of_list;
36928 + while (!err && p < e) {
36929 + err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
36931 + p += strlen(p) + 1;
36933 + au_kfree_try_rcu(value);
36936 + au_kfree_try_rcu(o);
36939 + inode_unlock_shared(h_isrc);
36944 +/* ---------------------------------------------------------------------- */
36946 +static int au_smack_reentering(struct super_block *sb)
36948 +#if IS_ENABLED(CONFIG_SECURITY_SMACK) || IS_ENABLED(CONFIG_SECURITY_SELINUX)
36950 + * as a part of lookup, smack_d_instantiate() is called, and it calls
36951 + * i_op->getxattr(). ouch.
36953 + return si_pid_test(sb);
36964 +struct au_lgxattr {
36972 + const char *name;
36979 +static ssize_t au_lgxattr(struct dentry *dentry, struct inode *inode,
36980 + struct au_lgxattr *arg)
36984 + struct path h_path;
36985 + struct super_block *sb;
36987 + sb = dentry->d_sb;
36988 + reenter = au_smack_reentering(sb);
36990 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
36991 + if (unlikely(err))
36994 + err = au_h_path_getattr(dentry, inode, /*force*/1, &h_path, reenter);
36995 + if (unlikely(err))
36997 + if (unlikely(!h_path.dentry))
36998 + /* illegally overlapped or something */
36999 + goto out_di; /* pretending success */
37001 + /* always topmost entry only */
37002 + switch (arg->type) {
37003 + case AU_XATTR_LIST:
37004 + err = vfs_listxattr(h_path.dentry,
37005 + arg->u.list.list, arg->u.list.size);
37007 + case AU_XATTR_GET:
37008 + AuDebugOn(d_is_negative(h_path.dentry));
37009 + err = vfs_getxattr(mnt_idmap(h_path.mnt), h_path.dentry,
37010 + arg->u.get.name, arg->u.get.value,
37011 + arg->u.get.size);
37017 + di_read_unlock(dentry, AuLock_IR);
37020 + si_read_unlock(sb);
37026 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
37028 + struct au_lgxattr arg = {
37029 + .type = AU_XATTR_LIST,
37036 + return au_lgxattr(dentry, /*inode*/NULL, &arg);
37039 +static ssize_t au_getxattr(struct dentry *dentry, struct inode *inode,
37040 + const char *name, void *value, size_t size)
37042 + struct au_lgxattr arg = {
37043 + .type = AU_XATTR_GET,
37051 + return au_lgxattr(dentry, inode, &arg);
37054 +static int au_setxattr(struct dentry *dentry, struct inode *inode,
37055 + const char *name, const void *value, size_t size,
37058 + struct au_sxattr arg = {
37059 + .type = AU_XATTR_SET,
37068 + return au_sxattr(dentry, inode, &arg);
37071 +/* ---------------------------------------------------------------------- */
37073 +static int au_xattr_get(const struct xattr_handler *handler,
37074 + struct dentry *dentry, struct inode *inode,
37075 + const char *name, void *buffer, size_t size)
37077 + return au_getxattr(dentry, inode, name, buffer, size);
37080 +static int au_xattr_set(const struct xattr_handler *handler,
37081 + struct mnt_idmap *idmap,
37082 + struct dentry *dentry, struct inode *inode,
37083 + const char *name, const void *value, size_t size,
37086 + return au_setxattr(dentry, inode, name, value, size, flags);
37089 +static const struct xattr_handler au_xattr_handler = {
37092 + .get = au_xattr_get,
37093 + .set = au_xattr_set
37096 +static const struct xattr_handler *au_xattr_handlers[] = {
37097 + &au_xattr_handler, /* must be last */
37101 +void au_xattr_init(struct super_block *sb)
37103 + sb->s_xattr = au_xattr_handlers;
37105 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
37106 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
37107 +++ linux/fs/aufs/xino.c 2023-04-24 08:13:40.972909448 +0200
37109 +// SPDX-License-Identifier: GPL-2.0
37111 + * Copyright (C) 2005-2022 Junjiro R. Okajima
37113 + * This program is free software; you can redistribute it and/or modify
37114 + * it under the terms of the GNU General Public License as published by
37115 + * the Free Software Foundation; either version 2 of the License, or
37116 + * (at your option) any later version.
37118 + * This program is distributed in the hope that it will be useful,
37119 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
37120 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
37121 + * GNU General Public License for more details.
37123 + * You should have received a copy of the GNU General Public License
37124 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
37128 + * external inode number translation table and bitmap
37130 + * things to consider
37132 + * + au_xino object
37133 + * + XINO files (xino, xib, xigen)
37134 + * + dynamic debugfs entries (xiN)
37135 + * + static debugfs entries (xib, xigen)
37136 + * + static sysfs entry (xi_path)
37137 + * - several entry points to handle them.
37138 + * + mount(2) without xino option (default)
37139 + * + mount(2) with xino option
37140 + * + mount(2) with noxino option
37142 + * + remount with add/del branches
37143 + * + remount with xino/noxino options
37146 +#include <linux/seq_file.h>
37147 +#include <linux/statfs.h>
37150 +static aufs_bindex_t sbr_find_shared(struct super_block *sb, aufs_bindex_t btop,
37151 + aufs_bindex_t bbot,
37152 + struct super_block *h_sb)
37154 + /* todo: try binary-search if the branches are many */
37155 + for (; btop <= bbot; btop++)
37156 + if (h_sb == au_sbr_sb(sb, btop))
37162 + * find another branch who is on the same filesystem of the specified
37163 + * branch{@btgt}. search until @bbot.
37165 +static aufs_bindex_t is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
37166 + aufs_bindex_t bbot)
37168 + aufs_bindex_t bindex;
37169 + struct super_block *tgt_sb;
37171 + tgt_sb = au_sbr_sb(sb, btgt);
37172 + bindex = sbr_find_shared(sb, /*btop*/0, btgt - 1, tgt_sb);
37174 + bindex = sbr_find_shared(sb, btgt + 1, bbot, tgt_sb);
37179 +/* ---------------------------------------------------------------------- */
37182 + * stop unnecessary notify events at creating xino files
37185 +aufs_bindex_t au_xi_root(struct super_block *sb, struct dentry *dentry)
37187 + aufs_bindex_t bfound, bindex, bbot;
37188 + struct dentry *parent;
37189 + struct au_branch *br;
37192 + parent = dentry->d_parent; /* safe d_parent access */
37193 + bbot = au_sbbot(sb);
37194 + for (bindex = 0; bindex <= bbot; bindex++) {
37195 + br = au_sbr(sb, bindex);
37196 + if (au_br_dentry(br) == parent) {
37202 + AuDbg("bfound b%d\n", bfound);
37206 +struct au_xino_lock_dir {
37207 + struct au_hinode *hdir;
37208 + struct dentry *parent;
37209 + struct inode *dir;
37212 +static struct dentry *au_dget_parent_lock(struct dentry *dentry,
37213 + unsigned int lsc)
37215 + struct dentry *parent;
37216 + struct inode *dir;
37218 + parent = dget_parent(dentry);
37219 + dir = d_inode(parent);
37220 + inode_lock_nested(dir, lsc);
37221 +#if 0 /* it should not happen */
37222 + spin_lock(&dentry->d_lock);
37223 + if (unlikely(dentry->d_parent != parent)) {
37224 + spin_unlock(&dentry->d_lock);
37225 + inode_unlock(dir);
37230 + spin_unlock(&dentry->d_lock);
37237 +static void au_xino_lock_dir(struct super_block *sb, struct path *xipath,
37238 + struct au_xino_lock_dir *ldir)
37240 + aufs_bindex_t bindex;
37242 + ldir->hdir = NULL;
37243 + bindex = au_xi_root(sb, xipath->dentry);
37244 + if (bindex >= 0) {
37245 + /* rw branch root */
37246 + ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
37247 + au_hn_inode_lock_nested(ldir->hdir, AuLsc_I_PARENT);
37250 + ldir->parent = au_dget_parent_lock(xipath->dentry,
37252 + ldir->dir = d_inode(ldir->parent);
37256 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
37259 + au_hn_inode_unlock(ldir->hdir);
37261 + inode_unlock(ldir->dir);
37262 + dput(ldir->parent);
37266 +/* ---------------------------------------------------------------------- */
37269 + * create and set a new xino file
37271 +struct file *au_xino_create(struct super_block *sb, char *fpath, int silent,
37274 + struct file *file;
37275 + struct dentry *h_parent, *d;
37276 + struct inode *h_dir, *inode;
37278 + static DEFINE_MUTEX(mtx);
37281 + * at mount-time, and the xino file is the default path,
37282 + * hnotify is disabled so we have no notify events to ignore.
37283 + * when a user specified the xino, we cannot get au_hdir to be ignored.
37286 + mutex_lock(&mtx);
37287 + file = vfsub_filp_open(fpath, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37288 + /* | __FMODE_NONOTIFY */,
37290 + if (IS_ERR(file)) {
37292 + mutex_unlock(&mtx);
37294 + pr_err("open %s(%ld)\n", fpath, PTR_ERR(file));
37298 + /* keep file count */
37300 + d = file->f_path.dentry;
37301 + h_parent = au_dget_parent_lock(d, AuLsc_I_PARENT);
37303 + mutex_unlock(&mtx);
37304 + /* mnt_want_write() is unnecessary here */
37305 + h_dir = d_inode(h_parent);
37306 + inode = file_inode(file);
37307 + /* no delegation since it is just created */
37308 + if (inode->i_nlink)
37309 + err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
37311 + inode_unlock(h_dir);
37313 + if (unlikely(err)) {
37315 + pr_err("unlink %s(%d)\n", fpath, err);
37320 + if (unlikely(sb && sb == d->d_sb)) {
37322 + pr_err("%s must be outside\n", fpath);
37325 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
37327 + pr_err("xino doesn't support %s(%s)\n",
37328 + fpath, au_sbtype(d->d_sb));
37331 + return file; /* success */
37335 + file = ERR_PTR(err);
37340 + * create a new xinofile at the same place/path as @base.
37342 +struct file *au_xino_create2(struct super_block *sb, struct path *base,
37343 + struct file *copy_src)
37345 + struct file *file;
37346 + struct dentry *dentry;
37347 + struct inode *dir, *delegated;
37348 + struct qstr *name;
37349 + struct path ppath, path;
37350 + int err, do_unlock;
37351 + struct au_xino_lock_dir ldir;
37354 + au_xino_lock_dir(sb, base, &ldir);
37355 + dentry = base->dentry;
37356 + ppath.dentry = dentry->d_parent; /* dir inode is locked */
37357 + ppath.mnt = base->mnt;
37358 + dir = d_inode(ppath.dentry);
37361 + name = &dentry->d_name;
37362 + path.dentry = vfsub_lookup_one_len(name->name, &ppath, name->len);
37363 + if (IS_ERR(path.dentry)) {
37364 + file = (void *)path.dentry;
37365 + pr_err("%pd lookup err %ld\n", dentry, PTR_ERR(path.dentry));
37369 + /* no need to mnt_want_write() since we call dentry_open() later */
37370 + err = vfs_create(mnt_idmap(base->mnt), dir, path.dentry, 0666, NULL);
37371 + if (unlikely(err)) {
37372 + file = ERR_PTR(err);
37373 + pr_err("%pd create err %d\n", dentry, err);
37377 + path.mnt = base->mnt;
37378 + file = vfsub_dentry_open(&path,
37379 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37380 + /* | __FMODE_NONOTIFY */);
37381 + if (IS_ERR(file)) {
37382 + pr_err("%pd open err %ld\n", dentry, PTR_ERR(file));
37386 + delegated = NULL;
37387 + err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
37388 + au_xino_unlock_dir(&ldir);
37390 + if (unlikely(err == -EWOULDBLOCK)) {
37391 + pr_warn("cannot retry for NFSv4 delegation"
37392 + " for an internal unlink\n");
37395 + if (unlikely(err)) {
37396 + pr_err("%pd unlink err %d\n", dentry, err);
37401 + /* no one can touch copy_src xino */
37402 + err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
37403 + if (unlikely(err)) {
37404 + pr_err("%pd copy err %d\n", dentry, err);
37408 + goto out_dput; /* success */
37412 + file = ERR_PTR(err);
37414 + dput(path.dentry);
37417 + au_xino_unlock_dir(&ldir);
37421 +struct file *au_xino_file1(struct au_xino *xi)
37423 + struct file *file;
37424 + unsigned int u, nfile;
37427 + nfile = xi->xi_nfile;
37428 + for (u = 0; u < nfile; u++) {
37429 + file = xi->xi_file[u];
37437 +static int au_xino_file_set(struct au_xino *xi, int idx, struct file *file)
37448 + if (idx < xi->xi_nfile) {
37449 + f = xi->xi_file[idx];
37453 + p = au_kzrealloc(xi->xi_file,
37454 + sizeof(*xi->xi_file) * xi->xi_nfile,
37455 + sizeof(*xi->xi_file) * (idx + 1),
37456 + GFP_NOFS, /*may_shrink*/0);
37458 + MtxMustLock(&xi->xi_mtx);
37460 + xi->xi_nfile = idx + 1;
37468 + xi->xi_file[idx] = file;
37475 + * if @xinew->xi is not set, then create new xigen file.
37477 +struct file *au_xi_new(struct super_block *sb, struct au_xi_new *xinew)
37479 + struct file *file;
37482 + SiMustAnyLock(sb);
37484 + file = au_xino_create2(sb, xinew->base, xinew->copy_src);
37485 + if (IS_ERR(file)) {
37486 + err = PTR_ERR(file);
37487 + pr_err("%s[%d], err %d\n",
37488 + xinew->xi ? "xino" : "xigen",
37489 + xinew->idx, err);
37494 + err = au_xino_file_set(xinew->xi, xinew->idx, file);
37497 + /* todo: make xigen file an array */
37498 + /* err = au_xigen_file_set(sb, xinew->idx, file); */
37501 + if (unlikely(err))
37502 + file = ERR_PTR(err);
37508 +/* ---------------------------------------------------------------------- */
37511 + * truncate xino files
37513 +static int au_xino_do_trunc(struct super_block *sb, aufs_bindex_t bindex,
37514 + int idx, struct kstatfs *st)
37518 + struct file *file, *new_xino;
37519 + struct au_xi_new xinew = {
37524 + xinew.xi = au_sbr(sb, bindex)->br_xino;
37525 + file = au_xino_file(xinew.xi, idx);
37529 + xinew.base = &file->f_path;
37530 + err = vfs_statfs(xinew.base, st);
37531 + if (unlikely(err)) {
37532 + AuErr1("statfs err %d, ignored\n", err);
37537 + blocks = file_inode(file)->i_blocks;
37538 + pr_info("begin truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37539 + bindex, idx, (u64)blocks, st->f_bfree, st->f_blocks);
37541 + xinew.copy_src = file;
37542 + new_xino = au_xi_new(sb, &xinew);
37543 + if (IS_ERR(new_xino)) {
37544 + err = PTR_ERR(new_xino);
37545 + pr_err("xino(b%d-%d), err %d, ignored\n", bindex, idx, err);
37549 + err = vfs_statfs(&new_xino->f_path, st);
37551 + pr_info("end truncating xino(b%d-%d), ib%llu, %llu/%llu free blks\n",
37552 + bindex, idx, (u64)file_inode(new_xino)->i_blocks,
37553 + st->f_bfree, st->f_blocks);
37555 + AuErr1("statfs err %d, ignored\n", err);
37563 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex, int idx_begin)
37566 + unsigned long jiffy;
37567 + aufs_bindex_t bbot;
37568 + struct kstatfs *st;
37569 + struct au_branch *br;
37570 + struct au_xino *xi;
37573 + st = kmalloc(sizeof(*st), GFP_NOFS);
37574 + if (unlikely(!st))
37578 + bbot = au_sbbot(sb);
37579 + if (unlikely(bindex < 0 || bbot < bindex))
37584 + br = au_sbr(sb, bindex);
37585 + xi = br->br_xino;
37586 + for (i = idx_begin; !err && i < xi->xi_nfile; i++)
37587 + err = au_xino_do_trunc(sb, bindex, i, st);
37589 + au_sbi(sb)->si_xino_jiffy = jiffy;
37592 + au_kfree_rcu(st);
37597 +struct xino_do_trunc_args {
37598 + struct super_block *sb;
37599 + struct au_branch *br;
37603 +static void xino_do_trunc(void *_args)
37605 + struct xino_do_trunc_args *args = _args;
37606 + struct super_block *sb;
37607 + struct au_branch *br;
37608 + struct inode *dir;
37610 + aufs_bindex_t bindex;
37614 + dir = d_inode(sb->s_root);
37618 + si_noflush_write_lock(sb);
37619 + ii_read_lock_parent(dir);
37620 + bindex = au_br_index(sb, br->br_id);
37621 + err = au_xino_trunc(sb, bindex, idx);
37622 + ii_read_unlock(dir);
37623 + if (unlikely(err))
37624 + pr_warn("err b%d, (%d)\n", bindex, err);
37625 + atomic_dec(&br->br_xino->xi_truncating);
37626 + au_lcnt_dec(&br->br_count);
37627 + si_write_unlock(sb);
37628 + au_nwt_done(&au_sbi(sb)->si_nowait);
37629 + au_kfree_rcu(args);
37633 + * returns the index in the xi_file array whose corresponding file is necessary
37634 + * to truncate, or -1 which means no need to truncate.
37636 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
37640 + struct kstatfs st;
37641 + struct au_sbinfo *sbinfo;
37642 + struct au_xino *xi;
37643 + struct file *file;
37645 + /* todo: si_xino_expire and the ratio should be customizable */
37646 + sbinfo = au_sbi(sb);
37647 + if (time_before(jiffies,
37648 + sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
37651 + /* truncation border */
37652 + xi = br->br_xino;
37653 + for (u = 0; u < xi->xi_nfile; u++) {
37654 + file = au_xino_file(xi, u);
37658 + err = vfs_statfs(&file->f_path, &st);
37659 + if (unlikely(err)) {
37660 + AuErr1("statfs err %d, ignored\n", err);
37663 + if (div64_u64(st.f_bfree * 100, st.f_blocks)
37664 + >= AUFS_XINO_DEF_TRUNC)
37671 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
37674 + struct xino_do_trunc_args *args;
37677 + idx = xino_trunc_test(sb, br);
37681 + if (atomic_inc_return(&br->br_xino->xi_truncating) > 1)
37684 + /* lock and kfree() will be called in trunc_xino() */
37685 + args = kmalloc(sizeof(*args), GFP_NOFS);
37686 + if (unlikely(!args)) {
37687 + AuErr1("no memory\n");
37691 + au_lcnt_inc(&br->br_count);
37695 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
37697 + return; /* success */
37699 + pr_err("wkq %d\n", wkq_err);
37700 + au_lcnt_dec(&br->br_count);
37701 + au_kfree_rcu(args);
37704 + atomic_dec(&br->br_xino->xi_truncating);
37707 +/* ---------------------------------------------------------------------- */
37709 +struct au_xi_calc {
37714 +static void au_xi_calc(struct super_block *sb, ino_t h_ino,
37715 + struct au_xi_calc *calc)
37719 + maxent = au_xi_maxent(sb);
37720 + calc->idx = div64_u64_rem(h_ino, maxent, &calc->pos);
37721 + calc->pos *= sizeof(ino_t);
37724 +static int au_xino_do_new_async(struct super_block *sb, struct au_branch *br,
37725 + struct au_xi_calc *calc)
37728 + struct file *file;
37729 + struct au_xino *xi = br->br_xino;
37730 + struct au_xi_new xinew = {
37734 + SiMustAnyLock(sb);
37740 + mutex_lock(&xi->xi_mtx);
37741 + file = au_xino_file(xi, calc->idx);
37745 + file = au_xino_file(xi, /*idx*/-1);
37746 + AuDebugOn(!file);
37747 + xinew.idx = calc->idx;
37748 + xinew.base = &file->f_path;
37749 + /* xinew.copy_src = NULL; */
37750 + file = au_xi_new(sb, &xinew);
37751 + if (IS_ERR(file))
37752 + err = PTR_ERR(file);
37755 + mutex_unlock(&xi->xi_mtx);
37760 +struct au_xino_do_new_async_args {
37761 + struct super_block *sb;
37762 + struct au_branch *br;
37763 + struct au_xi_calc calc;
37767 +struct au_xi_writing {
37768 + struct hlist_bl_node node;
37769 + ino_t h_ino, ino;
37772 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37775 +static void au_xino_call_do_new_async(void *args)
37777 + struct au_xino_do_new_async_args *a = args;
37778 + struct au_branch *br;
37779 + struct super_block *sb;
37780 + struct au_sbinfo *sbi;
37781 + struct inode *root;
37782 + struct file *file;
37783 + struct au_xi_writing *del, *p;
37784 + struct hlist_bl_head *hbl;
37785 + struct hlist_bl_node *pos;
37790 + sbi = au_sbi(sb);
37791 + si_noflush_read_lock(sb);
37792 + root = d_inode(sb->s_root);
37793 + ii_read_lock_child(root);
37794 + err = au_xino_do_new_async(sb, br, &a->calc);
37795 + if (unlikely(err)) {
37796 + AuIOErr("err %d\n", err);
37800 + file = au_xino_file(br->br_xino, a->calc.idx);
37801 + AuDebugOn(!file);
37802 + err = au_xino_do_write(file, &a->calc, a->ino);
37803 + if (unlikely(err)) {
37804 + AuIOErr("err %d\n", err);
37809 + hbl = &br->br_xino->xi_writing;
37810 + hlist_bl_lock(hbl);
37811 + au_hbl_for_each(pos, hbl) {
37812 + p = container_of(pos, struct au_xi_writing, node);
37813 + if (p->ino == a->ino) {
37815 + hlist_bl_del(&p->node);
37819 + hlist_bl_unlock(hbl);
37820 + au_kfree_rcu(del);
37823 + au_lcnt_dec(&br->br_count);
37824 + ii_read_unlock(root);
37825 + si_read_unlock(sb);
37826 + au_nwt_done(&sbi->si_nowait);
37831 + * create a new xino file asynchronously
37833 +static int au_xino_new_async(struct super_block *sb, struct au_branch *br,
37834 + struct au_xi_calc *calc, ino_t ino)
37837 + struct au_xino_do_new_async_args *arg;
37840 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
37841 + if (unlikely(!arg))
37846 + arg->calc = *calc;
37848 + au_lcnt_inc(&br->br_count);
37849 + err = au_wkq_nowait(au_xino_call_do_new_async, arg, sb, AuWkq_NEST);
37850 + if (unlikely(err)) {
37851 + pr_err("wkq %d\n", err);
37852 + au_lcnt_dec(&br->br_count);
37853 + au_kfree_rcu(arg);
37861 + * read @ino from xinofile for the specified branch{@sb, @bindex}
37862 + * at the position of @h_ino.
37864 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37869 + struct au_xi_calc calc;
37870 + struct au_sbinfo *sbinfo;
37871 + struct file *file;
37872 + struct au_xino *xi;
37873 + struct hlist_bl_head *hbl;
37874 + struct hlist_bl_node *pos;
37875 + struct au_xi_writing *p;
37878 + if (!au_opt_test(au_mntflags(sb), XINO))
37879 + return 0; /* no xino */
37882 + au_xi_calc(sb, h_ino, &calc);
37883 + xi = au_sbr(sb, bindex)->br_xino;
37884 + file = au_xino_file(xi, calc.idx);
37886 + hbl = &xi->xi_writing;
37887 + hlist_bl_lock(hbl);
37888 + au_hbl_for_each(pos, hbl) {
37889 + p = container_of(pos, struct au_xi_writing, node);
37890 + if (p->h_ino == h_ino) {
37891 + AuDbg("hi%llu, i%llu, found\n",
37892 + (u64)p->h_ino, (u64)p->ino);
37897 + hlist_bl_unlock(hbl);
37899 + } else if (vfsub_f_size_read(file) < calc.pos + sizeof(*ino))
37900 + return 0; /* no xino */
37902 + sbinfo = au_sbi(sb);
37903 + sz = xino_fread(file, ino, sizeof(*ino), &calc.pos);
37904 + if (sz == sizeof(*ino))
37905 + return 0; /* success */
37908 + if (unlikely(sz >= 0)) {
37910 + AuIOErr("xino read error (%zd)\n", sz);
37915 +static int au_xino_do_write(struct file *file, struct au_xi_calc *calc,
37920 + sz = xino_fwrite(file, &ino, sizeof(ino), &calc->pos);
37921 + if (sz == sizeof(ino))
37922 + return 0; /* success */
37924 + AuIOErr("write failed (%zd)\n", sz);
37929 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
37930 + * at the position of @h_ino.
37931 + * even if @ino is zero, it is written to the xinofile and means no entry.
37932 + * if the size of the xino file on a specific filesystem exceeds the watermark,
37933 + * try truncating it.
37935 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37939 + unsigned int mnt_flags;
37940 + struct au_xi_calc calc;
37941 + struct file *file;
37942 + struct au_branch *br;
37943 + struct au_xino *xi;
37944 + struct au_xi_writing *p;
37946 + SiMustAnyLock(sb);
37948 + mnt_flags = au_mntflags(sb);
37949 + if (!au_opt_test(mnt_flags, XINO))
37952 + au_xi_calc(sb, h_ino, &calc);
37953 + br = au_sbr(sb, bindex);
37954 + xi = br->br_xino;
37955 + file = au_xino_file(xi, calc.idx);
37957 + /* store the inum pair into the list */
37958 + p = kmalloc(sizeof(*p), GFP_NOFS | __GFP_NOFAIL);
37959 + p->h_ino = h_ino;
37961 + au_hbl_add(&p->node, &xi->xi_writing);
37963 + /* create and write a new xino file asynchronously */
37964 + err = au_xino_new_async(sb, br, &calc, ino);
37966 + return 0; /* success */
37970 + err = au_xino_do_write(file, &calc, ino);
37972 + br = au_sbr(sb, bindex);
37973 + if (au_opt_test(mnt_flags, TRUNC_XINO)
37974 + && au_test_fs_trunc_xino(au_br_sb(br)))
37975 + xino_try_trunc(sb, br);
37976 + return 0; /* success */
37980 + AuIOErr("write failed (%d)\n", err);
37984 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
37987 +/* todo: unnecessary to support mmap_sem since kernel-space? */
37988 +ssize_t xino_fread(struct file *file, void *kbuf, size_t size, loff_t *pos)
37992 + const int prevent_endless = 10;
37996 + err = vfsub_read_k(file, kbuf, size, pos);
37997 + if (err == -EINTR
37998 + && !au_wkq_test()
37999 + && fatal_signal_pending(current)) {
38000 + err = xino_fread_wkq(file, kbuf, size, pos);
38001 + BUG_ON(err == -EINTR);
38003 + } while (i++ < prevent_endless
38004 + && (err == -EAGAIN || err == -EINTR));
38006 +#if 0 /* reserved for future use */
38008 + fsnotify_access(file->f_path.dentry);
38014 +struct xino_fread_args {
38016 + struct file *file;
38022 +static void call_xino_fread(void *args)
38024 + struct xino_fread_args *a = args;
38025 + *a->errp = xino_fread(a->file, a->buf, a->size, a->pos);
38028 +static ssize_t xino_fread_wkq(struct file *file, void *buf, size_t size,
38033 + struct xino_fread_args args = {
38041 + wkq_err = au_wkq_wait(call_xino_fread, &args);
38042 + if (unlikely(wkq_err))
38048 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38051 +static ssize_t do_xino_fwrite(struct file *file, void *kbuf, size_t size,
38056 + const int prevent_endless = 10;
38060 + err = vfsub_write_k(file, kbuf, size, pos);
38061 + if (err == -EINTR
38062 + && !au_wkq_test()
38063 + && fatal_signal_pending(current)) {
38064 + err = xino_fwrite_wkq(file, kbuf, size, pos);
38065 + BUG_ON(err == -EINTR);
38067 + } while (i++ < prevent_endless
38068 + && (err == -EAGAIN || err == -EINTR));
38070 +#if 0 /* reserved for future use */
38072 + fsnotify_modify(file->f_path.dentry);
38078 +struct do_xino_fwrite_args {
38080 + struct file *file;
38086 +static void call_do_xino_fwrite(void *args)
38088 + struct do_xino_fwrite_args *a = args;
38089 + *a->errp = do_xino_fwrite(a->file, a->buf, a->size, a->pos);
38092 +static ssize_t xino_fwrite_wkq(struct file *file, void *buf, size_t size,
38097 + struct do_xino_fwrite_args args = {
38106 + * it breaks RLIMIT_FSIZE and normal user's limit,
38107 + * users should care about quota and real 'filesystem full.'
38109 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
38110 + if (unlikely(wkq_err))
38116 +ssize_t xino_fwrite(struct file *file, void *buf, size_t size, loff_t *pos)
38120 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
38122 + err = do_xino_fwrite(file, buf, size, pos);
38126 + err = xino_fwrite_wkq(file, buf, size, pos);
38133 +/* ---------------------------------------------------------------------- */
38136 + * inode number bitmap
38138 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
38139 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
38143 + AuDebugOn(bit < 0 || page_bits <= bit);
38144 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
38148 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
38150 + AuDebugOn(ino < AUFS_FIRST_INO);
38151 + ino -= AUFS_FIRST_INO;
38152 + *pindex = ino / page_bits;
38153 + *bit = ino % page_bits;
38156 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
38161 + struct au_sbinfo *sbinfo;
38162 + struct file *xib;
38163 + unsigned long *p;
38165 + sbinfo = au_sbi(sb);
38166 + MtxMustLock(&sbinfo->si_xib_mtx);
38167 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
38168 + || !au_opt_test(sbinfo->si_mntflags, XINO));
38170 + if (pindex == sbinfo->si_xib_last_pindex)
38173 + xib = sbinfo->si_xib;
38174 + p = sbinfo->si_xib_buf;
38175 + pos = sbinfo->si_xib_last_pindex;
38176 + pos *= PAGE_SIZE;
38177 + sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38178 + if (unlikely(sz != PAGE_SIZE))
38182 + pos *= PAGE_SIZE;
38183 + if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
38184 + sz = xino_fread(xib, p, PAGE_SIZE, &pos);
38186 + memset(p, 0, PAGE_SIZE);
38187 + sz = xino_fwrite(xib, p, PAGE_SIZE, &pos);
38189 + if (sz == PAGE_SIZE) {
38190 + sbinfo->si_xib_last_pindex = pindex;
38191 + return 0; /* success */
38195 + AuIOErr1("write failed (%zd)\n", sz);
38202 +static void au_xib_clear_bit(struct inode *inode)
38205 + unsigned long pindex;
38206 + struct super_block *sb;
38207 + struct au_sbinfo *sbinfo;
38209 + AuDebugOn(inode->i_nlink);
38211 + sb = inode->i_sb;
38212 + xib_calc_bit(inode->i_ino, &pindex, &bit);
38213 + AuDebugOn(page_bits <= bit);
38214 + sbinfo = au_sbi(sb);
38215 + mutex_lock(&sbinfo->si_xib_mtx);
38216 + err = xib_pindex(sb, pindex);
38218 + clear_bit(bit, sbinfo->si_xib_buf);
38219 + sbinfo->si_xib_next_bit = bit;
38221 + mutex_unlock(&sbinfo->si_xib_mtx);
38224 +/* ---------------------------------------------------------------------- */
38227 + * truncate a xino bitmap file
38231 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
38235 + unsigned long pindex;
38236 + loff_t pos, pend;
38237 + struct au_sbinfo *sbinfo;
38239 + unsigned long *p;
38242 + sbinfo = au_sbi(sb);
38243 + MtxMustLock(&sbinfo->si_xib_mtx);
38244 + p = sbinfo->si_xib_buf;
38245 + pend = vfsub_f_size_read(file);
38247 + while (pos < pend) {
38248 + sz = xino_fread(file, page, PAGE_SIZE, &pos);
38250 + if (unlikely(sz <= 0))
38254 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
38255 + if (unlikely(*ino < AUFS_FIRST_INO))
38258 + xib_calc_bit(*ino, &pindex, &bit);
38259 + AuDebugOn(page_bits <= bit);
38260 + err = xib_pindex(sb, pindex);
38272 +static int xib_restore(struct super_block *sb)
38275 + unsigned int nfile;
38276 + aufs_bindex_t bindex, bbot;
38278 + struct au_branch *br;
38279 + struct au_xino *xi;
38280 + struct file *file;
38283 + page = (void *)__get_free_page(GFP_NOFS);
38284 + if (unlikely(!page))
38288 + bbot = au_sbbot(sb);
38289 + for (bindex = 0; !err && bindex <= bbot; bindex++)
38290 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0) {
38291 + br = au_sbr(sb, bindex);
38292 + xi = br->br_xino;
38293 + nfile = xi->xi_nfile;
38294 + for (i = 0; i < nfile; i++) {
38295 + file = au_xino_file(xi, i);
38297 + err = do_xib_restore(sb, file, page);
38300 + AuDbg("skip shared b%d\n", bindex);
38301 + free_page((unsigned long)page);
38307 +int au_xib_trunc(struct super_block *sb)
38312 + struct au_sbinfo *sbinfo;
38313 + unsigned long *p;
38314 + struct file *file;
38316 + SiMustWriteLock(sb);
38319 + sbinfo = au_sbi(sb);
38320 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
38323 + file = sbinfo->si_xib;
38324 + if (vfsub_f_size_read(file) <= PAGE_SIZE)
38327 + file = au_xino_create2(sb, &sbinfo->si_xib->f_path, NULL);
38328 + err = PTR_ERR(file);
38329 + if (IS_ERR(file))
38331 + fput(sbinfo->si_xib);
38332 + sbinfo->si_xib = file;
38334 + p = sbinfo->si_xib_buf;
38335 + memset(p, 0, PAGE_SIZE);
38337 + sz = xino_fwrite(sbinfo->si_xib, p, PAGE_SIZE, &pos);
38338 + if (unlikely(sz != PAGE_SIZE)) {
38340 + AuIOErr("err %d\n", err);
38346 + mutex_lock(&sbinfo->si_xib_mtx);
38347 + /* mnt_want_write() is unnecessary here */
38348 + err = xib_restore(sb);
38349 + mutex_unlock(&sbinfo->si_xib_mtx);
38355 +/* ---------------------------------------------------------------------- */
38357 +struct au_xino *au_xino_alloc(unsigned int nfile)
38359 + struct au_xino *xi;
38361 + xi = kzalloc(sizeof(*xi), GFP_NOFS);
38362 + if (unlikely(!xi))
38364 + xi->xi_nfile = nfile;
38365 + xi->xi_file = kcalloc(nfile, sizeof(*xi->xi_file), GFP_NOFS);
38366 + if (unlikely(!xi->xi_file))
38369 + xi->xi_nondir.total = 8; /* initial size */
38370 + xi->xi_nondir.array = kcalloc(xi->xi_nondir.total, sizeof(ino_t),
38372 + if (unlikely(!xi->xi_nondir.array))
38375 + spin_lock_init(&xi->xi_nondir.spin);
38376 + init_waitqueue_head(&xi->xi_nondir.wqh);
38377 + mutex_init(&xi->xi_mtx);
38378 + INIT_HLIST_BL_HEAD(&xi->xi_writing);
38379 + atomic_set(&xi->xi_truncating, 0);
38380 + kref_init(&xi->xi_kref);
38381 + goto out; /* success */
38384 + au_kfree_try_rcu(xi->xi_file);
38386 + au_kfree_rcu(xi);
38392 +static int au_xino_init(struct au_branch *br, int idx, struct file *file)
38395 + struct au_xino *xi;
38398 + xi = au_xino_alloc(idx + 1);
38399 + if (unlikely(!xi)) {
38406 + xi->xi_file[idx] = file;
38407 + AuDebugOn(br->br_xino);
38408 + br->br_xino = xi;
38414 +static void au_xino_release(struct kref *kref)
38416 + struct au_xino *xi;
38418 + unsigned long ul;
38419 + struct hlist_bl_head *hbl;
38420 + struct hlist_bl_node *pos, *n;
38421 + struct au_xi_writing *p;
38423 + xi = container_of(kref, struct au_xino, xi_kref);
38424 + for (i = 0; i < xi->xi_nfile; i++)
38425 + if (xi->xi_file[i])
38426 + fput(xi->xi_file[i]);
38427 + for (i = xi->xi_nondir.total - 1; i >= 0; i--)
38428 + AuDebugOn(xi->xi_nondir.array[i]);
38429 + mutex_destroy(&xi->xi_mtx);
38430 + hbl = &xi->xi_writing;
38431 + ul = au_hbl_count(hbl);
38432 + if (unlikely(ul)) {
38433 + pr_warn("xi_writing %lu\n", ul);
38434 + hlist_bl_lock(hbl);
38435 + hlist_bl_for_each_entry_safe(p, pos, n, hbl, node) {
38436 + hlist_bl_del(&p->node);
38437 + /* kmemleak reported au_kfree_rcu() doesn't free it */
38440 + hlist_bl_unlock(hbl);
38442 + au_kfree_try_rcu(xi->xi_file);
38443 + au_kfree_try_rcu(xi->xi_nondir.array);
38444 + au_kfree_rcu(xi);
38447 +int au_xino_put(struct au_branch *br)
38450 + struct au_xino *xi;
38453 + xi = br->br_xino;
38455 + br->br_xino = NULL;
38456 + ret = kref_put(&xi->xi_kref, au_xino_release);
38462 +/* ---------------------------------------------------------------------- */
38465 + * xino mount option handlers
38469 +static void xino_clear_xib(struct super_block *sb)
38471 + struct au_sbinfo *sbinfo;
38473 + SiMustWriteLock(sb);
38475 + sbinfo = au_sbi(sb);
38476 + if (sbinfo->si_xib)
38477 + fput(sbinfo->si_xib);
38478 + sbinfo->si_xib = NULL;
38479 + if (sbinfo->si_xib_buf)
38480 + free_page((unsigned long)sbinfo->si_xib_buf);
38481 + sbinfo->si_xib_buf = NULL;
38484 +static int au_xino_set_xib(struct super_block *sb, struct path *path)
38488 + struct au_sbinfo *sbinfo;
38489 + struct file *file;
38490 + struct super_block *xi_sb;
38492 + SiMustWriteLock(sb);
38494 + sbinfo = au_sbi(sb);
38495 + file = au_xino_create2(sb, path, sbinfo->si_xib);
38496 + err = PTR_ERR(file);
38497 + if (IS_ERR(file))
38499 + if (sbinfo->si_xib)
38500 + fput(sbinfo->si_xib);
38501 + sbinfo->si_xib = file;
38502 + xi_sb = file_inode(file)->i_sb;
38503 + sbinfo->si_ximaxent = xi_sb->s_maxbytes;
38504 + if (unlikely(sbinfo->si_ximaxent < PAGE_SIZE)) {
38506 + pr_err("s_maxbytes(%llu) on %s is too small\n",
38507 + (u64)sbinfo->si_ximaxent, au_sbtype(xi_sb));
38510 + sbinfo->si_ximaxent /= sizeof(ino_t);
38513 + if (!sbinfo->si_xib_buf)
38514 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
38515 + if (unlikely(!sbinfo->si_xib_buf))
38518 + sbinfo->si_xib_last_pindex = 0;
38519 + sbinfo->si_xib_next_bit = 0;
38520 + if (vfsub_f_size_read(file) < PAGE_SIZE) {
38522 + err = xino_fwrite(file, sbinfo->si_xib_buf, PAGE_SIZE, &pos);
38523 + if (unlikely(err != PAGE_SIZE))
38527 + goto out; /* success */
38530 + if (sbinfo->si_xib_buf)
38531 + free_page((unsigned long)sbinfo->si_xib_buf);
38532 + sbinfo->si_xib_buf = NULL;
38536 + fput(sbinfo->si_xib);
38537 + sbinfo->si_xib = NULL;
38543 +/* xino for each branch */
38544 +static void xino_clear_br(struct super_block *sb)
38546 + aufs_bindex_t bindex, bbot;
38547 + struct au_branch *br;
38549 + bbot = au_sbbot(sb);
38550 + for (bindex = 0; bindex <= bbot; bindex++) {
38551 + br = au_sbr(sb, bindex);
38557 +static void au_xino_set_br_shared(struct super_block *sb, struct au_branch *br,
38558 + aufs_bindex_t bshared)
38560 + struct au_branch *brshared;
38562 + brshared = au_sbr(sb, bshared);
38563 + AuDebugOn(!brshared->br_xino);
38564 + AuDebugOn(!brshared->br_xino->xi_file);
38565 + if (br->br_xino != brshared->br_xino) {
38566 + au_xino_get(brshared);
38568 + br->br_xino = brshared->br_xino;
38572 +struct au_xino_do_set_br {
38573 + struct au_branch *br;
38575 + aufs_bindex_t bshared;
38578 +static int au_xino_do_set_br(struct super_block *sb, struct path *path,
38579 + struct au_xino_do_set_br *args)
38582 + struct au_xi_calc calc;
38583 + struct file *file;
38584 + struct au_branch *br;
38585 + struct au_xi_new xinew = {
38590 + xinew.xi = br->br_xino;
38591 + au_xi_calc(sb, args->h_ino, &calc);
38592 + xinew.copy_src = au_xino_file(xinew.xi, calc.idx);
38593 + if (args->bshared >= 0)
38594 + /* shared xino */
38595 + au_xino_set_br_shared(sb, br, args->bshared);
38596 + else if (!xinew.xi) {
38598 + err = au_xino_init(br, calc.idx, xinew.copy_src);
38599 + if (unlikely(err))
38603 + /* force re-creating */
38604 + xinew.xi = br->br_xino;
38605 + xinew.idx = calc.idx;
38606 + mutex_lock(&xinew.xi->xi_mtx);
38607 + file = au_xi_new(sb, &xinew);
38608 + mutex_unlock(&xinew.xi->xi_mtx);
38609 + err = PTR_ERR(file);
38610 + if (IS_ERR(file))
38612 + AuDebugOn(!file);
38614 + err = au_xino_do_write(file, &calc, AUFS_ROOT_INO);
38615 + if (unlikely(err))
38623 +static int au_xino_set_br(struct super_block *sb, struct path *path)
38626 + aufs_bindex_t bindex, bbot;
38627 + struct au_xino_do_set_br args;
38628 + struct inode *inode;
38630 + SiMustWriteLock(sb);
38632 + bbot = au_sbbot(sb);
38633 + inode = d_inode(sb->s_root);
38634 + for (bindex = 0; bindex <= bbot; bindex++) {
38635 + args.h_ino = au_h_iptr(inode, bindex)->i_ino;
38636 + args.br = au_sbr(sb, bindex);
38637 + args.bshared = is_sb_shared(sb, bindex, bindex - 1);
38638 + err = au_xino_do_set_br(sb, path, &args);
38639 + if (unlikely(err))
38647 +void au_xino_clr(struct super_block *sb)
38649 + struct au_sbinfo *sbinfo;
38651 + au_xigen_clr(sb);
38652 + xino_clear_xib(sb);
38653 + xino_clear_br(sb);
38654 + dbgaufs_brs_del(sb, 0);
38655 + sbinfo = au_sbi(sb);
38656 + /* lvalue, do not call au_mntflags() */
38657 + au_opt_clr(sbinfo->si_mntflags, XINO);
38660 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xiopt, int remount)
38663 + struct dentry *dentry, *parent, *cur_dentry, *cur_parent;
38664 + struct qstr *dname, *cur_name;
38665 + struct file *cur_xino;
38666 + struct au_sbinfo *sbinfo;
38667 + struct path *path, *cur_path;
38669 + SiMustWriteLock(sb);
38672 + sbinfo = au_sbi(sb);
38673 + path = &xiopt->file->f_path;
38674 + dentry = path->dentry;
38675 + parent = dget_parent(dentry);
38678 + cur_xino = sbinfo->si_xib;
38680 + cur_path = &cur_xino->f_path;
38681 + cur_dentry = cur_path->dentry;
38682 + cur_parent = dget_parent(cur_dentry);
38683 + cur_name = &cur_dentry->d_name;
38684 + dname = &dentry->d_name;
38685 + skip = (cur_parent == parent
38686 + && au_qstreq(dname, cur_name));
38687 + dput(cur_parent);
38693 + au_opt_set(sbinfo->si_mntflags, XINO);
38694 + err = au_xino_set_xib(sb, path);
38695 + /* si_x{read,write} are set */
38697 + err = au_xigen_set(sb, path);
38699 + err = au_xino_set_br(sb, path);
38701 + dbgaufs_brs_add(sb, 0, /*topdown*/1);
38702 + goto out; /* success */
38706 + AuIOErr("failed setting xino(%d).\n", err);
38715 + * create a xinofile at the default place/path.
38717 +struct file *au_xino_def(struct super_block *sb)
38719 + struct file *file;
38721 + struct au_branch *br;
38722 + struct super_block *h_sb;
38723 + struct path path;
38724 + aufs_bindex_t bbot, bindex, bwr;
38727 + bbot = au_sbbot(sb);
38729 + for (bindex = 0; bindex <= bbot; bindex++) {
38730 + br = au_sbr(sb, bindex);
38731 + if (au_br_writable(br->br_perm)
38732 + && !au_test_fs_bad_xino(au_br_sb(br))) {
38739 + file = ERR_PTR(-ENOMEM);
38740 + page = (void *)__get_free_page(GFP_NOFS);
38741 + if (unlikely(!page))
38743 + path.mnt = au_br_mnt(br);
38744 + path.dentry = au_h_dptr(sb->s_root, bwr);
38745 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
38746 + file = (void *)p;
38747 + if (!IS_ERR(p)) {
38748 + strcat(p, "/" AUFS_XINO_FNAME);
38749 + AuDbg("%s\n", p);
38750 + file = au_xino_create(sb, p, /*silent*/0, /*wbrtop*/1);
38752 + free_page((unsigned long)page);
38754 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0,
38756 + if (IS_ERR(file))
38758 + h_sb = file->f_path.dentry->d_sb;
38759 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
38760 + pr_err("xino doesn't support %s(%s)\n",
38761 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
38763 + file = ERR_PTR(-EINVAL);
38771 +/* ---------------------------------------------------------------------- */
38774 + * initialize the xinofile for the specified branch @br
38775 + * at the place/path where @base_file indicates.
38776 + * test whether another branch is on the same filesystem or not,
38777 + * if found then share the xinofile with another branch.
38779 +int au_xino_init_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
38780 + struct path *base)
38783 + struct au_xino_do_set_br args = {
38788 + args.bshared = sbr_find_shared(sb, /*btop*/0, au_sbbot(sb),
38790 + err = au_xino_do_set_br(sb, base, &args);
38791 + if (unlikely(err))
38797 +/* ---------------------------------------------------------------------- */
38800 + * get an unused inode number from bitmap
38802 +ino_t au_xino_new_ino(struct super_block *sb)
38805 + unsigned long *p, pindex, ul, pend;
38806 + struct au_sbinfo *sbinfo;
38807 + struct file *file;
38808 + int free_bit, err;
38810 + if (!au_opt_test(au_mntflags(sb), XINO))
38811 + return iunique(sb, AUFS_FIRST_INO);
38813 + sbinfo = au_sbi(sb);
38814 + mutex_lock(&sbinfo->si_xib_mtx);
38815 + p = sbinfo->si_xib_buf;
38816 + free_bit = sbinfo->si_xib_next_bit;
38817 + if (free_bit < page_bits && !test_bit(free_bit, p))
38818 + goto out; /* success */
38819 + free_bit = find_first_zero_bit(p, page_bits);
38820 + if (free_bit < page_bits)
38821 + goto out; /* success */
38823 + pindex = sbinfo->si_xib_last_pindex;
38824 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
38825 + err = xib_pindex(sb, ul);
38826 + if (unlikely(err))
38828 + free_bit = find_first_zero_bit(p, page_bits);
38829 + if (free_bit < page_bits)
38830 + goto out; /* success */
38833 + file = sbinfo->si_xib;
38834 + pend = vfsub_f_size_read(file) / PAGE_SIZE;
38835 + for (ul = pindex + 1; ul <= pend; ul++) {
38836 + err = xib_pindex(sb, ul);
38837 + if (unlikely(err))
38839 + free_bit = find_first_zero_bit(p, page_bits);
38840 + if (free_bit < page_bits)
38841 + goto out; /* success */
38846 + set_bit(free_bit, p);
38847 + sbinfo->si_xib_next_bit = free_bit + 1;
38848 + pindex = sbinfo->si_xib_last_pindex;
38849 + mutex_unlock(&sbinfo->si_xib_mtx);
38850 + ino = xib_calc_ino(pindex, free_bit);
38851 + AuDbg("i%lu\n", (unsigned long)ino);
38854 + mutex_unlock(&sbinfo->si_xib_mtx);
38859 +/* for s_op->delete_inode() */
38860 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
38863 + unsigned int mnt_flags;
38864 + aufs_bindex_t bindex, bbot, bi;
38865 + unsigned char try_trunc;
38866 + struct au_iinfo *iinfo;
38867 + struct super_block *sb;
38868 + struct au_hinode *hi;
38869 + struct inode *h_inode;
38870 + struct au_branch *br;
38871 + struct au_xi_calc calc;
38872 + struct file *file;
38874 + AuDebugOn(au_is_bad_inode(inode));
38876 + sb = inode->i_sb;
38877 + mnt_flags = au_mntflags(sb);
38878 + if (!au_opt_test(mnt_flags, XINO)
38879 + || inode->i_ino == AUFS_ROOT_INO)
38883 + au_xigen_inc(inode);
38884 + au_xib_clear_bit(inode);
38887 + iinfo = au_ii(inode);
38888 + bindex = iinfo->ii_btop;
38892 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
38893 + hi = au_hinode(iinfo, bindex);
38894 + bbot = iinfo->ii_bbot;
38895 + for (; bindex <= bbot; bindex++, hi++) {
38896 + h_inode = hi->hi_inode;
38898 + || (!unlinked && h_inode->i_nlink))
38901 + /* inode may not be revalidated */
38902 + bi = au_br_index(sb, hi->hi_id);
38906 + br = au_sbr(sb, bi);
38907 + au_xi_calc(sb, h_inode->i_ino, &calc);
38908 + file = au_xino_file(br->br_xino, calc.idx);
38909 + if (IS_ERR_OR_NULL(file))
38912 + err = au_xino_do_write(file, &calc, /*ino*/0);
38913 + if (!err && try_trunc
38914 + && au_test_fs_trunc_xino(au_br_sb(br)))
38915 + xino_try_trunc(sb, br);
38919 +/* ---------------------------------------------------------------------- */
38921 +static int au_xinondir_find(struct au_xino *xi, ino_t h_ino)
38923 + int found, total, i;
38926 + total = xi->xi_nondir.total;
38927 + for (i = 0; i < total; i++) {
38928 + if (xi->xi_nondir.array[i] != h_ino)
38937 +static int au_xinondir_expand(struct au_xino *xi)
38942 + BUILD_BUG_ON(KMALLOC_MAX_SIZE > INT_MAX);
38945 + sz = xi->xi_nondir.total * sizeof(ino_t);
38946 + if (unlikely(sz > KMALLOC_MAX_SIZE / 2))
38948 + p = au_kzrealloc(xi->xi_nondir.array, sz, sz << 1, GFP_ATOMIC,
38949 + /*may_shrink*/0);
38951 + xi->xi_nondir.array = p;
38952 + xi->xi_nondir.total <<= 1;
38953 + AuDbg("xi_nondir.total %d\n", xi->xi_nondir.total);
38961 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
38962 + ino_t h_ino, int idx)
38964 + struct au_xino *xi;
38966 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
38967 + xi = au_sbr(sb, bindex)->br_xino;
38968 + AuDebugOn(idx < 0 || xi->xi_nondir.total <= idx);
38970 + spin_lock(&xi->xi_nondir.spin);
38971 + AuDebugOn(xi->xi_nondir.array[idx] != h_ino);
38972 + xi->xi_nondir.array[idx] = 0;
38973 + spin_unlock(&xi->xi_nondir.spin);
38974 + wake_up_all(&xi->xi_nondir.wqh);
38977 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
38980 + int err, found, empty;
38981 + struct au_xino *xi;
38985 + if (!au_opt_test(au_mntflags(sb), XINO))
38986 + goto out; /* no xino */
38988 + xi = au_sbr(sb, bindex)->br_xino;
38991 + spin_lock(&xi->xi_nondir.spin);
38992 + found = au_xinondir_find(xi, h_ino);
38993 + if (found == -1) {
38994 + empty = au_xinondir_find(xi, /*h_ino*/0);
38995 + if (empty == -1) {
38996 + empty = xi->xi_nondir.total;
38997 + err = au_xinondir_expand(xi);
38998 + if (unlikely(err))
39001 + xi->xi_nondir.array[empty] = h_ino;
39004 + spin_unlock(&xi->xi_nondir.spin);
39005 + wait_event(xi->xi_nondir.wqh,
39006 + xi->xi_nondir.array[found] != h_ino);
39011 + spin_unlock(&xi->xi_nondir.spin);
39016 +/* ---------------------------------------------------------------------- */
39018 +int au_xino_path(struct seq_file *seq, struct file *file)
39022 + err = au_seq_path(seq, &file->f_path);
39023 + if (unlikely(err))
39026 +#define Deleted "\\040(deleted)"
39027 + seq->count -= sizeof(Deleted) - 1;
39028 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
39029 + sizeof(Deleted) - 1));
39035 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
39036 --- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
39037 +++ linux/include/uapi/linux/aufs_type.h 2023-07-04 23:43:55.849649534 +0200
39039 +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
39041 + * Copyright (C) 2005-2022 Junjiro R. Okajima
39043 + * This program is free software; you can redistribute it and/or modify
39044 + * it under the terms of the GNU General Public License as published by
39045 + * the Free Software Foundation; either version 2 of the License, or
39046 + * (at your option) any later version.
39048 + * This program is distributed in the hope that it will be useful,
39049 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
39050 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
39051 + * GNU General Public License for more details.
39053 + * You should have received a copy of the GNU General Public License
39054 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
39057 +#ifndef __AUFS_TYPE_H__
39058 +#define __AUFS_TYPE_H__
39060 +#define AUFS_NAME "aufs"
39064 + * define it before including all other headers.
39065 + * sched.h may use pr_* macros before defining "current", so define the
39066 + * no-current version first, and re-define later.
39068 +#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
39069 +#include <linux/sched.h>
39071 +#define pr_fmt(fmt) \
39072 + AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
39073 + (int)sizeof(current->comm), current->comm, current->pid
39074 +#include <linux/limits.h>
39076 +#include <stdint.h>
39077 +#include <sys/types.h>
39078 +#include <limits.h>
39079 +#endif /* __KERNEL__ */
39081 +#define AUFS_VERSION "6.x-rcN-20230515"
39083 +/* todo? move this to linux-2.6.19/include/magic.h */
39084 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
39086 +/* ---------------------------------------------------------------------- */
39089 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
39090 +typedef int8_t aufs_bindex_t;
39091 +#define AUFS_BRANCH_MAX 127
39093 +typedef int16_t aufs_bindex_t;
39094 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
39095 +#define AUFS_BRANCH_MAX 511
39096 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
39097 +#define AUFS_BRANCH_MAX 1023
39098 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
39099 +#define AUFS_BRANCH_MAX 32767
39103 +#ifndef AUFS_BRANCH_MAX
39104 +#error unknown CONFIG_AUFS_BRANCH_MAX value
39106 +#endif /* __KERNEL__ */
39108 +/* ---------------------------------------------------------------------- */
39110 +#define AUFS_FSTYPE AUFS_NAME
39112 +#define AUFS_ROOT_INO 2
39113 +#define AUFS_FIRST_INO 11
39115 +#define AUFS_WH_PFX ".wh."
39116 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
39117 +#define AUFS_WH_TMP_LEN 4
39118 +/* a limit for rmdir/rename a dir and copyup */
39119 +#define AUFS_MAX_NAMELEN (NAME_MAX \
39120 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
39122 + - AUFS_WH_TMP_LEN) /* hex */
39123 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
39124 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
39125 +#define AUFS_XINO_DEF_SEC 30 /* seconds */
39126 +#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
39127 +#define AUFS_DIRWH_DEF 3
39128 +#define AUFS_RDCACHE_DEF 10 /* seconds */
39129 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
39130 +#define AUFS_RDBLK_DEF 512 /* bytes */
39131 +#define AUFS_RDHASH_DEF 32
39132 +#define AUFS_WKQ_NAME AUFS_NAME "d"
39133 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
39134 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
39135 +#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
39136 +#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
39138 +/* pseudo-link maintenace under /proc */
39139 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
39140 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
39141 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
39143 +/* dirren, renamed dir */
39144 +#define AUFS_DR_INFO_PFX AUFS_WH_PFX ".dr."
39145 +#define AUFS_DR_BRHINO_NAME AUFS_WH_PFX "hino"
39146 +/* whiteouted doubly */
39147 +#define AUFS_WH_DR_INFO_PFX AUFS_WH_PFX AUFS_DR_INFO_PFX
39148 +#define AUFS_WH_DR_BRHINO AUFS_WH_PFX AUFS_DR_BRHINO_NAME
39150 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
39151 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
39153 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
39154 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
39155 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
39157 +/* doubly whiteouted */
39158 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
39159 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
39160 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
39162 +/* branch permissions and attributes */
39163 +#define AUFS_BRPERM_RW "rw"
39164 +#define AUFS_BRPERM_RO "ro"
39165 +#define AUFS_BRPERM_RR "rr"
39166 +#define AUFS_BRATTR_COO_REG "coo_reg"
39167 +#define AUFS_BRATTR_COO_ALL "coo_all"
39168 +#define AUFS_BRATTR_FHSM "fhsm"
39169 +#define AUFS_BRATTR_UNPIN "unpin"
39170 +#define AUFS_BRATTR_ICEX "icex"
39171 +#define AUFS_BRATTR_ICEX_SEC "icexsec"
39172 +#define AUFS_BRATTR_ICEX_SYS "icexsys"
39173 +#define AUFS_BRATTR_ICEX_TR "icextr"
39174 +#define AUFS_BRATTR_ICEX_USR "icexusr"
39175 +#define AUFS_BRATTR_ICEX_OTH "icexoth"
39176 +#define AUFS_BRRATTR_WH "wh"
39177 +#define AUFS_BRWATTR_NLWH "nolwh"
39178 +#define AUFS_BRWATTR_MOO "moo"
39180 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
39181 +#define AuBrPerm_RO (1 << 1) /* readonly */
39182 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
39183 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
39185 +#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
39186 +#define AuBrAttr_COO_ALL (1 << 4)
39187 +#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
39189 +#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
39190 +#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
39191 + branch. meaningless since
39192 + linux-3.18-rc1 */
39194 +/* ignore error in copying XATTR */
39195 +#define AuBrAttr_ICEX_SEC (1 << 7)
39196 +#define AuBrAttr_ICEX_SYS (1 << 8)
39197 +#define AuBrAttr_ICEX_TR (1 << 9)
39198 +#define AuBrAttr_ICEX_USR (1 << 10)
39199 +#define AuBrAttr_ICEX_OTH (1 << 11)
39200 +#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
39201 + | AuBrAttr_ICEX_SYS \
39202 + | AuBrAttr_ICEX_TR \
39203 + | AuBrAttr_ICEX_USR \
39204 + | AuBrAttr_ICEX_OTH)
39206 +#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
39207 +#define AuBrRAttr_Mask AuBrRAttr_WH
39209 +#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
39210 +#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
39211 +#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
39213 +#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
39215 +/* #warning test userspace */
39217 +#ifndef CONFIG_AUFS_FHSM
39218 +#undef AuBrAttr_FHSM
39219 +#define AuBrAttr_FHSM 0
39221 +#ifndef CONFIG_AUFS_XATTR
39222 +#undef AuBrAttr_ICEX
39223 +#define AuBrAttr_ICEX 0
39224 +#undef AuBrAttr_ICEX_SEC
39225 +#define AuBrAttr_ICEX_SEC 0
39226 +#undef AuBrAttr_ICEX_SYS
39227 +#define AuBrAttr_ICEX_SYS 0
39228 +#undef AuBrAttr_ICEX_TR
39229 +#define AuBrAttr_ICEX_TR 0
39230 +#undef AuBrAttr_ICEX_USR
39231 +#define AuBrAttr_ICEX_USR 0
39232 +#undef AuBrAttr_ICEX_OTH
39233 +#define AuBrAttr_ICEX_OTH 0
39237 +/* the longest combination */
39238 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
39239 +#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
39240 + "+" AUFS_BRATTR_COO_REG \
39241 + "+" AUFS_BRATTR_FHSM \
39242 + "+" AUFS_BRATTR_UNPIN \
39243 + "+" AUFS_BRATTR_ICEX_SEC \
39244 + "+" AUFS_BRATTR_ICEX_SYS \
39245 + "+" AUFS_BRATTR_ICEX_USR \
39246 + "+" AUFS_BRATTR_ICEX_OTH \
39247 + "+" AUFS_BRWATTR_NLWH)
39250 + char a[AuBrPermStrSz];
39251 +} au_br_perm_str_t;
39253 +static inline int au_br_writable(int brperm)
39255 + return brperm & AuBrPerm_RW;
39258 +static inline int au_br_whable(int brperm)
39260 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
39263 +static inline int au_br_wh_linkable(int brperm)
39265 + return !(brperm & AuBrWAttr_NoLinkWH);
39268 +static inline int au_br_cmoo(int brperm)
39270 + return brperm & AuBrAttr_CMOO_Mask;
39273 +static inline int au_br_fhsm(int brperm)
39275 + return brperm & AuBrAttr_FHSM;
39278 +/* ---------------------------------------------------------------------- */
39282 + /* readdir in userspace */
39286 + AuCtl_WBR_FD, /* pathconf wrapper */
39287 + AuCtl_IBUSY, /* busy inode */
39288 + AuCtl_MVDOWN, /* move-down */
39289 + AuCtl_BR, /* info about branches */
39290 + AuCtl_FHSM_FD /* connection for fhsm */
39293 +/* borrowed from linux/include/linux/kernel.h */
39295 +#ifdef _GNU_SOURCE
39296 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
39298 +#define ALIGN(x, a) (((x) + (a) - 1) & ~((a) - 1))
39300 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
39303 +/* borrowed from linux/include/linux/compiler-gcc3.h */
39305 +#define __aligned(x) __attribute__((aligned(x)))
39310 +#define __packed __attribute__((packed))
39314 +struct au_rdu_cookie {
39319 + uint32_t generation;
39322 +struct au_rdu_ent {
39331 +static inline int au_rdu_len(int nlen)
39333 + /* include the terminating NULL */
39334 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
39335 + sizeof(uint64_t));
39338 +union au_rdu_ent_ul {
39339 + struct au_rdu_ent __user *e;
39351 + uint64_t sz; /* AuCtl_RDU */
39352 + uint64_t nent; /* AuCtl_RDU_INO */
39354 + union au_rdu_ent_ul ent;
39355 + uint16_t verify[AufsCtlRduV_End];
39357 + /* input/output */
39361 + union au_rdu_ent_ul tail;
39362 + /* number of entries which were added in a single call */
39367 + struct au_rdu_cookie cookie;
39370 +/* ---------------------------------------------------------------------- */
39372 +/* dirren. the branch is identified by the filename who contains this */
39373 +struct au_drinfo {
39376 + uint8_t oldnamelen;
39377 + uint64_t _padding;
39379 + uint8_t oldname[];
39382 +struct au_drinfo_fdata {
39384 + struct au_drinfo drinfo;
39387 +#define AUFS_DRINFO_MAGIC_V1 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x01)
39389 +#define AUFS_DRINFO_MAGIC_V2 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x02)
39391 +/* ---------------------------------------------------------------------- */
39393 +struct aufs_wbr_fd {
39398 +/* ---------------------------------------------------------------------- */
39400 +struct aufs_ibusy {
39401 + uint64_t ino, h_ino;
39405 +/* ---------------------------------------------------------------------- */
39407 +/* error code for move-down */
39408 +/* the actual message strings are implemented in aufs-util.git */
39410 + EAU_MVDOWN_OPAQUE = 1,
39411 + EAU_MVDOWN_WHITEOUT,
39412 + EAU_MVDOWN_UPPER,
39413 + EAU_MVDOWN_BOTTOM,
39414 + EAU_MVDOWN_NOUPPER,
39415 + EAU_MVDOWN_NOLOWERBR,
39419 +/* flags for move-down */
39420 +#define AUFS_MVDOWN_DMSG 1
39421 +#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
39422 +#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
39423 +#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
39424 +#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
39425 +#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
39426 +#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
39427 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
39428 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
39429 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
39430 +#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
39431 +#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
39432 +#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
39434 +/* index for move-down */
39436 + AUFS_MVDOWN_UPPER,
39437 + AUFS_MVDOWN_LOWER,
39438 + AUFS_MVDOWN_NARRAY
39442 + * additional info of move-down
39443 + * number of free blocks and inodes.
39444 + * subset of struct kstatfs, but smaller and always 64bit.
39446 +struct aufs_stfs {
39447 + uint64_t f_blocks;
39448 + uint64_t f_bavail;
39449 + uint64_t f_files;
39450 + uint64_t f_ffree;
39453 +struct aufs_stbr {
39454 + int16_t brid; /* optional input */
39455 + int16_t bindex; /* output */
39456 + struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
39459 +struct aufs_mvdown {
39460 + uint32_t flags; /* input/output */
39461 + struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
39462 + int8_t au_errno; /* output */
39465 +/* ---------------------------------------------------------------------- */
39467 +union aufs_brinfo {
39468 + /* PATH_MAX may differ between kernel-space and user-space */
39469 + char _spacer[4096];
39477 +/* ---------------------------------------------------------------------- */
39479 +#define AuCtlType 'A'
39480 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
39481 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
39482 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
39483 + struct aufs_wbr_fd)
39484 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
39485 +#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
39486 + struct aufs_mvdown)
39487 +#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
39488 +#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
39490 +#endif /* __AUFS_TYPE_H__ */
39491 SPDX-License-Identifier: GPL-2.0
39492 aufs6.x-rcN loopback patch
39494 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
39495 index 131294601819..56016fc9c9d8 100644
39496 --- a/drivers/block/loop.c
39497 +++ b/drivers/block/loop.c
39498 @@ -54,7 +54,7 @@ struct loop_device {
39500 char lo_file_name[LO_NAME_SIZE];
39502 - struct file * lo_backing_file;
39503 + struct file *lo_backing_file, *lo_backing_virt_file;
39504 struct block_device *lo_device;
39506 gfp_t old_gfp_mask;
39507 @@ -510,6 +510,15 @@ static inline void loop_update_dio(struct loop_device *lo)
39511 +static struct file *loop_real_file(struct file *file)
39513 + struct file *f = NULL;
39515 + if (file->f_path.dentry->d_sb->s_op->real_loop)
39516 + f = file->f_path.dentry->d_sb->s_op->real_loop(file);
39520 static void loop_reread_partitions(struct loop_device *lo)
39523 @@ -567,6 +576,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39525 struct file *file = fget(arg);
39526 struct file *old_file;
39527 + struct file *f, *virt_file = NULL, *old_virt_file;
39531 @@ -590,11 +600,19 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39532 if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
39535 + f = loop_real_file(file);
39537 + virt_file = file;
39542 error = loop_validate_file(file, bdev);
39546 old_file = lo->lo_backing_file;
39547 + old_virt_file = lo->lo_backing_virt_file;
39551 @@ -607,6 +625,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39552 blk_mq_freeze_queue(lo->lo_queue);
39553 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
39554 lo->lo_backing_file = file;
39555 + lo->lo_backing_virt_file = virt_file;
39556 lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
39557 mapping_set_gfp_mask(file->f_mapping,
39558 lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39559 @@ -629,6 +648,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39563 + if (old_virt_file)
39564 + fput(old_virt_file);
39566 loop_reread_partitions(lo);
39568 @@ -642,6 +663,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
39569 loop_global_unlock(lo, is_loop);
39577 @@ -1013,6 +1036,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39578 const struct loop_config *config)
39580 struct file *file = fget(config->fd);
39581 + struct file *f, *virt_file = NULL;
39582 struct inode *inode;
39583 struct address_space *mapping;
39585 @@ -1028,6 +1052,13 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39586 /* This is safe, since we have a reference from open(). */
39587 __module_get(THIS_MODULE);
39589 + f = loop_real_file(file);
39591 + virt_file = file;
39597 * If we don't hold exclusive handle for the device, upgrade to it
39598 * here to avoid changing device under exclusive owner.
39599 @@ -1091,6 +1122,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39600 lo->use_dio = lo->lo_flags & LO_FLAGS_DIRECT_IO;
39601 lo->lo_device = bdev;
39602 lo->lo_backing_file = file;
39603 + lo->lo_backing_virt_file = virt_file;
39604 lo->old_gfp_mask = mapping_gfp_mask(mapping);
39605 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
39607 @@ -1146,6 +1178,8 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39608 bd_abort_claiming(bdev, loop_configure);
39613 /* This is safe: open() is still holding a reference. */
39614 module_put(THIS_MODULE);
39616 @@ -1154,6 +1188,7 @@ static int loop_configure(struct loop_device *lo, fmode_t mode,
39617 static void __loop_clr_fd(struct loop_device *lo, bool release)
39620 + struct file *virt_filp = lo->lo_backing_virt_file;
39621 gfp_t gfp = lo->old_gfp_mask;
39623 if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags))
39624 @@ -1170,6 +1205,7 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39625 spin_lock_irq(&lo->lo_lock);
39626 filp = lo->lo_backing_file;
39627 lo->lo_backing_file = NULL;
39628 + lo->lo_backing_virt_file = NULL;
39629 spin_unlock_irq(&lo->lo_lock);
39631 lo->lo_device = NULL;
39632 @@ -1232,6 +1268,8 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
39633 * fput can take open_mutex which is usually taken before lo_mutex.
39640 static int loop_clr_fd(struct loop_device *lo)
39641 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
39642 index 1ed300047a41..ea63191d6473 100644
39643 --- a/fs/aufs/f_op.c
39644 +++ b/fs/aufs/f_op.c
39645 @@ -309,7 +309,7 @@ static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
39646 if (IS_ERR(h_file))
39649 - if (au_test_loopback_kthread()) {
39650 + if (0 && au_test_loopback_kthread()) {
39651 au_warn_loopback(h_file->f_path.dentry->d_sb);
39652 if (file->f_mapping != h_file->f_mapping) {
39653 file->f_mapping = h_file->f_mapping;
39654 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
39655 index 58043e31e5f3..e2bfae6f9d59 100644
39656 --- a/fs/aufs/loop.c
39657 +++ b/fs/aufs/loop.c
39658 @@ -133,3 +133,19 @@ void au_loopback_fin(void)
39659 symbol_put(loop_backing_file);
39660 au_kfree_try_rcu(au_warn_loopback_array);
39663 +/* ---------------------------------------------------------------------- */
39665 +/* support the loopback block device insude aufs */
39667 +struct file *aufs_real_loop(struct file *file)
39671 + BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
39672 + fi_read_lock(file);
39673 + f = au_hf_top(file);
39674 + fi_read_unlock(file);
39678 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
39679 index 03d4908a6c03..34d356e181d5 100644
39680 --- a/fs/aufs/loop.h
39681 +++ b/fs/aufs/loop.h
39682 @@ -26,6 +26,8 @@ void au_warn_loopback(struct super_block *h_sb);
39684 int au_loopback_init(void);
39685 void au_loopback_fin(void);
39687 +struct file *aufs_real_loop(struct file *file);
39689 AuStub(struct file *, loop_backing_file, return NULL, struct super_block *sb)
39691 @@ -36,6 +38,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
39693 AuStubInt0(au_loopback_init, void)
39694 AuStubVoid(au_loopback_fin, void)
39696 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
39697 #endif /* BLK_DEV_LOOP */
39699 #endif /* __KERNEL__ */
39700 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
39701 index 81922d4faf54..c8a62c267d72 100644
39702 --- a/fs/aufs/super.c
39703 +++ b/fs/aufs/super.c
39704 @@ -758,7 +758,10 @@ const struct super_operations aufs_sop = {
39705 .show_options = aufs_show_options,
39706 .statfs = aufs_statfs,
39707 .put_super = aufs_put_super,
39708 - .sync_fs = aufs_sync_fs
39709 + .sync_fs = aufs_sync_fs,
39710 +#ifdef CONFIG_AUFS_BDEV_LOOP
39711 + .real_loop = aufs_real_loop
39715 /* ---------------------------------------------------------------------- */
39716 diff --git a/include/linux/fs.h b/include/linux/fs.h
39717 index 1e3054e5367d..2a780fcbdc75 100644
39718 --- a/include/linux/fs.h
39719 +++ b/include/linux/fs.h
39720 @@ -1934,6 +1934,10 @@ struct super_operations {
39721 struct shrink_control *);
39722 long (*free_cached_objects)(struct super_block *,
39723 struct shrink_control *);
39724 +#if IS_ENABLED(CONFIG_BLK_DEV_LOOP) || IS_ENABLED(CONFIG_BLK_DEV_LOOP_MODULE)
39726 + struct file *(*real_loop)(struct file *);